External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-12-30 20:27:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #412 from recp/sse_only

Browse Source 
separate SSE and SSE2
This commit is contained in:
Recep Aslantas
2024-04-01 17:52:06 +03:00
committed by GitHub
parent 9ad7dd3fbc 568634a79e
commit 1796cc5ce2
7 changed files with 106 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/cglm/common.h
View File

@@ -45,6 +45,10 @@
# define CGLM_LIKELY(expr) (expr)
#endif
#if defined(_M_FP_FAST) || defined(__FAST_MATH__)
# define CGLM_FAST_MATH
#endif
#define GLM_SHUFFLE4(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
#define GLM_SHUFFLE3(z, y, x) (((z) << 4) | ((y) << 2) | (x))

14
include/cglm/simd/intrin.h
View File

@@ -10,6 +10,9 @@
#if defined( _MSC_VER )
# if (defined(_M_AMD64) || defined(_M_X64)) || _M_IX86_FP == 2
# ifndef __SSE__
# define __SSE__
# endif
# ifndef __SSE2__
# define __SSE2__
# endif
@@ -24,15 +27,22 @@
# endif
#endif
#if defined( __SSE__ ) || defined( __SSE2__ )
#if defined(__SSE__)
# include <xmmintrin.h>
# include <emmintrin.h>
# define CGLM_SSE_FP 1
# ifndef CGLM_SIMD_x86
# define CGLM_SIMD_x86
# endif
#endif
#if defined(__SSE2__)
# include <emmintrin.h>
# define CGLM_SSE2_FP 1
# ifndef CGLM_SIMD_x86
# define CGLM_SIMD_x86
# endif
#endif
#if defined(__SSE3__)
# include <pmmintrin.h>
# ifndef CGLM_SIMD_x86

39
include/cglm/simd/x86.h
View File

@@ -21,7 +21,7 @@
#define glmm_set1(x) _mm_set1_ps(x)
#define glmm_128 __m128
#ifdef CGLM_USE_INT_DOMAIN
#if defined(CGLM_USE_INT_DOMAIN) && defined(__SSE2__)
# define glmm_shuff1(xmm, z, y, x, w) \
_mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(xmm), \
_MM_SHUFFLE(z, y, x, w)))
@@ -55,17 +55,40 @@
#endif
/* Note that `0x80000000` corresponds to `INT_MIN` for a 32-bit int. */
#define GLMM_NEGZEROf ((int)0x80000000) /* 0x80000000 ---> -0.0f */
#define GLMM__SIGNMASKf(X, Y, Z, W) \
#if defined(__SSE2__)
# define GLMM_NEGZEROf ((int)0x80000000) /* 0x80000000 ---> -0.0f */
# define GLMM_POSZEROf ((int)0x00000000) /* 0x00000000 ---> +0.0f */
#else
# ifdef CGLM_FAST_MATH
union { int i; float f; } static GLMM_NEGZEROf_TU = { .i = (int)0x80000000 };
# define GLMM_NEGZEROf GLMM_NEGZEROf_TU.f
# define GLMM_POSZEROf 0.0f
# else
# define GLMM_NEGZEROf -0.0f
# define GLMM_POSZEROf 0.0f
# endif
#endif
#if defined(__SSE2__)
# define GLMM__SIGNMASKf(X, Y, Z, W) \
_mm_castsi128_ps(_mm_set_epi32(X, Y, Z, W))
/* _mm_set_ps(X, Y, Z, W); */
#else
# define GLMM__SIGNMASKf(X, Y, Z, W) _mm_set_ps(X, Y, Z, W)
#endif
#define glmm_float32x4_SIGNMASK_PNPN GLMM__SIGNMASKf(0, GLMM_NEGZEROf, 0, GLMM_NEGZEROf)
#define glmm_float32x4_SIGNMASK_NPNP GLMM__SIGNMASKf(GLMM_NEGZEROf, 0, GLMM_NEGZEROf, 0)
#define glmm_float32x4_SIGNMASK_NPPN GLMM__SIGNMASKf(GLMM_NEGZEROf, 0, 0, GLMM_NEGZEROf)
#define glmm_float32x4_SIGNMASK_PNPN GLMM__SIGNMASKf(GLMM_POSZEROf, GLMM_NEGZEROf, GLMM_POSZEROf, GLMM_NEGZEROf)
#define glmm_float32x4_SIGNMASK_NPNP GLMM__SIGNMASKf(GLMM_NEGZEROf, GLMM_POSZEROf, GLMM_NEGZEROf, GLMM_POSZEROf)
#define glmm_float32x4_SIGNMASK_NPPN GLMM__SIGNMASKf(GLMM_NEGZEROf, GLMM_POSZEROf, GLMM_POSZEROf, GLMM_NEGZEROf)
/* fasth math prevents -0.0f to work */
#if defined(__SSE2__)
# define glmm_float32x4_SIGNMASK_NEG _mm_castsi128_ps(_mm_set1_epi32(GLMM_NEGZEROf)) /* _mm_set1_ps(-0.0f) */
#else
# define glmm_float32x4_SIGNMASK_NEG _mm_set1_ps(GLMM_NEGZEROf)
#endif
#define glmm_float32x4_SIGNMASK_NEG _mm_castsi128_ps(_mm_set1_epi32(GLMM_NEGZEROf)) /* _mm_set1_ps(-0.0f) */
#define glmm_float32x8_SIGNMASK_NEG _mm256_castsi256_ps(_mm256_set1_epi32(GLMM_NEGZEROf))
static inline
@@ -207,6 +230,7 @@ glmm_norm_inf(__m128 a) {
return _mm_cvtss_f32(glmm_vhmax(glmm_abs(a)));
}
#if defined(__SSE2__)
static inline
__m128
glmm_load3(float v[3]) {
@@ -225,6 +249,7 @@ glmm_store3(float v[3], __m128 vx) {
_mm_storel_pi(CGLM_CASTPTR_ASSUME_ALIGNED(v, __m64), vx);
_mm_store_ss(&v[2], glmm_shuff1(vx, 2, 2, 2, 2));
}
#endif
static inline
__m128