[Bf-blender-cvs] SVN commit: /data/svn/bf-blender [54498] trunk/blender/intern/cycles/device /device_cpu.cpp: Fix cycles not using SSE3 kernel after recent, order with SSE2 should be switched,
Brecht Van Lommel
brechtvanlommel at pandora.be
Tue Feb 12 15:58:46 CET 2013
Revision: 54498
http://projects.blender.org/scm/viewvc.php?view=rev&root=bf-blender&revision=54498
Author: blendix
Date: 2013-02-12 14:58:46 +0000 (Tue, 12 Feb 2013)
Log Message:
-----------
Fix cycles not using SSE3 kernel after recent, order with SSE2 should be switched,
pointed out by Chad Fraleigh.
Modified Paths:
--------------
trunk/blender/intern/cycles/device/device_cpu.cpp
Modified: trunk/blender/intern/cycles/device/device_cpu.cpp
===================================================================
--- trunk/blender/intern/cycles/device/device_cpu.cpp 2013-02-12 14:45:59 UTC (rev 54497)
+++ trunk/blender/intern/cycles/device/device_cpu.cpp 2013-02-12 14:58:46 UTC (rev 54498)
@@ -171,7 +171,7 @@
int end_sample = tile.start_sample + tile.num_samples;
#ifdef WITH_OPTIMIZED_KERNEL
- if(system_cpu_support_sse2()) {
+ if(system_cpu_support_sse3()) {
for(int sample = start_sample; sample < end_sample; sample++) {
if (task.get_cancel() || task_pool.cancelled()) {
if(task.need_finish_queue == false)
@@ -180,7 +180,7 @@
for(int y = tile.y; y < tile.y + tile.h; y++) {
for(int x = tile.x; x < tile.x + tile.w; x++) {
- kernel_cpu_sse2_path_trace(&kg, render_buffer, rng_state,
+ kernel_cpu_sse3_path_trace(&kg, render_buffer, rng_state,
sample, x, y, tile.offset, tile.stride);
}
}
@@ -190,7 +190,7 @@
task.update_progress(tile);
}
}
- else if(system_cpu_support_sse3()) {
+ else if(system_cpu_support_sse2()) {
for(int sample = start_sample; sample < end_sample; sample++) {
if (task.get_cancel() || task_pool.cancelled()) {
if(task.need_finish_queue == false)
@@ -199,7 +199,7 @@
for(int y = tile.y; y < tile.y + tile.h; y++) {
for(int x = tile.x; x < tile.x + tile.w; x++) {
- kernel_cpu_sse3_path_trace(&kg, render_buffer, rng_state,
+ kernel_cpu_sse2_path_trace(&kg, render_buffer, rng_state,
sample, x, y, tile.offset, tile.stride);
}
}
@@ -247,16 +247,16 @@
void thread_tonemap(DeviceTask& task)
{
#ifdef WITH_OPTIMIZED_KERNEL
- if(system_cpu_support_sse2()) {
+ if(system_cpu_support_sse3()) {
for(int y = task.y; y < task.y + task.h; y++)
for(int x = task.x; x < task.x + task.w; x++)
- kernel_cpu_sse2_tonemap(&kernel_globals, (uchar4*)task.rgba, (float*)task.buffer,
+ kernel_cpu_sse3_tonemap(&kernel_globals, (uchar4*)task.rgba, (float*)task.buffer,
task.sample, task.resolution, x, y, task.offset, task.stride);
}
- else if(system_cpu_support_sse3()) {
+ else if(system_cpu_support_sse2()) {
for(int y = task.y; y < task.y + task.h; y++)
for(int x = task.x; x < task.x + task.w; x++)
- kernel_cpu_sse3_tonemap(&kernel_globals, (uchar4*)task.rgba, (float*)task.buffer,
+ kernel_cpu_sse2_tonemap(&kernel_globals, (uchar4*)task.rgba, (float*)task.buffer,
task.sample, task.resolution, x, y, task.offset, task.stride);
}
else
@@ -278,17 +278,17 @@
#endif
#ifdef WITH_OPTIMIZED_KERNEL
- if(system_cpu_support_sse2()) {
+ if(system_cpu_support_sse3()) {
for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++) {
- kernel_cpu_sse2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x);
+ kernel_cpu_sse3_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x);
if(task_pool.cancelled())
break;
}
}
- else if(system_cpu_support_sse3()) {
+ else if(system_cpu_support_sse2()) {
for(int x = task.shader_x; x < task.shader_x + task.shader_w; x++) {
- kernel_cpu_sse3_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x);
+ kernel_cpu_sse2_shader(&kg, (uint4*)task.shader_input, (float4*)task.shader_output, task.shader_eval_type, x);
if(task_pool.cancelled())
break;
More information about the Bf-blender-cvs
mailing list