From 7c2f36509bbc7e565cdb38416924bd6a08ba0f7a Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Sun, 13 Dec 2020 21:13:07 +1100 Subject: [PATCH] Have std/bmp decode 1, 2 or 4 bits per pixel --- release/c/wuffs-unsupported-snapshot.c | 277 +++++++++++++++++++++++-- std/bmp/decode_bmp.wuffs | 270 ++++++++++++++++++++++-- 2 files changed, 509 insertions(+), 38 deletions(-) diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index e87e854a..ec528c8d 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -16129,6 +16129,8 @@ const char wuffs_bmp__note__internal_note_short_read[] = "@bmp: internal note: s #define WUFFS_BMP__COMPRESSION_ALPHABITFIELDS 6 +#define WUFFS_BMP__COMPRESSION_LOW_BIT_DEPTH 256 + #define WUFFS_BMP__RLE_STATE_NEUTRAL 0 #define WUFFS_BMP__RLE_STATE_RUN 1 @@ -16163,6 +16165,12 @@ wuffs_bmp__decoder__swizzle_bitfields( wuffs_base__pixel_buffer* a_dst, wuffs_base__io_buffer* a_src); +static wuffs_base__status +wuffs_bmp__decoder__swizzle_low_bit_depth( + wuffs_bmp__decoder* self, + wuffs_base__pixel_buffer* a_dst, + wuffs_base__io_buffer* a_src); + static wuffs_base__status wuffs_bmp__decoder__skip_frame( wuffs_bmp__decoder* self, @@ -16327,6 +16335,7 @@ wuffs_bmp__decoder__decode_image_config( uint32_t v_height = 0; uint32_t v_planes = 0; uint32_t v_dst_pixfmt = 0; + uint32_t v_byte_width = 0; const uint8_t* iop_a_src = NULL; const uint8_t* io0_a_src WUFFS_BASE__POTENTIALLY_UNUSED = NULL; @@ -16838,7 +16847,10 @@ wuffs_bmp__decoder__decode_image_config( goto exit; } if (self->private_impl.f_compression == 0) { - if (self->private_impl.f_bits_per_pixel == 8) { + if ((self->private_impl.f_bits_per_pixel == 1) || (self->private_impl.f_bits_per_pixel == 2) || (self->private_impl.f_bits_per_pixel == 4)) { + self->private_impl.f_src_pixfmt = 2198077448; + self->private_impl.f_compression = 256; + } else if (self->private_impl.f_bits_per_pixel == 8) { self->private_impl.f_src_pixfmt = 2198077448; } else if (self->private_impl.f_bits_per_pixel == 16) { self->private_impl.f_compression = 3; @@ -16871,6 +16883,13 @@ wuffs_bmp__decoder__decode_image_config( status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); goto exit; } + } else if (self->private_impl.f_compression == 2) { + if (self->private_impl.f_bits_per_pixel == 4) { + self->private_impl.f_src_pixfmt = 2198077448; + } else { + status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); + goto exit; + } } else if (self->private_impl.f_compression == 3) { if ((self->private_impl.f_bits_per_pixel == 16) || (self->private_impl.f_bits_per_pixel == 32)) { self->private_impl.f_src_pixfmt = 2164308923; @@ -16882,7 +16901,19 @@ wuffs_bmp__decoder__decode_image_config( status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); goto exit; } - if (self->private_impl.f_bits_per_pixel == 8) { + if (self->private_impl.f_bits_per_pixel == 1) { + v_byte_width = ((self->private_impl.f_width >> 3) + (((self->private_impl.f_width & 7) + 7) >> 3)); + self->private_impl.f_bytes_per_row = ((((((uint64_t)(v_byte_width)) * 1) + 3) >> 2) << 2); + self->private_impl.f_pad_per_row = ((4 - (v_byte_width & 3)) & 3); + } else if (self->private_impl.f_bits_per_pixel == 2) { + v_byte_width = ((self->private_impl.f_width >> 2) + (((self->private_impl.f_width & 3) + 3) >> 2)); + self->private_impl.f_bytes_per_row = ((((((uint64_t)(v_byte_width)) * 1) + 3) >> 2) << 2); + self->private_impl.f_pad_per_row = ((4 - (v_byte_width & 3)) & 3); + } else if (self->private_impl.f_bits_per_pixel == 4) { + v_byte_width = ((self->private_impl.f_width >> 1) + (self->private_impl.f_width & 1)); + self->private_impl.f_bytes_per_row = ((((((uint64_t)(v_byte_width)) * 1) + 3) >> 2) << 2); + self->private_impl.f_pad_per_row = ((4 - (v_byte_width & 3)) & 3); + } else if (self->private_impl.f_bits_per_pixel == 8) { self->private_impl.f_bytes_per_row = ((((((uint64_t)(self->private_impl.f_width)) * 1) + 3) >> 2) << 2); self->private_impl.f_pad_per_row = ((4 - (self->private_impl.f_width & 3)) & 3); } else if (self->private_impl.f_bits_per_pixel == 16) { @@ -17170,7 +17201,7 @@ wuffs_bmp__decoder__decode_frame( if (a_src) { iop_a_src = a_src->data.ptr + a_src->meta.ri; } - } else if (self->private_impl.f_compression == 1) { + } else if (self->private_impl.f_compression < 3) { if (a_src) { a_src->meta.ri = ((size_t)(iop_a_src - a_src->data.ptr)); } @@ -17178,7 +17209,7 @@ wuffs_bmp__decoder__decode_frame( if (a_src) { iop_a_src = a_src->data.ptr + a_src->meta.ri; } - } else { + } else if (self->private_impl.f_compression == 3) { if (a_src) { a_src->meta.ri = ((size_t)(iop_a_src - a_src->data.ptr)); } @@ -17186,6 +17217,14 @@ wuffs_bmp__decoder__decode_frame( if (a_src) { iop_a_src = a_src->data.ptr + a_src->meta.ri; } + } else { + if (a_src) { + a_src->meta.ri = ((size_t)(iop_a_src - a_src->data.ptr)); + } + v_status = wuffs_bmp__decoder__swizzle_low_bit_depth(self, a_dst, a_src); + if (a_src) { + iop_a_src = a_src->data.ptr + a_src->meta.ri; + } } if (wuffs_base__status__is_ok(&v_status)) { goto label__0__break; @@ -17349,7 +17388,10 @@ wuffs_bmp__decoder__swizzle_rle( uint64_t v_n = 0; uint32_t v_p0 = 0; uint8_t v_code = 0; + uint8_t v_indexes[2] = {0}; uint32_t v_rle_state = 0; + uint32_t v_chunk_bits = 0; + uint32_t v_chunk_count = 0; const uint8_t* iop_a_src = NULL; const uint8_t* io0_a_src WUFFS_BASE__POTENTIALLY_UNUSED = NULL; @@ -17409,10 +17451,21 @@ wuffs_bmp__decoder__swizzle_rle( } v_code = wuffs_base__load_u8be__no_bounds_check(iop_a_src); (iop_a_src += 1, wuffs_base__make_empty_struct()); - v_p0 = 0; - while (v_p0 < self->private_impl.f_rle_length) { - self->private_data.f_scratch[v_p0] = v_code; - v_p0 += 1; + if (self->private_impl.f_bits_per_pixel == 8) { + v_p0 = 0; + while (v_p0 < self->private_impl.f_rle_length) { + self->private_data.f_scratch[v_p0] = v_code; + v_p0 += 1; + } + } else { + v_indexes[0] = (v_code >> 4); + v_indexes[1] = (v_code & 15); + v_p0 = 0; + while (v_p0 < self->private_impl.f_rle_length) { + self->private_data.f_scratch[(v_p0 + 0)] = v_indexes[0]; + self->private_data.f_scratch[(v_p0 + 1)] = v_indexes[1]; + v_p0 += 2; + } } wuffs_base__pixel_swizzler__swizzle_interleaved_from_slice(&self->private_impl.f_swizzler, v_dst, v_dst_palette, wuffs_base__slice_u8__subslice_j(wuffs_base__make_slice_u8(self->private_data.f_scratch, 2048), self->private_impl.f_rle_length)); wuffs_base__u32__sat_add_indirect(&self->private_impl.f_dst_x, self->private_impl.f_rle_length); @@ -17438,19 +17491,38 @@ wuffs_bmp__decoder__swizzle_rle( goto label__inner__continue; } self->private_impl.f_rle_length = ((uint32_t)(v_code)); - self->private_impl.f_rle_padded = ((v_code & 1) != 0); + self->private_impl.f_rle_padded = ((self->private_impl.f_bits_per_pixel == 8) && ((v_code & 1) != 0)); v_rle_state = 3; goto label__inner__continue; } else if (v_rle_state == 3) { - v_n = wuffs_base__pixel_swizzler__limited_swizzle_u32_interleaved_from_reader( - &self->private_impl.f_swizzler, - self->private_impl.f_rle_length, - v_dst, - v_dst_palette, - &iop_a_src, - io2_a_src); - wuffs_base__u32__sat_add_indirect(&self->private_impl.f_dst_x, ((uint32_t)((v_n & 4294967295)))); - wuffs_base__u32__sat_sub_indirect(&self->private_impl.f_rle_length, ((uint32_t)((v_n & 4294967295)))); + if (self->private_impl.f_bits_per_pixel == 8) { + v_n = wuffs_base__pixel_swizzler__limited_swizzle_u32_interleaved_from_reader( + &self->private_impl.f_swizzler, + self->private_impl.f_rle_length, + v_dst, + v_dst_palette, + &iop_a_src, + io2_a_src); + wuffs_base__u32__sat_add_indirect(&self->private_impl.f_dst_x, ((uint32_t)((v_n & 4294967295)))); + wuffs_base__u32__sat_sub_indirect(&self->private_impl.f_rle_length, ((uint32_t)((v_n & 4294967295)))); + } else { + v_chunk_count = ((self->private_impl.f_rle_length + 3) / 4); + v_p0 = 0; + while ((v_chunk_count > 0) && (((uint64_t)(io2_a_src - iop_a_src)) >= 2)) { + v_chunk_bits = ((uint32_t)(wuffs_base__load_u16be__no_bounds_check(iop_a_src))); + (iop_a_src += 2, wuffs_base__make_empty_struct()); + self->private_data.f_scratch[(v_p0 + 0)] = ((uint8_t)((15 & (v_chunk_bits >> 12)))); + self->private_data.f_scratch[(v_p0 + 1)] = ((uint8_t)((15 & (v_chunk_bits >> 8)))); + self->private_data.f_scratch[(v_p0 + 2)] = ((uint8_t)((15 & (v_chunk_bits >> 4)))); + self->private_data.f_scratch[(v_p0 + 3)] = ((uint8_t)((15 & (v_chunk_bits >> 0)))); + v_p0 = ((v_p0 & 255) + 4); + v_chunk_count -= 1; + } + v_p0 = wuffs_base__u32__min(v_p0, self->private_impl.f_rle_length); + wuffs_base__pixel_swizzler__swizzle_interleaved_from_slice(&self->private_impl.f_swizzler, v_dst, v_dst_palette, wuffs_base__slice_u8__subslice_j(wuffs_base__make_slice_u8(self->private_data.f_scratch, 2048), v_p0)); + wuffs_base__u32__sat_add_indirect(&self->private_impl.f_dst_x, v_p0); + wuffs_base__u32__sat_sub_indirect(&self->private_impl.f_rle_length, v_p0); + } if (self->private_impl.f_rle_length > 0) { goto label__goto_suspend__break; } @@ -17684,6 +17756,175 @@ wuffs_bmp__decoder__swizzle_bitfields( return status; } +// -------- func bmp.decoder.swizzle_low_bit_depth + +static wuffs_base__status +wuffs_bmp__decoder__swizzle_low_bit_depth( + wuffs_bmp__decoder* self, + wuffs_base__pixel_buffer* a_dst, + wuffs_base__io_buffer* a_src) { + wuffs_base__status status = wuffs_base__make_status(NULL); + + wuffs_base__pixel_format v_dst_pixfmt = {0}; + uint32_t v_dst_bits_per_pixel = 0; + uint64_t v_dst_bytes_per_pixel = 0; + uint64_t v_dst_bytes_per_row = 0; + wuffs_base__slice_u8 v_dst_palette = {0}; + wuffs_base__table_u8 v_tab = {0}; + wuffs_base__slice_u8 v_dst = {0}; + uint64_t v_i = 0; + uint64_t v_n = 0; + uint32_t v_p0 = 0; + uint32_t v_chunk_bits = 0; + uint32_t v_chunk_count = 0; + + const uint8_t* iop_a_src = NULL; + const uint8_t* io0_a_src WUFFS_BASE__POTENTIALLY_UNUSED = NULL; + const uint8_t* io1_a_src WUFFS_BASE__POTENTIALLY_UNUSED = NULL; + const uint8_t* io2_a_src WUFFS_BASE__POTENTIALLY_UNUSED = NULL; + if (a_src) { + io0_a_src = a_src->data.ptr; + io1_a_src = io0_a_src + a_src->meta.ri; + iop_a_src = io1_a_src; + io2_a_src = io0_a_src + a_src->meta.wi; + } + + v_dst_pixfmt = wuffs_base__pixel_buffer__pixel_format(a_dst); + v_dst_bits_per_pixel = wuffs_base__pixel_format__bits_per_pixel(&v_dst_pixfmt); + if ((v_dst_bits_per_pixel & 7) != 0) { + status = wuffs_base__make_status(wuffs_base__error__unsupported_option); + goto exit; + } + v_dst_bytes_per_pixel = ((uint64_t)((v_dst_bits_per_pixel / 8))); + v_dst_bytes_per_row = (((uint64_t)(self->private_impl.f_width)) * v_dst_bytes_per_pixel); + v_dst_palette = wuffs_base__pixel_buffer__palette_or_else(a_dst, wuffs_base__make_slice_u8((self->private_data.f_scratch) + 1024, 1024)); + v_tab = wuffs_base__pixel_buffer__plane(a_dst, 0); + label__loop__continue:; + while (true) { + if (self->private_impl.f_dst_x == self->private_impl.f_width) { + self->private_impl.f_dst_x = 0; + self->private_impl.f_dst_y += self->private_impl.f_dst_y_inc; + if (self->private_impl.f_dst_y == self->private_impl.f_dst_y_end) { + goto label__loop__break; + } + } + v_dst = wuffs_base__table_u8__row(v_tab, self->private_impl.f_dst_y); + if (v_dst_bytes_per_row < ((uint64_t)(v_dst.len))) { + v_dst = wuffs_base__slice_u8__subslice_j(v_dst, v_dst_bytes_per_row); + } + v_i = (((uint64_t)(self->private_impl.f_dst_x)) * v_dst_bytes_per_pixel); + if (v_i >= ((uint64_t)(v_dst.len))) { + goto label__loop__continue; + } + v_dst = wuffs_base__slice_u8__subslice_i(v_dst, v_i); + v_p0 = 0; + if (self->private_impl.f_bits_per_pixel == 1) { + v_chunk_count = ((wuffs_base__u32__sat_sub(self->private_impl.f_width, self->private_impl.f_dst_x) + 31) / 32); + v_chunk_count = wuffs_base__u32__min(v_chunk_count, 16); + while ((v_chunk_count > 0) && (((uint64_t)(io2_a_src - iop_a_src)) >= 4)) { + v_chunk_bits = wuffs_base__load_u32be__no_bounds_check(iop_a_src); + (iop_a_src += 4, wuffs_base__make_empty_struct()); + self->private_data.f_scratch[(v_p0 + 0)] = ((uint8_t)((1 & (v_chunk_bits >> 31)))); + self->private_data.f_scratch[(v_p0 + 1)] = ((uint8_t)((1 & (v_chunk_bits >> 30)))); + self->private_data.f_scratch[(v_p0 + 2)] = ((uint8_t)((1 & (v_chunk_bits >> 29)))); + self->private_data.f_scratch[(v_p0 + 3)] = ((uint8_t)((1 & (v_chunk_bits >> 28)))); + self->private_data.f_scratch[(v_p0 + 4)] = ((uint8_t)((1 & (v_chunk_bits >> 27)))); + self->private_data.f_scratch[(v_p0 + 5)] = ((uint8_t)((1 & (v_chunk_bits >> 26)))); + self->private_data.f_scratch[(v_p0 + 6)] = ((uint8_t)((1 & (v_chunk_bits >> 25)))); + self->private_data.f_scratch[(v_p0 + 7)] = ((uint8_t)((1 & (v_chunk_bits >> 24)))); + self->private_data.f_scratch[(v_p0 + 8)] = ((uint8_t)((1 & (v_chunk_bits >> 23)))); + self->private_data.f_scratch[(v_p0 + 9)] = ((uint8_t)((1 & (v_chunk_bits >> 22)))); + self->private_data.f_scratch[(v_p0 + 10)] = ((uint8_t)((1 & (v_chunk_bits >> 21)))); + self->private_data.f_scratch[(v_p0 + 11)] = ((uint8_t)((1 & (v_chunk_bits >> 20)))); + self->private_data.f_scratch[(v_p0 + 12)] = ((uint8_t)((1 & (v_chunk_bits >> 19)))); + self->private_data.f_scratch[(v_p0 + 13)] = ((uint8_t)((1 & (v_chunk_bits >> 18)))); + self->private_data.f_scratch[(v_p0 + 14)] = ((uint8_t)((1 & (v_chunk_bits >> 17)))); + self->private_data.f_scratch[(v_p0 + 15)] = ((uint8_t)((1 & (v_chunk_bits >> 16)))); + self->private_data.f_scratch[(v_p0 + 16)] = ((uint8_t)((1 & (v_chunk_bits >> 15)))); + self->private_data.f_scratch[(v_p0 + 17)] = ((uint8_t)((1 & (v_chunk_bits >> 14)))); + self->private_data.f_scratch[(v_p0 + 18)] = ((uint8_t)((1 & (v_chunk_bits >> 13)))); + self->private_data.f_scratch[(v_p0 + 19)] = ((uint8_t)((1 & (v_chunk_bits >> 12)))); + self->private_data.f_scratch[(v_p0 + 20)] = ((uint8_t)((1 & (v_chunk_bits >> 11)))); + self->private_data.f_scratch[(v_p0 + 21)] = ((uint8_t)((1 & (v_chunk_bits >> 10)))); + self->private_data.f_scratch[(v_p0 + 22)] = ((uint8_t)((1 & (v_chunk_bits >> 9)))); + self->private_data.f_scratch[(v_p0 + 23)] = ((uint8_t)((1 & (v_chunk_bits >> 8)))); + self->private_data.f_scratch[(v_p0 + 24)] = ((uint8_t)((1 & (v_chunk_bits >> 7)))); + self->private_data.f_scratch[(v_p0 + 25)] = ((uint8_t)((1 & (v_chunk_bits >> 6)))); + self->private_data.f_scratch[(v_p0 + 26)] = ((uint8_t)((1 & (v_chunk_bits >> 5)))); + self->private_data.f_scratch[(v_p0 + 27)] = ((uint8_t)((1 & (v_chunk_bits >> 4)))); + self->private_data.f_scratch[(v_p0 + 28)] = ((uint8_t)((1 & (v_chunk_bits >> 3)))); + self->private_data.f_scratch[(v_p0 + 29)] = ((uint8_t)((1 & (v_chunk_bits >> 2)))); + self->private_data.f_scratch[(v_p0 + 30)] = ((uint8_t)((1 & (v_chunk_bits >> 1)))); + self->private_data.f_scratch[(v_p0 + 31)] = ((uint8_t)((1 & (v_chunk_bits >> 0)))); + v_p0 = ((v_p0 & 511) + 32); + v_chunk_count -= 1; + } + } else if (self->private_impl.f_bits_per_pixel == 2) { + v_chunk_count = ((wuffs_base__u32__sat_sub(self->private_impl.f_width, self->private_impl.f_dst_x) + 15) / 16); + v_chunk_count = wuffs_base__u32__min(v_chunk_count, 32); + while ((v_chunk_count > 0) && (((uint64_t)(io2_a_src - iop_a_src)) >= 4)) { + v_chunk_bits = wuffs_base__load_u32be__no_bounds_check(iop_a_src); + (iop_a_src += 4, wuffs_base__make_empty_struct()); + self->private_data.f_scratch[(v_p0 + 0)] = ((uint8_t)((3 & (v_chunk_bits >> 30)))); + self->private_data.f_scratch[(v_p0 + 1)] = ((uint8_t)((3 & (v_chunk_bits >> 28)))); + self->private_data.f_scratch[(v_p0 + 2)] = ((uint8_t)((3 & (v_chunk_bits >> 26)))); + self->private_data.f_scratch[(v_p0 + 3)] = ((uint8_t)((3 & (v_chunk_bits >> 24)))); + self->private_data.f_scratch[(v_p0 + 4)] = ((uint8_t)((3 & (v_chunk_bits >> 22)))); + self->private_data.f_scratch[(v_p0 + 5)] = ((uint8_t)((3 & (v_chunk_bits >> 20)))); + self->private_data.f_scratch[(v_p0 + 6)] = ((uint8_t)((3 & (v_chunk_bits >> 18)))); + self->private_data.f_scratch[(v_p0 + 7)] = ((uint8_t)((3 & (v_chunk_bits >> 16)))); + self->private_data.f_scratch[(v_p0 + 8)] = ((uint8_t)((3 & (v_chunk_bits >> 14)))); + self->private_data.f_scratch[(v_p0 + 9)] = ((uint8_t)((3 & (v_chunk_bits >> 12)))); + self->private_data.f_scratch[(v_p0 + 10)] = ((uint8_t)((3 & (v_chunk_bits >> 10)))); + self->private_data.f_scratch[(v_p0 + 11)] = ((uint8_t)((3 & (v_chunk_bits >> 8)))); + self->private_data.f_scratch[(v_p0 + 12)] = ((uint8_t)((3 & (v_chunk_bits >> 6)))); + self->private_data.f_scratch[(v_p0 + 13)] = ((uint8_t)((3 & (v_chunk_bits >> 4)))); + self->private_data.f_scratch[(v_p0 + 14)] = ((uint8_t)((3 & (v_chunk_bits >> 2)))); + self->private_data.f_scratch[(v_p0 + 15)] = ((uint8_t)((3 & (v_chunk_bits >> 0)))); + v_p0 = ((v_p0 & 511) + 16); + v_chunk_count -= 1; + } + } else if (self->private_impl.f_bits_per_pixel == 4) { + v_chunk_count = ((wuffs_base__u32__sat_sub(self->private_impl.f_width, self->private_impl.f_dst_x) + 7) / 8); + v_chunk_count = wuffs_base__u32__min(v_chunk_count, 64); + while ((v_chunk_count > 0) && (((uint64_t)(io2_a_src - iop_a_src)) >= 4)) { + v_chunk_bits = wuffs_base__load_u32be__no_bounds_check(iop_a_src); + (iop_a_src += 4, wuffs_base__make_empty_struct()); + self->private_data.f_scratch[(v_p0 + 0)] = ((uint8_t)((15 & (v_chunk_bits >> 28)))); + self->private_data.f_scratch[(v_p0 + 1)] = ((uint8_t)((15 & (v_chunk_bits >> 24)))); + self->private_data.f_scratch[(v_p0 + 2)] = ((uint8_t)((15 & (v_chunk_bits >> 20)))); + self->private_data.f_scratch[(v_p0 + 3)] = ((uint8_t)((15 & (v_chunk_bits >> 16)))); + self->private_data.f_scratch[(v_p0 + 4)] = ((uint8_t)((15 & (v_chunk_bits >> 12)))); + self->private_data.f_scratch[(v_p0 + 5)] = ((uint8_t)((15 & (v_chunk_bits >> 8)))); + self->private_data.f_scratch[(v_p0 + 6)] = ((uint8_t)((15 & (v_chunk_bits >> 4)))); + self->private_data.f_scratch[(v_p0 + 7)] = ((uint8_t)((15 & (v_chunk_bits >> 0)))); + v_p0 = ((v_p0 & 511) + 8); + v_chunk_count -= 1; + } + } + v_p0 = wuffs_base__u32__min(v_p0, wuffs_base__u32__sat_sub(self->private_impl.f_width, self->private_impl.f_dst_x)); + v_n = wuffs_base__pixel_swizzler__swizzle_interleaved_from_slice(&self->private_impl.f_swizzler, v_dst, v_dst_palette, wuffs_base__slice_u8__subslice_j(wuffs_base__make_slice_u8(self->private_data.f_scratch, 2048), v_p0)); + if (v_n == 0) { + status = wuffs_base__make_status(wuffs_bmp__note__internal_note_short_read); + goto ok; + } + wuffs_base__u32__sat_add_indirect(&self->private_impl.f_dst_x, ((uint32_t)((v_n & 4294967295)))); + } + label__loop__break:; + status = wuffs_base__make_status(NULL); + goto ok; + + goto ok; + ok: + goto exit; + exit: + if (a_src) { + a_src->meta.ri = ((size_t)(iop_a_src - a_src->data.ptr)); + } + + return status; +} + // -------- func bmp.decoder.skip_frame static wuffs_base__status diff --git a/std/bmp/decode_bmp.wuffs b/std/bmp/decode_bmp.wuffs index 942fde5e..de2b8168 100644 --- a/std/bmp/decode_bmp.wuffs +++ b/std/bmp/decode_bmp.wuffs @@ -26,6 +26,7 @@ pri const COMPRESSION_BITFIELDS : base.u32 = 3 pri const COMPRESSION_JPEG : base.u32 = 4 pri const COMPRESSION_PNG : base.u32 = 5 pri const COMPRESSION_ALPHABITFIELDS : base.u32 = 6 +pri const COMPRESSION_LOW_BIT_DEPTH : base.u32 = 0x100 pub struct decoder? implements base.image_decoder( width : base.u32[..= 0x7FFF_FFFF], @@ -88,6 +89,7 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ var height : base.u32 var planes : base.u32 var dst_pixfmt : base.u32 + var byte_width : base.u32 if (this.call_sequence <> 0) or (this.io_redirect_fourcc == 1) { return base."#bad call sequence" @@ -221,7 +223,12 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ } if this.compression == COMPRESSION_NONE { - if this.bits_per_pixel == 8 { + if (this.bits_per_pixel == 1) or + (this.bits_per_pixel == 2) or + (this.bits_per_pixel == 4) { + this.src_pixfmt = base.PIXEL_FORMAT__INDEXED__BGRA_BINARY + this.compression = COMPRESSION_LOW_BIT_DEPTH + } else if this.bits_per_pixel == 8 { this.src_pixfmt = base.PIXEL_FORMAT__INDEXED__BGRA_BINARY } else if this.bits_per_pixel == 16 { // BMP's 16-bit default is BGRX_5551. @@ -251,6 +258,13 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ return "#unsupported BMP file" } + } else if this.compression == COMPRESSION_RLE4 { + if this.bits_per_pixel == 4 { + this.src_pixfmt = base.PIXEL_FORMAT__INDEXED__BGRA_BINARY + } else { + return "#unsupported BMP file" + } + } else if this.compression == COMPRESSION_BITFIELDS { if (this.bits_per_pixel == 16) or (this.bits_per_pixel == 32) { this.src_pixfmt = base.PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE @@ -263,7 +277,22 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ } // The "((x + 3) >> 2) << 2" dance rounds x up to a multiple of 4. - if this.bits_per_pixel == 8 { + if this.bits_per_pixel == 1 { + // byte_width is this.width divided by 8, rounding up. + byte_width = (this.width >> 3) + (((this.width & 7) + 7) >> 3) + this.bytes_per_row = ((((byte_width as base.u64) * 1) + 3) >> 2) << 2 + this.pad_per_row = (4 - (byte_width & 3)) & 3 + } else if this.bits_per_pixel == 2 { + // byte_width is this.width divided by 4, rounding up. + byte_width = (this.width >> 2) + (((this.width & 3) + 3) >> 2) + this.bytes_per_row = ((((byte_width as base.u64) * 1) + 3) >> 2) << 2 + this.pad_per_row = (4 - (byte_width & 3)) & 3 + } else if this.bits_per_pixel == 4 { + // byte_width is this.width divided by 2, rounding up. + byte_width = (this.width >> 1) + (this.width & 1) + this.bytes_per_row = ((((byte_width as base.u64) * 1) + 3) >> 2) << 2 + this.pad_per_row = (4 - (byte_width & 3)) & 3 + } else if this.bits_per_pixel == 8 { this.bytes_per_row = ((((this.width as base.u64) * 1) + 3) >> 2) << 2 this.pad_per_row = (4 - (this.width & 3)) & 3 } else if this.bits_per_pixel == 16 { @@ -370,10 +399,12 @@ pub func decoder.decode_frame?(dst: ptr base.pixel_buffer, src: base.io_reader, while true { if this.compression == COMPRESSION_NONE { status = this.swizzle_none!(dst: args.dst, src: args.src) - } else if this.compression == COMPRESSION_RLE8 { + } else if this.compression < COMPRESSION_BITFIELDS { status = this.swizzle_rle!(dst: args.dst, src: args.src) - } else { + } else if this.compression == COMPRESSION_BITFIELDS { status = this.swizzle_bitfields!(dst: args.dst, src: args.src) + } else { + status = this.swizzle_low_bit_depth!(dst: args.dst, src: args.src) } if status.is_ok() { @@ -475,11 +506,15 @@ pri func decoder.swizzle_rle!(dst: ptr base.pixel_buffer, src: base.io_reader) b var i : base.u64 var n : base.u64 - var p0 : base.u32[..= 255] - var code : base.u8 + var p0 : base.u32[..= 259] + var code : base.u8 + var indexes : array[2] base.u8 var rle_state : base.u32 + var chunk_bits : base.u32 + var chunk_count : base.u32[..= 64] + // TODO: the dst_pixfmt variable shouldn't be necessary. We should be able // to chain the two calls: "args.dst.pixel_format().bits_per_pixel()". dst_pixfmt = args.dst.pixel_format() @@ -530,12 +565,24 @@ pri func decoder.swizzle_rle!(dst: ptr base.pixel_buffer, src: base.io_reader) b } code = args.src.peek_u8() args.src.skip_u32_fast!(actual: 1, worst_case: 1) - p0 = 0 - while p0 < this.rle_length { - assert p0 < 255 via "a < b: a < c; c <= b"(c: this.rle_length) - this.scratch[p0] = code - p0 += 1 - } endwhile + if this.bits_per_pixel == 8 { + p0 = 0 + while p0 < this.rle_length { + assert p0 < 255 via "a < b: a < c; c <= b"(c: this.rle_length) + this.scratch[p0] = code + p0 += 1 + } endwhile + } else { + indexes[0] = code >> 4 + indexes[1] = code & 0x0F + p0 = 0 + while p0 < this.rle_length { + assert p0 < 255 via "a < b: a < c; c <= b"(c: this.rle_length) + this.scratch[p0 + 0] = indexes[0] + this.scratch[p0 + 1] = indexes[1] + p0 += 2 + } endwhile + } this.swizzler.swizzle_interleaved_from_slice!( dst: dst, dst_palette: dst_palette, @@ -567,18 +614,44 @@ pri func decoder.swizzle_rle!(dst: ptr base.pixel_buffer, src: base.io_reader) b continue.inner } this.rle_length = code as base.u32 - this.rle_padded = (code & 1) <> 0 + this.rle_padded = (this.bits_per_pixel == 8) and ((code & 1) <> 0) rle_state = RLE_STATE_LITERAL continue.inner } else if rle_state == RLE_STATE_LITERAL { - n = this.swizzler.limited_swizzle_u32_interleaved_from_reader!( - up_to_num_pixels: this.rle_length, - dst: dst, - dst_palette: dst_palette, - src: args.src) - this.dst_x ~sat+= (n & 0xFFFF_FFFF) as base.u32 - this.rle_length ~sat-= (n & 0xFFFF_FFFF) as base.u32 + if this.bits_per_pixel == 8 { + n = this.swizzler.limited_swizzle_u32_interleaved_from_reader!( + up_to_num_pixels: this.rle_length, + dst: dst, + dst_palette: dst_palette, + src: args.src) + this.dst_x ~sat+= (n & 0xFFFF_FFFF) as base.u32 + this.rle_length ~sat-= (n & 0xFFFF_FFFF) as base.u32 + } else { + // Calculate the remaining number of 16-bit chunks. At + // 4 bits per pixel there are 4 pixels per chunk. + // Division rounds up. + chunk_count = (this.rle_length + 3) / 4 + p0 = 0 + while (chunk_count > 0) and (args.src.length() >= 2) { + chunk_bits = args.src.peek_u16be_as_u32() + args.src.skip_u32_fast!(actual: 2, worst_case: 2) + this.scratch[p0 + 0x00] = (0x0F & (chunk_bits >> 0x0C)) as base.u8 + this.scratch[p0 + 0x01] = (0x0F & (chunk_bits >> 0x08)) as base.u8 + this.scratch[p0 + 0x02] = (0x0F & (chunk_bits >> 0x04)) as base.u8 + this.scratch[p0 + 0x03] = (0x0F & (chunk_bits >> 0x00)) as base.u8 + p0 = (p0 & 255) + 0x04 + chunk_count -= 1 + } endwhile + p0 = p0.min(a: this.rle_length) + this.swizzler.swizzle_interleaved_from_slice!( + dst: dst, + dst_palette: dst_palette, + src: this.scratch[.. p0]) + this.dst_x ~sat+= p0 + this.rle_length ~sat-= p0 + } + if this.rle_length > 0 { break.goto_suspend } @@ -794,6 +867,163 @@ pri func decoder.swizzle_bitfields!(dst: ptr base.pixel_buffer, src: base.io_rea return ok } +pri func decoder.swizzle_low_bit_depth!(dst: ptr base.pixel_buffer, src: base.io_reader) base.status { + var dst_pixfmt : base.pixel_format + var dst_bits_per_pixel : base.u32[..= 256] + var dst_bytes_per_pixel : base.u64[..= 32] + var dst_bytes_per_row : base.u64 + var dst_palette : slice base.u8 + var tab : table base.u8 + var dst : slice base.u8 + var i : base.u64 + var n : base.u64 + + var p0 : base.u32[..= 543] + + var chunk_bits : base.u32 + var chunk_count : base.u32 + + // TODO: the dst_pixfmt variable shouldn't be necessary. We should be able + // to chain the two calls: "args.dst.pixel_format().bits_per_pixel()". + dst_pixfmt = args.dst.pixel_format() + dst_bits_per_pixel = dst_pixfmt.bits_per_pixel() + if (dst_bits_per_pixel & 7) <> 0 { + return base."#unsupported option" + } + dst_bytes_per_pixel = (dst_bits_per_pixel / 8) as base.u64 + dst_bytes_per_row = (this.width as base.u64) * dst_bytes_per_pixel + dst_palette = args.dst.palette_or_else(fallback: this.scratch[1024 ..]) + tab = args.dst.plane(p: 0) + + while.loop true { + if this.dst_x == this.width { + this.dst_x = 0 + this.dst_y ~mod+= this.dst_y_inc + if this.dst_y == this.dst_y_end { + break.loop + } + } + + dst = tab.row(y: this.dst_y) + if dst_bytes_per_row < dst.length() { + dst = dst[.. dst_bytes_per_row] + } + i = (this.dst_x as base.u64) * dst_bytes_per_pixel + if i >= dst.length() { + // TODO: advance args.src if the dst pixel_buffer bounds is + // smaller than this BMP's image bounds? + continue.loop + } + dst = dst[i ..] + p0 = 0 + + if this.bits_per_pixel == 1 { + // Calculate the remaining number of 32-bit chunks. At 1 bit per + // pixel there are 32 pixels per chunk. Division rounds up. + chunk_count = ((this.width ~sat- this.dst_x) + 31) / 32 + chunk_count = chunk_count.min(a: 16) // Keep p0 <= 512. + while (chunk_count > 0) and (args.src.length() >= 4) { + chunk_bits = args.src.peek_u32be() + args.src.skip_u32_fast!(actual: 4, worst_case: 4) + this.scratch[p0 + 0x00] = (0x01 & (chunk_bits >> 0x1F)) as base.u8 + this.scratch[p0 + 0x01] = (0x01 & (chunk_bits >> 0x1E)) as base.u8 + this.scratch[p0 + 0x02] = (0x01 & (chunk_bits >> 0x1D)) as base.u8 + this.scratch[p0 + 0x03] = (0x01 & (chunk_bits >> 0x1C)) as base.u8 + this.scratch[p0 + 0x04] = (0x01 & (chunk_bits >> 0x1B)) as base.u8 + this.scratch[p0 + 0x05] = (0x01 & (chunk_bits >> 0x1A)) as base.u8 + this.scratch[p0 + 0x06] = (0x01 & (chunk_bits >> 0x19)) as base.u8 + this.scratch[p0 + 0x07] = (0x01 & (chunk_bits >> 0x18)) as base.u8 + this.scratch[p0 + 0x08] = (0x01 & (chunk_bits >> 0x17)) as base.u8 + this.scratch[p0 + 0x09] = (0x01 & (chunk_bits >> 0x16)) as base.u8 + this.scratch[p0 + 0x0A] = (0x01 & (chunk_bits >> 0x15)) as base.u8 + this.scratch[p0 + 0x0B] = (0x01 & (chunk_bits >> 0x14)) as base.u8 + this.scratch[p0 + 0x0C] = (0x01 & (chunk_bits >> 0x13)) as base.u8 + this.scratch[p0 + 0x0D] = (0x01 & (chunk_bits >> 0x12)) as base.u8 + this.scratch[p0 + 0x0E] = (0x01 & (chunk_bits >> 0x11)) as base.u8 + this.scratch[p0 + 0x0F] = (0x01 & (chunk_bits >> 0x10)) as base.u8 + this.scratch[p0 + 0x10] = (0x01 & (chunk_bits >> 0x0F)) as base.u8 + this.scratch[p0 + 0x11] = (0x01 & (chunk_bits >> 0x0E)) as base.u8 + this.scratch[p0 + 0x12] = (0x01 & (chunk_bits >> 0x0D)) as base.u8 + this.scratch[p0 + 0x13] = (0x01 & (chunk_bits >> 0x0C)) as base.u8 + this.scratch[p0 + 0x14] = (0x01 & (chunk_bits >> 0x0B)) as base.u8 + this.scratch[p0 + 0x15] = (0x01 & (chunk_bits >> 0x0A)) as base.u8 + this.scratch[p0 + 0x16] = (0x01 & (chunk_bits >> 0x09)) as base.u8 + this.scratch[p0 + 0x17] = (0x01 & (chunk_bits >> 0x08)) as base.u8 + this.scratch[p0 + 0x18] = (0x01 & (chunk_bits >> 0x07)) as base.u8 + this.scratch[p0 + 0x19] = (0x01 & (chunk_bits >> 0x06)) as base.u8 + this.scratch[p0 + 0x1A] = (0x01 & (chunk_bits >> 0x05)) as base.u8 + this.scratch[p0 + 0x1B] = (0x01 & (chunk_bits >> 0x04)) as base.u8 + this.scratch[p0 + 0x1C] = (0x01 & (chunk_bits >> 0x03)) as base.u8 + this.scratch[p0 + 0x1D] = (0x01 & (chunk_bits >> 0x02)) as base.u8 + this.scratch[p0 + 0x1E] = (0x01 & (chunk_bits >> 0x01)) as base.u8 + this.scratch[p0 + 0x1F] = (0x01 & (chunk_bits >> 0x00)) as base.u8 + p0 = (p0 & 511) + 0x20 + chunk_count -= 1 + } endwhile + + } else if this.bits_per_pixel == 2 { + // Calculate the remaining number of 32-bit chunks. At 2 bits per + // pixel there are 16 pixels per chunk. Division rounds up. + chunk_count = ((this.width ~sat- this.dst_x) + 15) / 16 + chunk_count = chunk_count.min(a: 32) // Keep p0 <= 512. + while (chunk_count > 0) and (args.src.length() >= 4) { + chunk_bits = args.src.peek_u32be() + args.src.skip_u32_fast!(actual: 4, worst_case: 4) + this.scratch[p0 + 0x00] = (0x03 & (chunk_bits >> 0x1E)) as base.u8 + this.scratch[p0 + 0x01] = (0x03 & (chunk_bits >> 0x1C)) as base.u8 + this.scratch[p0 + 0x02] = (0x03 & (chunk_bits >> 0x1A)) as base.u8 + this.scratch[p0 + 0x03] = (0x03 & (chunk_bits >> 0x18)) as base.u8 + this.scratch[p0 + 0x04] = (0x03 & (chunk_bits >> 0x16)) as base.u8 + this.scratch[p0 + 0x05] = (0x03 & (chunk_bits >> 0x14)) as base.u8 + this.scratch[p0 + 0x06] = (0x03 & (chunk_bits >> 0x12)) as base.u8 + this.scratch[p0 + 0x07] = (0x03 & (chunk_bits >> 0x10)) as base.u8 + this.scratch[p0 + 0x08] = (0x03 & (chunk_bits >> 0x0E)) as base.u8 + this.scratch[p0 + 0x09] = (0x03 & (chunk_bits >> 0x0C)) as base.u8 + this.scratch[p0 + 0x0A] = (0x03 & (chunk_bits >> 0x0A)) as base.u8 + this.scratch[p0 + 0x0B] = (0x03 & (chunk_bits >> 0x08)) as base.u8 + this.scratch[p0 + 0x0C] = (0x03 & (chunk_bits >> 0x06)) as base.u8 + this.scratch[p0 + 0x0D] = (0x03 & (chunk_bits >> 0x04)) as base.u8 + this.scratch[p0 + 0x0E] = (0x03 & (chunk_bits >> 0x02)) as base.u8 + this.scratch[p0 + 0x0F] = (0x03 & (chunk_bits >> 0x00)) as base.u8 + p0 = (p0 & 511) + 0x10 + chunk_count -= 1 + } endwhile + + } else if this.bits_per_pixel == 4 { + // Calculate the remaining number of 32-bit chunks. At 4 bits per + // pixel there are 8 pixels per chunk. Division rounds up. + chunk_count = ((this.width ~sat- this.dst_x) + 7) / 8 + chunk_count = chunk_count.min(a: 64) // Keep p0 <= 512. + while (chunk_count > 0) and (args.src.length() >= 4) { + chunk_bits = args.src.peek_u32be() + args.src.skip_u32_fast!(actual: 4, worst_case: 4) + this.scratch[p0 + 0x00] = (0x0F & (chunk_bits >> 0x1C)) as base.u8 + this.scratch[p0 + 0x01] = (0x0F & (chunk_bits >> 0x18)) as base.u8 + this.scratch[p0 + 0x02] = (0x0F & (chunk_bits >> 0x14)) as base.u8 + this.scratch[p0 + 0x03] = (0x0F & (chunk_bits >> 0x10)) as base.u8 + this.scratch[p0 + 0x04] = (0x0F & (chunk_bits >> 0x0C)) as base.u8 + this.scratch[p0 + 0x05] = (0x0F & (chunk_bits >> 0x08)) as base.u8 + this.scratch[p0 + 0x06] = (0x0F & (chunk_bits >> 0x04)) as base.u8 + this.scratch[p0 + 0x07] = (0x0F & (chunk_bits >> 0x00)) as base.u8 + p0 = (p0 & 511) + 0x08 + chunk_count -= 1 + } endwhile + } + + p0 = p0.min(a: this.width ~sat- this.dst_x) + n = this.swizzler.swizzle_interleaved_from_slice!( + dst: dst, + dst_palette: dst_palette, + src: this.scratch[.. p0]) + if n == 0 { + return "@internal note: short read" + } + this.dst_x ~sat+= (n & 0xFFFF_FFFF) as base.u32 + } endwhile.loop + + return ok +} + pri func decoder.skip_frame?(src: base.io_reader) { args.src.skip_u32?(n: this.padding) args.src.skip?(n: this.bytes_per_row * (this.height as base.u64))