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

starded work on gather/scatter instructions

Browse Source
This commit is contained in:
Anton Mitrokhin
2014-09-11 19:06:07 +04:00
parent 0881463d69
commit 8b8e313dc6
1 changed files with 29 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
examples/intrinsics/knc.h
View File

@@ -865,10 +865,6 @@ static FORCEINLINE __vec16_i64 __select(__vec16_i1 mask,
return ret;
}
// static FORCEINLINE int64_t __extract_element(__vec16_i64 v, uint32_t index) {
// return (uint64_t(((int32_t *)&v.v_hi)[index])<<32) | (uint64_t(((int32_t *)&v.v_lo)[index]));
// }
template <class RetVecType> static RetVecType __smear_i64(const int64_t &l);
template <> FORCEINLINE __vec16_i64 __smear_i64<__vec16_i64>(const int64_t &l) {
const int *i = (const int*)&l;
@@ -1921,6 +1917,34 @@ __gather_base_offsets32_double(uint8_t *base, uint32_t scale, __vec16_i32 offset
return ret;
}
// TODO: Test this implementation
static FORCEINLINE __vec16_f
__gather64_float(__vec16_i64 addr, __vec16_i1 mask)
{
__vec16_f ret;
// There is no gather instruction with 64-bit offsets in KNC.
// We have to manually iterate over the upper 32 bits ;-)
__vec16_i1 still_to_do = mask;
const __vec16_i32 signed_offsets = _mm512_add_epi32(addr.v_lo, __smear_i32<__vec16_i32>((int32_t)INT_MIN));
while (still_to_do) {
int first_active_lane = _mm_tzcnt_32((int)still_to_do);
const uint &hi32 = ((uint*)&addr.v_hi)[first_active_lane];
__vec16_i1 match = _mm512_mask_cmp_epi32_mask(still_to_do,addr.v_hi,
__smear_i32<__vec16_i32>((int32_t)hi32),
_MM_CMPINT_EQ);
void * base = (void*)((((unsigned long)hi32) << 32) + (unsigned long)(-(long)INT_MIN));
ret.v = _mm512_mask_i32extgather_ps(ret.v, match, signed_offsets,
base, _MM_UPCONV_PS_NONE, 1,
_MM_HINT_NONE);
still_to_do = _mm512_kxor(match, still_to_do);
}
return ret;
}
/*! gather with 64-bit offsets.
\todo add optimization that falls back to 32-bit offset gather if
@@ -1932,7 +1956,7 @@ static FORCEINLINE __vec16_f
__gather_base_offsets64_float(uint8_t *_base, uint32_t scale, __vec16_i64 offsets,
__vec16_i1 mask) {
const __vec16_i32 signed_offsets = _mm512_add_epi32(offsets.v_lo, __smear_i32<__vec16_i32>((int32_t)INT_MIN));
const __vec16_i32 signed_offsets = _mm512_add_epi32(offsets.v_lo, __smear_i32<__vec16_i32>((int32_t)INT_MIN));
// There is no gather instruction with 64-bit offsets in KNC.
// We have to manually iterate over the upper 32 bits ;-)
__vec16_i1 still_to_do = mask;