diff --git a/internal/cgen/base/all-impl.c b/internal/cgen/base/all-impl.c index 0ba6e6e2..894207d8 100644 --- a/internal/cgen/base/all-impl.c +++ b/internal/cgen/base/all-impl.c @@ -115,6 +115,20 @@ const uint64_t wuffs_base__low_bits_mask__u64[65] = { // !! INSERT base/strconv-impl.c. +#endif // !defined(WUFFS_CONFIG__MODULES) || + // defined(WUFFS_CONFIG__MODULE__BASE) + +#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) + +// !! INSERT base/f64conv-submodule.c. + +#endif // !defined(WUFFS_CONFIG__MODULES) || + // defined(WUFFS_CONFIG__MODULE__BASE) + +#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) + +// !! INSERT base/pixconv-submodule.c. + #endif // !defined(WUFFS_CONFIG__MODULES) || // defined(WUFFS_CONFIG__MODULE__BASE) diff --git a/internal/cgen/base/f64conv-submodule.c b/internal/cgen/base/f64conv-submodule.c new file mode 100644 index 00000000..3bea12ce --- /dev/null +++ b/internal/cgen/base/f64conv-submodule.c @@ -0,0 +1,1264 @@ +// After editing this file, run "go generate" in the parent directory. + +// Copyright 2020 The Wuffs Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// ---------------- IEEE 754 Floating Point + +#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE 1023 +#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION 500 + +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL is the largest N +// such that ((10 << N) < (1 << 64)). +#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL 60 + +// wuffs_base__private_implementation__high_prec_dec (abbreviated as HPD) is a +// fixed precision floating point decimal number, augmented with ±infinity +// values, but it cannot represent NaN (Not a Number). +// +// "High precision" means that the mantissa holds 500 decimal digits. 500 is +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION. +// +// An HPD isn't for general purpose arithmetic, only for conversions to and +// from IEEE 754 double-precision floating point, where the largest and +// smallest positive, finite values are approximately 1.8e+308 and 4.9e-324. +// HPD exponents above +1023 mean infinity, below -1023 mean zero. The ±1023 +// bounds are further away from zero than ±(324 + 500), where 500 and 1023 is +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION and +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE. +// +// digits[.. num_digits] are the number's digits in big-endian order. The +// uint8_t values are in the range [0 ..= 9], not ['0' ..= '9'], where e.g. '7' +// is the ASCII value 0x37. +// +// decimal_point is the index (within digits) of the decimal point. It may be +// negative or be larger than num_digits, in which case the explicit digits are +// padded with implicit zeroes. +// +// For example, if num_digits is 3 and digits is "\x07\x08\x09": +// - A decimal_point of -2 means ".00789" +// - A decimal_point of -1 means ".0789" +// - A decimal_point of +0 means ".789" +// - A decimal_point of +1 means "7.89" +// - A decimal_point of +2 means "78.9" +// - A decimal_point of +3 means "789." +// - A decimal_point of +4 means "7890." +// - A decimal_point of +5 means "78900." +// +// As above, a decimal_point higher than +1023 means that the overall value is +// infinity, lower than -1023 means zero. +// +// negative is a sign bit. An HPD can distinguish positive and negative zero. +// +// truncated is whether there are more than +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION digits, and at +// least one of those extra digits are non-zero. The existence of long-tail +// digits can affect rounding. +// +// The "all fields are zero" value is valid, and represents the number +0. +typedef struct { + uint32_t num_digits; + int32_t decimal_point; + bool negative; + bool truncated; + uint8_t digits[WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION]; +} wuffs_base__private_implementation__high_prec_dec; + +// wuffs_base__private_implementation__high_prec_dec__trim trims trailing +// zeroes from the h->digits[.. h->num_digits] slice. They have no benefit, +// since we explicitly track h->decimal_point. +// +// Preconditions: +// - h is non-NULL. +static inline void // +wuffs_base__private_implementation__high_prec_dec__trim( + wuffs_base__private_implementation__high_prec_dec* h) { + while ((h->num_digits > 0) && (h->digits[h->num_digits - 1] == 0)) { + h->num_digits--; + } +} + +static wuffs_base__status // +wuffs_base__private_implementation__high_prec_dec__parse( + wuffs_base__private_implementation__high_prec_dec* h, + wuffs_base__slice_u8 s) { + if (!h) { + return wuffs_base__make_status(wuffs_base__error__bad_receiver); + } + h->num_digits = 0; + h->decimal_point = 0; + h->negative = false; + h->truncated = false; + + uint8_t* p = s.ptr; + uint8_t* q = s.ptr + s.len; + + for (; (p < q) && (*p == '_'); p++) { + } + if (p >= q) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + + // Parse sign. + do { + if (*p == '+') { + p++; + } else if (*p == '-') { + h->negative = true; + p++; + } else { + break; + } + for (; (p < q) && (*p == '_'); p++) { + } + } while (0); + + // Parse digits. + uint32_t nd = 0; + int32_t dp = 0; + bool saw_digits = false; + bool saw_non_zero_digits = false; + bool saw_dot = false; + for (; p < q; p++) { + if (*p == '_') { + // No-op. + + } else if ((*p == '.') || (*p == ',')) { + // As per https://en.wikipedia.org/wiki/Decimal_separator, both '.' or + // ',' are commonly used. We just parse either, regardless of LOCALE. + if (saw_dot) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + saw_dot = true; + dp = (int32_t)nd; + + } else if ('0' == *p) { + if (!saw_dot && !saw_non_zero_digits && saw_digits) { + // We don't allow unnecessary leading zeroes: "000123" or "0644". + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + saw_digits = true; + if (nd == 0) { + // Track leading zeroes implicitly. + dp--; + } else if (nd < + WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->digits[nd++] = 0; + } else { + // Long-tail zeroes are ignored. + } + + } else if (('0' < *p) && (*p <= '9')) { + if (!saw_dot && !saw_non_zero_digits && saw_digits) { + // We don't allow unnecessary leading zeroes: "000123" or "0644". + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + saw_digits = true; + saw_non_zero_digits = true; + if (nd < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->digits[nd++] = (uint8_t)(*p - '0'); + } else { + // Long-tail non-zeroes set the truncated bit. + h->truncated = true; + } + + } else { + break; + } + } + + if (!saw_digits) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + if (!saw_dot) { + dp = (int32_t)nd; + } + + // Parse exponent. + if ((p < q) && ((*p == 'E') || (*p == 'e'))) { + p++; + for (; (p < q) && (*p == '_'); p++) { + } + if (p >= q) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + + int32_t exp_sign = +1; + if (*p == '+') { + p++; + } else if (*p == '-') { + exp_sign = -1; + p++; + } + + int32_t exp = 0; + const int32_t exp_large = + WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE + + WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION; + bool saw_exp_digits = false; + for (; p < q; p++) { + if (*p == '_') { + // No-op. + } else if (('0' <= *p) && (*p <= '9')) { + saw_exp_digits = true; + if (exp < exp_large) { + exp = (10 * exp) + ((int32_t)(*p - '0')); + } + } else { + break; + } + } + if (!saw_exp_digits) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + dp += exp_sign * exp; + } + + // Finish. + if (p != q) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + h->num_digits = nd; + if (nd == 0) { + h->decimal_point = 0; + } else if (dp < + -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { + h->decimal_point = + -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE - 1; + } else if (dp > + +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { + h->decimal_point = + +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE + 1; + } else { + h->decimal_point = dp; + } + wuffs_base__private_implementation__high_prec_dec__trim(h); + return wuffs_base__make_status(NULL); +} + +// -------- + +// The etc__hpd_left_shift and etc__powers_of_5 tables were printed by +// script/print-hpd-left-shift.go. That script has an optional -comments flag, +// whose output is not copied here, which prints further detail. +// +// These tables are used in +// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits. + +// wuffs_base__private_implementation__hpd_left_shift[i] encodes the number of +// new digits created after multiplying a positive integer by (1 << i): the +// additional length in the decimal representation. For example, shifting "234" +// by 3 (equivalent to multiplying by 8) will produce "1872". Going from a +// 3-length string to a 4-length string means that 1 new digit was added (and +// existing digits may have changed). +// +// Shifting by i can add either N or N-1 new digits, depending on whether the +// original positive integer compares >= or < to the i'th power of 5 (as 10 +// equals 2 * 5). Comparison is lexicographic, not numerical. +// +// For example, shifting by 4 (i.e. multiplying by 16) can add 1 or 2 new +// digits, depending on a lexicographic comparison to (5 ** 4), i.e. "625": +// - ("1" << 4) is "16", which adds 1 new digit. +// - ("5678" << 4) is "90848", which adds 1 new digit. +// - ("624" << 4) is "9984", which adds 1 new digit. +// - ("62498" << 4) is "999968", which adds 1 new digit. +// - ("625" << 4) is "10000", which adds 2 new digits. +// - ("625001" << 4) is "10000016", which adds 2 new digits. +// - ("7008" << 4) is "112128", which adds 2 new digits. +// - ("99" << 4) is "1584", which adds 2 new digits. +// +// Thus, when i is 4, N is 2 and (5 ** i) is "625". This etc__hpd_left_shift +// array encodes this as: +// - etc__hpd_left_shift[4] is 0x1006 = (2 << 11) | 0x0006. +// - etc__hpd_left_shift[5] is 0x1009 = (? << 11) | 0x0009. +// where the ? isn't relevant for i == 4. +// +// The high 5 bits of etc__hpd_left_shift[i] is N, the higher of the two +// possible number of new digits. The low 11 bits are an offset into the +// etc__powers_of_5 array (of length 0x051C, so offsets fit in 11 bits). When i +// is 4, its offset and the next one is 6 and 9, and etc__powers_of_5[6 .. 9] +// is the string "\x06\x02\x05", so the relevant power of 5 is "625". +// +// Thanks to Ken Thompson for the original idea. +static const uint16_t wuffs_base__private_implementation__hpd_left_shift[65] = { + 0x0000, 0x0800, 0x0801, 0x0803, 0x1006, 0x1009, 0x100D, 0x1812, 0x1817, + 0x181D, 0x2024, 0x202B, 0x2033, 0x203C, 0x2846, 0x2850, 0x285B, 0x3067, + 0x3073, 0x3080, 0x388E, 0x389C, 0x38AB, 0x38BB, 0x40CC, 0x40DD, 0x40EF, + 0x4902, 0x4915, 0x4929, 0x513E, 0x5153, 0x5169, 0x5180, 0x5998, 0x59B0, + 0x59C9, 0x61E3, 0x61FD, 0x6218, 0x6A34, 0x6A50, 0x6A6D, 0x6A8B, 0x72AA, + 0x72C9, 0x72E9, 0x7B0A, 0x7B2B, 0x7B4D, 0x8370, 0x8393, 0x83B7, 0x83DC, + 0x8C02, 0x8C28, 0x8C4F, 0x9477, 0x949F, 0x94C8, 0x9CF2, 0x051C, 0x051C, + 0x051C, 0x051C, +}; + +// wuffs_base__private_implementation__powers_of_5 contains the powers of 5, +// concatenated together: "5", "25", "125", "625", "3125", etc. +static const uint8_t wuffs_base__private_implementation__powers_of_5[0x051C] = { + 5, 2, 5, 1, 2, 5, 6, 2, 5, 3, 1, 2, 5, 1, 5, 6, 2, 5, 7, 8, 1, 2, 5, 3, 9, + 0, 6, 2, 5, 1, 9, 5, 3, 1, 2, 5, 9, 7, 6, 5, 6, 2, 5, 4, 8, 8, 2, 8, 1, 2, + 5, 2, 4, 4, 1, 4, 0, 6, 2, 5, 1, 2, 2, 0, 7, 0, 3, 1, 2, 5, 6, 1, 0, 3, 5, + 1, 5, 6, 2, 5, 3, 0, 5, 1, 7, 5, 7, 8, 1, 2, 5, 1, 5, 2, 5, 8, 7, 8, 9, 0, + 6, 2, 5, 7, 6, 2, 9, 3, 9, 4, 5, 3, 1, 2, 5, 3, 8, 1, 4, 6, 9, 7, 2, 6, 5, + 6, 2, 5, 1, 9, 0, 7, 3, 4, 8, 6, 3, 2, 8, 1, 2, 5, 9, 5, 3, 6, 7, 4, 3, 1, + 6, 4, 0, 6, 2, 5, 4, 7, 6, 8, 3, 7, 1, 5, 8, 2, 0, 3, 1, 2, 5, 2, 3, 8, 4, + 1, 8, 5, 7, 9, 1, 0, 1, 5, 6, 2, 5, 1, 1, 9, 2, 0, 9, 2, 8, 9, 5, 5, 0, 7, + 8, 1, 2, 5, 5, 9, 6, 0, 4, 6, 4, 4, 7, 7, 5, 3, 9, 0, 6, 2, 5, 2, 9, 8, 0, + 2, 3, 2, 2, 3, 8, 7, 6, 9, 5, 3, 1, 2, 5, 1, 4, 9, 0, 1, 1, 6, 1, 1, 9, 3, + 8, 4, 7, 6, 5, 6, 2, 5, 7, 4, 5, 0, 5, 8, 0, 5, 9, 6, 9, 2, 3, 8, 2, 8, 1, + 2, 5, 3, 7, 2, 5, 2, 9, 0, 2, 9, 8, 4, 6, 1, 9, 1, 4, 0, 6, 2, 5, 1, 8, 6, + 2, 6, 4, 5, 1, 4, 9, 2, 3, 0, 9, 5, 7, 0, 3, 1, 2, 5, 9, 3, 1, 3, 2, 2, 5, + 7, 4, 6, 1, 5, 4, 7, 8, 5, 1, 5, 6, 2, 5, 4, 6, 5, 6, 6, 1, 2, 8, 7, 3, 0, + 7, 7, 3, 9, 2, 5, 7, 8, 1, 2, 5, 2, 3, 2, 8, 3, 0, 6, 4, 3, 6, 5, 3, 8, 6, + 9, 6, 2, 8, 9, 0, 6, 2, 5, 1, 1, 6, 4, 1, 5, 3, 2, 1, 8, 2, 6, 9, 3, 4, 8, + 1, 4, 4, 5, 3, 1, 2, 5, 5, 8, 2, 0, 7, 6, 6, 0, 9, 1, 3, 4, 6, 7, 4, 0, 7, + 2, 2, 6, 5, 6, 2, 5, 2, 9, 1, 0, 3, 8, 3, 0, 4, 5, 6, 7, 3, 3, 7, 0, 3, 6, + 1, 3, 2, 8, 1, 2, 5, 1, 4, 5, 5, 1, 9, 1, 5, 2, 2, 8, 3, 6, 6, 8, 5, 1, 8, + 0, 6, 6, 4, 0, 6, 2, 5, 7, 2, 7, 5, 9, 5, 7, 6, 1, 4, 1, 8, 3, 4, 2, 5, 9, + 0, 3, 3, 2, 0, 3, 1, 2, 5, 3, 6, 3, 7, 9, 7, 8, 8, 0, 7, 0, 9, 1, 7, 1, 2, + 9, 5, 1, 6, 6, 0, 1, 5, 6, 2, 5, 1, 8, 1, 8, 9, 8, 9, 4, 0, 3, 5, 4, 5, 8, + 5, 6, 4, 7, 5, 8, 3, 0, 0, 7, 8, 1, 2, 5, 9, 0, 9, 4, 9, 4, 7, 0, 1, 7, 7, + 2, 9, 2, 8, 2, 3, 7, 9, 1, 5, 0, 3, 9, 0, 6, 2, 5, 4, 5, 4, 7, 4, 7, 3, 5, + 0, 8, 8, 6, 4, 6, 4, 1, 1, 8, 9, 5, 7, 5, 1, 9, 5, 3, 1, 2, 5, 2, 2, 7, 3, + 7, 3, 6, 7, 5, 4, 4, 3, 2, 3, 2, 0, 5, 9, 4, 7, 8, 7, 5, 9, 7, 6, 5, 6, 2, + 5, 1, 1, 3, 6, 8, 6, 8, 3, 7, 7, 2, 1, 6, 1, 6, 0, 2, 9, 7, 3, 9, 3, 7, 9, + 8, 8, 2, 8, 1, 2, 5, 5, 6, 8, 4, 3, 4, 1, 8, 8, 6, 0, 8, 0, 8, 0, 1, 4, 8, + 6, 9, 6, 8, 9, 9, 4, 1, 4, 0, 6, 2, 5, 2, 8, 4, 2, 1, 7, 0, 9, 4, 3, 0, 4, + 0, 4, 0, 0, 7, 4, 3, 4, 8, 4, 4, 9, 7, 0, 7, 0, 3, 1, 2, 5, 1, 4, 2, 1, 0, + 8, 5, 4, 7, 1, 5, 2, 0, 2, 0, 0, 3, 7, 1, 7, 4, 2, 2, 4, 8, 5, 3, 5, 1, 5, + 6, 2, 5, 7, 1, 0, 5, 4, 2, 7, 3, 5, 7, 6, 0, 1, 0, 0, 1, 8, 5, 8, 7, 1, 1, + 2, 4, 2, 6, 7, 5, 7, 8, 1, 2, 5, 3, 5, 5, 2, 7, 1, 3, 6, 7, 8, 8, 0, 0, 5, + 0, 0, 9, 2, 9, 3, 5, 5, 6, 2, 1, 3, 3, 7, 8, 9, 0, 6, 2, 5, 1, 7, 7, 6, 3, + 5, 6, 8, 3, 9, 4, 0, 0, 2, 5, 0, 4, 6, 4, 6, 7, 7, 8, 1, 0, 6, 6, 8, 9, 4, + 5, 3, 1, 2, 5, 8, 8, 8, 1, 7, 8, 4, 1, 9, 7, 0, 0, 1, 2, 5, 2, 3, 2, 3, 3, + 8, 9, 0, 5, 3, 3, 4, 4, 7, 2, 6, 5, 6, 2, 5, 4, 4, 4, 0, 8, 9, 2, 0, 9, 8, + 5, 0, 0, 6, 2, 6, 1, 6, 1, 6, 9, 4, 5, 2, 6, 6, 7, 2, 3, 6, 3, 2, 8, 1, 2, + 5, 2, 2, 2, 0, 4, 4, 6, 0, 4, 9, 2, 5, 0, 3, 1, 3, 0, 8, 0, 8, 4, 7, 2, 6, + 3, 3, 3, 6, 1, 8, 1, 6, 4, 0, 6, 2, 5, 1, 1, 1, 0, 2, 2, 3, 0, 2, 4, 6, 2, + 5, 1, 5, 6, 5, 4, 0, 4, 2, 3, 6, 3, 1, 6, 6, 8, 0, 9, 0, 8, 2, 0, 3, 1, 2, + 5, 5, 5, 5, 1, 1, 1, 5, 1, 2, 3, 1, 2, 5, 7, 8, 2, 7, 0, 2, 1, 1, 8, 1, 5, + 8, 3, 4, 0, 4, 5, 4, 1, 0, 1, 5, 6, 2, 5, 2, 7, 7, 5, 5, 5, 7, 5, 6, 1, 5, + 6, 2, 8, 9, 1, 3, 5, 1, 0, 5, 9, 0, 7, 9, 1, 7, 0, 2, 2, 7, 0, 5, 0, 7, 8, + 1, 2, 5, 1, 3, 8, 7, 7, 7, 8, 7, 8, 0, 7, 8, 1, 4, 4, 5, 6, 7, 5, 5, 2, 9, + 5, 3, 9, 5, 8, 5, 1, 1, 3, 5, 2, 5, 3, 9, 0, 6, 2, 5, 6, 9, 3, 8, 8, 9, 3, + 9, 0, 3, 9, 0, 7, 2, 2, 8, 3, 7, 7, 6, 4, 7, 6, 9, 7, 9, 2, 5, 5, 6, 7, 6, + 2, 6, 9, 5, 3, 1, 2, 5, 3, 4, 6, 9, 4, 4, 6, 9, 5, 1, 9, 5, 3, 6, 1, 4, 1, + 8, 8, 8, 2, 3, 8, 4, 8, 9, 6, 2, 7, 8, 3, 8, 1, 3, 4, 7, 6, 5, 6, 2, 5, 1, + 7, 3, 4, 7, 2, 3, 4, 7, 5, 9, 7, 6, 8, 0, 7, 0, 9, 4, 4, 1, 1, 9, 2, 4, 4, + 8, 1, 3, 9, 1, 9, 0, 6, 7, 3, 8, 2, 8, 1, 2, 5, 8, 6, 7, 3, 6, 1, 7, 3, 7, + 9, 8, 8, 4, 0, 3, 5, 4, 7, 2, 0, 5, 9, 6, 2, 2, 4, 0, 6, 9, 5, 9, 5, 3, 3, + 6, 9, 1, 4, 0, 6, 2, 5, +}; + +// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits +// returns the number of additional decimal digits when left-shifting by shift. +// +// See below for preconditions. +static uint32_t // +wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits( + wuffs_base__private_implementation__high_prec_dec* h, + uint32_t shift) { + // Masking with 0x3F should be unnecessary (assuming the preconditions) but + // it's cheap and ensures that we don't overflow the + // wuffs_base__private_implementation__hpd_left_shift array. + shift &= 63; + + uint32_t x_a = wuffs_base__private_implementation__hpd_left_shift[shift]; + uint32_t x_b = wuffs_base__private_implementation__hpd_left_shift[shift + 1]; + uint32_t num_new_digits = x_a >> 11; + uint32_t pow5_a = 0x7FF & x_a; + uint32_t pow5_b = 0x7FF & x_b; + + const uint8_t* pow5 = + &wuffs_base__private_implementation__powers_of_5[pow5_a]; + uint32_t i = 0; + uint32_t n = pow5_b - pow5_a; + for (; i < n; i++) { + if (i >= h->num_digits) { + return num_new_digits - 1; + } else if (h->digits[i] == pow5[i]) { + continue; + } else if (h->digits[i] < pow5[i]) { + return num_new_digits - 1; + } else { + return num_new_digits; + } + } + return num_new_digits; +} + +// -------- + +// wuffs_base__private_implementation__high_prec_dec__rounded_integer returns +// the integral (non-fractional) part of h, provided that it is 18 or fewer +// decimal digits. For 19 or more digits, it returns UINT64_MAX. Note that: +// - (1 << 53) is 9007199254740992, which has 16 decimal digits. +// - (1 << 56) is 72057594037927936, which has 17 decimal digits. +// - (1 << 59) is 576460752303423488, which has 18 decimal digits. +// - (1 << 63) is 9223372036854775808, which has 19 decimal digits. +// and that IEEE 754 double precision has 52 mantissa bits. +// +// That integral part is rounded-to-even: rounding 7.5 or 8.5 both give 8. +// +// h's negative bit is ignored: rounding -8.6 returns 9. +// +// See below for preconditions. +static uint64_t // +wuffs_base__private_implementation__high_prec_dec__rounded_integer( + wuffs_base__private_implementation__high_prec_dec* h) { + if ((h->num_digits == 0) || (h->decimal_point < 0)) { + return 0; + } else if (h->decimal_point > 18) { + return UINT64_MAX; + } + + uint32_t dp = (uint32_t)(h->decimal_point); + uint64_t n = 0; + uint32_t i = 0; + for (; i < dp; i++) { + n = (10 * n) + ((i < h->num_digits) ? h->digits[i] : 0); + } + + bool round_up = false; + if (dp < h->num_digits) { + round_up = h->digits[dp] >= 5; + if ((h->digits[dp] == 5) && (dp + 1 == h->num_digits)) { + // We are exactly halfway. If we're truncated, round up, otherwise round + // to even. + round_up = h->truncated || // + ((dp > 0) && (1 & h->digits[dp - 1])); + } + } + if (round_up) { + n++; + } + + return n; +} + +// wuffs_base__private_implementation__high_prec_dec__small_xshift shifts h's +// number (where 'x' is 'l' or 'r' for left or right) by a small shift value. +// +// Preconditions: +// - h is non-NULL. +// - h->decimal_point is "not extreme". +// - shift is non-zero. +// - shift is "a small shift". +// +// "Not extreme" means within +// ±WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE. +// +// "A small shift" means not more than +// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL. +// +// wuffs_base__private_implementation__high_prec_dec__rounded_integer and +// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits +// have the same preconditions. + +static void // +wuffs_base__private_implementation__high_prec_dec__small_lshift( + wuffs_base__private_implementation__high_prec_dec* h, + uint32_t shift) { + if (h->num_digits == 0) { + return; + } + uint32_t num_new_digits = + wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits( + h, shift); + uint32_t rx = h->num_digits - 1; // Read index. + uint32_t wx = h->num_digits - 1 + num_new_digits; // Write index. + uint64_t n = 0; + + // Repeat: pick up a digit, put down a digit, right to left. + while (((int32_t)rx) >= 0) { + n += ((uint64_t)(h->digits[rx])) << shift; + uint64_t quo = n / 10; + uint64_t rem = n - (10 * quo); + if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->digits[wx] = (uint8_t)rem; + } else if (rem > 0) { + h->truncated = true; + } + n = quo; + wx--; + rx--; + } + + // Put down leading digits, right to left. + while (n > 0) { + uint64_t quo = n / 10; + uint64_t rem = n - (10 * quo); + if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->digits[wx] = (uint8_t)rem; + } else if (rem > 0) { + h->truncated = true; + } + n = quo; + wx--; + } + + // Finish. + h->num_digits += num_new_digits; + if (h->num_digits > + WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->num_digits = WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION; + } + h->decimal_point += (int32_t)num_new_digits; + wuffs_base__private_implementation__high_prec_dec__trim(h); +} + +static void // +wuffs_base__private_implementation__high_prec_dec__small_rshift( + wuffs_base__private_implementation__high_prec_dec* h, + uint32_t shift) { + uint32_t rx = 0; // Read index. + uint32_t wx = 0; // Write index. + uint64_t n = 0; + + // Pick up enough leading digits to cover the first shift. + while ((n >> shift) == 0) { + if (rx < h->num_digits) { + // Read a digit. + n = (10 * n) + h->digits[rx++]; + } else if (n == 0) { + // h's number used to be zero and remains zero. + return; + } else { + // Read sufficient implicit trailing zeroes. + while ((n >> shift) == 0) { + n = 10 * n; + rx++; + } + break; + } + } + h->decimal_point -= ((int32_t)(rx - 1)); + if (h->decimal_point < + -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { + // After the shift, h's number is effectively zero. + h->num_digits = 0; + h->decimal_point = 0; + h->negative = false; + h->truncated = false; + return; + } + + // Repeat: pick up a digit, put down a digit, left to right. + uint64_t mask = (((uint64_t)(1)) << shift) - 1; + while (rx < h->num_digits) { + uint8_t new_digit = ((uint8_t)(n >> shift)); + n = (10 * (n & mask)) + h->digits[rx++]; + h->digits[wx++] = new_digit; + } + + // Put down trailing digits, left to right. + while (n > 0) { + uint8_t new_digit = ((uint8_t)(n >> shift)); + n = 10 * (n & mask); + if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { + h->digits[wx++] = new_digit; + } else if (new_digit > 0) { + h->truncated = true; + } + } + + // Finish. + h->num_digits = wx; + wuffs_base__private_implementation__high_prec_dec__trim(h); +} + +// -------- + +// The wuffs_base__private_implementation__etc_powers_of_10 tables were printed +// by script/print-mpb-powers-of-10.go. That script has an optional -comments +// flag, whose output is not copied here, which prints further detail. +// +// These tables are used in +// wuffs_base__private_implementation__medium_prec_bin__assign_from_hpd. + +// wuffs_base__private_implementation__big_powers_of_10 contains approximations +// to the powers of 10, ranging from 1e-348 to 1e+340, with the exponent +// stepping by 8: -348, -340, -332, ..., -12, -4, +4, +12, ..., +340. Each step +// consists of three uint32_t elements. There are 87 triples, 87 * 3 = 261. +// +// For example, the third approximation, for 1e-332, consists of the uint32_t +// triple (0x3055AC76, 0x8B16FB20, 0xFFFFFB72). The first two of that triple +// are a little-endian uint64_t value: 0x8B16FB203055AC76. The last one is an +// int32_t value: -1166. Together, they represent the approximation: +// 1e-332 ≈ 0x8B16FB203055AC76 * (2 ** -1166) +// Similarly, the (0x00000000, 0x9C400000, 0xFFFFFFCE) uint32_t triple means: +// 1e+4 ≈ 0x9C40000000000000 * (2 ** -50) // This approx'n is exact. +// Similarly, the (0xD4C4FB27, 0xED63A231, 0x000000A2) uint32_t triple means: +// 1e+68 ≈ 0xED63A231D4C4FB27 * (2 ** 162) +static const uint32_t + wuffs_base__private_implementation__big_powers_of_10[261] = { + 0x081C0288, 0xFA8FD5A0, 0xFFFFFB3C, 0xA23EBF76, 0xBAAEE17F, 0xFFFFFB57, + 0x3055AC76, 0x8B16FB20, 0xFFFFFB72, 0x5DCE35EA, 0xCF42894A, 0xFFFFFB8C, + 0x55653B2D, 0x9A6BB0AA, 0xFFFFFBA7, 0x3D1A45DF, 0xE61ACF03, 0xFFFFFBC1, + 0xC79AC6CA, 0xAB70FE17, 0xFFFFFBDC, 0xBEBCDC4F, 0xFF77B1FC, 0xFFFFFBF6, + 0x416BD60C, 0xBE5691EF, 0xFFFFFC11, 0x907FFC3C, 0x8DD01FAD, 0xFFFFFC2C, + 0x31559A83, 0xD3515C28, 0xFFFFFC46, 0xADA6C9B5, 0x9D71AC8F, 0xFFFFFC61, + 0x23EE8BCB, 0xEA9C2277, 0xFFFFFC7B, 0x4078536D, 0xAECC4991, 0xFFFFFC96, + 0x5DB6CE57, 0x823C1279, 0xFFFFFCB1, 0x4DFB5637, 0xC2109436, 0xFFFFFCCB, + 0x3848984F, 0x9096EA6F, 0xFFFFFCE6, 0x25823AC7, 0xD77485CB, 0xFFFFFD00, + 0x97BF97F4, 0xA086CFCD, 0xFFFFFD1B, 0x172AACE5, 0xEF340A98, 0xFFFFFD35, + 0x2A35B28E, 0xB23867FB, 0xFFFFFD50, 0xD2C63F3B, 0x84C8D4DF, 0xFFFFFD6B, + 0x1AD3CDBA, 0xC5DD4427, 0xFFFFFD85, 0xBB25C996, 0x936B9FCE, 0xFFFFFDA0, + 0x7D62A584, 0xDBAC6C24, 0xFFFFFDBA, 0x0D5FDAF6, 0xA3AB6658, 0xFFFFFDD5, + 0xDEC3F126, 0xF3E2F893, 0xFFFFFDEF, 0xAAFF80B8, 0xB5B5ADA8, 0xFFFFFE0A, + 0x6C7C4A8B, 0x87625F05, 0xFFFFFE25, 0x34C13053, 0xC9BCFF60, 0xFFFFFE3F, + 0x91BA2655, 0x964E858C, 0xFFFFFE5A, 0x70297EBD, 0xDFF97724, 0xFFFFFE74, + 0xB8E5B88F, 0xA6DFBD9F, 0xFFFFFE8F, 0x88747D94, 0xF8A95FCF, 0xFFFFFEA9, + 0x8FA89BCF, 0xB9447093, 0xFFFFFEC4, 0xBF0F156B, 0x8A08F0F8, 0xFFFFFEDF, + 0x653131B6, 0xCDB02555, 0xFFFFFEF9, 0xD07B7FAC, 0x993FE2C6, 0xFFFFFF14, + 0x2A2B3B06, 0xE45C10C4, 0xFFFFFF2E, 0x697392D3, 0xAA242499, 0xFFFFFF49, + 0x8300CA0E, 0xFD87B5F2, 0xFFFFFF63, 0x92111AEB, 0xBCE50864, 0xFFFFFF7E, + 0x6F5088CC, 0x8CBCCC09, 0xFFFFFF99, 0xE219652C, 0xD1B71758, 0xFFFFFFB3, + 0x00000000, 0x9C400000, 0xFFFFFFCE, 0x00000000, 0xE8D4A510, 0xFFFFFFE8, + 0xAC620000, 0xAD78EBC5, 0x00000003, 0xF8940984, 0x813F3978, 0x0000001E, + 0xC90715B3, 0xC097CE7B, 0x00000038, 0x7BEA5C70, 0x8F7E32CE, 0x00000053, + 0xABE98068, 0xD5D238A4, 0x0000006D, 0x179A2245, 0x9F4F2726, 0x00000088, + 0xD4C4FB27, 0xED63A231, 0x000000A2, 0x8CC8ADA8, 0xB0DE6538, 0x000000BD, + 0x1AAB65DB, 0x83C7088E, 0x000000D8, 0x42711D9A, 0xC45D1DF9, 0x000000F2, + 0xA61BE758, 0x924D692C, 0x0000010D, 0x1A708DEA, 0xDA01EE64, 0x00000127, + 0x9AEF774A, 0xA26DA399, 0x00000142, 0xB47D6B85, 0xF209787B, 0x0000015C, + 0x79DD1877, 0xB454E4A1, 0x00000177, 0x5B9BC5C2, 0x865B8692, 0x00000192, + 0xC8965D3D, 0xC83553C5, 0x000001AC, 0xFA97A0B3, 0x952AB45C, 0x000001C7, + 0x99A05FE3, 0xDE469FBD, 0x000001E1, 0xDB398C25, 0xA59BC234, 0x000001FC, + 0xA3989F5C, 0xF6C69A72, 0x00000216, 0x54E9BECE, 0xB7DCBF53, 0x00000231, + 0xF22241E2, 0x88FCF317, 0x0000024C, 0xD35C78A5, 0xCC20CE9B, 0x00000266, + 0x7B2153DF, 0x98165AF3, 0x00000281, 0x971F303A, 0xE2A0B5DC, 0x0000029B, + 0x5CE3B396, 0xA8D9D153, 0x000002B6, 0xA4A7443C, 0xFB9B7CD9, 0x000002D0, + 0xA7A44410, 0xBB764C4C, 0x000002EB, 0xB6409C1A, 0x8BAB8EEF, 0x00000306, + 0xA657842C, 0xD01FEF10, 0x00000320, 0xE9913129, 0x9B10A4E5, 0x0000033B, + 0xA19C0C9D, 0xE7109BFB, 0x00000355, 0x623BF429, 0xAC2820D9, 0x00000370, + 0x7AA7CF85, 0x80444B5E, 0x0000038B, 0x03ACDD2D, 0xBF21E440, 0x000003A5, + 0x5E44FF8F, 0x8E679C2F, 0x000003C0, 0x9C8CB841, 0xD433179D, 0x000003DA, + 0xB4E31BA9, 0x9E19DB92, 0x000003F5, 0xBADF77D9, 0xEB96BF6E, 0x0000040F, + 0x9BF0EE6B, 0xAF87023B, 0x0000042A, +}; + +// wuffs_base__private_implementation__small_powers_of_10 contains +// approximations to the powers of 10, ranging from 1e+0 to 1e+7, with the +// exponent stepping by 1. Each step consists of three uint32_t elements. +// +// For example, the third approximation, for 1e+2, consists of the uint32_t +// triple (0x00000000, 0xC8000000, 0xFFFFFFC7). The first two of that triple +// are a little-endian uint64_t value: 0xC800000000000000. The last one is an +// int32_t value: -57. Together, they represent the approximation: +// 1e+2 ≈ 0xC800000000000000 * (2 ** -57) // This approx'n is exact. +// Similarly, the (0x00000000, 0x9C400000, 0xFFFFFFCE) uint32_t triple means: +// 1e+4 ≈ 0x9C40000000000000 * (2 ** -50) // This approx'n is exact. +static const uint32_t + wuffs_base__private_implementation__small_powers_of_10[24] = { + 0x00000000, 0x80000000, 0xFFFFFFC1, 0x00000000, 0xA0000000, 0xFFFFFFC4, + 0x00000000, 0xC8000000, 0xFFFFFFC7, 0x00000000, 0xFA000000, 0xFFFFFFCA, + 0x00000000, 0x9C400000, 0xFFFFFFCE, 0x00000000, 0xC3500000, 0xFFFFFFD1, + 0x00000000, 0xF4240000, 0xFFFFFFD4, 0x00000000, 0x98968000, 0xFFFFFFD8, +}; + +// wuffs_base__private_implementation__f64_powers_of_10 holds powers of 10 that +// can be exactly represented by a float64 (what C calls a double). +static const double wuffs_base__private_implementation__f64_powers_of_10[23] = { + 1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10, 1e11, + 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22, +}; + +// -------- + +// wuffs_base__private_implementation__medium_prec_bin (abbreviated as MPB) is +// a fixed precision floating point binary number. Unlike IEEE 754 Floating +// Point, it cannot represent infinity or NaN (Not a Number). +// +// "Medium precision" means that the mantissa holds 64 binary digits, a little +// more than "double precision", and sizeof(MPB) > sizeof(double). 64 is +// obviously the number of bits in a uint64_t. +// +// An MPB isn't for general purpose arithmetic, only for conversions to and +// from IEEE 754 double-precision floating point. +// +// There is no implicit mantissa bit. The mantissa field is zero if and only if +// the overall floating point value is ±0. An MPB is normalized if the mantissa +// is zero or its high bit (the 1<<63 bit) is set. +// +// There is no negative bit. An MPB can only represent non-negative numbers. +// +// The "all fields are zero" value is valid, and represents the number +0. +// +// This is the "Do It Yourself Floating Point" data structure from Loitsch, +// "Printing Floating-Point Numbers Quickly and Accurately with Integers" +// (https://www.cs.tufts.edu/~nr/cs257/archive/florian-loitsch/printf.pdf). +// +// Florian Loitsch is also the primary contributor to +// https://github.com/google/double-conversion +typedef struct { + uint64_t mantissa; + int32_t exp2; +} wuffs_base__private_implementation__medium_prec_bin; + +static uint32_t // +wuffs_base__private_implementation__medium_prec_bin__normalize( + wuffs_base__private_implementation__medium_prec_bin* m) { + if (m->mantissa == 0) { + return 0; + } + uint32_t shift = wuffs_base__count_leading_zeroes_u64(m->mantissa); + m->mantissa <<= shift; + m->exp2 -= (int32_t)shift; + return shift; +} + +// wuffs_base__private_implementation__medium_prec_bin__mul_pow_10 sets m to be +// (m * pow), where pow comes from an etc_powers_of_10 triple starting at p. +// +// The result is rounded, but not necessarily normalized. +// +// Preconditions: +// - m is non-NULL. +// - m->mantissa is non-zero. +// - m->mantissa's high bit is set (i.e. m is normalized). +// +// The etc_powers_of_10 triple is already normalized. +static void // +wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( + wuffs_base__private_implementation__medium_prec_bin* m, + const uint32_t* p) { + uint64_t p_mantissa = ((uint64_t)p[0]) | (((uint64_t)p[1]) << 32); + int32_t p_exp2 = (int32_t)p[2]; + + wuffs_base__multiply_u64__output o = + wuffs_base__multiply_u64(m->mantissa, p_mantissa); + // Round the mantissa up. It cannot overflow because the maximum possible + // value of o.hi is 0xFFFFFFFFFFFFFFFE. + m->mantissa = o.hi + (o.lo >> 63); + m->exp2 = m->exp2 + p_exp2 + 64; +} + +// wuffs_base__private_implementation__medium_prec_bin__as_f64 converts m to a +// double (what C calls a double-precision float64). +// +// Preconditions: +// - m is non-NULL. +// - m->mantissa is non-zero. +// - m->mantissa's high bit is set (i.e. m is normalized). +static double // +wuffs_base__private_implementation__medium_prec_bin__as_f64( + const wuffs_base__private_implementation__medium_prec_bin* m, + bool negative) { + uint64_t mantissa64 = m->mantissa; + // An mpb's mantissa has the implicit (binary) decimal point at the right + // hand end of the mantissa's explicit digits. A double-precision's mantissa + // has that decimal point near the left hand end. There's also an explicit + // versus implicit leading 1 bit (binary digit). Together, the difference in + // semantics corresponds to adding 63. + int32_t exp2 = m->exp2 + 63; + + // Ensure that exp2 is at least -1022, the minimum double-precision exponent + // for normal (as opposed to subnormal) numbers. + if (-1022 > exp2) { + uint32_t n = (uint32_t)(-1022 - exp2); + mantissa64 >>= n; + exp2 += (int32_t)n; + } + + // Extract the (1 + 52) bits from the 64-bit mantissa64. 52 is the number of + // explicit mantissa bits in a double-precision f64. + // + // Before, we have 64 bits and due to normalization, the high bit 'H' is 1. + // 63 55 47 etc 15 7 + // H210_9876_5432_1098_7654_etc_etc_etc_5432_1098_7654_3210 + // ++++_++++_++++_++++_++++_etc_etc_etc_++++_+..._...._.... Kept bits. + // ...._...._...H_2109_8765_etc_etc_etc_6543_2109_8765_4321 After shifting. + // After, we have 53 bits (and bit #52 is this 'H' bit). + uint64_t mantissa53 = mantissa64 >> 11; + + // Round up if the old bit #10 (the highest bit dropped by shifting) was set. + // We also fix any overflow from rounding up. + if (mantissa64 & 1024) { + mantissa53++; + if ((mantissa53 >> 53) != 0) { + mantissa53 >>= 1; + exp2++; + } + } + + // Handle double-precision infinity (a nominal exponent of 1024) and + // subnormals (an exponent of -1023 and no implicit mantissa bit, bit #52). + if (exp2 >= 1024) { + mantissa53 = 0; + exp2 = 1024; + } else if ((mantissa53 >> 52) == 0) { + exp2 = -1023; + } + + // Pack the bits and return. + const int32_t f64_bias = -1023; + uint64_t exp2_bits = + (uint64_t)((exp2 - f64_bias) & 0x07FF); // (1 << 11) - 1. + uint64_t bits = (mantissa53 & 0x000FFFFFFFFFFFFF) | // (1 << 52) - 1. + (exp2_bits << 52) | // + (negative ? 0x8000000000000000 : 0); // (1 << 63). + return wuffs_base__ieee_754_bit_representation__to_f64(bits); +} + +// wuffs_base__private_implementation__medium_prec_bin__parse_number_f64 +// converts from an HPD to a double, using an MPB as scratch space. It returns +// a NULL status.repr if there is no ambiguity in the truncation or rounding to +// a float64 (an IEEE 754 double-precision floating point value). +// +// It may modify m even if it returns a non-NULL status.repr. +static wuffs_base__result_f64 // +wuffs_base__private_implementation__medium_prec_bin__parse_number_f64( + wuffs_base__private_implementation__medium_prec_bin* m, + const wuffs_base__private_implementation__high_prec_dec* h, + bool skip_fast_path_for_tests) { + do { + // m->mantissa is a uint64_t, which is an integer approximation to a + // rational value - h's underlying digits after m's normalization. This + // error is an upper bound on the difference between the approximate and + // actual value. + // + // The DiyFpStrtod function in https://github.com/google/double-conversion + // uses a finer grain (1/8th of the ULP, Unit in the Last Place) when + // tracking error. This implementation is coarser (1 ULP) but simpler. + // + // It is an error in the "numerical approximation" sense, not in the + // typical programming sense (as in "bad input" or "a result type"). + uint64_t error = 0; + + // Convert up to 19 decimal digits (in h->digits) to 64 binary digits (in + // m->mantissa): (1e19 < (1<<64)) and ((1<<64) < 1e20). If we have more + // than 19 digits, we're truncating (with error). + uint32_t i; + uint32_t i_end = h->num_digits; + if (i_end > 19) { + i_end = 19; + error = 1; + } + uint64_t mantissa = 0; + for (i = 0; i < i_end; i++) { + mantissa = (10 * mantissa) + h->digits[i]; + } + m->mantissa = mantissa; + m->exp2 = 0; + + // Check that exp10 lies in the (big_powers_of_10 + small_powers_of_10) + // range, -348 ..= +347, stepping big_powers_of_10 by 8 (which is 87 + // triples) and small_powers_of_10 by 1 (which is 8 triples). + int32_t exp10 = h->decimal_point - ((int32_t)(i_end)); + if (exp10 < -348) { + goto fail; + } + uint32_t bpo10 = ((uint32_t)(exp10 + 348)) / 8; + uint32_t spo10 = ((uint32_t)(exp10 + 348)) % 8; + if (bpo10 >= 87) { + goto fail; + } + + // Try a fast path, if float64 math would be exact. + // + // 15 is such that 1e15 can be losslessly represented in a float64 + // mantissa: (1e15 < (1<<53)) and ((1<<53) < 1e16). + // + // 22 is the maximum valid index for the + // wuffs_base__private_implementation__f64_powers_of_10 array. + do { + if (skip_fast_path_for_tests || ((mantissa >> 52) != 0)) { + break; + } + double d = (double)mantissa; + + if (exp10 == 0) { + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = h->negative ? -d : +d; + return ret; + + } else if (exp10 > 0) { + if (exp10 > 22) { + if (exp10 > (15 + 22)) { + break; + } + // If exp10 is in the range 23 ..= 37, try moving a few of the zeroes + // from the exponent to the mantissa. If we're still under 1e15, we + // haven't truncated any mantissa bits. + if (exp10 > 22) { + d *= wuffs_base__private_implementation__f64_powers_of_10[exp10 - + 22]; + exp10 = 22; + if (d >= 1e15) { + break; + } + } + } + d *= wuffs_base__private_implementation__f64_powers_of_10[exp10]; + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = h->negative ? -d : +d; + return ret; + + } else { // "if (exp10 < 0)" is effectively "if (true)" here. + if (exp10 < -22) { + break; + } + d /= wuffs_base__private_implementation__f64_powers_of_10[-exp10]; + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = h->negative ? -d : +d; + return ret; + } + } while (0); + + // Normalize (and scale the error). + error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); + + // Multiplying two MPB values nominally multiplies two mantissas, call them + // A and B, which are integer approximations to the precise values (A+a) + // and (B+b) for some error terms a and b. + // + // MPB multiplication calculates (((A+a) * (B+b)) >> 64) to be ((A*B) >> + // 64). Shifting (truncating) and rounding introduces further error. The + // difference between the calculated result: + // ((A*B ) >> 64) + // and the true result: + // ((A*B + A*b + a*B + a*b) >> 64) + rounding_error + // is: + // (( A*b + a*B + a*b) >> 64) + rounding_error + // which can be re-grouped as: + // ((A*b) >> 64) + ((a*(B+b)) >> 64) + rounding_error + // + // Now, let A and a be "m->mantissa" and "error", and B and b be the + // pre-calculated power of 10. A and B are both less than (1 << 64), a is + // the "error" local variable and b is less than 1. + // + // An upper bound (in absolute value) on ((A*b) >> 64) is therefore 1. + // + // An upper bound on ((a*(B+b)) >> 64) is a, also known as error. + // + // Finally, the rounding_error is at most 1. + // + // In total, calling mpb__mul_pow_10 will raise the worst-case error by 2. + // The subsequent re-normalization can multiply that by a further factor. + + // Multiply by small_powers_of_10[etc]. + wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( + m, &wuffs_base__private_implementation__small_powers_of_10[3 * spo10]); + error += 2; + error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); + + // Multiply by big_powers_of_10[etc]. + wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( + m, &wuffs_base__private_implementation__big_powers_of_10[3 * bpo10]); + error += 2; + error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); + + // We have a good approximation of h, but we still have to check whether + // the error is small enough. Equivalently, whether the number of surplus + // mantissa bits (the bits dropped when going from m's 64 mantissa bits to + // the smaller number of double-precision mantissa bits) would always round + // up or down, even when perturbed by ±error. We start at 11 surplus bits + // (m has 64, double-precision has 1+52), but it can be higher for + // subnormals. + // + // In many cases, the error is small enough and we return true. + const int32_t f64_bias = -1023; + int32_t subnormal_exp2 = f64_bias - 63; + uint32_t surplus_bits = 11; + if (subnormal_exp2 >= m->exp2) { + surplus_bits += 1 + ((uint32_t)(subnormal_exp2 - m->exp2)); + } + + uint64_t surplus_mask = + (((uint64_t)1) << surplus_bits) - 1; // e.g. 0x07FF. + uint64_t surplus = m->mantissa & surplus_mask; + uint64_t halfway = ((uint64_t)1) << (surplus_bits - 1); // e.g. 0x0400. + + // Do the final calculation in *signed* arithmetic. + int64_t i_surplus = (int64_t)surplus; + int64_t i_halfway = (int64_t)halfway; + int64_t i_error = (int64_t)error; + + if ((i_surplus > (i_halfway - i_error)) && + (i_surplus < (i_halfway + i_error))) { + goto fail; + } + + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = wuffs_base__private_implementation__medium_prec_bin__as_f64( + m, h->negative); + return ret; + } while (0); + +fail: + do { + wuffs_base__result_f64 ret; + ret.status.repr = "#base: mpb__parse_number_f64 failed"; + ret.value = 0; + return ret; + } while (0); +} + +// -------- + +wuffs_base__result_f64 // +wuffs_base__parse_number_f64_special(wuffs_base__slice_u8 s, + const char* fallback_status_repr) { + do { + uint8_t* p = s.ptr; + uint8_t* q = s.ptr + s.len; + + for (; (p < q) && (*p == '_'); p++) { + } + if (p >= q) { + goto fallback; + } + + // Parse sign. + bool negative = false; + do { + if (*p == '+') { + p++; + } else if (*p == '-') { + negative = true; + p++; + } else { + break; + } + for (; (p < q) && (*p == '_'); p++) { + } + } while (0); + if (p >= q) { + goto fallback; + } + + bool nan = false; + switch (p[0]) { + case 'I': + case 'i': + if (((q - p) < 3) || // + ((p[1] != 'N') && (p[1] != 'n')) || // + ((p[2] != 'F') && (p[2] != 'f'))) { + goto fallback; + } + p += 3; + + if ((p >= q) || (*p == '_')) { + break; + } else if (((q - p) < 5) || // + ((p[0] != 'I') && (p[0] != 'i')) || // + ((p[1] != 'N') && (p[1] != 'n')) || // + ((p[2] != 'I') && (p[2] != 'i')) || // + ((p[3] != 'T') && (p[3] != 't')) || // + ((p[4] != 'Y') && (p[4] != 'y'))) { + goto fallback; + } + p += 5; + + if ((p >= q) || (*p == '_')) { + break; + } + goto fallback; + + case 'N': + case 'n': + if (((q - p) < 3) || // + ((p[1] != 'A') && (p[1] != 'a')) || // + ((p[2] != 'N') && (p[2] != 'n'))) { + goto fallback; + } + p += 3; + + if ((p >= q) || (*p == '_')) { + nan = true; + break; + } + goto fallback; + + default: + goto fallback; + } + + // Finish. + for (; (p < q) && (*p == '_'); p++) { + } + if (p != q) { + goto fallback; + } + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = wuffs_base__ieee_754_bit_representation__to_f64( + (nan ? 0x7FFFFFFFFFFFFFFF : 0x7FF0000000000000) | + (negative ? 0x8000000000000000 : 0)); + return ret; + } while (0); + +fallback: + do { + wuffs_base__result_f64 ret; + ret.status.repr = fallback_status_repr; + ret.value = 0; + return ret; + } while (0); +} + +wuffs_base__result_f64 // +wuffs_base__parse_number_f64(wuffs_base__slice_u8 s) { + wuffs_base__private_implementation__medium_prec_bin m; + wuffs_base__private_implementation__high_prec_dec h; + + do { + // powers converts decimal powers of 10 to binary powers of 2. For example, + // (10000 >> 13) is 1. It stops before the elements exceed 60, also known + // as WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL. + static const uint32_t num_powers = 19; + static const uint8_t powers[19] = { + 0, 3, 6, 9, 13, 16, 19, 23, 26, 29, // + 33, 36, 39, 43, 46, 49, 53, 56, 59, // + }; + + wuffs_base__status status = + wuffs_base__private_implementation__high_prec_dec__parse(&h, s); + if (status.repr) { + return wuffs_base__parse_number_f64_special(s, status.repr); + } + + // Handle zero and obvious extremes. The largest and smallest positive + // finite f64 values are approximately 1.8e+308 and 4.9e-324. + if ((h.num_digits == 0) || (h.decimal_point < -326)) { + goto zero; + } else if (h.decimal_point > 310) { + goto infinity; + } + + wuffs_base__result_f64 mpb_result = + wuffs_base__private_implementation__medium_prec_bin__parse_number_f64( + &m, &h, false); + if (mpb_result.status.repr == NULL) { + return mpb_result; + } + + // Scale by powers of 2 until we're in the range [½ .. 1], which gives us + // our exponent (in base-2). First we shift right, possibly a little too + // far, ending with a value certainly below 1 and possibly below ½... + const int32_t f64_bias = -1023; + int32_t exp2 = 0; + while (h.decimal_point > 0) { + uint32_t n = (uint32_t)(+h.decimal_point); + uint32_t shift = + (n < num_powers) + ? powers[n] + : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; + + wuffs_base__private_implementation__high_prec_dec__small_rshift(&h, + shift); + if (h.decimal_point < + -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { + goto zero; + } + exp2 += (int32_t)shift; + } + // ...then we shift left, putting us in [½ .. 1]. + while (h.decimal_point <= 0) { + uint32_t shift; + if (h.decimal_point == 0) { + if (h.digits[0] >= 5) { + break; + } + shift = (h.digits[0] <= 2) ? 2 : 1; + } else { + uint32_t n = (uint32_t)(-h.decimal_point); + shift = (n < num_powers) + ? powers[n] + : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; + } + + wuffs_base__private_implementation__high_prec_dec__small_lshift(&h, + shift); + if (h.decimal_point > + +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { + goto infinity; + } + exp2 -= (int32_t)shift; + } + + // We're in the range [½ .. 1] but f64 uses [1 .. 2]. + exp2--; + + // The minimum normal exponent is (f64_bias + 1). + while ((f64_bias + 1) > exp2) { + uint32_t n = (uint32_t)((f64_bias + 1) - exp2); + if (n > WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL) { + n = WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; + } + wuffs_base__private_implementation__high_prec_dec__small_rshift(&h, n); + exp2 += (int32_t)n; + } + + // Check for overflow. + if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1. + goto infinity; + } + + // Extract 53 bits for the mantissa (in base-2). + wuffs_base__private_implementation__high_prec_dec__small_lshift(&h, 53); + uint64_t man2 = + wuffs_base__private_implementation__high_prec_dec__rounded_integer(&h); + + // Rounding might have added one bit. If so, shift and re-check overflow. + if ((man2 >> 53) != 0) { + man2 >>= 1; + exp2++; + if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1. + goto infinity; + } + } + + // Handle subnormal numbers. + if ((man2 >> 52) == 0) { + exp2 = f64_bias; + } + + // Pack the bits and return. + uint64_t exp2_bits = + (uint64_t)((exp2 - f64_bias) & 0x07FF); // (1 << 11) - 1. + uint64_t bits = (man2 & 0x000FFFFFFFFFFFFF) | // (1 << 52) - 1. + (exp2_bits << 52) | // + (h.negative ? 0x8000000000000000 : 0); // (1 << 63). + + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); + return ret; + } while (0); + +zero: + do { + uint64_t bits = h.negative ? 0x8000000000000000 : 0; + + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); + return ret; + } while (0); + +infinity: + do { + uint64_t bits = h.negative ? 0xFFF0000000000000 : 0x7FF0000000000000; + + wuffs_base__result_f64 ret; + ret.status.repr = NULL; + ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); + return ret; + } while (0); +} diff --git a/internal/cgen/base/image-impl.c b/internal/cgen/base/image-impl.c index 413eec59..6f4baa17 100644 --- a/internal/cgen/base/image-impl.c +++ b/internal/cgen/base/image-impl.c @@ -23,51 +23,6 @@ const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { // -------- -static uint64_t // -wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - - size_t n = len4; - while (n--) { - uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); - uint32_t b5 = 0x1F & (argb >> (8 - 5)); - uint32_t g6 = 0x3F & (argb >> (16 - 6)); - uint32_t r5 = 0x1F & (argb >> (24 - 5)); - uint32_t alpha = argb & 0xFF000000; - wuffs_base__store_u32le__no_bounds_check( - d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0)); - s += 4; - d += 4; - } - return len4 * 4; -} - -static uint64_t // -wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - - size_t n = len4; - while (n--) { - uint8_t b0 = s[0]; - uint8_t b1 = s[1]; - uint8_t b2 = s[2]; - uint8_t b3 = s[3]; - d[0] = b2; - d[1] = b1; - d[2] = b0; - d[3] = b3; - s += 4; - d += 4; - } - return len4 * 4; -} - static inline uint32_t // wuffs_base__swap_u32_argb_abgr(uint32_t u) { uint32_t o = u & 0xFF00FF00; @@ -78,152 +33,6 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { // -------- -static inline uint32_t // -wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, - uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); - - // Convert dst from nonpremul to premul. - dr = (dr * da) / 0xFFFF; - dg = (dg * da) / 0xFFFF; - db = (db * da) / 0xFFFF; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert dst from premul to nonpremul. - if (da != 0) { - dr = (dr * 0xFFFF) / da; - dg = (dg * 0xFFFF) / da; - db = (db * 0xFFFF) / da; - } - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, - uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); - - // Convert dst from nonpremul to premul. - dr = (dr * da) / 0xFFFF; - dg = (dg * da) / 0xFFFF; - db = (db * da) / 0xFFFF; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (premul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = sr + ((dr * ia) / 0xFFFF); - dg = sg + ((dg * ia) / 0xFFFF); - db = sb + ((db * ia) / 0xFFFF); - - // Convert dst from premul to nonpremul. - if (da != 0) { - dr = (dr * 0xFFFF) / da; - dg = (dg * 0xFFFF) / da; - db = (db * 0xFFFF) / da; - } - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, - uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, - uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (premul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = sr + ((dr * ia) / 0xFFFF); - dg = sg + ((dg * ia) / 0xFFFF); - db = sb + ((db * ia) / 0xFFFF); - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -// -------- - wuffs_base__color_u32_argb_premul // wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, uint32_t x, @@ -447,985 +256,3 @@ wuffs_base__pixel_palette__closest_element( return (uint8_t)best_index; } - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - return wuffs_base__slice_u8__copy_from_slice(dst, src); -} - -static uint64_t // -wuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len3 = src.len / 3; - size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3; - if (len > 0) { - memmove(dst.ptr, src.ptr, len * 3); - } - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - if (len > 0) { - memmove(dst.ptr, src.ptr, len * 4); - } - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len3 = src.len / 3; - size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t b5 = s[0] >> 3; - uint32_t g6 = s[1] >> 2; - uint32_t r5 = s[2] >> 3; - uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); - - s += 1 * 3; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len4 = src.len / 4; - size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), - wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))))); - - s += 1 * 4; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len4 = src.len / 4; - size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * ((uint32_t)s[3]); - uint32_t sr = 0x101 * ((uint32_t)s[2]); - uint32_t sg = 0x101 * ((uint32_t)s[1]); - uint32_t sb = 0x101 * ((uint32_t)s[0]); - - // Convert from 565 color to 16-bit color. - uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); - uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); - uint32_t dr = (0x8421 * old_r5) >> 4; - uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); - uint32_t dg = (0x1041 * old_g6) >> 2; - uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); - uint32_t db = (0x8421 * old_b5) >> 4; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 565 color and combine the components. - uint32_t new_r5 = 0x1F & (dr >> 11); - uint32_t new_g6 = 0x3F & (dg >> 10); - uint32_t new_b5 = 0x1F & (db >> 11); - uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), - (uint16_t)new_rgb_565); - - s += 1 * 4; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t y5 = s[0] >> 3; - uint32_t y6 = s[0] >> 2; - uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__index__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (1 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (2 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (3 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 2; - n -= loop_unroll_count; - } - - while (n >= 1) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0); - } - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len4 = src.len / 4; - size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))); - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - - s += 1 * 4; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len4 = src.len / 4; - size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * ((uint32_t)s[3]); - uint32_t sr = 0x101 * ((uint32_t)s[2]); - uint32_t sg = 0x101 * ((uint32_t)s[1]); - uint32_t sb = 0x101 * ((uint32_t)s[0]); - uint32_t dr = 0x101 * ((uint32_t)d[2]); - uint32_t dg = 0x101 * ((uint32_t)d[1]); - uint32_t db = 0x101 * ((uint32_t)d[0]); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 8-bit color. - d[0] = (uint8_t)(db >> 8); - d[1] = (uint8_t)(dg >> 8); - d[2] = (uint8_t)(dr >> 8); - - s += 1 * 4; - d += 1 * 3; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - // The comparison in the while condition is ">", not ">=", because with - // ">=", the last 4-byte store could write past the end of the dst slice. - // - // Each 4-byte store writes one too many bytes, but a subsequent store - // will overwrite that with the correct byte. There is always another - // store, whether a 4-byte store in this loop or a 1-byte store in the - // next loop. - while (n > loop_unroll_count) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (1 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (2 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (3 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 3; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - } - uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[1] * 4)); - if (s1) { - wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1); - } - uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[2] * 4)); - if (s2) { - wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2); - } - uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[3] * 4)); - if (s3) { - wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3); - } - - s += loop_unroll_count * 1; - d += loop_unroll_count * 3; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - } - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint8_t s0 = s[0]; - d[0] = s0; - d[1] = s0; - d[2] = s0; - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (1 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (2 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (3 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 4; - n -= loop_unroll_count; - } - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); - } - uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[1] * 4)); - if (s1) { - wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1); - } - uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[2] * 4)); - if (s2) { - wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2); - } - uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[3] * 4)); - if (s3) { - wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3); - } - - s += loop_unroll_count * 1; - d += loop_unroll_count * 4; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); - } - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len3 = src.len / 3; - size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3))); - - s += 1 * 3; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0])); - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__pixel_swizzler__bgr_565__y; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - case WUFFS_BASE__PIXEL_FORMAT__RGB: - return wuffs_base__pixel_swizzler__xxx__y; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - return wuffs_base__pixel_swizzler__xxxx__y; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( - wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__copy_1_1; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr_565__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; - } - return NULL; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__pixel_swizzler__bgr_565__bgr; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - return wuffs_base__pixel_swizzler__copy_3_3; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - return wuffs_base__pixel_swizzler__xxxx__xxx; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - // TODO. - break; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( - wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__copy_4_4; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - // TODO. - break; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - // TODO. - break; - } - return NULL; -} - -// -------- - -wuffs_base__status // -wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__pixel_format src_format, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - if (!p) { - return wuffs_base__make_status(wuffs_base__error__bad_receiver); - } - - // TODO: support many more formats. - - wuffs_base__pixel_swizzler__func func = NULL; - - switch (src_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__Y: - func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette, - src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( - p, dst_format, dst_palette, src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - func = wuffs_base__pixel_swizzler__prepare__bgr( - p, dst_format, dst_palette, src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( - p, dst_format, dst_palette, src_palette, blend); - break; - } - - p->private_impl.func = func; - return wuffs_base__make_status( - func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option); -} - -uint64_t // -wuffs_base__pixel_swizzler__swizzle_interleaved( - const wuffs_base__pixel_swizzler* p, - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (p && p->private_impl.func) { - return (*p->private_impl.func)(dst, dst_palette, src); - } - return 0; -} diff --git a/internal/cgen/base/pixconv-submodule.c b/internal/cgen/base/pixconv-submodule.c new file mode 100644 index 00000000..1e065bd3 --- /dev/null +++ b/internal/cgen/base/pixconv-submodule.c @@ -0,0 +1,1190 @@ +// After editing this file, run "go generate" in the parent directory. + +// Copyright 2017 The Wuffs Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// https://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// ---------------- Pixel Swizzler + +static inline uint32_t // +wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_nonpremul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +static inline uint32_t // +wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +static inline uint32_t // +wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, + uint32_t src_nonpremul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +static inline uint32_t // +wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + + size_t n = len4; + while (n--) { + uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); + uint32_t b5 = 0x1F & (argb >> (8 - 5)); + uint32_t g6 = 0x3F & (argb >> (16 - 6)); + uint32_t r5 = 0x1F & (argb >> (24 - 5)); + uint32_t alpha = argb & 0xFF000000; + wuffs_base__store_u32le__no_bounds_check( + d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0)); + s += 4; + d += 4; + } + return len4 * 4; +} + +static uint64_t // +wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + + size_t n = len4; + while (n--) { + uint8_t b0 = s[0]; + uint8_t b1 = s[1]; + uint8_t b2 = s[2]; + uint8_t b3 = s[3]; + d[0] = b2; + d[1] = b1; + d[2] = b0; + d[3] = b3; + s += 4; + d += 4; + } + return len4 * 4; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + return wuffs_base__slice_u8__copy_from_slice(dst, src); +} + +static uint64_t // +wuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len3 = src.len / 3; + size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 3); + } + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 4); + } + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len3 = src.len / 3; + size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t b5 = s[0] >> 3; + uint32_t g6 = s[1] >> 2; + uint32_t r5 = s[2] >> 3; + uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); + + s += 1 * 3; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len4 = src.len / 4; + size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))))); + + s += 1 * 4; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len4 = src.len / 4; + size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * ((uint32_t)s[3]); + uint32_t sr = 0x101 * ((uint32_t)s[2]); + uint32_t sg = 0x101 * ((uint32_t)s[1]); + uint32_t sb = 0x101 * ((uint32_t)s[0]); + + // Convert from 565 color to 16-bit color. + uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); + uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); + uint32_t dr = (0x8421 * old_r5) >> 4; + uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); + uint32_t dg = (0x1041 * old_g6) >> 2; + uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); + uint32_t db = (0x8421 * old_b5) >> 4; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 565 color and combine the components. + uint32_t new_r5 = 0x1F & (dr >> 11); + uint32_t new_g6 = 0x3F & (dg >> 10); + uint32_t new_b5 = 0x1F & (db >> 11); + uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), + (uint16_t)new_rgb_565); + + s += 1 * 4; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t y5 = s[0] >> 3; + uint32_t y6 = s[0] >> 2; + uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__index__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (1 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (2 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (3 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 2; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0); + } + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len4 = src.len / 4; + size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 4; + d += 1 * 3; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len4 = src.len / 4; + size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * ((uint32_t)s[3]); + uint32_t sr = 0x101 * ((uint32_t)s[2]); + uint32_t sg = 0x101 * ((uint32_t)s[1]); + uint32_t sb = 0x101 * ((uint32_t)s[0]); + uint32_t dr = 0x101 * ((uint32_t)d[2]); + uint32_t dg = 0x101 * ((uint32_t)d[1]); + uint32_t db = 0x101 * ((uint32_t)d[0]); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color. + d[0] = (uint8_t)(db >> 8); + d[1] = (uint8_t)(dg >> 8); + d[2] = (uint8_t)(dr >> 8); + + s += 1 * 4; + d += 1 * 3; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + // The comparison in the while condition is ">", not ">=", because with + // ">=", the last 4-byte store could write past the end of the dst slice. + // + // Each 4-byte store writes one too many bytes, but a subsequent store + // will overwrite that with the correct byte. There is always another + // store, whether a 4-byte store in this loop or a 1-byte store in the + // next loop. + while (n > loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 3; + n -= loop_unroll_count; + } + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + } + uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[1] * 4)); + if (s1) { + wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1); + } + uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[2] * 4)); + if (s2) { + wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2); + } + uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[3] * 4)); + if (s3) { + wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3); + } + + s += loop_unroll_count * 1; + d += loop_unroll_count * 3; + n -= loop_unroll_count; + } + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + } + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint8_t s0 = s[0]; + d[0] = s0; + d[1] = s0; + d[2] = s0; + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); + } + uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[1] * 4)); + if (s1) { + wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1); + } + uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[2] * 4)); + if (s2) { + wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2); + } + uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[3] * 4)); + if (s3) { + wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3); + } + + s += loop_unroll_count * 1; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); + } + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len3 = src.len / 3; + size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3))); + + s += 1 * 3; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0])); + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__pixel_swizzler__bgr_565__y; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + case WUFFS_BASE__PIXEL_FORMAT__RGB: + return wuffs_base__pixel_swizzler__xxx__y; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + return wuffs_base__pixel_swizzler__xxxx__y; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_1_1; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; + } + return NULL; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__pixel_swizzler__bgr_565__bgr; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + return wuffs_base__pixel_swizzler__copy_3_3; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + return wuffs_base__pixel_swizzler__xxxx__xxx; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_4_4; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + +// -------- + +wuffs_base__status // +wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__pixel_format src_format, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + if (!p) { + return wuffs_base__make_status(wuffs_base__error__bad_receiver); + } + + // TODO: support many more formats. + + wuffs_base__pixel_swizzler__func func = NULL; + + switch (src_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__Y: + func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette, + src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( + p, dst_format, dst_palette, src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + func = wuffs_base__pixel_swizzler__prepare__bgr( + p, dst_format, dst_palette, src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + p, dst_format, dst_palette, src_palette, blend); + break; + } + + p->private_impl.func = func; + return wuffs_base__make_status( + func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option); +} + +uint64_t // +wuffs_base__pixel_swizzler__swizzle_interleaved( + const wuffs_base__pixel_swizzler* p, + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (p && p->private_impl.func) { + return (*p->private_impl.func)(dst, dst_palette, src); + } + return 0; +} diff --git a/internal/cgen/base/strconv-impl.c b/internal/cgen/base/strconv-impl.c index 06710a1e..9d290409 100644 --- a/internal/cgen/base/strconv-impl.c +++ b/internal/cgen/base/strconv-impl.c @@ -299,1255 +299,6 @@ fail_out_of_bounds: } while (0); } - // ---------------- IEEE 754 Floating Point - -#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE 1023 -#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION 500 - -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL is the largest N -// such that ((10 << N) < (1 << 64)). -#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL 60 - -// wuffs_base__private_implementation__high_prec_dec (abbreviated as HPD) is a -// fixed precision floating point decimal number, augmented with ±infinity -// values, but it cannot represent NaN (Not a Number). -// -// "High precision" means that the mantissa holds 500 decimal digits. 500 is -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION. -// -// An HPD isn't for general purpose arithmetic, only for conversions to and -// from IEEE 754 double-precision floating point, where the largest and -// smallest positive, finite values are approximately 1.8e+308 and 4.9e-324. -// HPD exponents above +1023 mean infinity, below -1023 mean zero. The ±1023 -// bounds are further away from zero than ±(324 + 500), where 500 and 1023 is -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION and -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE. -// -// digits[.. num_digits] are the number's digits in big-endian order. The -// uint8_t values are in the range [0 ..= 9], not ['0' ..= '9'], where e.g. '7' -// is the ASCII value 0x37. -// -// decimal_point is the index (within digits) of the decimal point. It may be -// negative or be larger than num_digits, in which case the explicit digits are -// padded with implicit zeroes. -// -// For example, if num_digits is 3 and digits is "\x07\x08\x09": -// - A decimal_point of -2 means ".00789" -// - A decimal_point of -1 means ".0789" -// - A decimal_point of +0 means ".789" -// - A decimal_point of +1 means "7.89" -// - A decimal_point of +2 means "78.9" -// - A decimal_point of +3 means "789." -// - A decimal_point of +4 means "7890." -// - A decimal_point of +5 means "78900." -// -// As above, a decimal_point higher than +1023 means that the overall value is -// infinity, lower than -1023 means zero. -// -// negative is a sign bit. An HPD can distinguish positive and negative zero. -// -// truncated is whether there are more than -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION digits, and at -// least one of those extra digits are non-zero. The existence of long-tail -// digits can affect rounding. -// -// The "all fields are zero" value is valid, and represents the number +0. -typedef struct { - uint32_t num_digits; - int32_t decimal_point; - bool negative; - bool truncated; - uint8_t digits[WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION]; -} wuffs_base__private_implementation__high_prec_dec; - -// wuffs_base__private_implementation__high_prec_dec__trim trims trailing -// zeroes from the h->digits[.. h->num_digits] slice. They have no benefit, -// since we explicitly track h->decimal_point. -// -// Preconditions: -// - h is non-NULL. -static inline void // -wuffs_base__private_implementation__high_prec_dec__trim( - wuffs_base__private_implementation__high_prec_dec* h) { - while ((h->num_digits > 0) && (h->digits[h->num_digits - 1] == 0)) { - h->num_digits--; - } -} - -static wuffs_base__status // -wuffs_base__private_implementation__high_prec_dec__parse( - wuffs_base__private_implementation__high_prec_dec* h, - wuffs_base__slice_u8 s) { - if (!h) { - return wuffs_base__make_status(wuffs_base__error__bad_receiver); - } - h->num_digits = 0; - h->decimal_point = 0; - h->negative = false; - h->truncated = false; - - uint8_t* p = s.ptr; - uint8_t* q = s.ptr + s.len; - - for (; (p < q) && (*p == '_'); p++) { - } - if (p >= q) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - - // Parse sign. - do { - if (*p == '+') { - p++; - } else if (*p == '-') { - h->negative = true; - p++; - } else { - break; - } - for (; (p < q) && (*p == '_'); p++) { - } - } while (0); - - // Parse digits. - uint32_t nd = 0; - int32_t dp = 0; - bool saw_digits = false; - bool saw_non_zero_digits = false; - bool saw_dot = false; - for (; p < q; p++) { - if (*p == '_') { - // No-op. - - } else if ((*p == '.') || (*p == ',')) { - // As per https://en.wikipedia.org/wiki/Decimal_separator, both '.' or - // ',' are commonly used. We just parse either, regardless of LOCALE. - if (saw_dot) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - saw_dot = true; - dp = (int32_t)nd; - - } else if ('0' == *p) { - if (!saw_dot && !saw_non_zero_digits && saw_digits) { - // We don't allow unnecessary leading zeroes: "000123" or "0644". - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - saw_digits = true; - if (nd == 0) { - // Track leading zeroes implicitly. - dp--; - } else if (nd < - WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->digits[nd++] = 0; - } else { - // Long-tail zeroes are ignored. - } - - } else if (('0' < *p) && (*p <= '9')) { - if (!saw_dot && !saw_non_zero_digits && saw_digits) { - // We don't allow unnecessary leading zeroes: "000123" or "0644". - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - saw_digits = true; - saw_non_zero_digits = true; - if (nd < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->digits[nd++] = (uint8_t)(*p - '0'); - } else { - // Long-tail non-zeroes set the truncated bit. - h->truncated = true; - } - - } else { - break; - } - } - - if (!saw_digits) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - if (!saw_dot) { - dp = (int32_t)nd; - } - - // Parse exponent. - if ((p < q) && ((*p == 'E') || (*p == 'e'))) { - p++; - for (; (p < q) && (*p == '_'); p++) { - } - if (p >= q) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - - int32_t exp_sign = +1; - if (*p == '+') { - p++; - } else if (*p == '-') { - exp_sign = -1; - p++; - } - - int32_t exp = 0; - const int32_t exp_large = - WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE + - WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION; - bool saw_exp_digits = false; - for (; p < q; p++) { - if (*p == '_') { - // No-op. - } else if (('0' <= *p) && (*p <= '9')) { - saw_exp_digits = true; - if (exp < exp_large) { - exp = (10 * exp) + ((int32_t)(*p - '0')); - } - } else { - break; - } - } - if (!saw_exp_digits) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - dp += exp_sign * exp; - } - - // Finish. - if (p != q) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - h->num_digits = nd; - if (nd == 0) { - h->decimal_point = 0; - } else if (dp < - -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { - h->decimal_point = - -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE - 1; - } else if (dp > - +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { - h->decimal_point = - +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE + 1; - } else { - h->decimal_point = dp; - } - wuffs_base__private_implementation__high_prec_dec__trim(h); - return wuffs_base__make_status(NULL); -} - -// -------- - -// The etc__hpd_left_shift and etc__powers_of_5 tables were printed by -// script/print-hpd-left-shift.go. That script has an optional -comments flag, -// whose output is not copied here, which prints further detail. -// -// These tables are used in -// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits. - -// wuffs_base__private_implementation__hpd_left_shift[i] encodes the number of -// new digits created after multiplying a positive integer by (1 << i): the -// additional length in the decimal representation. For example, shifting "234" -// by 3 (equivalent to multiplying by 8) will produce "1872". Going from a -// 3-length string to a 4-length string means that 1 new digit was added (and -// existing digits may have changed). -// -// Shifting by i can add either N or N-1 new digits, depending on whether the -// original positive integer compares >= or < to the i'th power of 5 (as 10 -// equals 2 * 5). Comparison is lexicographic, not numerical. -// -// For example, shifting by 4 (i.e. multiplying by 16) can add 1 or 2 new -// digits, depending on a lexicographic comparison to (5 ** 4), i.e. "625": -// - ("1" << 4) is "16", which adds 1 new digit. -// - ("5678" << 4) is "90848", which adds 1 new digit. -// - ("624" << 4) is "9984", which adds 1 new digit. -// - ("62498" << 4) is "999968", which adds 1 new digit. -// - ("625" << 4) is "10000", which adds 2 new digits. -// - ("625001" << 4) is "10000016", which adds 2 new digits. -// - ("7008" << 4) is "112128", which adds 2 new digits. -// - ("99" << 4) is "1584", which adds 2 new digits. -// -// Thus, when i is 4, N is 2 and (5 ** i) is "625". This etc__hpd_left_shift -// array encodes this as: -// - etc__hpd_left_shift[4] is 0x1006 = (2 << 11) | 0x0006. -// - etc__hpd_left_shift[5] is 0x1009 = (? << 11) | 0x0009. -// where the ? isn't relevant for i == 4. -// -// The high 5 bits of etc__hpd_left_shift[i] is N, the higher of the two -// possible number of new digits. The low 11 bits are an offset into the -// etc__powers_of_5 array (of length 0x051C, so offsets fit in 11 bits). When i -// is 4, its offset and the next one is 6 and 9, and etc__powers_of_5[6 .. 9] -// is the string "\x06\x02\x05", so the relevant power of 5 is "625". -// -// Thanks to Ken Thompson for the original idea. -static const uint16_t wuffs_base__private_implementation__hpd_left_shift[65] = { - 0x0000, 0x0800, 0x0801, 0x0803, 0x1006, 0x1009, 0x100D, 0x1812, 0x1817, - 0x181D, 0x2024, 0x202B, 0x2033, 0x203C, 0x2846, 0x2850, 0x285B, 0x3067, - 0x3073, 0x3080, 0x388E, 0x389C, 0x38AB, 0x38BB, 0x40CC, 0x40DD, 0x40EF, - 0x4902, 0x4915, 0x4929, 0x513E, 0x5153, 0x5169, 0x5180, 0x5998, 0x59B0, - 0x59C9, 0x61E3, 0x61FD, 0x6218, 0x6A34, 0x6A50, 0x6A6D, 0x6A8B, 0x72AA, - 0x72C9, 0x72E9, 0x7B0A, 0x7B2B, 0x7B4D, 0x8370, 0x8393, 0x83B7, 0x83DC, - 0x8C02, 0x8C28, 0x8C4F, 0x9477, 0x949F, 0x94C8, 0x9CF2, 0x051C, 0x051C, - 0x051C, 0x051C, -}; - -// wuffs_base__private_implementation__powers_of_5 contains the powers of 5, -// concatenated together: "5", "25", "125", "625", "3125", etc. -static const uint8_t wuffs_base__private_implementation__powers_of_5[0x051C] = { - 5, 2, 5, 1, 2, 5, 6, 2, 5, 3, 1, 2, 5, 1, 5, 6, 2, 5, 7, 8, 1, 2, 5, 3, 9, - 0, 6, 2, 5, 1, 9, 5, 3, 1, 2, 5, 9, 7, 6, 5, 6, 2, 5, 4, 8, 8, 2, 8, 1, 2, - 5, 2, 4, 4, 1, 4, 0, 6, 2, 5, 1, 2, 2, 0, 7, 0, 3, 1, 2, 5, 6, 1, 0, 3, 5, - 1, 5, 6, 2, 5, 3, 0, 5, 1, 7, 5, 7, 8, 1, 2, 5, 1, 5, 2, 5, 8, 7, 8, 9, 0, - 6, 2, 5, 7, 6, 2, 9, 3, 9, 4, 5, 3, 1, 2, 5, 3, 8, 1, 4, 6, 9, 7, 2, 6, 5, - 6, 2, 5, 1, 9, 0, 7, 3, 4, 8, 6, 3, 2, 8, 1, 2, 5, 9, 5, 3, 6, 7, 4, 3, 1, - 6, 4, 0, 6, 2, 5, 4, 7, 6, 8, 3, 7, 1, 5, 8, 2, 0, 3, 1, 2, 5, 2, 3, 8, 4, - 1, 8, 5, 7, 9, 1, 0, 1, 5, 6, 2, 5, 1, 1, 9, 2, 0, 9, 2, 8, 9, 5, 5, 0, 7, - 8, 1, 2, 5, 5, 9, 6, 0, 4, 6, 4, 4, 7, 7, 5, 3, 9, 0, 6, 2, 5, 2, 9, 8, 0, - 2, 3, 2, 2, 3, 8, 7, 6, 9, 5, 3, 1, 2, 5, 1, 4, 9, 0, 1, 1, 6, 1, 1, 9, 3, - 8, 4, 7, 6, 5, 6, 2, 5, 7, 4, 5, 0, 5, 8, 0, 5, 9, 6, 9, 2, 3, 8, 2, 8, 1, - 2, 5, 3, 7, 2, 5, 2, 9, 0, 2, 9, 8, 4, 6, 1, 9, 1, 4, 0, 6, 2, 5, 1, 8, 6, - 2, 6, 4, 5, 1, 4, 9, 2, 3, 0, 9, 5, 7, 0, 3, 1, 2, 5, 9, 3, 1, 3, 2, 2, 5, - 7, 4, 6, 1, 5, 4, 7, 8, 5, 1, 5, 6, 2, 5, 4, 6, 5, 6, 6, 1, 2, 8, 7, 3, 0, - 7, 7, 3, 9, 2, 5, 7, 8, 1, 2, 5, 2, 3, 2, 8, 3, 0, 6, 4, 3, 6, 5, 3, 8, 6, - 9, 6, 2, 8, 9, 0, 6, 2, 5, 1, 1, 6, 4, 1, 5, 3, 2, 1, 8, 2, 6, 9, 3, 4, 8, - 1, 4, 4, 5, 3, 1, 2, 5, 5, 8, 2, 0, 7, 6, 6, 0, 9, 1, 3, 4, 6, 7, 4, 0, 7, - 2, 2, 6, 5, 6, 2, 5, 2, 9, 1, 0, 3, 8, 3, 0, 4, 5, 6, 7, 3, 3, 7, 0, 3, 6, - 1, 3, 2, 8, 1, 2, 5, 1, 4, 5, 5, 1, 9, 1, 5, 2, 2, 8, 3, 6, 6, 8, 5, 1, 8, - 0, 6, 6, 4, 0, 6, 2, 5, 7, 2, 7, 5, 9, 5, 7, 6, 1, 4, 1, 8, 3, 4, 2, 5, 9, - 0, 3, 3, 2, 0, 3, 1, 2, 5, 3, 6, 3, 7, 9, 7, 8, 8, 0, 7, 0, 9, 1, 7, 1, 2, - 9, 5, 1, 6, 6, 0, 1, 5, 6, 2, 5, 1, 8, 1, 8, 9, 8, 9, 4, 0, 3, 5, 4, 5, 8, - 5, 6, 4, 7, 5, 8, 3, 0, 0, 7, 8, 1, 2, 5, 9, 0, 9, 4, 9, 4, 7, 0, 1, 7, 7, - 2, 9, 2, 8, 2, 3, 7, 9, 1, 5, 0, 3, 9, 0, 6, 2, 5, 4, 5, 4, 7, 4, 7, 3, 5, - 0, 8, 8, 6, 4, 6, 4, 1, 1, 8, 9, 5, 7, 5, 1, 9, 5, 3, 1, 2, 5, 2, 2, 7, 3, - 7, 3, 6, 7, 5, 4, 4, 3, 2, 3, 2, 0, 5, 9, 4, 7, 8, 7, 5, 9, 7, 6, 5, 6, 2, - 5, 1, 1, 3, 6, 8, 6, 8, 3, 7, 7, 2, 1, 6, 1, 6, 0, 2, 9, 7, 3, 9, 3, 7, 9, - 8, 8, 2, 8, 1, 2, 5, 5, 6, 8, 4, 3, 4, 1, 8, 8, 6, 0, 8, 0, 8, 0, 1, 4, 8, - 6, 9, 6, 8, 9, 9, 4, 1, 4, 0, 6, 2, 5, 2, 8, 4, 2, 1, 7, 0, 9, 4, 3, 0, 4, - 0, 4, 0, 0, 7, 4, 3, 4, 8, 4, 4, 9, 7, 0, 7, 0, 3, 1, 2, 5, 1, 4, 2, 1, 0, - 8, 5, 4, 7, 1, 5, 2, 0, 2, 0, 0, 3, 7, 1, 7, 4, 2, 2, 4, 8, 5, 3, 5, 1, 5, - 6, 2, 5, 7, 1, 0, 5, 4, 2, 7, 3, 5, 7, 6, 0, 1, 0, 0, 1, 8, 5, 8, 7, 1, 1, - 2, 4, 2, 6, 7, 5, 7, 8, 1, 2, 5, 3, 5, 5, 2, 7, 1, 3, 6, 7, 8, 8, 0, 0, 5, - 0, 0, 9, 2, 9, 3, 5, 5, 6, 2, 1, 3, 3, 7, 8, 9, 0, 6, 2, 5, 1, 7, 7, 6, 3, - 5, 6, 8, 3, 9, 4, 0, 0, 2, 5, 0, 4, 6, 4, 6, 7, 7, 8, 1, 0, 6, 6, 8, 9, 4, - 5, 3, 1, 2, 5, 8, 8, 8, 1, 7, 8, 4, 1, 9, 7, 0, 0, 1, 2, 5, 2, 3, 2, 3, 3, - 8, 9, 0, 5, 3, 3, 4, 4, 7, 2, 6, 5, 6, 2, 5, 4, 4, 4, 0, 8, 9, 2, 0, 9, 8, - 5, 0, 0, 6, 2, 6, 1, 6, 1, 6, 9, 4, 5, 2, 6, 6, 7, 2, 3, 6, 3, 2, 8, 1, 2, - 5, 2, 2, 2, 0, 4, 4, 6, 0, 4, 9, 2, 5, 0, 3, 1, 3, 0, 8, 0, 8, 4, 7, 2, 6, - 3, 3, 3, 6, 1, 8, 1, 6, 4, 0, 6, 2, 5, 1, 1, 1, 0, 2, 2, 3, 0, 2, 4, 6, 2, - 5, 1, 5, 6, 5, 4, 0, 4, 2, 3, 6, 3, 1, 6, 6, 8, 0, 9, 0, 8, 2, 0, 3, 1, 2, - 5, 5, 5, 5, 1, 1, 1, 5, 1, 2, 3, 1, 2, 5, 7, 8, 2, 7, 0, 2, 1, 1, 8, 1, 5, - 8, 3, 4, 0, 4, 5, 4, 1, 0, 1, 5, 6, 2, 5, 2, 7, 7, 5, 5, 5, 7, 5, 6, 1, 5, - 6, 2, 8, 9, 1, 3, 5, 1, 0, 5, 9, 0, 7, 9, 1, 7, 0, 2, 2, 7, 0, 5, 0, 7, 8, - 1, 2, 5, 1, 3, 8, 7, 7, 7, 8, 7, 8, 0, 7, 8, 1, 4, 4, 5, 6, 7, 5, 5, 2, 9, - 5, 3, 9, 5, 8, 5, 1, 1, 3, 5, 2, 5, 3, 9, 0, 6, 2, 5, 6, 9, 3, 8, 8, 9, 3, - 9, 0, 3, 9, 0, 7, 2, 2, 8, 3, 7, 7, 6, 4, 7, 6, 9, 7, 9, 2, 5, 5, 6, 7, 6, - 2, 6, 9, 5, 3, 1, 2, 5, 3, 4, 6, 9, 4, 4, 6, 9, 5, 1, 9, 5, 3, 6, 1, 4, 1, - 8, 8, 8, 2, 3, 8, 4, 8, 9, 6, 2, 7, 8, 3, 8, 1, 3, 4, 7, 6, 5, 6, 2, 5, 1, - 7, 3, 4, 7, 2, 3, 4, 7, 5, 9, 7, 6, 8, 0, 7, 0, 9, 4, 4, 1, 1, 9, 2, 4, 4, - 8, 1, 3, 9, 1, 9, 0, 6, 7, 3, 8, 2, 8, 1, 2, 5, 8, 6, 7, 3, 6, 1, 7, 3, 7, - 9, 8, 8, 4, 0, 3, 5, 4, 7, 2, 0, 5, 9, 6, 2, 2, 4, 0, 6, 9, 5, 9, 5, 3, 3, - 6, 9, 1, 4, 0, 6, 2, 5, -}; - -// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits -// returns the number of additional decimal digits when left-shifting by shift. -// -// See below for preconditions. -static uint32_t // -wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits( - wuffs_base__private_implementation__high_prec_dec* h, - uint32_t shift) { - // Masking with 0x3F should be unnecessary (assuming the preconditions) but - // it's cheap and ensures that we don't overflow the - // wuffs_base__private_implementation__hpd_left_shift array. - shift &= 63; - - uint32_t x_a = wuffs_base__private_implementation__hpd_left_shift[shift]; - uint32_t x_b = wuffs_base__private_implementation__hpd_left_shift[shift + 1]; - uint32_t num_new_digits = x_a >> 11; - uint32_t pow5_a = 0x7FF & x_a; - uint32_t pow5_b = 0x7FF & x_b; - - const uint8_t* pow5 = - &wuffs_base__private_implementation__powers_of_5[pow5_a]; - uint32_t i = 0; - uint32_t n = pow5_b - pow5_a; - for (; i < n; i++) { - if (i >= h->num_digits) { - return num_new_digits - 1; - } else if (h->digits[i] == pow5[i]) { - continue; - } else if (h->digits[i] < pow5[i]) { - return num_new_digits - 1; - } else { - return num_new_digits; - } - } - return num_new_digits; -} - -// -------- - -// wuffs_base__private_implementation__high_prec_dec__rounded_integer returns -// the integral (non-fractional) part of h, provided that it is 18 or fewer -// decimal digits. For 19 or more digits, it returns UINT64_MAX. Note that: -// - (1 << 53) is 9007199254740992, which has 16 decimal digits. -// - (1 << 56) is 72057594037927936, which has 17 decimal digits. -// - (1 << 59) is 576460752303423488, which has 18 decimal digits. -// - (1 << 63) is 9223372036854775808, which has 19 decimal digits. -// and that IEEE 754 double precision has 52 mantissa bits. -// -// That integral part is rounded-to-even: rounding 7.5 or 8.5 both give 8. -// -// h's negative bit is ignored: rounding -8.6 returns 9. -// -// See below for preconditions. -static uint64_t // -wuffs_base__private_implementation__high_prec_dec__rounded_integer( - wuffs_base__private_implementation__high_prec_dec* h) { - if ((h->num_digits == 0) || (h->decimal_point < 0)) { - return 0; - } else if (h->decimal_point > 18) { - return UINT64_MAX; - } - - uint32_t dp = (uint32_t)(h->decimal_point); - uint64_t n = 0; - uint32_t i = 0; - for (; i < dp; i++) { - n = (10 * n) + ((i < h->num_digits) ? h->digits[i] : 0); - } - - bool round_up = false; - if (dp < h->num_digits) { - round_up = h->digits[dp] >= 5; - if ((h->digits[dp] == 5) && (dp + 1 == h->num_digits)) { - // We are exactly halfway. If we're truncated, round up, otherwise round - // to even. - round_up = h->truncated || // - ((dp > 0) && (1 & h->digits[dp - 1])); - } - } - if (round_up) { - n++; - } - - return n; -} - -// wuffs_base__private_implementation__high_prec_dec__small_xshift shifts h's -// number (where 'x' is 'l' or 'r' for left or right) by a small shift value. -// -// Preconditions: -// - h is non-NULL. -// - h->decimal_point is "not extreme". -// - shift is non-zero. -// - shift is "a small shift". -// -// "Not extreme" means within -// ±WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE. -// -// "A small shift" means not more than -// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL. -// -// wuffs_base__private_implementation__high_prec_dec__rounded_integer and -// wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits -// have the same preconditions. - -static void // -wuffs_base__private_implementation__high_prec_dec__small_lshift( - wuffs_base__private_implementation__high_prec_dec* h, - uint32_t shift) { - if (h->num_digits == 0) { - return; - } - uint32_t num_new_digits = - wuffs_base__private_implementation__high_prec_dec__lshift_num_new_digits( - h, shift); - uint32_t rx = h->num_digits - 1; // Read index. - uint32_t wx = h->num_digits - 1 + num_new_digits; // Write index. - uint64_t n = 0; - - // Repeat: pick up a digit, put down a digit, right to left. - while (((int32_t)rx) >= 0) { - n += ((uint64_t)(h->digits[rx])) << shift; - uint64_t quo = n / 10; - uint64_t rem = n - (10 * quo); - if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->digits[wx] = (uint8_t)rem; - } else if (rem > 0) { - h->truncated = true; - } - n = quo; - wx--; - rx--; - } - - // Put down leading digits, right to left. - while (n > 0) { - uint64_t quo = n / 10; - uint64_t rem = n - (10 * quo); - if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->digits[wx] = (uint8_t)rem; - } else if (rem > 0) { - h->truncated = true; - } - n = quo; - wx--; - } - - // Finish. - h->num_digits += num_new_digits; - if (h->num_digits > - WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->num_digits = WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION; - } - h->decimal_point += (int32_t)num_new_digits; - wuffs_base__private_implementation__high_prec_dec__trim(h); -} - -static void // -wuffs_base__private_implementation__high_prec_dec__small_rshift( - wuffs_base__private_implementation__high_prec_dec* h, - uint32_t shift) { - uint32_t rx = 0; // Read index. - uint32_t wx = 0; // Write index. - uint64_t n = 0; - - // Pick up enough leading digits to cover the first shift. - while ((n >> shift) == 0) { - if (rx < h->num_digits) { - // Read a digit. - n = (10 * n) + h->digits[rx++]; - } else if (n == 0) { - // h's number used to be zero and remains zero. - return; - } else { - // Read sufficient implicit trailing zeroes. - while ((n >> shift) == 0) { - n = 10 * n; - rx++; - } - break; - } - } - h->decimal_point -= ((int32_t)(rx - 1)); - if (h->decimal_point < - -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { - // After the shift, h's number is effectively zero. - h->num_digits = 0; - h->decimal_point = 0; - h->negative = false; - h->truncated = false; - return; - } - - // Repeat: pick up a digit, put down a digit, left to right. - uint64_t mask = (((uint64_t)(1)) << shift) - 1; - while (rx < h->num_digits) { - uint8_t new_digit = ((uint8_t)(n >> shift)); - n = (10 * (n & mask)) + h->digits[rx++]; - h->digits[wx++] = new_digit; - } - - // Put down trailing digits, left to right. - while (n > 0) { - uint8_t new_digit = ((uint8_t)(n >> shift)); - n = 10 * (n & mask); - if (wx < WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION) { - h->digits[wx++] = new_digit; - } else if (new_digit > 0) { - h->truncated = true; - } - } - - // Finish. - h->num_digits = wx; - wuffs_base__private_implementation__high_prec_dec__trim(h); -} - -// -------- - -// The wuffs_base__private_implementation__etc_powers_of_10 tables were printed -// by script/print-mpb-powers-of-10.go. That script has an optional -comments -// flag, whose output is not copied here, which prints further detail. -// -// These tables are used in -// wuffs_base__private_implementation__medium_prec_bin__assign_from_hpd. - -// wuffs_base__private_implementation__big_powers_of_10 contains approximations -// to the powers of 10, ranging from 1e-348 to 1e+340, with the exponent -// stepping by 8: -348, -340, -332, ..., -12, -4, +4, +12, ..., +340. Each step -// consists of three uint32_t elements. There are 87 triples, 87 * 3 = 261. -// -// For example, the third approximation, for 1e-332, consists of the uint32_t -// triple (0x3055AC76, 0x8B16FB20, 0xFFFFFB72). The first two of that triple -// are a little-endian uint64_t value: 0x8B16FB203055AC76. The last one is an -// int32_t value: -1166. Together, they represent the approximation: -// 1e-332 ≈ 0x8B16FB203055AC76 * (2 ** -1166) -// Similarly, the (0x00000000, 0x9C400000, 0xFFFFFFCE) uint32_t triple means: -// 1e+4 ≈ 0x9C40000000000000 * (2 ** -50) // This approx'n is exact. -// Similarly, the (0xD4C4FB27, 0xED63A231, 0x000000A2) uint32_t triple means: -// 1e+68 ≈ 0xED63A231D4C4FB27 * (2 ** 162) -static const uint32_t - wuffs_base__private_implementation__big_powers_of_10[261] = { - 0x081C0288, 0xFA8FD5A0, 0xFFFFFB3C, 0xA23EBF76, 0xBAAEE17F, 0xFFFFFB57, - 0x3055AC76, 0x8B16FB20, 0xFFFFFB72, 0x5DCE35EA, 0xCF42894A, 0xFFFFFB8C, - 0x55653B2D, 0x9A6BB0AA, 0xFFFFFBA7, 0x3D1A45DF, 0xE61ACF03, 0xFFFFFBC1, - 0xC79AC6CA, 0xAB70FE17, 0xFFFFFBDC, 0xBEBCDC4F, 0xFF77B1FC, 0xFFFFFBF6, - 0x416BD60C, 0xBE5691EF, 0xFFFFFC11, 0x907FFC3C, 0x8DD01FAD, 0xFFFFFC2C, - 0x31559A83, 0xD3515C28, 0xFFFFFC46, 0xADA6C9B5, 0x9D71AC8F, 0xFFFFFC61, - 0x23EE8BCB, 0xEA9C2277, 0xFFFFFC7B, 0x4078536D, 0xAECC4991, 0xFFFFFC96, - 0x5DB6CE57, 0x823C1279, 0xFFFFFCB1, 0x4DFB5637, 0xC2109436, 0xFFFFFCCB, - 0x3848984F, 0x9096EA6F, 0xFFFFFCE6, 0x25823AC7, 0xD77485CB, 0xFFFFFD00, - 0x97BF97F4, 0xA086CFCD, 0xFFFFFD1B, 0x172AACE5, 0xEF340A98, 0xFFFFFD35, - 0x2A35B28E, 0xB23867FB, 0xFFFFFD50, 0xD2C63F3B, 0x84C8D4DF, 0xFFFFFD6B, - 0x1AD3CDBA, 0xC5DD4427, 0xFFFFFD85, 0xBB25C996, 0x936B9FCE, 0xFFFFFDA0, - 0x7D62A584, 0xDBAC6C24, 0xFFFFFDBA, 0x0D5FDAF6, 0xA3AB6658, 0xFFFFFDD5, - 0xDEC3F126, 0xF3E2F893, 0xFFFFFDEF, 0xAAFF80B8, 0xB5B5ADA8, 0xFFFFFE0A, - 0x6C7C4A8B, 0x87625F05, 0xFFFFFE25, 0x34C13053, 0xC9BCFF60, 0xFFFFFE3F, - 0x91BA2655, 0x964E858C, 0xFFFFFE5A, 0x70297EBD, 0xDFF97724, 0xFFFFFE74, - 0xB8E5B88F, 0xA6DFBD9F, 0xFFFFFE8F, 0x88747D94, 0xF8A95FCF, 0xFFFFFEA9, - 0x8FA89BCF, 0xB9447093, 0xFFFFFEC4, 0xBF0F156B, 0x8A08F0F8, 0xFFFFFEDF, - 0x653131B6, 0xCDB02555, 0xFFFFFEF9, 0xD07B7FAC, 0x993FE2C6, 0xFFFFFF14, - 0x2A2B3B06, 0xE45C10C4, 0xFFFFFF2E, 0x697392D3, 0xAA242499, 0xFFFFFF49, - 0x8300CA0E, 0xFD87B5F2, 0xFFFFFF63, 0x92111AEB, 0xBCE50864, 0xFFFFFF7E, - 0x6F5088CC, 0x8CBCCC09, 0xFFFFFF99, 0xE219652C, 0xD1B71758, 0xFFFFFFB3, - 0x00000000, 0x9C400000, 0xFFFFFFCE, 0x00000000, 0xE8D4A510, 0xFFFFFFE8, - 0xAC620000, 0xAD78EBC5, 0x00000003, 0xF8940984, 0x813F3978, 0x0000001E, - 0xC90715B3, 0xC097CE7B, 0x00000038, 0x7BEA5C70, 0x8F7E32CE, 0x00000053, - 0xABE98068, 0xD5D238A4, 0x0000006D, 0x179A2245, 0x9F4F2726, 0x00000088, - 0xD4C4FB27, 0xED63A231, 0x000000A2, 0x8CC8ADA8, 0xB0DE6538, 0x000000BD, - 0x1AAB65DB, 0x83C7088E, 0x000000D8, 0x42711D9A, 0xC45D1DF9, 0x000000F2, - 0xA61BE758, 0x924D692C, 0x0000010D, 0x1A708DEA, 0xDA01EE64, 0x00000127, - 0x9AEF774A, 0xA26DA399, 0x00000142, 0xB47D6B85, 0xF209787B, 0x0000015C, - 0x79DD1877, 0xB454E4A1, 0x00000177, 0x5B9BC5C2, 0x865B8692, 0x00000192, - 0xC8965D3D, 0xC83553C5, 0x000001AC, 0xFA97A0B3, 0x952AB45C, 0x000001C7, - 0x99A05FE3, 0xDE469FBD, 0x000001E1, 0xDB398C25, 0xA59BC234, 0x000001FC, - 0xA3989F5C, 0xF6C69A72, 0x00000216, 0x54E9BECE, 0xB7DCBF53, 0x00000231, - 0xF22241E2, 0x88FCF317, 0x0000024C, 0xD35C78A5, 0xCC20CE9B, 0x00000266, - 0x7B2153DF, 0x98165AF3, 0x00000281, 0x971F303A, 0xE2A0B5DC, 0x0000029B, - 0x5CE3B396, 0xA8D9D153, 0x000002B6, 0xA4A7443C, 0xFB9B7CD9, 0x000002D0, - 0xA7A44410, 0xBB764C4C, 0x000002EB, 0xB6409C1A, 0x8BAB8EEF, 0x00000306, - 0xA657842C, 0xD01FEF10, 0x00000320, 0xE9913129, 0x9B10A4E5, 0x0000033B, - 0xA19C0C9D, 0xE7109BFB, 0x00000355, 0x623BF429, 0xAC2820D9, 0x00000370, - 0x7AA7CF85, 0x80444B5E, 0x0000038B, 0x03ACDD2D, 0xBF21E440, 0x000003A5, - 0x5E44FF8F, 0x8E679C2F, 0x000003C0, 0x9C8CB841, 0xD433179D, 0x000003DA, - 0xB4E31BA9, 0x9E19DB92, 0x000003F5, 0xBADF77D9, 0xEB96BF6E, 0x0000040F, - 0x9BF0EE6B, 0xAF87023B, 0x0000042A, -}; - -// wuffs_base__private_implementation__small_powers_of_10 contains -// approximations to the powers of 10, ranging from 1e+0 to 1e+7, with the -// exponent stepping by 1. Each step consists of three uint32_t elements. -// -// For example, the third approximation, for 1e+2, consists of the uint32_t -// triple (0x00000000, 0xC8000000, 0xFFFFFFC7). The first two of that triple -// are a little-endian uint64_t value: 0xC800000000000000. The last one is an -// int32_t value: -57. Together, they represent the approximation: -// 1e+2 ≈ 0xC800000000000000 * (2 ** -57) // This approx'n is exact. -// Similarly, the (0x00000000, 0x9C400000, 0xFFFFFFCE) uint32_t triple means: -// 1e+4 ≈ 0x9C40000000000000 * (2 ** -50) // This approx'n is exact. -static const uint32_t - wuffs_base__private_implementation__small_powers_of_10[24] = { - 0x00000000, 0x80000000, 0xFFFFFFC1, 0x00000000, 0xA0000000, 0xFFFFFFC4, - 0x00000000, 0xC8000000, 0xFFFFFFC7, 0x00000000, 0xFA000000, 0xFFFFFFCA, - 0x00000000, 0x9C400000, 0xFFFFFFCE, 0x00000000, 0xC3500000, 0xFFFFFFD1, - 0x00000000, 0xF4240000, 0xFFFFFFD4, 0x00000000, 0x98968000, 0xFFFFFFD8, -}; - -// wuffs_base__private_implementation__f64_powers_of_10 holds powers of 10 that -// can be exactly represented by a float64 (what C calls a double). -static const double wuffs_base__private_implementation__f64_powers_of_10[23] = { - 1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10, 1e11, - 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22, -}; - -// -------- - -// wuffs_base__private_implementation__medium_prec_bin (abbreviated as MPB) is -// a fixed precision floating point binary number. Unlike IEEE 754 Floating -// Point, it cannot represent infinity or NaN (Not a Number). -// -// "Medium precision" means that the mantissa holds 64 binary digits, a little -// more than "double precision", and sizeof(MPB) > sizeof(double). 64 is -// obviously the number of bits in a uint64_t. -// -// An MPB isn't for general purpose arithmetic, only for conversions to and -// from IEEE 754 double-precision floating point. -// -// There is no implicit mantissa bit. The mantissa field is zero if and only if -// the overall floating point value is ±0. An MPB is normalized if the mantissa -// is zero or its high bit (the 1<<63 bit) is set. -// -// There is no negative bit. An MPB can only represent non-negative numbers. -// -// The "all fields are zero" value is valid, and represents the number +0. -// -// This is the "Do It Yourself Floating Point" data structure from Loitsch, -// "Printing Floating-Point Numbers Quickly and Accurately with Integers" -// (https://www.cs.tufts.edu/~nr/cs257/archive/florian-loitsch/printf.pdf). -// -// Florian Loitsch is also the primary contributor to -// https://github.com/google/double-conversion -typedef struct { - uint64_t mantissa; - int32_t exp2; -} wuffs_base__private_implementation__medium_prec_bin; - -static uint32_t // -wuffs_base__private_implementation__medium_prec_bin__normalize( - wuffs_base__private_implementation__medium_prec_bin* m) { - if (m->mantissa == 0) { - return 0; - } - uint32_t shift = wuffs_base__count_leading_zeroes_u64(m->mantissa); - m->mantissa <<= shift; - m->exp2 -= (int32_t)shift; - return shift; -} - -// wuffs_base__private_implementation__medium_prec_bin__mul_pow_10 sets m to be -// (m * pow), where pow comes from an etc_powers_of_10 triple starting at p. -// -// The result is rounded, but not necessarily normalized. -// -// Preconditions: -// - m is non-NULL. -// - m->mantissa is non-zero. -// - m->mantissa's high bit is set (i.e. m is normalized). -// -// The etc_powers_of_10 triple is already normalized. -static void // -wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( - wuffs_base__private_implementation__medium_prec_bin* m, - const uint32_t* p) { - uint64_t p_mantissa = ((uint64_t)p[0]) | (((uint64_t)p[1]) << 32); - int32_t p_exp2 = (int32_t)p[2]; - - wuffs_base__multiply_u64__output o = - wuffs_base__multiply_u64(m->mantissa, p_mantissa); - // Round the mantissa up. It cannot overflow because the maximum possible - // value of o.hi is 0xFFFFFFFFFFFFFFFE. - m->mantissa = o.hi + (o.lo >> 63); - m->exp2 = m->exp2 + p_exp2 + 64; -} - -// wuffs_base__private_implementation__medium_prec_bin__as_f64 converts m to a -// double (what C calls a double-precision float64). -// -// Preconditions: -// - m is non-NULL. -// - m->mantissa is non-zero. -// - m->mantissa's high bit is set (i.e. m is normalized). -static double // -wuffs_base__private_implementation__medium_prec_bin__as_f64( - const wuffs_base__private_implementation__medium_prec_bin* m, - bool negative) { - uint64_t mantissa64 = m->mantissa; - // An mpb's mantissa has the implicit (binary) decimal point at the right - // hand end of the mantissa's explicit digits. A double-precision's mantissa - // has that decimal point near the left hand end. There's also an explicit - // versus implicit leading 1 bit (binary digit). Together, the difference in - // semantics corresponds to adding 63. - int32_t exp2 = m->exp2 + 63; - - // Ensure that exp2 is at least -1022, the minimum double-precision exponent - // for normal (as opposed to subnormal) numbers. - if (-1022 > exp2) { - uint32_t n = (uint32_t)(-1022 - exp2); - mantissa64 >>= n; - exp2 += (int32_t)n; - } - - // Extract the (1 + 52) bits from the 64-bit mantissa64. 52 is the number of - // explicit mantissa bits in a double-precision f64. - // - // Before, we have 64 bits and due to normalization, the high bit 'H' is 1. - // 63 55 47 etc 15 7 - // H210_9876_5432_1098_7654_etc_etc_etc_5432_1098_7654_3210 - // ++++_++++_++++_++++_++++_etc_etc_etc_++++_+..._...._.... Kept bits. - // ...._...._...H_2109_8765_etc_etc_etc_6543_2109_8765_4321 After shifting. - // After, we have 53 bits (and bit #52 is this 'H' bit). - uint64_t mantissa53 = mantissa64 >> 11; - - // Round up if the old bit #10 (the highest bit dropped by shifting) was set. - // We also fix any overflow from rounding up. - if (mantissa64 & 1024) { - mantissa53++; - if ((mantissa53 >> 53) != 0) { - mantissa53 >>= 1; - exp2++; - } - } - - // Handle double-precision infinity (a nominal exponent of 1024) and - // subnormals (an exponent of -1023 and no implicit mantissa bit, bit #52). - if (exp2 >= 1024) { - mantissa53 = 0; - exp2 = 1024; - } else if ((mantissa53 >> 52) == 0) { - exp2 = -1023; - } - - // Pack the bits and return. - const int32_t f64_bias = -1023; - uint64_t exp2_bits = - (uint64_t)((exp2 - f64_bias) & 0x07FF); // (1 << 11) - 1. - uint64_t bits = (mantissa53 & 0x000FFFFFFFFFFFFF) | // (1 << 52) - 1. - (exp2_bits << 52) | // - (negative ? 0x8000000000000000 : 0); // (1 << 63). - return wuffs_base__ieee_754_bit_representation__to_f64(bits); -} - -// wuffs_base__private_implementation__medium_prec_bin__parse_number_f64 -// converts from an HPD to a double, using an MPB as scratch space. It returns -// a NULL status.repr if there is no ambiguity in the truncation or rounding to -// a float64 (an IEEE 754 double-precision floating point value). -// -// It may modify m even if it returns a non-NULL status.repr. -static wuffs_base__result_f64 // -wuffs_base__private_implementation__medium_prec_bin__parse_number_f64( - wuffs_base__private_implementation__medium_prec_bin* m, - const wuffs_base__private_implementation__high_prec_dec* h, - bool skip_fast_path_for_tests) { - do { - // m->mantissa is a uint64_t, which is an integer approximation to a - // rational value - h's underlying digits after m's normalization. This - // error is an upper bound on the difference between the approximate and - // actual value. - // - // The DiyFpStrtod function in https://github.com/google/double-conversion - // uses a finer grain (1/8th of the ULP, Unit in the Last Place) when - // tracking error. This implementation is coarser (1 ULP) but simpler. - // - // It is an error in the "numerical approximation" sense, not in the - // typical programming sense (as in "bad input" or "a result type"). - uint64_t error = 0; - - // Convert up to 19 decimal digits (in h->digits) to 64 binary digits (in - // m->mantissa): (1e19 < (1<<64)) and ((1<<64) < 1e20). If we have more - // than 19 digits, we're truncating (with error). - uint32_t i; - uint32_t i_end = h->num_digits; - if (i_end > 19) { - i_end = 19; - error = 1; - } - uint64_t mantissa = 0; - for (i = 0; i < i_end; i++) { - mantissa = (10 * mantissa) + h->digits[i]; - } - m->mantissa = mantissa; - m->exp2 = 0; - - // Check that exp10 lies in the (big_powers_of_10 + small_powers_of_10) - // range, -348 ..= +347, stepping big_powers_of_10 by 8 (which is 87 - // triples) and small_powers_of_10 by 1 (which is 8 triples). - int32_t exp10 = h->decimal_point - ((int32_t)(i_end)); - if (exp10 < -348) { - goto fail; - } - uint32_t bpo10 = ((uint32_t)(exp10 + 348)) / 8; - uint32_t spo10 = ((uint32_t)(exp10 + 348)) % 8; - if (bpo10 >= 87) { - goto fail; - } - - // Try a fast path, if float64 math would be exact. - // - // 15 is such that 1e15 can be losslessly represented in a float64 - // mantissa: (1e15 < (1<<53)) and ((1<<53) < 1e16). - // - // 22 is the maximum valid index for the - // wuffs_base__private_implementation__f64_powers_of_10 array. - do { - if (skip_fast_path_for_tests || ((mantissa >> 52) != 0)) { - break; - } - double d = (double)mantissa; - - if (exp10 == 0) { - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = h->negative ? -d : +d; - return ret; - - } else if (exp10 > 0) { - if (exp10 > 22) { - if (exp10 > (15 + 22)) { - break; - } - // If exp10 is in the range 23 ..= 37, try moving a few of the zeroes - // from the exponent to the mantissa. If we're still under 1e15, we - // haven't truncated any mantissa bits. - if (exp10 > 22) { - d *= wuffs_base__private_implementation__f64_powers_of_10[exp10 - - 22]; - exp10 = 22; - if (d >= 1e15) { - break; - } - } - } - d *= wuffs_base__private_implementation__f64_powers_of_10[exp10]; - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = h->negative ? -d : +d; - return ret; - - } else { // "if (exp10 < 0)" is effectively "if (true)" here. - if (exp10 < -22) { - break; - } - d /= wuffs_base__private_implementation__f64_powers_of_10[-exp10]; - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = h->negative ? -d : +d; - return ret; - } - } while (0); - - // Normalize (and scale the error). - error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); - - // Multiplying two MPB values nominally multiplies two mantissas, call them - // A and B, which are integer approximations to the precise values (A+a) - // and (B+b) for some error terms a and b. - // - // MPB multiplication calculates (((A+a) * (B+b)) >> 64) to be ((A*B) >> - // 64). Shifting (truncating) and rounding introduces further error. The - // difference between the calculated result: - // ((A*B ) >> 64) - // and the true result: - // ((A*B + A*b + a*B + a*b) >> 64) + rounding_error - // is: - // (( A*b + a*B + a*b) >> 64) + rounding_error - // which can be re-grouped as: - // ((A*b) >> 64) + ((a*(B+b)) >> 64) + rounding_error - // - // Now, let A and a be "m->mantissa" and "error", and B and b be the - // pre-calculated power of 10. A and B are both less than (1 << 64), a is - // the "error" local variable and b is less than 1. - // - // An upper bound (in absolute value) on ((A*b) >> 64) is therefore 1. - // - // An upper bound on ((a*(B+b)) >> 64) is a, also known as error. - // - // Finally, the rounding_error is at most 1. - // - // In total, calling mpb__mul_pow_10 will raise the worst-case error by 2. - // The subsequent re-normalization can multiply that by a further factor. - - // Multiply by small_powers_of_10[etc]. - wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( - m, &wuffs_base__private_implementation__small_powers_of_10[3 * spo10]); - error += 2; - error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); - - // Multiply by big_powers_of_10[etc]. - wuffs_base__private_implementation__medium_prec_bin__mul_pow_10( - m, &wuffs_base__private_implementation__big_powers_of_10[3 * bpo10]); - error += 2; - error <<= wuffs_base__private_implementation__medium_prec_bin__normalize(m); - - // We have a good approximation of h, but we still have to check whether - // the error is small enough. Equivalently, whether the number of surplus - // mantissa bits (the bits dropped when going from m's 64 mantissa bits to - // the smaller number of double-precision mantissa bits) would always round - // up or down, even when perturbed by ±error. We start at 11 surplus bits - // (m has 64, double-precision has 1+52), but it can be higher for - // subnormals. - // - // In many cases, the error is small enough and we return true. - const int32_t f64_bias = -1023; - int32_t subnormal_exp2 = f64_bias - 63; - uint32_t surplus_bits = 11; - if (subnormal_exp2 >= m->exp2) { - surplus_bits += 1 + ((uint32_t)(subnormal_exp2 - m->exp2)); - } - - uint64_t surplus_mask = - (((uint64_t)1) << surplus_bits) - 1; // e.g. 0x07FF. - uint64_t surplus = m->mantissa & surplus_mask; - uint64_t halfway = ((uint64_t)1) << (surplus_bits - 1); // e.g. 0x0400. - - // Do the final calculation in *signed* arithmetic. - int64_t i_surplus = (int64_t)surplus; - int64_t i_halfway = (int64_t)halfway; - int64_t i_error = (int64_t)error; - - if ((i_surplus > (i_halfway - i_error)) && - (i_surplus < (i_halfway + i_error))) { - goto fail; - } - - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = wuffs_base__private_implementation__medium_prec_bin__as_f64( - m, h->negative); - return ret; - } while (0); - -fail: - do { - wuffs_base__result_f64 ret; - ret.status.repr = "#base: mpb__parse_number_f64 failed"; - ret.value = 0; - return ret; - } while (0); -} - -// -------- - -wuffs_base__result_f64 // -wuffs_base__parse_number_f64_special(wuffs_base__slice_u8 s, - const char* fallback_status_repr) { - do { - uint8_t* p = s.ptr; - uint8_t* q = s.ptr + s.len; - - for (; (p < q) && (*p == '_'); p++) { - } - if (p >= q) { - goto fallback; - } - - // Parse sign. - bool negative = false; - do { - if (*p == '+') { - p++; - } else if (*p == '-') { - negative = true; - p++; - } else { - break; - } - for (; (p < q) && (*p == '_'); p++) { - } - } while (0); - if (p >= q) { - goto fallback; - } - - bool nan = false; - switch (p[0]) { - case 'I': - case 'i': - if (((q - p) < 3) || // - ((p[1] != 'N') && (p[1] != 'n')) || // - ((p[2] != 'F') && (p[2] != 'f'))) { - goto fallback; - } - p += 3; - - if ((p >= q) || (*p == '_')) { - break; - } else if (((q - p) < 5) || // - ((p[0] != 'I') && (p[0] != 'i')) || // - ((p[1] != 'N') && (p[1] != 'n')) || // - ((p[2] != 'I') && (p[2] != 'i')) || // - ((p[3] != 'T') && (p[3] != 't')) || // - ((p[4] != 'Y') && (p[4] != 'y'))) { - goto fallback; - } - p += 5; - - if ((p >= q) || (*p == '_')) { - break; - } - goto fallback; - - case 'N': - case 'n': - if (((q - p) < 3) || // - ((p[1] != 'A') && (p[1] != 'a')) || // - ((p[2] != 'N') && (p[2] != 'n'))) { - goto fallback; - } - p += 3; - - if ((p >= q) || (*p == '_')) { - nan = true; - break; - } - goto fallback; - - default: - goto fallback; - } - - // Finish. - for (; (p < q) && (*p == '_'); p++) { - } - if (p != q) { - goto fallback; - } - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = wuffs_base__ieee_754_bit_representation__to_f64( - (nan ? 0x7FFFFFFFFFFFFFFF : 0x7FF0000000000000) | - (negative ? 0x8000000000000000 : 0)); - return ret; - } while (0); - -fallback: - do { - wuffs_base__result_f64 ret; - ret.status.repr = fallback_status_repr; - ret.value = 0; - return ret; - } while (0); -} - -wuffs_base__result_f64 // -wuffs_base__parse_number_f64(wuffs_base__slice_u8 s) { - wuffs_base__private_implementation__medium_prec_bin m; - wuffs_base__private_implementation__high_prec_dec h; - - do { - // powers converts decimal powers of 10 to binary powers of 2. For example, - // (10000 >> 13) is 1. It stops before the elements exceed 60, also known - // as WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL. - static const uint32_t num_powers = 19; - static const uint8_t powers[19] = { - 0, 3, 6, 9, 13, 16, 19, 23, 26, 29, // - 33, 36, 39, 43, 46, 49, 53, 56, 59, // - }; - - wuffs_base__status status = - wuffs_base__private_implementation__high_prec_dec__parse(&h, s); - if (status.repr) { - return wuffs_base__parse_number_f64_special(s, status.repr); - } - - // Handle zero and obvious extremes. The largest and smallest positive - // finite f64 values are approximately 1.8e+308 and 4.9e-324. - if ((h.num_digits == 0) || (h.decimal_point < -326)) { - goto zero; - } else if (h.decimal_point > 310) { - goto infinity; - } - - wuffs_base__result_f64 mpb_result = - wuffs_base__private_implementation__medium_prec_bin__parse_number_f64( - &m, &h, false); - if (mpb_result.status.repr == NULL) { - return mpb_result; - } - - // Scale by powers of 2 until we're in the range [½ .. 1], which gives us - // our exponent (in base-2). First we shift right, possibly a little too - // far, ending with a value certainly below 1 and possibly below ½... - const int32_t f64_bias = -1023; - int32_t exp2 = 0; - while (h.decimal_point > 0) { - uint32_t n = (uint32_t)(+h.decimal_point); - uint32_t shift = - (n < num_powers) - ? powers[n] - : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; - - wuffs_base__private_implementation__high_prec_dec__small_rshift(&h, - shift); - if (h.decimal_point < - -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { - goto zero; - } - exp2 += (int32_t)shift; - } - // ...then we shift left, putting us in [½ .. 1]. - while (h.decimal_point <= 0) { - uint32_t shift; - if (h.decimal_point == 0) { - if (h.digits[0] >= 5) { - break; - } - shift = (h.digits[0] <= 2) ? 2 : 1; - } else { - uint32_t n = (uint32_t)(-h.decimal_point); - shift = (n < num_powers) - ? powers[n] - : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; - } - - wuffs_base__private_implementation__high_prec_dec__small_lshift(&h, - shift); - if (h.decimal_point > - +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) { - goto infinity; - } - exp2 -= (int32_t)shift; - } - - // We're in the range [½ .. 1] but f64 uses [1 .. 2]. - exp2--; - - // The minimum normal exponent is (f64_bias + 1). - while ((f64_bias + 1) > exp2) { - uint32_t n = (uint32_t)((f64_bias + 1) - exp2); - if (n > WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL) { - n = WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL; - } - wuffs_base__private_implementation__high_prec_dec__small_rshift(&h, n); - exp2 += (int32_t)n; - } - - // Check for overflow. - if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1. - goto infinity; - } - - // Extract 53 bits for the mantissa (in base-2). - wuffs_base__private_implementation__high_prec_dec__small_lshift(&h, 53); - uint64_t man2 = - wuffs_base__private_implementation__high_prec_dec__rounded_integer(&h); - - // Rounding might have added one bit. If so, shift and re-check overflow. - if ((man2 >> 53) != 0) { - man2 >>= 1; - exp2++; - if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1. - goto infinity; - } - } - - // Handle subnormal numbers. - if ((man2 >> 52) == 0) { - exp2 = f64_bias; - } - - // Pack the bits and return. - uint64_t exp2_bits = - (uint64_t)((exp2 - f64_bias) & 0x07FF); // (1 << 11) - 1. - uint64_t bits = (man2 & 0x000FFFFFFFFFFFFF) | // (1 << 52) - 1. - (exp2_bits << 52) | // - (h.negative ? 0x8000000000000000 : 0); // (1 << 63). - - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); - return ret; - } while (0); - -zero: - do { - uint64_t bits = h.negative ? 0x8000000000000000 : 0; - - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); - return ret; - } while (0); - -infinity: - do { - uint64_t bits = h.negative ? 0xFFF0000000000000 : 0x7FF0000000000000; - - wuffs_base__result_f64 ret; - ret.status.repr = NULL; - ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits); - return ret; - } while (0); -} - // ---------------- Hexadecimal size_t // diff --git a/internal/cgen/cgen.go b/internal/cgen/cgen.go index 0a046498..7f6491d9 100644 --- a/internal/cgen/cgen.go +++ b/internal/cgen/cgen.go @@ -116,13 +116,15 @@ func Do(args []string) error { } buf := make(buffer, 0, 128*1024) if err := expandBangBangInsert(&buf, baseAllImplC, map[string]func(*buffer) error{ - "// !! INSERT InterfaceDeclarations.\n": insertInterfaceDeclarations, - "// !! INSERT InterfaceDefinitions.\n": insertInterfaceDefinitions, - "// !! INSERT base/all-private.h.\n": insertBaseAllPrivateH, - "// !! INSERT base/all-public.h.\n": insertBaseAllPublicH, - "// !! INSERT base/copyright\n": insertBaseCopyright, - "// !! INSERT base/image-impl.c.\n": insertBaseImageImplC, - "// !! INSERT base/strconv-impl.c.\n": insertBaseStrConvImplC, + "// !! INSERT InterfaceDeclarations.\n": insertInterfaceDeclarations, + "// !! INSERT InterfaceDefinitions.\n": insertInterfaceDefinitions, + "// !! INSERT base/all-private.h.\n": insertBaseAllPrivateH, + "// !! INSERT base/all-public.h.\n": insertBaseAllPublicH, + "// !! INSERT base/copyright\n": insertBaseCopyright, + "// !! INSERT base/f64conv-submodule.c.\n": insertBaseF64ConvSubmoduleC, + "// !! INSERT base/image-impl.c.\n": insertBaseImageImplC, + "// !! INSERT base/pixconv-submodule.c.\n": insertBasePixConvSubmoduleC, + "// !! INSERT base/strconv-impl.c.\n": insertBaseStrConvImplC, "// !! INSERT wuffs_base__status strings.\n": func(b *buffer) error { for _, z := range builtin.Statuses { msg, _ := t.Unescape(z) @@ -327,12 +329,24 @@ func insertBaseCopyright(buf *buffer) error { return nil } +func insertBaseF64ConvSubmoduleC(buf *buffer) error { + buf.writes(baseF64ConvSubmoduleC) + buf.writeb('\n') + return nil +} + func insertBaseImageImplC(buf *buffer) error { buf.writes(baseImageImplC) buf.writeb('\n') return nil } +func insertBasePixConvSubmoduleC(buf *buffer) error { + buf.writes(basePixConvSubmoduleC) + buf.writeb('\n') + return nil +} + func insertBaseStrConvImplC(buf *buffer) error { buf.writes(baseStrConvImplC) buf.writeb('\n') diff --git a/internal/cgen/data.go b/internal/cgen/data.go index 4f5bce9b..9d08a414 100644 --- a/internal/cgen/data.go +++ b/internal/cgen/data.go @@ -23,20 +23,14 @@ const baseAllImplC = "" + "" + "// ----------------\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE)\n\nconst uint8_t wuffs_base__low_bits_mask__u8[9] = {\n 0x00, 0x01, 0x03, 0x07, 0x0F, 0x1F, 0x3F, 0x7F, 0xFF,\n};\n\nconst uint16_t wuffs_base__low_bits_mask__u16[17] = {\n 0x0000, 0x0001, 0x0003, 0x0007, 0x000F, 0x001F, 0x003F, 0x007F, 0x00FF,\n 0x01FF, 0x03FF, 0x07FF, 0x0FFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF,\n};\n\nconst uint32_t wuffs_base__low_bits_mask__u32[33] = {\n 0x00000000, 0x00000001, 0x00000003, 0x00000007, 0x0000000F, 0x0000001F,\n 0x0000003F, 0x0000007F, 0x000000FF, 0x000001FF, 0x000003FF, 0x000007FF,\n 0x00000FFF, 0x00001FFF, 0x00003FFF, 0x00007FFF, 0x0000FFFF, 0x0001FFFF,\n 0x0003FFFF, 0x0007FFFF, 0x000FFFFF, 0x001FFFFF, 0x003FFFFF, 0x007FFFFF,\n 0x00FFFFFF, 0x01FFFFFF, 0x03FFFFFF, 0x07FFFFFF, 0x0FFFFFFF, 0x1FFFFFFF,\n 0x3FFFFFFF, 0x7FFFFFFF, 0xFFFFFFFF,\n};\n\nconst uint64_t wuffs_base__low_bits_mask__u64[65] = {\n 0x0000000000000000, 0x0000000000000001, 0x0000000000000003,\n 0x0000000000" + "000007, 0x000000000000000F, 0x000000000000001F,\n 0x000000000000003F, 0x000000000000007F, 0x00000000000000FF,\n 0x00000000000001FF, 0x00000000000003FF, 0x00000000000007FF,\n 0x0000000000000FFF, 0x0000000000001FFF, 0x0000000000003FFF,\n 0x0000000000007FFF, 0x000000000000FFFF, 0x000000000001FFFF,\n 0x000000000003FFFF, 0x000000000007FFFF, 0x00000000000FFFFF,\n 0x00000000001FFFFF, 0x00000000003FFFFF, 0x00000000007FFFFF,\n 0x0000000000FFFFFF, 0x0000000001FFFFFF, 0x0000000003FFFFFF,\n 0x0000000007FFFFFF, 0x000000000FFFFFFF, 0x000000001FFFFFFF,\n 0x000000003FFFFFFF, 0x000000007FFFFFFF, 0x00000000FFFFFFFF,\n 0x00000001FFFFFFFF, 0x00000003FFFFFFFF, 0x00000007FFFFFFFF,\n 0x0000000FFFFFFFFF, 0x0000001FFFFFFFFF, 0x0000003FFFFFFFFF,\n 0x0000007FFFFFFFFF, 0x000000FFFFFFFFFF, 0x000001FFFFFFFFFF,\n 0x000003FFFFFFFFFF, 0x000007FFFFFFFFFF, 0x00000FFFFFFFFFFF,\n 0x00001FFFFFFFFFFF, 0x00003FFFFFFFFFFF, 0x00007FFFFFFFFFFF,\n 0x0000FFFFFFFFFFFF, 0x0001FFFFFFFFFFFF, 0x0003FFFFFFFFFFFF,\n 0x0007FFFFFF" + - "FFFFFF, 0x000FFFFFFFFFFFFF, 0x001FFFFFFFFFFFFF,\n 0x003FFFFFFFFFFFFF, 0x007FFFFFFFFFFFFF, 0x00FFFFFFFFFFFFFF,\n 0x01FFFFFFFFFFFFFF, 0x03FFFFFFFFFFFFFF, 0x07FFFFFFFFFFFFFF,\n 0x0FFFFFFFFFFFFFFF, 0x1FFFFFFFFFFFFFFF, 0x3FFFFFFFFFFFFFFF,\n 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF,\n};\n\n// !! INSERT wuffs_base__status strings.\n\n// !! INSERT InterfaceDefinitions.\n\n// !! INSERT base/image-impl.c.\n\n// !! INSERT base/strconv-impl.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE)\n\n#ifdef __cplusplus\n} // extern \"C\"\n#endif\n\n#endif // WUFFS_IMPLEMENTATION\n\n// !! WUFFS MONOLITHIC RELEASE DISCARDS EVERYTHING BELOW.\n\n#endif // WUFFS_INCLUDE_GUARD__BASE\n" + + "FFFFFF, 0x000FFFFFFFFFFFFF, 0x001FFFFFFFFFFFFF,\n 0x003FFFFFFFFFFFFF, 0x007FFFFFFFFFFFFF, 0x00FFFFFFFFFFFFFF,\n 0x01FFFFFFFFFFFFFF, 0x03FFFFFFFFFFFFFF, 0x07FFFFFFFFFFFFFF,\n 0x0FFFFFFFFFFFFFFF, 0x1FFFFFFFFFFFFFFF, 0x3FFFFFFFFFFFFFFF,\n 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF,\n};\n\n// !! INSERT wuffs_base__status strings.\n\n// !! INSERT InterfaceDefinitions.\n\n// !! INSERT base/image-impl.c.\n\n// !! INSERT base/strconv-impl.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE)\n\n// !! INSERT base/f64conv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE)\n\n// !! INSERT base/pixconv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE)\n\n#ifdef __cplusplus\n} // extern \"C\"\n#endif\n\n#endif // WUFFS_IMPLEMENTATION\n" + + "\n// !! WUFFS MONOLITHIC RELEASE DISCARDS EVERYTHING BELOW.\n\n#endif // WUFFS_INCLUDE_GUARD__BASE\n" + "" const baseImageImplC = "" + "// ---------------- Images\n\nconst uint32_t wuffs_base__pixel_format__bits_per_channel[16] = {\n 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,\n 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40,\n};\n\n" + "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n\n size_t n = len4;\n while (n--) {\n uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s);\n uint32_t b5 = 0x1F & (argb >> (8 - 5));\n uint32_t g6 = 0x3F & (argb >> (16 - 6));\n uint32_t r5 = 0x1F & (argb >> (24 - 5));\n uint32_t alpha = argb & 0xFF000000;\n wuffs_base__store_u32le__no_bounds_check(\n d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0));\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n\n size_t n = len4;\n while (n--) {\n " + - " uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\nstatic inline uint32_t //\nwuffs_base__swap_u32_argb_abgr(uint32_t u) {\n uint32_t o = u & 0xFF00FF00;\n uint32_t r = u & 0x00FF0000;\n uint32_t b = u & 0x000000FF;\n return o | (r >> 16) | (b << 16);\n}\n\n" + - "" + - "// --------\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * i" + - "a)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8))" + - ";\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n " + - " uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul,\n " + - " uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\n" + + "// --------\n\nstatic inline uint32_t //\nwuffs_base__swap_u32_argb_abgr(uint32_t u) {\n uint32_t o = u & 0xFF00FF00;\n uint32_t r = u & 0x00FF0000;\n uint32_t b = u & 0x000000FF;\n return o | (r >> 16) | (b << 16);\n}\n\n" + "" + "// --------\n\nwuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y) {\n if (!pb || (x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (pb->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n uint8_t* palette = pb->private_impl.planes[3].ptr;\n return wu" + "ffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x])));\n }\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x])));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: {\n uint8_t* palette = pb->private_impl.planes[3].ptr;\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul(\n wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n re" + @@ -47,50 +41,7 @@ const baseImageImplC = "" + "gr(color));\n break;\n\n default:\n // TODO: support more formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n return wuffs_base__make_status(NULL);\n}\n\n" + "" + "// --------\n\nuint8_t //\nwuffs_base__pixel_palette__closest_element(\n wuffs_base__slice_u8 palette_slice,\n wuffs_base__pixel_format palette_format,\n wuffs_base__color_u32_argb_premul c) {\n size_t n = palette_slice.len / 4;\n if (n > 256) {\n n = 256;\n }\n size_t best_index = 0;\n uint64_t best_score = 0xFFFFFFFFFFFFFFFF;\n\n // Work in 16-bit color.\n uint32_t ca = 0x101 * (0xFF & (c >> 24));\n uint32_t cr = 0x101 * (0xFF & (c >> 16));\n uint32_t cg = 0x101 * (0xFF & (c >> 8));\n uint32_t cb = 0x101 * (0xFF & (c >> 0));\n\n switch (palette_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n bool nonpremul = palette_format.repr ==\n WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL;\n\n size_t i;\n for (i = 0; i < n; i++) {\n // Work in 16-bit color.\n uint32_t pb = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 0]))" + - ";\n uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1]));\n uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2]));\n uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3]));\n\n // Convert to premultiplied alpha.\n if (nonpremul && (pa != 0xFFFF)) {\n pb = (pb * pa) / 0xFFFF;\n pg = (pg * pa) / 0xFFFF;\n pr = (pr * pa) / 0xFFFF;\n }\n\n // These deltas are conceptually int32_t (signed) but after squaring,\n // it's equivalent to work in uint32_t (unsigned).\n pb -= cb;\n pg -= cg;\n pr -= cr;\n pa -= ca;\n uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) +\n ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa));\n if (best_score > score) {\n best_score = score;\n best_index = i;\n }\n }\n break;\n }\n }\n\n return (uint8_t)best_index;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n return wuffs_base__slice_u8__copy_from_slice(dst, src);\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3;\n if (len > 0) {\n memmove(dst.ptr, src.ptr, len * 3);\n }\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 <" + - " src_len4 ? dst_len4 : src_len4;\n if (len > 0) {\n memmove(dst.ptr, src.ptr, len * 4);\n }\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t b5 = s[0] >> 3;\n uint32_t g6 = s[1] >> 2;\n uint32_t r5 = s[2] >> 3;\n uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 3;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len4 = src.len / 4;\n size_t len = " + - "dst_len2 < src_len4 ? dst_len2 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)))));\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0" + - "x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n\n // Convert from 565 color to 16-bit color.\n uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2));\n uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11);\n uint32_t dr = (0x8421 * old_r5) >> 4;\n uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5);\n uint32_t dg = (0x1041 * old_g6) >> 2;\n uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0);\n uint32_t db = (0x8421 * old_b5) >> 4;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 565 color and combine the components.\n uint32_t new_r5 = 0x1F & (dr >> 11);\n uint32_t new_g6 = 0x3F & (dg >> 10);\n uint32_t new_b5 = 0x1F & (db >> 11);\n uint32_t new_rgb" + - "_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2),\n (uint16_t)new_rgb_565);\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t y5 = s[0] >> 3;\n uint32_t y6 = s[0] >> 2;\n uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__index__src(\n wuffs_base__slice_u8" + - " dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (1 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (2 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (3 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr" + - " + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 2;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u16le__no_bounds" + - "_check(d + (0 * 2), (uint16_t)s0);\n }\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 =\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4;\n uint8_t* d = d" + - "st.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n uint32_t dr = 0x101 * ((uint32_t)d[2]);\n uint32_t dg = 0x101 * ((uint32_t)d[1]);\n uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n d[0] = (uint8_t)(db >> 8);\n d[1] = (uint8_t)(dg >> 8);\n d[2] = (uint8_t)(dr >> 8);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n ui" + - "nt8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len3 = dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n // The comparison in the while condition is \">\", not \">=\", because with\n // \">=\", the last 4-byte store could write past the end of the dst slice.\n //\n // Each 4-byte store writes one too many bytes, but a subsequent store\n // will overwrite that with the correct byte. There is always another\n // store, whether a 4-byte store in this loop or a 1-byte store in the\n // next loop.\n while (n > loop_unroll_count) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 3), wuffs_base__load_u32le__no_bounds_c" + - "heck(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__" + - "pixel_swizzler__xxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len3 = dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n " + - " ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n }\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 =" + - " dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint8_t s0 = s[0];\n d[0] = s0;\n d[1] = s0;\n d[2] = s0;\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_pale" + - "tte.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count)" + - " {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count *" + - " 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n" + - "\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + - "" + - "// --------\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__pixel_swizzler__xxx__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WU" + - "FFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__pixel_swizzler__xxxx__y;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_1_1;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette,\n " + - " src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_pa" + - "lette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette,\n src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n if (wuffs_base__pixel_swizzler__" + - "swap_rgbx_bgrx(dst_palette,\n src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__B" + - "GR:\n return wuffs_base__pixel_swizzler__copy_3_3;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__xxxx__xxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_b" + - "ase__pixel_swizzler__bgr_565__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_4_4;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swiz" + - "zler__bgra_premul__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\n" + - "" + - "// --------\n\nwuffs_base__status //\nwuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__pixel_format src_format,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n if (!p) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n\n // TODO: support many more formats.\n\n wuffs_base__pixel_swizzler__func func = NULL;\n\n switch (src_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette,\n src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n p, dst_format, dst_" + - "palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n func = wuffs_base__pixel_swizzler__prepare__bgr(\n p, dst_format, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n p, dst_format, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_impl.func = func;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}\n\nuint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (p && p->private_impl.func) {\n return (*p->private_impl.func)(dst, dst_palette, src);\n }\n return 0;\n}\n" + + ";\n uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1]));\n uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2]));\n uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3]));\n\n // Convert to premultiplied alpha.\n if (nonpremul && (pa != 0xFFFF)) {\n pb = (pb * pa) / 0xFFFF;\n pg = (pg * pa) / 0xFFFF;\n pr = (pr * pa) / 0xFFFF;\n }\n\n // These deltas are conceptually int32_t (signed) but after squaring,\n // it's equivalent to work in uint32_t (unsigned).\n pb -= cb;\n pg -= cg;\n pr -= cr;\n pa -= ca;\n uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) +\n ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa));\n if (best_score > score) {\n best_score = score;\n best_index = i;\n }\n }\n break;\n }\n }\n\n return (uint8_t)best_index;\n}\n" + "" const baseStrConvImplC = "" + @@ -105,8 +56,21 @@ const baseStrConvImplC = "" + "alue = +(int64_t)(r.value);\n return ret;\n }\n } while (0);\n\nfail_bad_argument:\n do {\n wuffs_base__result_i64 ret;\n ret.status.repr = wuffs_base__error__bad_argument;\n ret.value = 0;\n return ret;\n } while (0);\n\nfail_out_of_bounds:\n do {\n wuffs_base__result_i64 ret;\n ret.status.repr = wuffs_base__error__out_of_bounds;\n ret.value = 0;\n return ret;\n } while (0);\n}\n\nwuffs_base__result_u64 //\nwuffs_base__parse_number_u64(wuffs_base__slice_u8 s) {\n uint8_t* p = s.ptr;\n uint8_t* q = s.ptr + s.len;\n\n for (; (p < q) && (*p == '_'); p++) {\n }\n\n if (p >= q) {\n goto fail_bad_argument;\n\n } else if (*p == '0') {\n p++;\n if (p >= q) {\n goto ok_zero;\n }\n if (*p == '_') {\n p++;\n for (; p < q; p++) {\n if (*p != '_') {\n goto fail_bad_argument;\n }\n }\n goto ok_zero;\n }\n\n if ((*p == 'x') || (*p == 'X')) {\n p++;\n for (; (p < q) && (*p == '_'); p++) {\n }\n if (p < q) {\n goto hexadecimal;\n }\n\n " + " } else if ((*p == 'd') || (*p == 'D')) {\n p++;\n for (; (p < q) && (*p == '_'); p++) {\n }\n if (p < q) {\n goto decimal;\n }\n }\n\n goto fail_bad_argument;\n }\n\ndecimal:\n do {\n uint64_t v = wuffs_base__parse_number__decimal_digits[*p++];\n if (v == 0) {\n goto fail_bad_argument;\n }\n v &= 0x0F;\n\n // UINT64_MAX is 18446744073709551615, which is ((10 * max10) + max1).\n const uint64_t max10 = 1844674407370955161;\n const uint8_t max1 = 5;\n\n for (; p < q; p++) {\n if (*p == '_') {\n continue;\n }\n uint8_t digit = wuffs_base__parse_number__decimal_digits[*p];\n if (digit == 0) {\n goto fail_bad_argument;\n }\n digit &= 0x0F;\n if ((v > max10) || ((v == max10) && (digit > max1))) {\n goto fail_out_of_bounds;\n }\n v = (10 * v) + ((uint64_t)(digit));\n }\n\n wuffs_base__result_u64 ret;\n ret.status.repr = NULL;\n ret.value = v;\n return ret;\n } while (0);\n\nhexadecimal:\n do {\n uint64_t v = " + "wuffs_base__parse_number__hexadecimal_digits[*p++];\n if (v == 0) {\n goto fail_bad_argument;\n }\n v &= 0x0F;\n\n for (; p < q; p++) {\n if (*p == '_') {\n continue;\n }\n uint8_t digit = wuffs_base__parse_number__hexadecimal_digits[*p];\n if (digit == 0) {\n goto fail_bad_argument;\n }\n digit &= 0x0F;\n if ((v >> 60) != 0) {\n goto fail_out_of_bounds;\n }\n v = (v << 4) | ((uint64_t)(digit));\n }\n\n wuffs_base__result_u64 ret;\n ret.status.repr = NULL;\n ret.value = v;\n return ret;\n } while (0);\n\nok_zero:\n do {\n wuffs_base__result_u64 ret;\n ret.status.repr = NULL;\n ret.value = 0;\n return ret;\n } while (0);\n\nfail_bad_argument:\n do {\n wuffs_base__result_u64 ret;\n ret.status.repr = wuffs_base__error__bad_argument;\n ret.value = 0;\n return ret;\n } while (0);\n\nfail_out_of_bounds:\n do {\n wuffs_base__result_u64 ret;\n ret.status.repr = wuffs_base__error__out_of_bounds;\n ret.value = 0;\n return ret;\n " + - " } while (0);\n}\n\n " + + " } while (0);\n}\n\n" + "" + + "// ---------------- Hexadecimal\n\nsize_t //\nwuffs_base__hexadecimal__decode2(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t src_len2 = src.len / 2;\n size_t len = dst.len < src_len2 ? dst.len : src_len2;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n while (n--) {\n *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[0]] << 4) |\n (wuffs_base__parse_number__hexadecimal_digits[s[1]] & 0x0F));\n d += 1;\n s += 2;\n }\n\n return len;\n}\n\nsize_t //\nwuffs_base__hexadecimal__decode4(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t src_len4 = src.len / 4;\n size_t len = dst.len < src_len4 ? dst.len : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n while (n--) {\n *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[2]] << 4) |\n (wuffs_base__parse_number__hexadecimal_digits[s[3]] & 0x0F));\n d += 1;\n s +" + + "= 4;\n }\n\n return len;\n}\n\n" + + "" + + "// ---------------- Unicode and UTF-8\n\nsize_t //\nwuffs_base__utf_8__encode(wuffs_base__slice_u8 dst, uint32_t code_point) {\n if (code_point <= 0x7F) {\n if (dst.len >= 1) {\n dst.ptr[0] = (uint8_t)(code_point);\n return 1;\n }\n\n } else if (code_point <= 0x07FF) {\n if (dst.len >= 2) {\n dst.ptr[0] = (uint8_t)(0xC0 | ((code_point >> 6)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 2;\n }\n\n } else if (code_point <= 0xFFFF) {\n if ((dst.len >= 3) && ((code_point < 0xD800) || (0xDFFF < code_point))) {\n dst.ptr[0] = (uint8_t)(0xE0 | ((code_point >> 12)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F));\n dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 3;\n }\n\n } else if (code_point <= 0x10FFFF) {\n if (dst.len >= 4) {\n dst.ptr[0] = (uint8_t)(0xF0 | ((code_point >> 18)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 12) & 0x3F));\n dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3" + + "F));\n dst.ptr[3] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 4;\n }\n }\n\n return 0;\n}\n\n// wuffs_base__utf_8__byte_length_minus_1 is the byte length (minus 1) of a\n// UTF-8 encoded code point, based on the encoding's initial byte.\n// - 0x00 is 1-byte UTF-8 (ASCII).\n// - 0x01 is the start of 2-byte UTF-8.\n// - 0x02 is the start of 3-byte UTF-8.\n// - 0x03 is the start of 4-byte UTF-8.\n// - 0x40 is a UTF-8 tail byte.\n// - 0x80 is invalid UTF-8.\n//\n// RFC 3629 (UTF-8) gives this grammar for valid UTF-8:\n// UTF8-1 = %x00-7F\n// UTF8-2 = %xC2-DF UTF8-tail\n// UTF8-3 = %xE0 %xA0-BF UTF8-tail / %xE1-EC 2( UTF8-tail ) /\n// %xED %x80-9F UTF8-tail / %xEE-EF 2( UTF8-tail )\n// UTF8-4 = %xF0 %x90-BF 2( UTF8-tail ) / %xF1-F3 3( UTF8-tail ) /\n// %xF4 %x80-8F 2( UTF8-tail )\n// UTF8-tail = %x80-BF\nstatic const uint8_t wuffs_base__utf_8__byte_length_minus_1[256] = {\n // 0 1 2 3 4 5 6 7\n // 8 9" + + " A B C D E F\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x00 ..= 0x07.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x08 ..= 0x0F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x10 ..= 0x17.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x18 ..= 0x1F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x20 ..= 0x27.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x28 ..= 0x2F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x30 ..= 0x37.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x38 ..= 0x3F.\n\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x40 ..= 0x47.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x48 ..= 0x4F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x50 ..= 0x57.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x58 ..= 0x5F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x60 ..= 0x67.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x68 .." + + "= 0x6F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x70 ..= 0x77.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x78 ..= 0x7F.\n\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x80 ..= 0x87.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x88 ..= 0x8F.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x90 ..= 0x97.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x98 ..= 0x9F.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA0 ..= 0xA7.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA8 ..= 0xAF.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB0 ..= 0xB7.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB8 ..= 0xBF.\n\n 0x80, 0x80, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC0 ..= 0xC7.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC8 ..= 0xCF.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD0 ..= 0xD7.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD8 ..= 0xDF.\n 0x02, 0x02, 0x02" + + ", 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE0 ..= 0xE7.\n 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE8 ..= 0xEF.\n 0x03, 0x03, 0x03, 0x03, 0x03, 0x80, 0x80, 0x80, // 0xF0 ..= 0xF7.\n 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, // 0xF8 ..= 0xFF.\n // 0 1 2 3 4 5 6 7\n // 8 9 A B C D E F\n};\n\nwuffs_base__utf_8__next__output //\nwuffs_base__utf_8__next(wuffs_base__slice_u8 s) {\n if (s.len == 0) {\n return wuffs_base__make_utf_8__next__output(0, 0);\n }\n uint32_t c = s.ptr[0];\n switch (wuffs_base__utf_8__byte_length_minus_1[c & 0xFF]) {\n case 0:\n return wuffs_base__make_utf_8__next__output(c, 1);\n\n case 1:\n if (s.len < 2) {\n break;\n }\n c = wuffs_base__load_u16le__no_bounds_check(s.ptr);\n if ((c & 0xC000) != 0x8000) {\n break;\n }\n c = (0x0007C0 & (c << 6)) | (0x00003F & (c >> 8));\n return wuffs_base__make_utf_8__next__output(c, 2);\n\n case 2:\n if (s.len < 3) {\n " + + " break;\n }\n c = wuffs_base__load_u24le__no_bounds_check(s.ptr);\n if ((c & 0xC0C000) != 0x808000) {\n break;\n }\n c = (0x00F000 & (c << 12)) | (0x000FC0 & (c >> 2)) |\n (0x00003F & (c >> 16));\n if ((c <= 0x07FF) || ((0xD800 <= c) && (c <= 0xDFFF))) {\n break;\n }\n return wuffs_base__make_utf_8__next__output(c, 3);\n\n case 3:\n if (s.len < 4) {\n break;\n }\n c = wuffs_base__load_u32le__no_bounds_check(s.ptr);\n if ((c & 0xC0C0C000) != 0x80808000) {\n break;\n }\n c = (0x1C0000 & (c << 18)) | (0x03F000 & (c << 4)) |\n (0x000FC0 & (c >> 10)) | (0x00003F & (c >> 24));\n if ((c <= 0xFFFF) || (0x110000 <= c)) {\n break;\n }\n return wuffs_base__make_utf_8__next__output(c, 4);\n }\n\n return wuffs_base__make_utf_8__next__output(\n WUFFS_BASE__UNICODE_REPLACEMENT_CHARACTER, 1);\n}\n\nsize_t //\nwuffs_base__utf_8__longest_valid_prefix(wuffs_base__slice_u8 s) {\n // TODO: possibly optimize the all-A" + + "SCII case (4 or 8 bytes at a time).\n //\n // TODO: possibly optimize this by manually inlining the\n // wuffs_base__utf_8__next calls.\n size_t original_len = s.len;\n while (s.len > 0) {\n wuffs_base__utf_8__next__output o = wuffs_base__utf_8__next(s);\n if ((o.code_point > 0x7F) && (o.byte_length == 1)) {\n break;\n }\n s.ptr += o.byte_length;\n s.len -= o.byte_length;\n }\n return original_len - s.len;\n}\n\nsize_t //\nwuffs_base__ascii__longest_valid_prefix(wuffs_base__slice_u8 s) {\n // TODO: possibly optimize this by checking 4 or 8 bytes at a time.\n uint8_t* original_ptr = s.ptr;\n uint8_t* p = s.ptr;\n uint8_t* q = s.ptr + s.len;\n for (; (p != q) && ((*p & 0x80) == 0); p++) {\n }\n return (size_t)(p - original_ptr);\n}\n" + + "" + +const baseF64ConvSubmoduleC = "" + "// ---------------- IEEE 754 Floating Point\n\n#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE 1023\n#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION 500\n\n// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL is the largest N\n// such that ((10 << N) < (1 << 64)).\n#define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL 60\n\n// wuffs_base__private_implementation__high_prec_dec (abbreviated as HPD) is a\n// fixed precision floating point decimal number, augmented with ±infinity\n// values, but it cannot represent NaN (Not a Number).\n//\n// \"High precision\" means that the mantissa holds 500 decimal digits. 500 is\n// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION.\n//\n// An HPD isn't for general purpose arithmetic, only for conversions to and\n// from IEEE 754 double-precision floating point, where the largest and\n// smallest positive, finite values are approximately 1.8e+308 and 4.9e-324.\n// HPD exponents above +1023 mean infinity, below -1023 mean zero. Th" + "e ±1023\n// bounds are further away from zero than ±(324 + 500), where 500 and 1023 is\n// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION and\n// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE.\n//\n// digits[.. num_digits] are the number's digits in big-endian order. The\n// uint8_t values are in the range [0 ..= 9], not ['0' ..= '9'], where e.g. '7'\n// is the ASCII value 0x37.\n//\n// decimal_point is the index (within digits) of the decimal point. It may be\n// negative or be larger than num_digits, in which case the explicit digits are\n// padded with implicit zeroes.\n//\n// For example, if num_digits is 3 and digits is \"\\x07\\x08\\x09\":\n// - A decimal_point of -2 means \".00789\"\n// - A decimal_point of -1 means \".0789\"\n// - A decimal_point of +0 means \".789\"\n// - A decimal_point of +1 means \"7.89\"\n// - A decimal_point of +2 means \"78.9\"\n// - A decimal_point of +3 means \"789.\"\n// - A decimal_point of +4 means \"7890.\"\n// - A decimal_point of +5 means \"78900.\"\n//\n// As above, a" + " decimal_point higher than +1023 means that the overall value is\n// infinity, lower than -1023 means zero.\n//\n// negative is a sign bit. An HPD can distinguish positive and negative zero.\n//\n// truncated is whether there are more than\n// WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION digits, and at\n// least one of those extra digits are non-zero. The existence of long-tail\n// digits can affect rounding.\n//\n// The \"all fields are zero\" value is valid, and represents the number +0.\ntypedef struct {\n uint32_t num_digits;\n int32_t decimal_point;\n bool negative;\n bool truncated;\n uint8_t digits[WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DIGITS_PRECISION];\n} wuffs_base__private_implementation__high_prec_dec;\n\n// wuffs_base__private_implementation__high_prec_dec__trim trims trailing\n// zeroes from the h->digits[.. h->num_digits] slice. They have no benefit,\n// since we explicitly track h->decimal_point.\n//\n// Preconditions:\n// - h is non-NULL.\nstatic inline void //\nwuffs_base__private_implementation_" + @@ -162,18 +126,62 @@ const baseStrConvImplC = "" + "shift =\n (n < num_powers)\n ? powers[n]\n : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL;\n\n wuffs_base__private_implementation__high_prec_dec__small_rshift(&h,\n shift);\n if (h.decimal_point <\n -WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) {\n goto zero;\n }\n exp2 += (int32_t)shift;\n }\n // ...then we shift left, putting us in [½ .. 1].\n while (h.decimal_point <= 0) {\n uint32_t shift;\n if (h.decimal_point == 0) {\n if (h.digits[0] >= 5) {\n break;\n }\n shift = (h.digits[0] <= 2) ? 2 : 1;\n } else {\n uint32_t n = (uint32_t)(-h.decimal_point);\n shift = (n < num_powers)\n ? powers[n]\n : WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL;\n }\n\n wuffs_base__private_implementation__high_prec_dec__small_lshift(&h,\n " + " shift);\n if (h.decimal_point >\n +WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE) {\n goto infinity;\n }\n exp2 -= (int32_t)shift;\n }\n\n // We're in the range [½ .. 1] but f64 uses [1 .. 2].\n exp2--;\n\n // The minimum normal exponent is (f64_bias + 1).\n while ((f64_bias + 1) > exp2) {\n uint32_t n = (uint32_t)((f64_bias + 1) - exp2);\n if (n > WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL) {\n n = WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__SHIFT__MAX_INCL;\n }\n wuffs_base__private_implementation__high_prec_dec__small_rshift(&h, n);\n exp2 += (int32_t)n;\n }\n\n // Check for overflow.\n if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1.\n goto infinity;\n }\n\n // Extract 53 bits for the mantissa (in base-2).\n wuffs_base__private_implementation__high_prec_dec__small_lshift(&h, 53);\n uint64_t man2 =\n wuffs_base__private_implementation__high_prec_dec_" + "_rounded_integer(&h);\n\n // Rounding might have added one bit. If so, shift and re-check overflow.\n if ((man2 >> 53) != 0) {\n man2 >>= 1;\n exp2++;\n if ((exp2 - f64_bias) >= 0x07FF) { // (1 << 11) - 1.\n goto infinity;\n }\n }\n\n // Handle subnormal numbers.\n if ((man2 >> 52) == 0) {\n exp2 = f64_bias;\n }\n\n // Pack the bits and return.\n uint64_t exp2_bits =\n (uint64_t)((exp2 - f64_bias) & 0x07FF); // (1 << 11) - 1.\n uint64_t bits = (man2 & 0x000FFFFFFFFFFFFF) | // (1 << 52) - 1.\n (exp2_bits << 52) | //\n (h.negative ? 0x8000000000000000 : 0); // (1 << 63).\n\n wuffs_base__result_f64 ret;\n ret.status.repr = NULL;\n ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits);\n return ret;\n } while (0);\n\nzero:\n do {\n uint64_t bits = h.negative ? 0x8000000000000000 : 0;\n\n wuffs_base__result_f64 ret;\n ret.status.repr = NULL;\n ret.value = wuffs_base" + - "__ieee_754_bit_representation__to_f64(bits);\n return ret;\n } while (0);\n\ninfinity:\n do {\n uint64_t bits = h.negative ? 0xFFF0000000000000 : 0x7FF0000000000000;\n\n wuffs_base__result_f64 ret;\n ret.status.repr = NULL;\n ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits);\n return ret;\n } while (0);\n}\n\n" + + "__ieee_754_bit_representation__to_f64(bits);\n return ret;\n } while (0);\n\ninfinity:\n do {\n uint64_t bits = h.negative ? 0xFFF0000000000000 : 0x7FF0000000000000;\n\n wuffs_base__result_f64 ret;\n ret.status.repr = NULL;\n ret.value = wuffs_base__ieee_754_bit_representation__to_f64(bits);\n return ret;\n } while (0);\n}\n" + + "" + +const basePixConvSubmoduleC = "" + + "// ---------------- Pixel Swizzler\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr" + + " = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF " + + "& (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit c" + + "olor to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_premul_u32_axxx(ui" + + "nt32_t dst_premul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}" + + "\n\n" + "" + - "// ---------------- Hexadecimal\n\nsize_t //\nwuffs_base__hexadecimal__decode2(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t src_len2 = src.len / 2;\n size_t len = dst.len < src_len2 ? dst.len : src_len2;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n while (n--) {\n *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[0]] << 4) |\n (wuffs_base__parse_number__hexadecimal_digits[s[1]] & 0x0F));\n d += 1;\n s += 2;\n }\n\n return len;\n}\n\nsize_t //\nwuffs_base__hexadecimal__decode4(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t src_len4 = src.len / 4;\n size_t len = dst.len < src_len4 ? dst.len : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n while (n--) {\n *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[2]] << 4) |\n (wuffs_base__parse_number__hexadecimal_digits[s[3]] & 0x0F));\n d += 1;\n s +" + - "= 4;\n }\n\n return len;\n}\n\n" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n\n size_t n = len4;\n while (n--) {\n uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s);\n uint32_t b5 = 0x1F & (argb >> (8 - 5));\n uint32_t g6 = 0x3F & (argb >> (16 - 6));\n uint32_t r5 = 0x1F & (argb >> (24 - 5));\n uint32_t alpha = argb & 0xFF000000;\n wuffs_base__store_u32le__no_bounds_check(\n d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0));\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n\n size_t n = len4;\n while (n--) {\n " + + " uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\n" + "" + - "// ---------------- Unicode and UTF-8\n\nsize_t //\nwuffs_base__utf_8__encode(wuffs_base__slice_u8 dst, uint32_t code_point) {\n if (code_point <= 0x7F) {\n if (dst.len >= 1) {\n dst.ptr[0] = (uint8_t)(code_point);\n return 1;\n }\n\n } else if (code_point <= 0x07FF) {\n if (dst.len >= 2) {\n dst.ptr[0] = (uint8_t)(0xC0 | ((code_point >> 6)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 2;\n }\n\n } else if (code_point <= 0xFFFF) {\n if ((dst.len >= 3) && ((code_point < 0xD800) || (0xDFFF < code_point))) {\n dst.ptr[0] = (uint8_t)(0xE0 | ((code_point >> 12)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F));\n dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 3;\n }\n\n } else if (code_point <= 0x10FFFF) {\n if (dst.len >= 4) {\n dst.ptr[0] = (uint8_t)(0xF0 | ((code_point >> 18)));\n dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 12) & 0x3F));\n dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3" + - "F));\n dst.ptr[3] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F));\n return 4;\n }\n }\n\n return 0;\n}\n\n// wuffs_base__utf_8__byte_length_minus_1 is the byte length (minus 1) of a\n// UTF-8 encoded code point, based on the encoding's initial byte.\n// - 0x00 is 1-byte UTF-8 (ASCII).\n// - 0x01 is the start of 2-byte UTF-8.\n// - 0x02 is the start of 3-byte UTF-8.\n// - 0x03 is the start of 4-byte UTF-8.\n// - 0x40 is a UTF-8 tail byte.\n// - 0x80 is invalid UTF-8.\n//\n// RFC 3629 (UTF-8) gives this grammar for valid UTF-8:\n// UTF8-1 = %x00-7F\n// UTF8-2 = %xC2-DF UTF8-tail\n// UTF8-3 = %xE0 %xA0-BF UTF8-tail / %xE1-EC 2( UTF8-tail ) /\n// %xED %x80-9F UTF8-tail / %xEE-EF 2( UTF8-tail )\n// UTF8-4 = %xF0 %x90-BF 2( UTF8-tail ) / %xF1-F3 3( UTF8-tail ) /\n// %xF4 %x80-8F 2( UTF8-tail )\n// UTF8-tail = %x80-BF\nstatic const uint8_t wuffs_base__utf_8__byte_length_minus_1[256] = {\n // 0 1 2 3 4 5 6 7\n // 8 9" + - " A B C D E F\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x00 ..= 0x07.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x08 ..= 0x0F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x10 ..= 0x17.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x18 ..= 0x1F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x20 ..= 0x27.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x28 ..= 0x2F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x30 ..= 0x37.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x38 ..= 0x3F.\n\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x40 ..= 0x47.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x48 ..= 0x4F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x50 ..= 0x57.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x58 ..= 0x5F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x60 ..= 0x67.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x68 .." + - "= 0x6F.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x70 ..= 0x77.\n 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x78 ..= 0x7F.\n\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x80 ..= 0x87.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x88 ..= 0x8F.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x90 ..= 0x97.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x98 ..= 0x9F.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA0 ..= 0xA7.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA8 ..= 0xAF.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB0 ..= 0xB7.\n 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB8 ..= 0xBF.\n\n 0x80, 0x80, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC0 ..= 0xC7.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC8 ..= 0xCF.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD0 ..= 0xD7.\n 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD8 ..= 0xDF.\n 0x02, 0x02, 0x02" + - ", 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE0 ..= 0xE7.\n 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE8 ..= 0xEF.\n 0x03, 0x03, 0x03, 0x03, 0x03, 0x80, 0x80, 0x80, // 0xF0 ..= 0xF7.\n 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, // 0xF8 ..= 0xFF.\n // 0 1 2 3 4 5 6 7\n // 8 9 A B C D E F\n};\n\nwuffs_base__utf_8__next__output //\nwuffs_base__utf_8__next(wuffs_base__slice_u8 s) {\n if (s.len == 0) {\n return wuffs_base__make_utf_8__next__output(0, 0);\n }\n uint32_t c = s.ptr[0];\n switch (wuffs_base__utf_8__byte_length_minus_1[c & 0xFF]) {\n case 0:\n return wuffs_base__make_utf_8__next__output(c, 1);\n\n case 1:\n if (s.len < 2) {\n break;\n }\n c = wuffs_base__load_u16le__no_bounds_check(s.ptr);\n if ((c & 0xC000) != 0x8000) {\n break;\n }\n c = (0x0007C0 & (c << 6)) | (0x00003F & (c >> 8));\n return wuffs_base__make_utf_8__next__output(c, 2);\n\n case 2:\n if (s.len < 3) {\n " + - " break;\n }\n c = wuffs_base__load_u24le__no_bounds_check(s.ptr);\n if ((c & 0xC0C000) != 0x808000) {\n break;\n }\n c = (0x00F000 & (c << 12)) | (0x000FC0 & (c >> 2)) |\n (0x00003F & (c >> 16));\n if ((c <= 0x07FF) || ((0xD800 <= c) && (c <= 0xDFFF))) {\n break;\n }\n return wuffs_base__make_utf_8__next__output(c, 3);\n\n case 3:\n if (s.len < 4) {\n break;\n }\n c = wuffs_base__load_u32le__no_bounds_check(s.ptr);\n if ((c & 0xC0C0C000) != 0x80808000) {\n break;\n }\n c = (0x1C0000 & (c << 18)) | (0x03F000 & (c << 4)) |\n (0x000FC0 & (c >> 10)) | (0x00003F & (c >> 24));\n if ((c <= 0xFFFF) || (0x110000 <= c)) {\n break;\n }\n return wuffs_base__make_utf_8__next__output(c, 4);\n }\n\n return wuffs_base__make_utf_8__next__output(\n WUFFS_BASE__UNICODE_REPLACEMENT_CHARACTER, 1);\n}\n\nsize_t //\nwuffs_base__utf_8__longest_valid_prefix(wuffs_base__slice_u8 s) {\n // TODO: possibly optimize the all-A" + - "SCII case (4 or 8 bytes at a time).\n //\n // TODO: possibly optimize this by manually inlining the\n // wuffs_base__utf_8__next calls.\n size_t original_len = s.len;\n while (s.len > 0) {\n wuffs_base__utf_8__next__output o = wuffs_base__utf_8__next(s);\n if ((o.code_point > 0x7F) && (o.byte_length == 1)) {\n break;\n }\n s.ptr += o.byte_length;\n s.len -= o.byte_length;\n }\n return original_len - s.len;\n}\n\nsize_t //\nwuffs_base__ascii__longest_valid_prefix(wuffs_base__slice_u8 s) {\n // TODO: possibly optimize this by checking 4 or 8 bytes at a time.\n uint8_t* original_ptr = s.ptr;\n uint8_t* p = s.ptr;\n uint8_t* q = s.ptr + s.len;\n for (; (p != q) && ((*p & 0x80) == 0); p++) {\n }\n return (size_t)(p - original_ptr);\n}\n" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n return wuffs_base__slice_u8__copy_from_slice(dst, src);\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3;\n if (len > 0) {\n memmove(dst.ptr, src.ptr, len * 3);\n }\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 <" + + " src_len4 ? dst_len4 : src_len4;\n if (len > 0) {\n memmove(dst.ptr, src.ptr, len * 4);\n }\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t b5 = s[0] >> 3;\n uint32_t g6 = s[1] >> 2;\n uint32_t r5 = s[2] >> 3;\n uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 3;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len4 = src.len / 4;\n size_t len = " + + "dst_len2 < src_len4 ? dst_len2 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)))));\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0" + + "x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n\n // Convert from 565 color to 16-bit color.\n uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2));\n uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11);\n uint32_t dr = (0x8421 * old_r5) >> 4;\n uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5);\n uint32_t dg = (0x1041 * old_g6) >> 2;\n uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0);\n uint32_t db = (0x8421 * old_b5) >> 4;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 565 color and combine the components.\n uint32_t new_r5 = 0x1F & (dr >> 11);\n uint32_t new_g6 = 0x3F & (dg >> 10);\n uint32_t new_b5 = 0x1F & (db >> 11);\n uint32_t new_rgb" + + "_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2),\n (uint16_t)new_rgb_565);\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t y5 = s[0] >> 3;\n uint32_t y6 = s[0] >> 2;\n uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__index__src(\n wuffs_base__slice_u8" + + " dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (1 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (2 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (3 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr" + + " + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 2;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len2 = dst.len / 2;\n size_t len = dst_len2 < src.len ? dst_len2 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u16le__no_bounds" + + "_check(d + (0 * 2), (uint16_t)s0);\n }\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 =\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 = dst.len / 3;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4;\n uint8_t* d = d" + + "st.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n uint32_t dr = 0x101 * ((uint32_t)d[2]);\n uint32_t dg = 0x101 * ((uint32_t)d[1]);\n uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n d[0] = (uint8_t)(db >> 8);\n d[1] = (uint8_t)(dg >> 8);\n d[2] = (uint8_t)(dr >> 8);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len4 = src.len / 4;\n size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4;\n ui" + + "nt8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len3 = dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n // The comparison in the while condition is \">\", not \">=\", because with\n // \">=\", the last 4-byte store could write past the end of the dst slice.\n //\n // Each 4-byte store writes one too many bytes, but a subsequent store\n // will overwrite that with the correct byte. There is always another\n // store, whether a 4-byte store in this loop or a 1-byte store in the\n // next loop.\n while (n > loop_unroll_count) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 3), wuffs_base__load_u32le__no_bounds_c" + + "heck(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__" + + "pixel_swizzler__xxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len3 = dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n " + + " ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n }\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len3 =" + + " dst.len / 3;\n size_t len = dst_len3 < src.len ? dst_len3 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint8_t s0 = s[0];\n d[0] = s0;\n d[1] = s0;\n d[2] = s0;\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_pale" + + "tte.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count)" + + " {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count *" + + " 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n" + + "\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "" + + "// --------\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__pixel_swizzler__xxx__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WU" + + "FFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__pixel_swizzler__xxxx__y;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_1_1;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette,\n " + + " src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_pa" + + "lette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette,\n src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n if (wuffs_base__pixel_swizzler__" + + "swap_rgbx_bgrx(dst_palette,\n src_palette) != 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__B" + + "GR:\n return wuffs_base__pixel_swizzler__copy_3_3;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__xxxx__xxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_b" + + "ase__pixel_swizzler__bgr_565__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_4_4;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swiz" + + "zler__bgra_premul__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\n" + + "" + + "// --------\n\nwuffs_base__status //\nwuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__pixel_format src_format,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n if (!p) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n\n // TODO: support many more formats.\n\n wuffs_base__pixel_swizzler__func func = NULL;\n\n switch (src_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette,\n src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n p, dst_format, dst_" + + "palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n func = wuffs_base__pixel_swizzler__prepare__bgr(\n p, dst_format, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n p, dst_format, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_impl.func = func;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}\n\nuint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (p && p->private_impl.func) {\n return (*p->private_impl.func)(dst, dst_palette, src);\n }\n return 0;\n}\n" + "" const baseFundamentalPrivateH = "" + diff --git a/internal/cgen/gen.go b/internal/cgen/gen.go index 21a4f6dd..b946828f 100644 --- a/internal/cgen/gen.go +++ b/internal/cgen/gen.go @@ -77,6 +77,9 @@ func genBase(out *bytes.Buffer) error { {"base/image-impl.c", "baseImageImplC"}, {"base/strconv-impl.c", "baseStrConvImplC"}, + {"base/f64conv-submodule.c", "baseF64ConvSubmoduleC"}, + {"base/pixconv-submodule.c", "basePixConvSubmoduleC"}, + {"base/fundamental-private.h", "baseFundamentalPrivateH"}, {"base/fundamental-public.h", "baseFundamentalPublicH"}, {"base/memory-private.h", "baseMemoryPrivateH"}, diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 72a4d016..b5161ef4 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -8531,51 +8531,6 @@ const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { // -------- -static uint64_t // -wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - - size_t n = len4; - while (n--) { - uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); - uint32_t b5 = 0x1F & (argb >> (8 - 5)); - uint32_t g6 = 0x3F & (argb >> (16 - 6)); - uint32_t r5 = 0x1F & (argb >> (24 - 5)); - uint32_t alpha = argb & 0xFF000000; - wuffs_base__store_u32le__no_bounds_check( - d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0)); - s += 4; - d += 4; - } - return len4 * 4; -} - -static uint64_t // -wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - - size_t n = len4; - while (n--) { - uint8_t b0 = s[0]; - uint8_t b1 = s[1]; - uint8_t b2 = s[2]; - uint8_t b3 = s[3]; - d[0] = b2; - d[1] = b1; - d[2] = b0; - d[3] = b3; - s += 4; - d += 4; - } - return len4 * 4; -} - static inline uint32_t // wuffs_base__swap_u32_argb_abgr(uint32_t u) { uint32_t o = u & 0xFF00FF00; @@ -8586,152 +8541,6 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { // -------- -static inline uint32_t // -wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, - uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); - - // Convert dst from nonpremul to premul. - dr = (dr * da) / 0xFFFF; - dg = (dg * da) / 0xFFFF; - db = (db * da) / 0xFFFF; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert dst from premul to nonpremul. - if (da != 0) { - dr = (dr * 0xFFFF) / da; - dg = (dg * 0xFFFF) / da; - db = (db * 0xFFFF) / da; - } - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, - uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); - - // Convert dst from nonpremul to premul. - dr = (dr * da) / 0xFFFF; - dg = (dg * da) / 0xFFFF; - db = (db * da) / 0xFFFF; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (premul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = sr + ((dr * ia) / 0xFFFF); - dg = sg + ((dg * ia) / 0xFFFF); - db = sb + ((db * ia) / 0xFFFF); - - // Convert dst from premul to nonpremul. - if (da != 0) { - dr = (dr * 0xFFFF) / da; - dg = (dg * 0xFFFF) / da; - db = (db * 0xFFFF) / da; - } - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, - uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -static inline uint32_t // -wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, - uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); - uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); - uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); - uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); - uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); - uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); - uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); - uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (premul) over dst (premul). - da = sa + ((da * ia) / 0xFFFF); - dr = sr + ((dr * ia) / 0xFFFF); - dg = sg + ((dg * ia) / 0xFFFF); - db = sb + ((db * ia) / 0xFFFF); - - // Convert from 16-bit color to 8-bit color and combine the components. - da >>= 8; - dr >>= 8; - dg >>= 8; - db >>= 8; - return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); -} - -// -------- - wuffs_base__color_u32_argb_premul // wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, uint32_t x, @@ -8956,988 +8765,6 @@ wuffs_base__pixel_palette__closest_element( return (uint8_t)best_index; } -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - return wuffs_base__slice_u8__copy_from_slice(dst, src); -} - -static uint64_t // -wuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len3 = src.len / 3; - size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3; - if (len > 0) { - memmove(dst.ptr, src.ptr, len * 3); - } - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - if (len > 0) { - memmove(dst.ptr, src.ptr, len * 4); - } - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len3 = src.len / 3; - size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t b5 = s[0] >> 3; - uint32_t g6 = s[1] >> 2; - uint32_t r5 = s[2] >> 3; - uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); - - s += 1 * 3; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len4 = src.len / 4; - size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), - wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))))); - - s += 1 * 4; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t src_len4 = src.len / 4; - size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * ((uint32_t)s[3]); - uint32_t sr = 0x101 * ((uint32_t)s[2]); - uint32_t sg = 0x101 * ((uint32_t)s[1]); - uint32_t sb = 0x101 * ((uint32_t)s[0]); - - // Convert from 565 color to 16-bit color. - uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); - uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); - uint32_t dr = (0x8421 * old_r5) >> 4; - uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); - uint32_t dg = (0x1041 * old_g6) >> 2; - uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); - uint32_t db = (0x8421 * old_b5) >> 4; - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 565 color and combine the components. - uint32_t new_r5 = 0x1F & (dr >> 11); - uint32_t new_g6 = 0x3F & (dg >> 10); - uint32_t new_b5 = 0x1F & (db >> 11); - uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), - (uint16_t)new_rgb_565); - - s += 1 * 4; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t y5 = s[0] >> 3; - uint32_t y6 = s[0] >> 2; - uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0); - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__index__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (1 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (2 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u16le__no_bounds_check( - d + (3 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 2; - n -= loop_unroll_count; - } - - while (n >= 1) { - wuffs_base__store_u16le__no_bounds_check( - d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len2 = dst.len / 2; - size_t len = dst_len2 < src.len ? dst_len2 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0); - } - - s += 1 * 1; - d += 1 * 2; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len4 = src.len / 4; - size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))); - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - - s += 1 * 4; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t src_len4 = src.len / 4; - size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - // Convert from 8-bit color to 16-bit color. - uint32_t sa = 0x101 * ((uint32_t)s[3]); - uint32_t sr = 0x101 * ((uint32_t)s[2]); - uint32_t sg = 0x101 * ((uint32_t)s[1]); - uint32_t sb = 0x101 * ((uint32_t)s[0]); - uint32_t dr = 0x101 * ((uint32_t)d[2]); - uint32_t dg = 0x101 * ((uint32_t)d[1]); - uint32_t db = 0x101 * ((uint32_t)d[0]); - - // Calculate the inverse of the src-alpha: how much of the dst to keep. - uint32_t ia = 0xFFFF - sa; - - // Composite src (nonpremul) over dst (premul). - dr = ((sr * sa) + (dr * ia)) / 0xFFFF; - dg = ((sg * sa) + (dg * ia)) / 0xFFFF; - db = ((sb * sa) + (db * ia)) / 0xFFFF; - - // Convert from 16-bit color to 8-bit color. - d[0] = (uint8_t)(db >> 8); - d[1] = (uint8_t)(dg >> 8); - d[2] = (uint8_t)(dr >> 8); - - s += 1 * 4; - d += 1 * 3; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len4 = src.len / 4; - size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); - - s += 1 * 4; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - // The comparison in the while condition is ">", not ">=", because with - // ">=", the last 4-byte store could write past the end of the dst slice. - // - // Each 4-byte store writes one too many bytes, but a subsequent store - // will overwrite that with the correct byte. There is always another - // store, whether a 4-byte store in this loop or a 1-byte store in the - // next loop. - while (n > loop_unroll_count) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (1 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (2 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (3 * 3), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 3; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - } - uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[1] * 4)); - if (s1) { - wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1); - } - uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[2] * 4)); - if (s2) { - wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2); - } - uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[3] * 4)); - if (s3) { - wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3); - } - - s += loop_unroll_count * 1; - d += loop_unroll_count * 3; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); - } - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len3 = dst.len / 3; - size_t len = dst_len3 < src.len ? dst_len3 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - uint8_t s0 = s[0]; - d[0] = s0; - d[1] = s0; - d[2] = s0; - - s += 1 * 1; - d += 1 * 3; - n -= 1; - } - - return len; -} - -// -------- - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (1 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[1] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (2 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[2] * 4))); - wuffs_base__store_u32le__no_bounds_check( - d + (3 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[3] * 4))); - - s += loop_unroll_count * 1; - d += loop_unroll_count * 4; - n -= loop_unroll_count; - } - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( - dst_palette.ptr + ((size_t)s[0] * 4))); - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (dst_palette.len != 1024) { - return 0; - } - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - const size_t loop_unroll_count = 4; - - while (n >= loop_unroll_count) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); - } - uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[1] * 4)); - if (s1) { - wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1); - } - uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[2] * 4)); - if (s2) { - wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2); - } - uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[3] * 4)); - if (s3) { - wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3); - } - - s += loop_unroll_count * 1; - d += loop_unroll_count * 4; - n -= loop_unroll_count; - } - - while (n >= 1) { - uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + - ((size_t)s[0] * 4)); - if (s0) { - wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); - } - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t src_len3 = src.len / 3; - size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), - 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3))); - - s += 1 * 3; - d += 1 * 4; - n -= 1; - } - - return len; -} - -static uint64_t // -wuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - size_t dst_len4 = dst.len / 4; - size_t len = dst_len4 < src.len ? dst_len4 : src.len; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; - - // TODO: unroll. - - while (n >= 1) { - wuffs_base__store_u32le__no_bounds_check( - d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0])); - - s += 1 * 1; - d += 1 * 4; - n -= 1; - } - - return len; -} - -// -------- - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__pixel_swizzler__bgr_565__y; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - case WUFFS_BASE__PIXEL_FORMAT__RGB: - return wuffs_base__pixel_swizzler__xxx__y; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - return wuffs_base__pixel_swizzler__xxxx__y; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( - wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__copy_1_1; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr_565__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != - 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { - return NULL; - } - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__xxxx__index__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; - } - return NULL; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__pixel_swizzler__bgr_565__bgr; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - return wuffs_base__pixel_swizzler__copy_3_3; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - return wuffs_base__pixel_swizzler__xxxx__xxx; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - // TODO. - break; - } - return NULL; -} - -static wuffs_base__pixel_swizzler__func // -wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( - wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - switch (dst_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__copy_4_4; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - switch (blend) { - case WUFFS_BASE__PIXEL_BLEND__SRC: - return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src; - case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: - return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; - } - return NULL; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - // TODO. - break; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - // TODO. - break; - } - return NULL; -} - -// -------- - -wuffs_base__status // -wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, - wuffs_base__pixel_format dst_format, - wuffs_base__slice_u8 dst_palette, - wuffs_base__pixel_format src_format, - wuffs_base__slice_u8 src_palette, - wuffs_base__pixel_blend blend) { - if (!p) { - return wuffs_base__make_status(wuffs_base__error__bad_receiver); - } - - // TODO: support many more formats. - - wuffs_base__pixel_swizzler__func func = NULL; - - switch (src_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__Y: - func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette, - src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( - p, dst_format, dst_palette, src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGR: - func = wuffs_base__pixel_swizzler__prepare__bgr( - p, dst_format, dst_palette, src_palette, blend); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( - p, dst_format, dst_palette, src_palette, blend); - break; - } - - p->private_impl.func = func; - return wuffs_base__make_status( - func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option); -} - -uint64_t // -wuffs_base__pixel_swizzler__swizzle_interleaved( - const wuffs_base__pixel_swizzler* p, - wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 dst_palette, - wuffs_base__slice_u8 src) { - if (p && p->private_impl.func) { - return (*p->private_impl.func)(dst, dst_palette, src); - } - return 0; -} - // ---------------- String Conversions // wuffs_base__parse_number__foo_digits entries are 0x00 for invalid digits, @@ -10223,6 +9050,233 @@ fail_out_of_bounds: } while (0); } +// ---------------- Hexadecimal + +size_t // +wuffs_base__hexadecimal__decode2(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t src_len2 = src.len / 2; + size_t len = dst.len < src_len2 ? dst.len : src_len2; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + while (n--) { + *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[0]] << 4) | + (wuffs_base__parse_number__hexadecimal_digits[s[1]] & 0x0F)); + d += 1; + s += 2; + } + + return len; +} + +size_t // +wuffs_base__hexadecimal__decode4(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t src_len4 = src.len / 4; + size_t len = dst.len < src_len4 ? dst.len : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + while (n--) { + *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[2]] << 4) | + (wuffs_base__parse_number__hexadecimal_digits[s[3]] & 0x0F)); + d += 1; + s += 4; + } + + return len; +} + +// ---------------- Unicode and UTF-8 + +size_t // +wuffs_base__utf_8__encode(wuffs_base__slice_u8 dst, uint32_t code_point) { + if (code_point <= 0x7F) { + if (dst.len >= 1) { + dst.ptr[0] = (uint8_t)(code_point); + return 1; + } + + } else if (code_point <= 0x07FF) { + if (dst.len >= 2) { + dst.ptr[0] = (uint8_t)(0xC0 | ((code_point >> 6))); + dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); + return 2; + } + + } else if (code_point <= 0xFFFF) { + if ((dst.len >= 3) && ((code_point < 0xD800) || (0xDFFF < code_point))) { + dst.ptr[0] = (uint8_t)(0xE0 | ((code_point >> 12))); + dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F)); + dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); + return 3; + } + + } else if (code_point <= 0x10FFFF) { + if (dst.len >= 4) { + dst.ptr[0] = (uint8_t)(0xF0 | ((code_point >> 18))); + dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 12) & 0x3F)); + dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F)); + dst.ptr[3] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); + return 4; + } + } + + return 0; +} + +// wuffs_base__utf_8__byte_length_minus_1 is the byte length (minus 1) of a +// UTF-8 encoded code point, based on the encoding's initial byte. +// - 0x00 is 1-byte UTF-8 (ASCII). +// - 0x01 is the start of 2-byte UTF-8. +// - 0x02 is the start of 3-byte UTF-8. +// - 0x03 is the start of 4-byte UTF-8. +// - 0x40 is a UTF-8 tail byte. +// - 0x80 is invalid UTF-8. +// +// RFC 3629 (UTF-8) gives this grammar for valid UTF-8: +// UTF8-1 = %x00-7F +// UTF8-2 = %xC2-DF UTF8-tail +// UTF8-3 = %xE0 %xA0-BF UTF8-tail / %xE1-EC 2( UTF8-tail ) / +// %xED %x80-9F UTF8-tail / %xEE-EF 2( UTF8-tail ) +// UTF8-4 = %xF0 %x90-BF 2( UTF8-tail ) / %xF1-F3 3( UTF8-tail ) / +// %xF4 %x80-8F 2( UTF8-tail ) +// UTF8-tail = %x80-BF +static const uint8_t wuffs_base__utf_8__byte_length_minus_1[256] = { + // 0 1 2 3 4 5 6 7 + // 8 9 A B C D E F + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x00 ..= 0x07. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x08 ..= 0x0F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x10 ..= 0x17. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x18 ..= 0x1F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x20 ..= 0x27. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x28 ..= 0x2F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x30 ..= 0x37. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x38 ..= 0x3F. + + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x40 ..= 0x47. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x48 ..= 0x4F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x50 ..= 0x57. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x58 ..= 0x5F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x60 ..= 0x67. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x68 ..= 0x6F. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x70 ..= 0x77. + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x78 ..= 0x7F. + + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x80 ..= 0x87. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x88 ..= 0x8F. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x90 ..= 0x97. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x98 ..= 0x9F. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA0 ..= 0xA7. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA8 ..= 0xAF. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB0 ..= 0xB7. + 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB8 ..= 0xBF. + + 0x80, 0x80, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC0 ..= 0xC7. + 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC8 ..= 0xCF. + 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD0 ..= 0xD7. + 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD8 ..= 0xDF. + 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE0 ..= 0xE7. + 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE8 ..= 0xEF. + 0x03, 0x03, 0x03, 0x03, 0x03, 0x80, 0x80, 0x80, // 0xF0 ..= 0xF7. + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, // 0xF8 ..= 0xFF. + // 0 1 2 3 4 5 6 7 + // 8 9 A B C D E F +}; + +wuffs_base__utf_8__next__output // +wuffs_base__utf_8__next(wuffs_base__slice_u8 s) { + if (s.len == 0) { + return wuffs_base__make_utf_8__next__output(0, 0); + } + uint32_t c = s.ptr[0]; + switch (wuffs_base__utf_8__byte_length_minus_1[c & 0xFF]) { + case 0: + return wuffs_base__make_utf_8__next__output(c, 1); + + case 1: + if (s.len < 2) { + break; + } + c = wuffs_base__load_u16le__no_bounds_check(s.ptr); + if ((c & 0xC000) != 0x8000) { + break; + } + c = (0x0007C0 & (c << 6)) | (0x00003F & (c >> 8)); + return wuffs_base__make_utf_8__next__output(c, 2); + + case 2: + if (s.len < 3) { + break; + } + c = wuffs_base__load_u24le__no_bounds_check(s.ptr); + if ((c & 0xC0C000) != 0x808000) { + break; + } + c = (0x00F000 & (c << 12)) | (0x000FC0 & (c >> 2)) | + (0x00003F & (c >> 16)); + if ((c <= 0x07FF) || ((0xD800 <= c) && (c <= 0xDFFF))) { + break; + } + return wuffs_base__make_utf_8__next__output(c, 3); + + case 3: + if (s.len < 4) { + break; + } + c = wuffs_base__load_u32le__no_bounds_check(s.ptr); + if ((c & 0xC0C0C000) != 0x80808000) { + break; + } + c = (0x1C0000 & (c << 18)) | (0x03F000 & (c << 4)) | + (0x000FC0 & (c >> 10)) | (0x00003F & (c >> 24)); + if ((c <= 0xFFFF) || (0x110000 <= c)) { + break; + } + return wuffs_base__make_utf_8__next__output(c, 4); + } + + return wuffs_base__make_utf_8__next__output( + WUFFS_BASE__UNICODE_REPLACEMENT_CHARACTER, 1); +} + +size_t // +wuffs_base__utf_8__longest_valid_prefix(wuffs_base__slice_u8 s) { + // TODO: possibly optimize the all-ASCII case (4 or 8 bytes at a time). + // + // TODO: possibly optimize this by manually inlining the + // wuffs_base__utf_8__next calls. + size_t original_len = s.len; + while (s.len > 0) { + wuffs_base__utf_8__next__output o = wuffs_base__utf_8__next(s); + if ((o.code_point > 0x7F) && (o.byte_length == 1)) { + break; + } + s.ptr += o.byte_length; + s.len -= o.byte_length; + } + return original_len - s.len; +} + +size_t // +wuffs_base__ascii__longest_valid_prefix(wuffs_base__slice_u8 s) { + // TODO: possibly optimize this by checking 4 or 8 bytes at a time. + uint8_t* original_ptr = s.ptr; + uint8_t* p = s.ptr; + uint8_t* q = s.ptr + s.len; + for (; (p != q) && ((*p & 0x80) == 0); p++) { + } + return (size_t)(p - original_ptr); +} + +#endif // !defined(WUFFS_CONFIG__MODULES) || + // defined(WUFFS_CONFIG__MODULE__BASE) + +#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) + // ---------------- IEEE 754 Floating Point #define WUFFS_BASE__PRIVATE_IMPLEMENTATION__HPD__DECIMAL_POINT__RANGE 1023 @@ -11472,226 +10526,1184 @@ infinity: } while (0); } -// ---------------- Hexadecimal +#endif // !defined(WUFFS_CONFIG__MODULES) || + // defined(WUFFS_CONFIG__MODULE__BASE) -size_t // -wuffs_base__hexadecimal__decode2(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t src_len2 = src.len / 2; - size_t len = dst.len < src_len2 ? dst.len : src_len2; - uint8_t* d = dst.ptr; - uint8_t* s = src.ptr; - size_t n = len; +#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) - while (n--) { - *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[0]] << 4) | - (wuffs_base__parse_number__hexadecimal_digits[s[1]] & 0x0F)); - d += 1; - s += 2; +// ---------------- Pixel Swizzler + +static inline uint32_t // +wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_nonpremul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; } - return len; + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } -size_t // -wuffs_base__hexadecimal__decode4(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t src_len4 = src.len / 4; - size_t len = dst.len < src_len4 ? dst.len : src_len4; +static inline uint32_t // +wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0)); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +static inline uint32_t // +wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, + uint32_t src_nonpremul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +static inline uint32_t // +wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 0)); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert from 16-bit color to 8-bit color and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; uint8_t* d = dst.ptr; uint8_t* s = src.ptr; - size_t n = len; + size_t n = len4; while (n--) { - *d = (uint8_t)((wuffs_base__parse_number__hexadecimal_digits[s[2]] << 4) | - (wuffs_base__parse_number__hexadecimal_digits[s[3]] & 0x0F)); - d += 1; + uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); + uint32_t b5 = 0x1F & (argb >> (8 - 5)); + uint32_t g6 = 0x3F & (argb >> (16 - 6)); + uint32_t r5 = 0x1F & (argb >> (24 - 5)); + uint32_t alpha = argb & 0xFF000000; + wuffs_base__store_u32le__no_bounds_check( + d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0)); s += 4; + d += 4; + } + return len4 * 4; +} + +static uint64_t // +wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + + size_t n = len4; + while (n--) { + uint8_t b0 = s[0]; + uint8_t b1 = s[1]; + uint8_t b2 = s[2]; + uint8_t b3 = s[3]; + d[0] = b2; + d[1] = b1; + d[2] = b0; + d[3] = b3; + s += 4; + d += 4; + } + return len4 * 4; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + return wuffs_base__slice_u8__copy_from_slice(dst, src); +} + +static uint64_t // +wuffs_base__pixel_swizzler__copy_3_3(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len3 = src.len / 3; + size_t len = dst_len3 < src_len3 ? dst_len3 : src_len3; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 3); + } + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 4); + } + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgr(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len3 = src.len / 3; + size_t len = dst_len2 < src_len3 ? dst_len2 : src_len3; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t b5 = s[0] >> 3; + uint32_t g6 = s[1] >> 2; + uint32_t r5 = s[2] >> 3; + uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); + + s += 1 * 3; + d += 1 * 2; + n -= 1; } return len; } -// ---------------- Unicode and UTF-8 +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len4 = src.len / 4; + size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; -size_t // -wuffs_base__utf_8__encode(wuffs_base__slice_u8 dst, uint32_t code_point) { - if (code_point <= 0x7F) { - if (dst.len >= 1) { - dst.ptr[0] = (uint8_t)(code_point); - return 1; - } + // TODO: unroll. - } else if (code_point <= 0x07FF) { - if (dst.len >= 2) { - dst.ptr[0] = (uint8_t)(0xC0 | ((code_point >> 6))); - dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); - return 2; - } + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))))); - } else if (code_point <= 0xFFFF) { - if ((dst.len >= 3) && ((code_point < 0xD800) || (0xDFFF < code_point))) { - dst.ptr[0] = (uint8_t)(0xE0 | ((code_point >> 12))); - dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F)); - dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); - return 3; - } - - } else if (code_point <= 0x10FFFF) { - if (dst.len >= 4) { - dst.ptr[0] = (uint8_t)(0xF0 | ((code_point >> 18))); - dst.ptr[1] = (uint8_t)(0x80 | ((code_point >> 12) & 0x3F)); - dst.ptr[2] = (uint8_t)(0x80 | ((code_point >> 6) & 0x3F)); - dst.ptr[3] = (uint8_t)(0x80 | ((code_point >> 0) & 0x3F)); - return 4; - } + s += 1 * 4; + d += 1 * 2; + n -= 1; } - return 0; + return len; } -// wuffs_base__utf_8__byte_length_minus_1 is the byte length (minus 1) of a -// UTF-8 encoded code point, based on the encoding's initial byte. -// - 0x00 is 1-byte UTF-8 (ASCII). -// - 0x01 is the start of 2-byte UTF-8. -// - 0x02 is the start of 3-byte UTF-8. -// - 0x03 is the start of 4-byte UTF-8. -// - 0x40 is a UTF-8 tail byte. -// - 0x80 is invalid UTF-8. -// -// RFC 3629 (UTF-8) gives this grammar for valid UTF-8: -// UTF8-1 = %x00-7F -// UTF8-2 = %xC2-DF UTF8-tail -// UTF8-3 = %xE0 %xA0-BF UTF8-tail / %xE1-EC 2( UTF8-tail ) / -// %xED %x80-9F UTF8-tail / %xEE-EF 2( UTF8-tail ) -// UTF8-4 = %xF0 %x90-BF 2( UTF8-tail ) / %xF1-F3 3( UTF8-tail ) / -// %xF4 %x80-8F 2( UTF8-tail ) -// UTF8-tail = %x80-BF -static const uint8_t wuffs_base__utf_8__byte_length_minus_1[256] = { - // 0 1 2 3 4 5 6 7 - // 8 9 A B C D E F - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x00 ..= 0x07. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x08 ..= 0x0F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x10 ..= 0x17. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x18 ..= 0x1F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x20 ..= 0x27. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x28 ..= 0x2F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x30 ..= 0x37. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x38 ..= 0x3F. +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t src_len4 = src.len / 4; + size_t len = dst_len2 < src_len4 ? dst_len2 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x40 ..= 0x47. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x48 ..= 0x4F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x50 ..= 0x57. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x58 ..= 0x5F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x60 ..= 0x67. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x68 ..= 0x6F. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x70 ..= 0x77. - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // 0x78 ..= 0x7F. + // TODO: unroll. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x80 ..= 0x87. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x88 ..= 0x8F. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x90 ..= 0x97. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0x98 ..= 0x9F. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA0 ..= 0xA7. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xA8 ..= 0xAF. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB0 ..= 0xB7. - 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, // 0xB8 ..= 0xBF. + while (n >= 1) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * ((uint32_t)s[3]); + uint32_t sr = 0x101 * ((uint32_t)s[2]); + uint32_t sg = 0x101 * ((uint32_t)s[1]); + uint32_t sb = 0x101 * ((uint32_t)s[0]); - 0x80, 0x80, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC0 ..= 0xC7. - 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xC8 ..= 0xCF. - 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD0 ..= 0xD7. - 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, // 0xD8 ..= 0xDF. - 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE0 ..= 0xE7. - 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, // 0xE8 ..= 0xEF. - 0x03, 0x03, 0x03, 0x03, 0x03, 0x80, 0x80, 0x80, // 0xF0 ..= 0xF7. - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, // 0xF8 ..= 0xFF. - // 0 1 2 3 4 5 6 7 - // 8 9 A B C D E F -}; + // Convert from 565 color to 16-bit color. + uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); + uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); + uint32_t dr = (0x8421 * old_r5) >> 4; + uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); + uint32_t dg = (0x1041 * old_g6) >> 2; + uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); + uint32_t db = (0x8421 * old_b5) >> 4; -wuffs_base__utf_8__next__output // -wuffs_base__utf_8__next(wuffs_base__slice_u8 s) { - if (s.len == 0) { - return wuffs_base__make_utf_8__next__output(0, 0); - } - uint32_t c = s.ptr[0]; - switch (wuffs_base__utf_8__byte_length_minus_1[c & 0xFF]) { - case 0: - return wuffs_base__make_utf_8__next__output(c, 1); + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; - case 1: - if (s.len < 2) { - break; - } - c = wuffs_base__load_u16le__no_bounds_check(s.ptr); - if ((c & 0xC000) != 0x8000) { - break; - } - c = (0x0007C0 & (c << 6)) | (0x00003F & (c >> 8)); - return wuffs_base__make_utf_8__next__output(c, 2); + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; - case 2: - if (s.len < 3) { - break; - } - c = wuffs_base__load_u24le__no_bounds_check(s.ptr); - if ((c & 0xC0C000) != 0x808000) { - break; - } - c = (0x00F000 & (c << 12)) | (0x000FC0 & (c >> 2)) | - (0x00003F & (c >> 16)); - if ((c <= 0x07FF) || ((0xD800 <= c) && (c <= 0xDFFF))) { - break; - } - return wuffs_base__make_utf_8__next__output(c, 3); + // Convert from 16-bit color to 565 color and combine the components. + uint32_t new_r5 = 0x1F & (dr >> 11); + uint32_t new_g6 = 0x3F & (dg >> 10); + uint32_t new_b5 = 0x1F & (db >> 11); + uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), + (uint16_t)new_rgb_565); - case 3: - if (s.len < 4) { - break; - } - c = wuffs_base__load_u32le__no_bounds_check(s.ptr); - if ((c & 0xC0C0C000) != 0x80808000) { - break; - } - c = (0x1C0000 & (c << 18)) | (0x03F000 & (c << 4)) | - (0x000FC0 & (c >> 10)) | (0x00003F & (c >> 24)); - if ((c <= 0xFFFF) || (0x110000 <= c)) { - break; - } - return wuffs_base__make_utf_8__next__output(c, 4); + s += 1 * 4; + d += 1 * 2; + n -= 1; } - return wuffs_base__make_utf_8__next__output( - WUFFS_BASE__UNICODE_REPLACEMENT_CHARACTER, 1); + return len; } -size_t // -wuffs_base__utf_8__longest_valid_prefix(wuffs_base__slice_u8 s) { - // TODO: possibly optimize the all-ASCII case (4 or 8 bytes at a time). +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t y5 = s[0] >> 3; + uint32_t y6 = s[0] >> 2; + uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565); + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__index__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (1 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (2 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u16le__no_bounds_check( + d + (3 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 2; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), wuffs_base__load_u16le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len2 = dst.len / 2; + size_t len = dst_len2 < src.len ? dst_len2 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0); + } + + s += 1 * 1; + d += 1 * 2; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len4 = src.len / 4; + size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(s + (0 * 4))); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 4; + d += 1 * 3; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t src_len4 = src.len / 4; + size_t len = dst_len3 < src_len4 ? dst_len3 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * ((uint32_t)s[3]); + uint32_t sr = 0x101 * ((uint32_t)s[2]); + uint32_t sg = 0x101 * ((uint32_t)s[1]); + uint32_t sb = 0x101 * ((uint32_t)s[0]); + uint32_t dr = 0x101 * ((uint32_t)d[2]); + uint32_t dg = 0x101 * ((uint32_t)d[1]); + uint32_t db = 0x101 * ((uint32_t)d[0]); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color. + d[0] = (uint8_t)(db >> 8); + d[1] = (uint8_t)(dg >> 8); + d[2] = (uint8_t)(dr >> 8); + + s += 1 * 4; + d += 1 * 3; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__xxx__index__src(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + // The comparison in the while condition is ">", not ">=", because with + // ">=", the last 4-byte store could write past the end of the dst slice. // - // TODO: possibly optimize this by manually inlining the - // wuffs_base__utf_8__next calls. - size_t original_len = s.len; - while (s.len > 0) { - wuffs_base__utf_8__next__output o = wuffs_base__utf_8__next(s); - if ((o.code_point > 0x7F) && (o.byte_length == 1)) { - break; - } - s.ptr += o.byte_length; - s.len -= o.byte_length; + // Each 4-byte store writes one too many bytes, but a subsequent store + // will overwrite that with the correct byte. There is always another + // store, whether a 4-byte store in this loop or a 1-byte store in the + // next loop. + while (n > loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 3), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 3; + n -= loop_unroll_count; } - return original_len - s.len; + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; } -size_t // -wuffs_base__ascii__longest_valid_prefix(wuffs_base__slice_u8 s) { - // TODO: possibly optimize this by checking 4 or 8 bytes at a time. - uint8_t* original_ptr = s.ptr; - uint8_t* p = s.ptr; - uint8_t* q = s.ptr + s.len; - for (; (p != q) && ((*p & 0x80) == 0); p++) { +static uint64_t // +wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; } - return (size_t)(p - original_ptr); + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + } + uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[1] * 4)); + if (s1) { + wuffs_base__store_u24le__no_bounds_check(d + (1 * 3), s1); + } + uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[2] * 4)); + if (s2) { + wuffs_base__store_u24le__no_bounds_check(d + (2 * 3), s2); + } + uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[3] * 4)); + if (s3) { + wuffs_base__store_u24le__no_bounds_check(d + (3 * 3), s3); + } + + s += loop_unroll_count * 1; + d += loop_unroll_count * 3; + n -= loop_unroll_count; + } + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + } + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxx__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len3 = dst.len / 3; + size_t len = dst_len3 < src.len ? dst_len3 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint8_t s0 = s[0]; + d[0] = s0; + d[1] = s0; + d[2] = s0; + + s += 1 * 1; + d += 1 * 3; + n -= 1; + } + + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[1] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[2] * 4))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[3] * 4))); + + s += loop_unroll_count * 1; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__load_u32le__no_bounds_check( + dst_palette.ptr + ((size_t)s[0] * 4))); + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (dst_palette.len != 1024) { + return 0; + } + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + const size_t loop_unroll_count = 4; + + while (n >= loop_unroll_count) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); + } + uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[1] * 4)); + if (s1) { + wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1); + } + uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[2] * 4)); + if (s2) { + wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2); + } + uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[3] * 4)); + if (s3) { + wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3); + } + + s += loop_unroll_count * 1; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr + + ((size_t)s[0] * 4)); + if (s0) { + wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0); + } + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len3 = src.len / 3; + size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3))); + + s += 1 * 3; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t len = dst_len4 < src.len ? dst_len4 : src.len; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0])); + + s += 1 * 1; + d += 1 * 4; + n -= 1; + } + + return len; +} + +// -------- + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__pixel_swizzler__bgr_565__y; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + case WUFFS_BASE__PIXEL_FORMAT__RGB: + return wuffs_base__pixel_swizzler__xxx__y; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + return wuffs_base__pixel_swizzler__xxxx__y; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_1_1; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) != + 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, + src_palette) != 1024) { + return NULL; + } + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__xxxx__index__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over; + } + return NULL; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__pixel_swizzler__bgr_565__bgr; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + return wuffs_base__pixel_swizzler__copy_3_3; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + return wuffs_base__pixel_swizzler__xxxx__xxx; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_4_4; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + +// -------- + +wuffs_base__status // +wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__pixel_format src_format, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + if (!p) { + return wuffs_base__make_status(wuffs_base__error__bad_receiver); + } + + // TODO: support many more formats. + + wuffs_base__pixel_swizzler__func func = NULL; + + switch (src_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__Y: + func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, dst_palette, + src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( + p, dst_format, dst_palette, src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + func = wuffs_base__pixel_swizzler__prepare__bgr( + p, dst_format, dst_palette, src_palette, blend); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + p, dst_format, dst_palette, src_palette, blend); + break; + } + + p->private_impl.func = func; + return wuffs_base__make_status( + func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option); +} + +uint64_t // +wuffs_base__pixel_swizzler__swizzle_interleaved( + const wuffs_base__pixel_swizzler* p, + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + if (p && p->private_impl.func) { + return (*p->private_impl.func)(dst, dst_palette, src); + } + return 0; } #endif // !defined(WUFFS_CONFIG__MODULES) ||