opencv/3rdparty/zlib-ng/arch/power/chunkset_power8.c
Letu Ren 0de26fd78e Add zlib-ng as an alternative zlib implementation
Zlib-ng is zlib replacement with optimizations for "next generation" systems. Its optimization may benifits image library decode and encode speed such as libpng. In our tests, if using zlib-ng and libpng combination on a x86_64 machine with AVX2, the time of `imdecode` amd `imencode` will drop 20% approximately. This patch enables zlib-ng's optimization if `CV_DISABLE_OPTIMIZATION` is OFF. Since Zlib-ng can dispatch intrinsics on the fly, port work is much easier.

Related discussion: https://github.com/opencv/opencv/issues/22573
2024-01-14 14:58:47 +08:00

56 lines
1.4 KiB
C

/* chunkset_power8.c -- VSX inline functions to copy small data chunks.
* For conditions of distribution and use, see copyright notice in zlib.h
*/
#ifdef POWER8_VSX
#include <altivec.h>
#include "../../zbuild.h"
typedef vector unsigned char chunk_t;
#define CHUNK_SIZE 16
#define HAVE_CHUNKMEMSET_2
#define HAVE_CHUNKMEMSET_4
#define HAVE_CHUNKMEMSET_8
static inline void chunkmemset_2(uint8_t *from, chunk_t *chunk) {
uint16_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = (vector unsigned char)vec_splats(tmp);
}
static inline void chunkmemset_4(uint8_t *from, chunk_t *chunk) {
uint32_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = (vector unsigned char)vec_splats(tmp);
}
static inline void chunkmemset_8(uint8_t *from, chunk_t *chunk) {
uint64_t tmp;
memcpy(&tmp, from, sizeof(tmp));
*chunk = (vector unsigned char)vec_splats((unsigned long long)tmp);
}
static inline void loadchunk(uint8_t const *s, chunk_t *chunk) {
*chunk = vec_xl(0, s);
}
static inline void storechunk(uint8_t *out, chunk_t *chunk) {
vec_xst(*chunk, 0, out);
}
#define CHUNKSIZE chunksize_power8
#define CHUNKCOPY chunkcopy_power8
#define CHUNKUNROLL chunkunroll_power8
#define CHUNKMEMSET chunkmemset_power8
#define CHUNKMEMSET_SAFE chunkmemset_safe_power8
#include "chunkset_tpl.h"
#define INFLATE_FAST inflate_fast_power8
#include "inffast_tpl.h"
#endif