opencv/3rdparty/zlib-ng/arch/x86/chunkset_ssse3.c
FantasqueX 85923c8f30
Merge pull request #26113 from FantasqueX:zlib-ng-2-2-1
Update zlib-ng to 2.2.1 #26113

Release: https://github.com/zlib-ng/zlib-ng/releases/tag/2.2.1
ARM diagnostics patch: https://github.com/zlib-ng/zlib-ng/pull/1774

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
2024-09-12 16:05:24 +03:00

94 lines
2.6 KiB
C

/* chunkset_ssse3.c -- SSSE3 inline functions to copy small data chunks.
* For conditions of distribution and use, see copyright notice in zlib.h
*/
#include "zbuild.h"
#if defined(X86_SSSE3)
#include <immintrin.h>
#include "../generic/chunk_permute_table.h"
typedef __m128i chunk_t;
#define CHUNK_SIZE 16
#define HAVE_CHUNKMEMSET_2
#define HAVE_CHUNKMEMSET_4
#define HAVE_CHUNKMEMSET_8
#define HAVE_CHUNK_MAG
static const lut_rem_pair perm_idx_lut[13] = {
{0, 1}, /* 3 */
{0, 0}, /* don't care */
{1 * 32, 1}, /* 5 */
{2 * 32, 4}, /* 6 */
{3 * 32, 2}, /* 7 */
{0 * 32, 0}, /* don't care */
{4 * 32, 7}, /* 9 */
{5 * 32, 6}, /* 10 */
{6 * 32, 5}, /* 11 */
{7 * 32, 4}, /* 12 */
{8 * 32, 3}, /* 13 */
{9 * 32, 2}, /* 14 */
{10 * 32, 1},/* 15 */
};
static inline void chunkmemset_2(uint8_t *from, chunk_t *chunk) {
int16_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = _mm_set1_epi16(tmp);
}
static inline void chunkmemset_4(uint8_t *from, chunk_t *chunk) {
int32_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = _mm_set1_epi32(tmp);
}
static inline void chunkmemset_8(uint8_t *from, chunk_t *chunk) {
int64_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = _mm_set1_epi64x(tmp);
}
static inline void loadchunk(uint8_t const *s, chunk_t *chunk) {
*chunk = _mm_loadu_si128((__m128i *)s);
}
static inline void storechunk(uint8_t *out, chunk_t *chunk) {
_mm_storeu_si128((__m128i *)out, *chunk);
}
static inline chunk_t GET_CHUNK_MAG(uint8_t *buf, uint32_t *chunk_rem, uint32_t dist) {
lut_rem_pair lut_rem = perm_idx_lut[dist - 3];
__m128i perm_vec, ret_vec;
/* Important to note:
* This is _not_ to subvert the memory sanitizer but to instead unpoison some
* bytes we willingly and purposefully load uninitialized that we swizzle over
* in a vector register, anyway. If what we assume is wrong about what is used,
* the memory sanitizer will still usefully flag it */
__msan_unpoison(buf + dist, 16 - dist);
ret_vec = _mm_loadu_si128((__m128i*)buf);
*chunk_rem = lut_rem.remval;
perm_vec = _mm_load_si128((__m128i*)(permute_table + lut_rem.idx));
ret_vec = _mm_shuffle_epi8(ret_vec, perm_vec);
return ret_vec;
}
#define CHUNKSIZE chunksize_ssse3
#define CHUNKMEMSET chunkmemset_ssse3
#define CHUNKMEMSET_SAFE chunkmemset_safe_ssse3
#define CHUNKCOPY chunkcopy_ssse3
#define CHUNKUNROLL chunkunroll_ssse3
#include "chunkset_tpl.h"
#define INFLATE_FAST inflate_fast_ssse3
#include "inffast_tpl.h"
#endif