aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

__not_equal_i8/16 was added

Browse Source
This commit is contained in:
Vsevolod Livinskiy
2014-11-06 15:02:28 +04:00
committed by Vsevolod Livinskij
parent 3b1445c660
commit 9316dd65c0
1 changed files with 12 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
examples/intrinsics/knc.h
View File

@@ -460,6 +460,12 @@ template <> FORCEINLINE __vec16_i8 __setzero_i8<__vec16_i8>() {
return __vec16_i8(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
}
static FORCEINLINE __vec16_i1 __not_equal_i8(__vec16_i8 a, __vec16_i8 b) {
__vec16_i32 tmp_a = _mm512_extload_epi32(&a, _MM_UPCONV_EPI32_SINT8, _MM_BROADCAST32_NONE, _MM_HINT_NONE);
__vec16_i32 tmp_b = _mm512_extload_epi32(&b, _MM_UPCONV_EPI32_SINT8, _MM_BROADCAST32_NONE, _MM_HINT_NONE);
return _mm512_cmpneq_epi32_mask(tmp_a, tmp_b);
}
template <class RetVecType> static RetVecType __smear_i8(int8_t i);
template <> FORCEINLINE __vec16_i8 __smear_i8<__vec16_i8>(int8_t i) {
__vec16_i32 tmp = _mm512_set1_epi32(i);
@@ -478,6 +484,12 @@ template <> FORCEINLINE __vec16_i16 __setzero_i16<__vec16_i16>() {
return __vec16_i16(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
}
static FORCEINLINE __vec16_i1 __not_equal_i16(__vec16_i16 a, __vec16_i16 b) {
__vec16_i32 tmp_a = _mm512_extload_epi32(&a, _MM_UPCONV_EPI32_SINT16, _MM_BROADCAST32_NONE, _MM_HINT_NONE);
__vec16_i32 tmp_b = _mm512_extload_epi32(&b, _MM_UPCONV_EPI32_SINT16, _MM_BROADCAST32_NONE, _MM_HINT_NONE);
return _mm512_cmpneq_epi32_mask(tmp_a, tmp_b);
}
template <class RetVecType> static RetVecType __smear_i16(int16_t i);
template <> FORCEINLINE __vec16_i16 __smear_i16<__vec16_i16>(int16_t i) {
__vec16_i32 tmp = _mm512_set1_epi32(i);