diff options
| author | Nathan Moinvaziri <nathan@nathanm.com> | 2022-08-28 20:27:37 -0700 |
|---|---|---|
| committer | Hans Kristian Rosbach <hk-github@circlestorm.org> | 2023-02-05 17:51:46 +0100 |
| commit | aa1109bb2ee79680b0ac966b1de8724ba5083093 (patch) | |
| tree | b3541be72e171ac12645d21b97d0be0d8aa0bd98 /cpu_features.h | |
| parent | 941ce547cc5e17d01bba861d3edc1f6b27e42a43 (diff) | |
| download | Project-Tick-aa1109bb2ee79680b0ac966b1de8724ba5083093.tar.gz Project-Tick-aa1109bb2ee79680b0ac966b1de8724ba5083093.zip | |
Use arch-specific versions of inflate_fast.
This should reduce the cost of indirection that occurs when calling functable
chunk copying functions inside inflate_fast. It should also allow the compiler
to optimize the inflate fast path for the specific architecture.
Diffstat (limited to 'cpu_features.h')
| -rw-r--r-- | cpu_features.h | 18 |
1 files changed, 18 insertions, 0 deletions
diff --git a/cpu_features.h b/cpu_features.h index 72e40a1652..d211cb112e 100644 --- a/cpu_features.h +++ b/cpu_features.h @@ -110,6 +110,24 @@ extern uint8_t* chunkmemset_power8(uint8_t *out, unsigned dist, unsigned len); extern uint8_t* chunkmemset_safe_power8(uint8_t *out, unsigned dist, unsigned len, unsigned left); #endif +/* inflate fast loop */ +extern void inflate_fast_c(void *strm, uint32_t start); +#ifdef X86_SSE2_CHUNKSET +extern void inflate_fast_sse2(void *strm, uint32_t start); +#endif +#ifdef X86_SSE41 +extern void inflate_fast_sse41(void *strm, uint32_t start); +#endif +#ifdef X86_AVX_CHUNKSET +extern void inflate_fast_avx(void *strm, uint32_t start); +#endif +#ifdef ARM_NEON_CHUNKSET +extern void inflate_fast_neon(void *strm, uint32_t start); +#endif +#ifdef POWER8_VSX_CHUNKSET +extern void inflate_fast_power8(void *strm, uint32_t start); +#endif + /* CRC32 */ typedef uint32_t (*crc32_func)(uint32_t crc32, const uint8_t *buf, size_t len); |