diff options
Diffstat (limited to 'numpy')
| -rw-r--r-- | numpy/core/src/multiarray/einsum.c.src | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/numpy/core/src/multiarray/einsum.c.src b/numpy/core/src/multiarray/einsum.c.src index b914e5bb3..198191db5 100644 --- a/numpy/core/src/multiarray/einsum.c.src +++ b/numpy/core/src/multiarray/einsum.c.src @@ -276,6 +276,8 @@ static void #if EINSUM_USE_SSE1 && @float32@ __m128 a, b; +#elif EINSUM_USE_SSE2 && @float64@ +__m128d a, b; #endif NPY_EINSUM_DBG_PRINT1("@name@_sum_of_products_contig_two (%d)\n", @@ -324,7 +326,29 @@ finish_after_unrolled_loop: /* Unroll the loop by 8 */ while (count >= 8) { count -= 8; +#elif EINSUM_USE_SSE2 && @float64@ + /* Use aligned instructions if possible */ + if (EINSUM_IS_SSE_ALIGNED(data0) && EINSUM_IS_SSE_ALIGNED(data1) && + EINSUM_IS_SSE_ALIGNED(data_out)) { + /* Unroll the loop by 8 */ + while (count >= 8) { + count -= 8; +/**begin repeat2 + * #i = 0, 2, 4, 6# + */ + a = _mm_mul_pd(_mm_load_pd(data0+@i@), _mm_load_pd(data1+@i@)); + b = _mm_add_pd(a, _mm_load_pd(data_out+@i@)); + _mm_store_pd(data_out+@i@, b); +/**end repeat2**/ + data0 += 8; + data1 += 8; + data_out += 8; + } + + /* Finish off the loop */ + goto finish_after_unrolled_loop; + } #if EINSUM_USE_SSE1 && @float32@ /**begin repeat2 * #i = 0, 4# @@ -333,6 +357,14 @@ finish_after_unrolled_loop: b = _mm_add_ps(a, _mm_loadu_ps(data_out+@i@)); _mm_storeu_ps(data_out+@i@, b); /**end repeat2**/ +#elif EINSUM_USE_SSE2 && @float64@ +/**begin repeat2 + * #i = 0, 2, 4, 6# + */ + a = _mm_mul_pd(_mm_loadu_pd(data0+@i@), _mm_loadu_pd(data1+@i@)); + b = _mm_add_pd(a, _mm_loadu_pd(data_out+@i@)); + _mm_storeu_pd(data_out+@i@, b); +/**end repeat2**/ #else /**begin repeat2 * #i = 0, 1, 2, 3, 4, 5, 6, 7# |