diff options
| author | Julian Taylor <jtaylor.debian@googlemail.com> | 2013-06-23 12:25:43 +0200 |
|---|---|---|
| committer | Julian Taylor <jtaylor.debian@googlemail.com> | 2013-06-23 12:25:43 +0200 |
| commit | e2a74ee4d49bf724917b26c0e11759fd877efaca (patch) | |
| tree | 6dd998a1d8f0f583849055669529c44f8e7c09b5 | |
| parent | 12e06a261e8ea6b08002de8e9933fd43e5465df9 (diff) | |
| download | numpy-e2a74ee4d49bf724917b26c0e11759fd877efaca.tar.gz | |
MAINT: make binary_scalar2 static and put __m128i into a template
| -rw-r--r-- | numpy/core/src/umath/simd.inc.src | 13 |
1 files changed, 6 insertions, 7 deletions
diff --git a/numpy/core/src/umath/simd.inc.src b/numpy/core/src/umath/simd.inc.src index 0382f2cf7..05f8710e5 100644 --- a/numpy/core/src/umath/simd.inc.src +++ b/numpy/core/src/umath/simd.inc.src @@ -363,7 +363,7 @@ sse2_binary_scalar1_@kind@_@TYPE@(@type@ * op, @type@ * ip1, @type@ * ip2, npy_i } -void +static void sse2_binary_scalar2_@kind@_@TYPE@(@type@ * op, @type@ * ip1, @type@ * ip2, npy_intp n) { const @vtype@ b = @vpre@_set1_@vsuf@(ip2[0]); @@ -518,7 +518,6 @@ sse2_@kind@_@TYPE@(@type@ * ip, @type@ * op, const npy_intp n) * # kind = logical_or, logical_and# * # and = 0, 1# * # op = ||, &&# - * # vop = or, and# * # vpre = _mm*2# * # vsuf = si128*2# * # vtype = __m128i*2# @@ -550,8 +549,8 @@ sse2_binary_@kind@_BOOL(npy_bool * op, npy_bool * ip1, npy_bool * ip2, npy_intp LOOP_BLOCK_ALIGN_VAR(op, @type@, 16) op[i] = ip1[i] @op@ ip2[i]; LOOP_BLOCKED(@type@, 16) { - @vtype@ a = @vloadu@((__m128i*)&ip1[i]); - @vtype@ b = @vloadu@((__m128i*)&ip2[i]); + @vtype@ a = @vloadu@((@vtype@*)&ip1[i]); + @vtype@ b = @vloadu@((@vtype@*)&ip2[i]); #if @and@ const @vtype@ zero = @vpre@_setzero_@vsuf@(); /* get 0xFF for non zeros*/ @@ -562,7 +561,7 @@ sse2_binary_@kind@_BOOL(npy_bool * op, npy_bool * ip1, npy_bool * ip2, npy_intp @vtype@ tmp = @vpre@_or_@vsuf@(a, b); #endif - @vstore@((__m128i*)&op[i], byte_to_true(tmp)); + @vstore@((@vtype@*)&op[i], byte_to_true(tmp)); } LOOP_BLOCKED_END { op[i] = (ip1[i] @op@ ip2[i]); @@ -589,7 +588,7 @@ sse2_@kind@_BOOL(@type@ * op, @type@ * ip, const npy_intp n) LOOP_BLOCK_ALIGN_VAR(op, @type@, 16) op[i] = (ip[i] @op@ 0); LOOP_BLOCKED(@type@, 16) { - @vtype@ a = @vloadu@((__m128i*)&ip[i]); + @vtype@ a = @vloadu@((@vtype@*)&ip[i]); #if @not@ const @vtype@ zero = @vpre@_setzero_@vsuf@(); const @vtype@ truemask = @vpre@_set1_epi8(1 == 1); @@ -600,7 +599,7 @@ sse2_@kind@_BOOL(@type@ * op, @type@ * ip, const npy_intp n) /* abs is kind of pointless but maybe its used for byte_to_true */ a = byte_to_true(a); #endif - @vstore@((__m128i*)&op[i], a); + @vstore@((@vtype@*)&op[i], a); } LOOP_BLOCKED_END { op[i] = (ip[i] @op@ 0); |