diff options
| author | Qiyu8 <fangchunlin@huawei.com> | 2021-01-22 16:43:47 +0800 |
|---|---|---|
| committer | Qiyu8 <fangchunlin@huawei.com> | 2021-01-22 16:43:47 +0800 |
| commit | 7a0a6e36c474ba4849aa67166649f48400467b76 (patch) | |
| tree | 5a19faed51dcd8727941bcfaa5e462043cebced4 /numpy | |
| parent | 617193440550cd50b7426a2c2008fdb60944c226 (diff) | |
| download | numpy-7a0a6e36c474ba4849aa67166649f48400467b76.tar.gz | |
correct the input paramater of f32.
Diffstat (limited to 'numpy')
| -rw-r--r-- | numpy/core/src/common/simd/avx2/arithmetic.h | 6 | ||||
| -rw-r--r-- | numpy/core/src/common/simd/sse/arithmetic.h | 4 |
2 files changed, 5 insertions, 5 deletions
diff --git a/numpy/core/src/common/simd/avx2/arithmetic.h b/numpy/core/src/common/simd/avx2/arithmetic.h index 770d4230c..9e13d6324 100644 --- a/numpy/core/src/common/simd/avx2/arithmetic.h +++ b/numpy/core/src/common/simd/avx2/arithmetic.h @@ -128,7 +128,7 @@ NPY_FINLINE npy_uint16 npyv_sumup_u8(npyv_u8 a) return (npy_uint16)_mm_cvtsi128_si32(one); } -NPY_FINLINE npy_uint32 npyv_sum_u32(__m256i a) +NPY_FINLINE npy_uint32 npyv_sum_u32(npyv_u32 a) { __m256i s0 = _mm256_hadd_epi32(a, a); s0 = _mm256_hadd_epi32(s0, s0); @@ -153,7 +153,7 @@ NPY_FINLINE npy_uint64 npyv_sum_u64(npyv_u64 a) return (npy_uint64)npyv128_cvtsi128_si64(one); } -NPY_FINLINE float npyv_sum_f32(npyv_u32 a) +NPY_FINLINE float npyv_sum_f32(npyv_f32 a) { __m256 sum_halves = _mm256_hadd_ps(a, a); sum_halves = _mm256_hadd_ps(sum_halves, sum_halves); @@ -163,7 +163,7 @@ NPY_FINLINE float npyv_sum_f32(npyv_u32 a) return _mm_cvtss_f32(sum); } -NPY_FINLINE double npyv_sum_f64(__m256d a) +NPY_FINLINE double npyv_sum_f64(npyv_f64 a) { __m256d sum_halves = _mm256_hadd_pd(a, a); __m128d lo = _mm256_castpd256_pd128(sum_halves); diff --git a/numpy/core/src/common/simd/sse/arithmetic.h b/numpy/core/src/common/simd/sse/arithmetic.h index 968de7545..92a53e630 100644 --- a/numpy/core/src/common/simd/sse/arithmetic.h +++ b/numpy/core/src/common/simd/sse/arithmetic.h @@ -176,7 +176,7 @@ NPY_FINLINE npy_uint64 npyv_sum_u64(npyv_u64 a) return idx[0] + idx[1]; } -NPY_FINLINE float npyv_sum_f32(__m128 a) +NPY_FINLINE float npyv_sum_f32(npyv_f32 a) { #ifdef NPY_HAVE_SSE3 __m128 sum_halves = _mm_hadd_ps(a, a); @@ -190,7 +190,7 @@ NPY_FINLINE float npyv_sum_f32(__m128 a) #endif } -NPY_FINLINE double npyv_sum_f64(__m128d a) +NPY_FINLINE double npyv_sum_f64(npyv_f64 a) { #ifdef NPY_HAVE_SSE3 return _mm_cvtsd_f64(_mm_hadd_pd(a, a)); |