mewin/glm
1
0
Fork 0
Code Releases Activity

Fixed SIMD code path selection

Browse Source
This commit is contained in:
Christophe Riccio
2016-05-28 17:54:37 +02:00
parent ab159770b3
commit 9f00ba86cb
15 changed files with 92 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
glm/simd/common.h
View File

@@ -3,12 +3,12 @@
#pragma once
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
//mad
GLM_FUNC_QUALIFIER __m128 glm_f32v1_mad(__m128 a, __m128 b, __m128 c)
{
# if GLM_ARCH & GLM_ARCH_AVX2
# if GLM_ARCH & GLM_ARCH_AVX2_FLAG
return _mm_fmadd_ss(a, b, c);
# else
return _mm_add_ss(_mm_mul_ss(a, b), c);
@@ -18,7 +18,7 @@ GLM_FUNC_QUALIFIER __m128 glm_f32v1_mad(__m128 a, __m128 b, __m128 c)
//mad
GLM_FUNC_QUALIFIER __m128 glm_f32v4_mad(__m128 a, __m128 b, __m128 c)
{
# if GLM_ARCH & GLM_ARCH_AVX2
# if GLM_ARCH & GLM_ARCH_AVX2_FLAG
return _mm_fmadd_ps(a, b, c);
# else
return _mm_add_ps(_mm_mul_ps(a, b), c);
@@ -33,7 +33,7 @@ GLM_FUNC_QUALIFIER __m128 glm_f32v4_abs(__m128 x)
GLM_FUNC_QUALIFIER __m128i glm_i32v4_abs(__m128i x)
{
# if GLM_ARCH & GLM_ARCH_SSSE3
# if GLM_ARCH & GLM_ARCH_SSSE3_FLAG
return _mm_sign_epi32(x, x);
# else
__m128i const sgn0 = _mm_srai_epi32(x, 31);
@@ -202,4 +202,4 @@ GLM_FUNC_QUALIFIER __m128 glm_f32v4_sqrt_wip(__m128 x)
return Mul3;
}
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG

10
glm/simd/geometric.h
View File

@@ -5,13 +5,13 @@
#include "common.h"
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
GLM_FUNC_QUALIFIER __m128 glm_f32v4_dot(__m128 v1, __m128 v2)
{
# if GLM_ARCH & GLM_ARCH_AVX
# if GLM_ARCH & GLM_ARCH_AVX_FLAG
return _mm_dp_ps(v1, v2, 0xff);
# elif GLM_ARCH & GLM_ARCH_SSE3
# elif GLM_ARCH & GLM_ARCH_SSE3_FLAG
__m128 const Mul0 = _mm_mul_ps(v1, v2);
__m128 const Hadd0 = _mm_hadd_ps(Mul0, Mul0);
__m128 const Hadd1 = _mm_hadd_ps(Hadd0, Hadd0);
@@ -28,9 +28,9 @@ GLM_FUNC_QUALIFIER __m128 glm_f32v4_dot(__m128 v1, __m128 v2)
GLM_FUNC_QUALIFIER __m128 glm_f32v1_dot(__m128 v1, __m128 v2)
{
# if GLM_ARCH & GLM_ARCH_AVX
# if GLM_ARCH & GLM_ARCH_AVX_FLAG
return _mm_dp_ps(v1, v2, 0xff);
# elif GLM_ARCH & GLM_ARCH_SSE3
# elif GLM_ARCH & GLM_ARCH_SSE3_FLAG
__m128 const mul0 = _mm_mul_ps(v1, v2);
__m128 const had0 = _mm_hadd_ps(mul0, mul0);
__m128 const had1 = _mm_hadd_ps(had0, had0);

4
glm/simd/integer.h
View File

@@ -3,7 +3,7 @@
#pragma once
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
GLM_FUNC_QUALIFIER __m128i glm_i128_interleave(__m128i x)
{
@@ -112,4 +112,4 @@ GLM_FUNC_QUALIFIER __m128i glm_i128_interleave2(__m128i x, __m128i y)
return Reg1;
}
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG

4
glm/simd/matrix.h
View File

@@ -5,7 +5,7 @@
#include "geometric.h"
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
static const __m128 GLM_VAR_USED _m128_rad_ps = _mm_set_ps1(3.141592653589793238462643383279f / 180.f);
static const __m128 GLM_VAR_USED _m128_deg_ps = _mm_set_ps1(180.f / 3.141592653589793238462643383279f);
@@ -1029,4 +1029,4 @@ GLM_FUNC_QUALIFIER void glm_f32m4_outer(__m128 const & c, __m128 const & r, __m1
out[3] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(3, 3, 3, 3)));
}
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG

4
glm/simd/packing.h
View File

@@ -3,6 +3,6 @@
#pragma once
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG

4
glm/simd/trigonometric.h
View File

@@ -3,7 +3,7 @@
#pragma once
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG

4
glm/simd/vector_relational.h
View File

@@ -3,6 +3,6 @@
#pragma once
#if GLM_ARCH & GLM_ARCH_SSE2
#if GLM_ARCH & GLM_ARCH_SSE2_FLAG
#endif//GLM_ARCH & GLM_ARCH_SSE2
#endif//GLM_ARCH & GLM_ARCH_SSE2_FLAG