External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-10-04 17:09:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

simd: use new glmm_min/max in vec4 where possible

Browse Source
This commit is contained in:
Recep Aslantas
2024-01-11 00:14:28 +03:00
parent dab86796a4
commit 6d8dd42ac2
1 changed files with 29 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

79
include/cglm/vec4.h
View File

@@ -653,17 +653,14 @@ CGLM_INLINE
void void
glm_vec4_maxadd(vec4 a, vec4 b, vec4 dest) { glm_vec4_maxadd(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(__wasm__) && defined(__wasm_simd128__)
glmm_store(dest, wasm_f32x4_add( glmm_store(dest, wasm_f32x4_add(glmm_load(dest),
glmm_load(dest), glmm_max(glmm_load(a), glmm_load(b))));
wasm_f32x4_pmax(glmm_load(a), glmm_load(b))));
#elif defined( __SSE__ ) || defined( __SSE2__ ) #elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_add_ps(glmm_load(dest), glmm_store(dest, _mm_add_ps(glmm_load(dest),
_mm_max_ps(glmm_load(a), glmm_max(glmm_load(a), glmm_load(b))));
glmm_load(b))));
#elif defined(CGLM_NEON_FP) #elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vaddq_f32(vld1q_f32(dest), glmm_store(dest, vaddq_f32(glmm_load(dest),
vmaxq_f32(vld1q_f32(a), glmm_max(glmm_load(a), glmm_load(b))));
vld1q_f32(b))));
#else #else
dest[0] += glm_max(a[0], b[0]); dest[0] += glm_max(a[0], b[0]);
dest[1] += glm_max(a[1], b[1]); dest[1] += glm_max(a[1], b[1]);
@@ -685,17 +682,14 @@ CGLM_INLINE
void void
glm_vec4_minadd(vec4 a, vec4 b, vec4 dest) { glm_vec4_minadd(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(__wasm__) && defined(__wasm_simd128__)
glmm_store(dest, wasm_f32x4_add( glmm_store(dest, wasm_f32x4_add(glmm_load(dest),
glmm_load(dest), glmm_min(glmm_load(a), glmm_load(b))));
wasm_f32x4_pmin(glmm_load(a), glmm_load(b))));
#elif defined( __SSE__ ) || defined( __SSE2__ ) #elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_add_ps(glmm_load(dest), glmm_store(dest, _mm_add_ps(glmm_load(dest),
_mm_min_ps(glmm_load(a), glmm_min(glmm_load(a), glmm_load(b))));
glmm_load(b))));
#elif defined(CGLM_NEON_FP) #elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vaddq_f32(vld1q_f32(dest), glmm_store(dest, vaddq_f32(glmm_load(dest),
vminq_f32(vld1q_f32(a), glmm_min(glmm_load(a), glmm_load(b))));
vld1q_f32(b))));
#else #else
dest[0] += glm_min(a[0], b[0]); dest[0] += glm_min(a[0], b[0]);
dest[1] += glm_min(a[1], b[1]); dest[1] += glm_min(a[1], b[1]);
@@ -825,17 +819,14 @@ CGLM_INLINE
void void
glm_vec4_maxsub(vec4 a, vec4 b, vec4 dest) { glm_vec4_maxsub(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(__wasm__) && defined(__wasm_simd128__)
glmm_store(dest, wasm_f32x4_sub( glmm_store(dest, wasm_f32x4_sub(glmm_load(dest),
glmm_load(dest), glmm_max(glmm_load(a), glmm_load(b))));
wasm_f32x4_pmax(glmm_load(a), glmm_load(b))));
#elif defined( __SSE__ ) || defined( __SSE2__ ) #elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_sub_ps(glmm_load(dest), glmm_store(dest, _mm_sub_ps(glmm_load(dest),
_mm_max_ps(glmm_load(a), glmm_max(glmm_load(a), glmm_load(b))));
glmm_load(b))));
#elif defined(CGLM_NEON_FP) #elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vsubq_f32(vld1q_f32(dest), glmm_store(dest, vsubq_f32(glmm_load(dest),
vmaxq_f32(vld1q_f32(a), glmm_max(glmm_load(a), glmm_load(b))));
vld1q_f32(b))));
#else #else
dest[0] -= glm_max(a[0], b[0]); dest[0] -= glm_max(a[0], b[0]);
dest[1] -= glm_max(a[1], b[1]); dest[1] -= glm_max(a[1], b[1]);
@@ -857,17 +848,14 @@ CGLM_INLINE
void void
glm_vec4_minsub(vec4 a, vec4 b, vec4 dest) { glm_vec4_minsub(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(__wasm__) && defined(__wasm_simd128__)
glmm_store(dest, wasm_f32x4_sub( glmm_store(dest, wasm_f32x4_sub(glmm_load(dest),
glmm_load(dest), glmm_min(glmm_load(a), glmm_load(b))));
wasm_f32x4_pmin(glmm_load(a), glmm_load(b))));
#elif defined( __SSE__ ) || defined( __SSE2__ ) #elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_sub_ps(glmm_load(dest), glmm_store(dest, _mm_sub_ps(glmm_load(dest),
_mm_min_ps(glmm_load(a), glmm_min(glmm_load(a), glmm_load(b))));
glmm_load(b))));
#elif defined(CGLM_NEON_FP) #elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vsubq_f32(vld1q_f32(dest), glmm_store(dest, vsubq_f32(vld1q_f32(dest),
vminq_f32(vld1q_f32(a), glmm_min(glmm_load(a), glmm_load(b))));
vld1q_f32(b))));
#else #else
dest[0] -= glm_min(a[0], b[0]); dest[0] -= glm_min(a[0], b[0]);
dest[1] -= glm_min(a[1], b[1]); dest[1] -= glm_min(a[1], b[1]);
@@ -1031,12 +1019,8 @@ glm_vec4_distance2(vec4 a, vec4 b) {
CGLM_INLINE CGLM_INLINE
void void
glm_vec4_maxv(vec4 a, vec4 b, vec4 dest) { glm_vec4_maxv(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(CGLM_SIMD)
glmm_store(dest, wasm_f32x4_pmax(glmm_load(a), glmm_load(b))); glmm_store(dest, glmm_max(glmm_load(a), glmm_load(b)));
#elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_max_ps(glmm_load(a), glmm_load(b)));
#elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vmaxq_f32(vld1q_f32(a), vld1q_f32(b)));
#else #else
dest[0] = glm_max(a[0], b[0]); dest[0] = glm_max(a[0], b[0]);
dest[1] = glm_max(a[1], b[1]); dest[1] = glm_max(a[1], b[1]);
@@ -1055,12 +1039,8 @@ glm_vec4_maxv(vec4 a, vec4 b, vec4 dest) {
CGLM_INLINE CGLM_INLINE
void void
glm_vec4_minv(vec4 a, vec4 b, vec4 dest) { glm_vec4_minv(vec4 a, vec4 b, vec4 dest) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(CGLM_SIMD)
glmm_store(dest, wasm_f32x4_pmin(glmm_load(a), glmm_load(b))); glmm_store(dest, glmm_min(glmm_load(a), glmm_load(b)));
#elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(dest, _mm_min_ps(glmm_load(a), glmm_load(b)));
#elif defined(CGLM_NEON_FP)
vst1q_f32(dest, vminq_f32(vld1q_f32(a), vld1q_f32(b)));
#else #else
dest[0] = glm_min(a[0], b[0]); dest[0] = glm_min(a[0], b[0]);
dest[1] = glm_min(a[1], b[1]); dest[1] = glm_min(a[1], b[1]);
@@ -1080,14 +1060,13 @@ CGLM_INLINE
void void
glm_vec4_clamp(vec4 v, float minVal, float maxVal) { glm_vec4_clamp(vec4 v, float minVal, float maxVal) {
#if defined(__wasm__) && defined(__wasm_simd128__) #if defined(__wasm__) && defined(__wasm_simd128__)
glmm_store(v, wasm_f32x4_pmin( glmm_store(v, glmm_min(glmm_max(glmm_load(v), wasm_f32x4_splat(minVal)),
wasm_f32x4_pmax(glmm_load(v), wasm_f32x4_splat(minVal)), wasm_f32x4_splat(maxVal)));
wasm_f32x4_splat(maxVal)));
#elif defined( __SSE__ ) || defined( __SSE2__ ) #elif defined( __SSE__ ) || defined( __SSE2__ )
glmm_store(v, _mm_min_ps(_mm_max_ps(glmm_load(v), _mm_set1_ps(minVal)), glmm_store(v, glmm_min(glmm_max(glmm_load(v), _mm_set1_ps(minVal)),
_mm_set1_ps(maxVal))); _mm_set1_ps(maxVal)));
#elif defined(CGLM_NEON_FP) #elif defined(CGLM_NEON_FP)
vst1q_f32(v, vminq_f32(vmaxq_f32(vld1q_f32(v), vdupq_n_f32(minVal)), glmm_store(v, glmm_min(glmm_max(vld1q_f32(v), vdupq_n_f32(minVal)),
vdupq_n_f32(maxVal))); vdupq_n_f32(maxVal)));
#else #else
v[0] = glm_clamp(v[0], minVal, maxVal); v[0] = glm_clamp(v[0], minVal, maxVal);