daiduo/TurboPFor-Integer-Compression
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add vrbitq function
Some checks failed
build / linux (clang, ubuntu-latest) (push) Has been cancelled
Details
build / linux (gcc, ubuntu-latest) (push) Has been cancelled
Details
build / macos (clang, macos-latest) (push) Has been cancelled
Details

Browse Source 
Signed-off-by: yangwenqing <1552539019@qq.com>
This commit is contained in:
yangwenqing
2025-06-11 16:39:27 +08:00
parent 6e00e1abcf
commit 2c942860f9
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
lib/include_/sse_neon.h
View File

@ -480,7 +480,7 @@ static ALWAYS_INLINE __m128i mm_testnz_epu8(_u_){
#define mm_cvtsi64_si128p(_u64p_, _u_) mm_loadu_epi64p(_u64p_,_v_)
#define _mm_cvtsi64_si128(_u_) (__m128i)__lsx_vreplgr2vr_d(_u_)
//---------------------------------------------- Reverse bits/bytes ---------------------------------------------------------------
static ALWAYS_INLINE __m128i mm_rbit_epi8(_v_) {
static ALWAYS_INLINE __m128i mm_rbit_epi8(__m128i _v_) {
uint64_t low_src = __lsx_vpickve2gr_du(_v_, 0);
uint64_t low = 0;
asm volatile(
@ -549,8 +549,8 @@ static ALWAYS_INLINE uint16_t _mm_movemask_epi8(__m128i v) {
__m128i sum32 = __lsx_vhaddw_wu_hu(sum16, sum16);
__m128i sum64 = __lsx_vhaddw_du_wu(sum32, sum32);
// 步骤5：提取低16位结果
return __lsx_vpickve2gr_hu(sum64, 0) | __lsx_vpickve2gr_hu(sum64, 8)<< 8;
// 步骤5：提取结果
return (uint16_t)__lsx_vpickve2gr_bu(sum64, 0) | (((uint16_t)__lsx_vpickve2gr_bu(sum64, 8))<< 8);
}