Cr8-xyz/blender
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Cycles: Minor optimization (~1%) for texture access on CPU

Browse Source
This commit is contained in:
Sv. Lockal
2014-01-06 22:05:31 +04:00
parent 4817da0df4
commit acc90b40bf
1 changed files with 11 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
intern/cycles/kernel/svm/svm_image.h
View File

@@ -116,6 +116,9 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
#ifdef __KERNEL_CPU__ #ifdef __KERNEL_CPU__
r = kernel_tex_image_interp(id, x, y); r = kernel_tex_image_interp(id, x, y);
#ifdef __KERNEL_SSE2__
__m128 *rv = (__m128 *)&r;
#endif
#else #else
/* not particularly proud of this massive switch, what are the /* not particularly proud of this massive switch, what are the
* alternatives? * alternatives?
@@ -234,6 +237,13 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
#endif #endif
if(use_alpha && r.w != 1.0f && r.w != 0.0f) { if(use_alpha && r.w != 1.0f && r.w != 0.0f) {
#ifdef __KERNEL_SSE2__
float alpha = r.w;
*rv = _mm_div_ps(*rv, _mm_set1_ps(alpha));
if(id >= TEX_NUM_FLOAT_IMAGES)
*rv = _mm_min_ps(*rv, _mm_set1_ps(1.0f));
r.w = alpha;
#else
float invw = 1.0f/r.w; float invw = 1.0f/r.w;
r.x *= invw; r.x *= invw;
r.y *= invw; r.y *= invw;
@@ -244,12 +254,12 @@ ccl_device float4 svm_image_texture(KernelGlobals *kg, int id, float x, float y,
r.y = min(r.y, 1.0f); r.y = min(r.y, 1.0f);
r.z = min(r.z, 1.0f); r.z = min(r.z, 1.0f);
} }
#endif
} }
if(srgb) { if(srgb) {
#ifdef __KERNEL_SSE2__ #ifdef __KERNEL_SSE2__
float alpha = r.w; float alpha = r.w;
__m128 *rv = (__m128 *)&r;
*rv = color_srgb_to_scene_linear(*rv); *rv = color_srgb_to_scene_linear(*rv);
r.w = alpha; r.w = alpha;
#else #else