Rename PUFFS to WUFFS, part 11 of N

This commit is contained in:
Nigel Tao committed 2017-11-30 18:18:40 +11:00
1 parent b9fd20bfe4
commit 024dc7f97a
6 files changed
+60 -57

No files matched your search

+1 -1
View File
@@ -90,7 +90,7 @@ func do(w *bufio.Writer, dir string, filename string, depth int) error {
continue
} else if os.IsNotExist(err) {
// This is probably an #include of a system header, like
// `#include "zlib.h"`, and not of Puffs code. Fall through
// `#include "zlib.h"`, and not of Wuffs code. Fall through
// and print the #include line as normal.
} else {
return err
+51 -48
View File
@@ -13,15 +13,15 @@
// limitations under the License.
// This file contains a hand-written C implementation of gen/c/std/flate.c's
// generated puffs_flate__flate_decoder__decode_huffman_fast function.
// generated wuffs_flate__flate_decoder__decode_huffman_fast function.
//
// It is not intended to be used in production settings, on untrusted data. Its
// purpose is to give a rough upper bound on how fast Puffs' generated C code
// can be, with a sufficiently smart Puffs compiler.
// purpose is to give a rough upper bound on how fast Wuffs' generated C code
// can be, with a sufficiently smart Wuffs compiler.
//
// To repeat, substituting in this C implementation is NOT SAFE, and may result
// in buffer overflows. This code exists only to aid optimization of the (safe)
// std/flate/*.puffs code and the Puffs compiler's code generation.
// std/flate/*.wuffs code and the Wuffs compiler's code generation.
//
// ----------------
//
@@ -32,16 +32,16 @@
//
// After that, add this line:
//
// #include "../../../script/puffs-flate-decoder-decode-huffman.c"
// #include "../../../script/wuffs-flate-decoder-decode-huffman.c"
//
// Then find the call to puffs_flate__flate_decoder__decode_huffman_fast. It
// should be inside the puffs_flate__flate_decoder__decode_blocks function body,
// Then find the call to wuffs_flate__flate_decoder__decode_huffman_fast. It
// should be inside the wuffs_flate__flate_decoder__decode_blocks function body,
// and the lines of code should look something like
//
// status =
// puffs_flate__flate_decoder__decode_huffman_fast(self, a_dst, a_src);
// wuffs_flate__flate_decoder__decode_huffman_fast(self, a_dst, a_src);
//
// Change the "puffs" to "c_puffs", i.e. add a "c_" prefix. The net result
// Change the "wuffs" to "c_wuffs", i.e. add a "c_" prefix. The net result
// should look something like:
//
// ----------------
@@ -51,20 +51,23 @@
// index 47cb0a9..e7e12b6 100644
// --- a/gen/c/std/flate.c
// +++ b/gen/c/std/flate.c
// @@ -336,6 +336,7 @@ puffs_flate__status puffs_flate__zlib_decoder__decode(
// #endif // PUFFS_FLATE_H
// @@ -336,6 +336,7 @@ wuffs_flate__status wuffs_flate__zlib_decoder__decode(
// #endif // WUFFS_FLATE_H
//
// // C HEADER ENDS HERE.
// +#include "../../../script/puffs-flate-decoder-decode-huffman.c"
// +#include "../../../script/wuffs-flate-decoder-decode-huffman.c"
//
// #ifndef PUFFS_BASE_IMPL_H
// #define PUFFS_BASE_IMPL_H
// @@ -1119,7 +1120,7 @@ static puffs_flate__status puffs_flate__flate_decoder__decode_blocks(
// #ifndef WUFFS_BASE_IMPL_H
// #define WUFFS_BASE_IMPL_H
// @@ -1119,7 +1120,7 @@ static wuffs_flate__status
// wuffs_flate__flate_decoder__decode_blocks(
// }
// }
// status =
// - puffs_flate__flate_decoder__decode_huffman_fast(self, a_dst, a_src);
// + c_puffs_flate__flate_decoder__decode_huffman_fast(self, a_dst, a_src);
// - wuffs_flate__flate_decoder__decode_huffman_fast(self, a_dst,
// a_src);
// + c_wuffs_flate__flate_decoder__decode_huffman_fast(self, a_dst,
// a_src);
// if (a_src.buf) {
// b_rptr_src = a_src.buf->ptr + a_src.buf->ri;
// }
@@ -72,16 +75,16 @@
// ----------------
//
// That concludes the two edits to gen/c/std/flate.c. Run the tests and
// benchmarks with the "-skipgen" flag, otherwise the "puffs" tool will
// benchmarks with the "-skipgen" flag, otherwise the "wuffs" tool will
// re-generate the C code and override your gen/c/std/flate.c edit:
//
// puffs test -skipgen std/flate
// puffs bench -skipgen std/flate
// wuffs test -skipgen std/flate
// wuffs bench -skipgen std/flate
//
// You may also want to focus on one specific test, e.g.:
//
// puffs test -skipgen -focus=puffs_flate_decode_midsummer std/flate
// puffs bench -skipgen -focus=puffs_flate_decode_100k std/flate
// wuffs test -skipgen -focus=wuffs_flate_decode_midsummer std/flate
// wuffs bench -skipgen -focus=wuffs_flate_decode_100k std/flate
#include <stddef.h>
#include <stdio.h> // For manual printf debugging.
@@ -89,53 +92,53 @@
// Define this macro to further speed up this C implementation, using two
// techniques that are not used by the zlib-the-library C implementation (as of
// version 1.2.11, current as of November 2017). Doing so gives data for "how
// fast can Puffs' zlib-the-format implementation be" instead of "is Puffs'
// fast can Wuffs' zlib-the-format implementation be" instead of "is Wuffs'
// generated C code as fast as zlib-the-library's hand-written C code".
//
// Whether the same techniques could apply to zlib-the-library is discussed at
// https://github.com/madler/zlib/pull/292
#ifdef __x86_64__
//#define PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
//#define WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#endif
// This is the generated function that we are explicitly overriding. Note that
// the function name is "puffs_etc", not "c_puffs_etc".
static puffs_flate__status puffs_flate__flate_decoder__decode_huffman_fast(
puffs_flate__flate_decoder* self,
puffs_base__writer1 a_dst,
puffs_base__reader1 a_src);
// the function name is "wuffs_etc", not "c_wuffs_etc".
static wuffs_flate__status wuffs_flate__flate_decoder__decode_huffman_fast(
wuffs_flate__flate_decoder* self,
wuffs_base__writer1 a_dst,
wuffs_base__reader1 a_src);
// This is the overriding implementation.
puffs_flate__status c_puffs_flate__flate_decoder__decode_huffman_fast(
puffs_flate__flate_decoder* self,
puffs_base__writer1 a_dst,
puffs_base__reader1 a_src) {
wuffs_flate__status c_wuffs_flate__flate_decoder__decode_huffman_fast(
wuffs_flate__flate_decoder* self,
wuffs_base__writer1 a_dst,
wuffs_base__reader1 a_src) {
// Avoid the -Werror=unused-function warning for the now-unused
// overridden puffs_flate__flate_decoder__decode_huffman_fast.
(void)(puffs_flate__flate_decoder__decode_huffman_fast);
// overridden wuffs_flate__flate_decoder__decode_huffman_fast.
(void)(wuffs_flate__flate_decoder__decode_huffman_fast);
if (!a_dst.buf || !a_src.buf) {
return PUFFS_FLATE__ERROR_BAD_ARGUMENT;
return WUFFS_FLATE__ERROR_BAD_ARGUMENT;
}
puffs_flate__status status = PUFFS_FLATE__STATUS_OK;
wuffs_flate__status status = WUFFS_FLATE__STATUS_OK;
// Load contextual state. Prepare to check that pdst and psrc remain within
// a_dst's and a_src's bounds.
uint8_t* pdst = a_dst.buf->ptr + a_dst.buf->wi;
uint8_t* qdst = a_dst.buf->ptr + a_dst.buf->len;
if ((qdst - pdst) < 258) {
return PUFFS_FLATE__STATUS_OK;
return WUFFS_FLATE__STATUS_OK;
} else {
qdst -= 258;
}
uint8_t* psrc = a_src.buf->ptr + a_src.buf->ri;
uint8_t* qsrc = a_src.buf->ptr + a_src.buf->wi;
if ((qsrc - psrc) < 12) {
return PUFFS_FLATE__STATUS_OK;
return WUFFS_FLATE__STATUS_OK;
} else {
qsrc -= 12;
}
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
uint64_t bits = self->private_impl.f_bits;
#else
uint32_t bits = self->private_impl.f_bits;
@@ -153,7 +156,7 @@ outer_loop:
while ((pdst <= qdst) && (psrc <= qsrc)) {
// Ensure that we have at least 15 bits of input.
if (n_bits < 15) {
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
bits |= *((uint64_t*)psrc) << n_bits;
psrc += 6;
n_bits += 48;
@@ -187,7 +190,7 @@ outer_loop:
}
if ((table_entry >> 24) != 0x10) {
status =
PUFFS_FLATE__ERROR_INTERNAL_ERROR_INCONSISTENT_HUFFMAN_DECODER_STATE;
WUFFS_FLATE__ERROR_INTERNAL_ERROR_INCONSISTENT_HUFFMAN_DECODER_STATE;
goto end;
}
uint32_t top = (table_entry >> 8) & 0xFFFF;
@@ -201,7 +204,7 @@ outer_loop:
uint32_t n = (table_entry >> 4) & 0x0F;
if (n) {
if (n_bits < n) {
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
bits |= *((uint64_t*)psrc) << n_bits;
psrc += 6;
n_bits += 48;
@@ -218,7 +221,7 @@ outer_loop:
// Ensure that we have at least 15 bits of input.
if (n_bits < 15) {
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
bits |= *((uint64_t*)psrc) << n_bits;
psrc += 6;
n_bits += 48;
@@ -242,7 +245,7 @@ outer_loop:
}
if ((table_entry >> 24) != 0x10) {
status =
PUFFS_FLATE__ERROR_INTERNAL_ERROR_INCONSISTENT_HUFFMAN_DECODER_STATE;
WUFFS_FLATE__ERROR_INTERNAL_ERROR_INCONSISTENT_HUFFMAN_DECODER_STATE;
goto end;
}
uint32_t top = (table_entry >> 8) & 0xFFFF;
@@ -256,7 +259,7 @@ outer_loop:
uint32_t n = (table_entry >> 4) & 0x0F;
if (n) {
if (n_bits < 15) {
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
bits |= *((uint64_t*)psrc) << n_bits;
psrc += 6;
n_bits += 48;
@@ -275,13 +278,13 @@ outer_loop:
// TODO: look at a sliding window, not just output written so far to dst.
if ((ptrdiff_t)(distance) > (pdst - pdst0)) {
status = PUFFS_FLATE__ERROR_BAD_ARGUMENT;
status = WUFFS_FLATE__ERROR_BAD_ARGUMENT;
goto end;
}
uint8_t* pback = pdst - distance;
#ifdef PUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
#ifdef WUFFS_FLATE__HAVE_64_BIT_UNALIGNED_LITTLE_ENDIAN_LOADS
// Back-copy fast path, copying 8 instead of 1 bytes at a time.
//
// This always copies 8*N bytes (where N is the smallest integer such that