blender
/
blender
134
424
Fork 606
Code Issues 6.3k Pull Requests 769 Projects 19 Wiki Activity

Cycles / CUDA: Increase sm_2x registers to 40. 

This fixes the ptaxs "ACCESS_VIOLATION" error and should allow our Linux and Windows build bots to compile again.
Unfortunately this comes with a performance penalty on sm_2x cards, so this is only a workaround for now. Branched Path is still globally disabled on GPU.
This commit is contained in:
Thomas Dinges 2014-04-08 23:25:54 +02:00
parent 5580afb5df
commit 297a2223b5
3 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
intern/cycles/device/device_cuda.cpp
View File

@ -320,7 +320,7 @@ public:
/* CUDA 5.x build flags for different archs */
if(major == 2) {
/* sm_2x */
arch_flags = "--maxrregcount=32 --use_fast_math";
arch_flags = "--maxrregcount=40 --use_fast_math";
}
else if(major == 3) {
/* sm_3x */

2
intern/cycles/kernel/CMakeLists.txt
View File

@ -165,7 +165,7 @@ if(WITH_CYCLES_CUDA_BINARIES)
# CUDA 5.x build flags for different archs
if(${arch} MATCHES "sm_2[0-9]")
# sm_2x
set(cuda_arch_flags "--maxrregcount=32")
set(cuda_arch_flags "--maxrregcount=40")
elseif(${arch} MATCHES "sm_3[0-9]")
# sm_3x
set(cuda_arch_flags "--maxrregcount=32")

2
intern/cycles/kernel/SConscript
View File

@ -90,7 +90,7 @@ if env['WITH_BF_CYCLES_CUDA_BINARIES']:
# CUDA 5.x build flags for different archs
if arch.startswith("sm_2"):
# sm_2x
cuda_arch_flags = "--maxrregcount=32 --use_fast_math"
cuda_arch_flags = "--maxrregcount=40 --use_fast_math"
elif arch.startswith("sm_3"):
# sm_3x
cuda_arch_flags = "--maxrregcount=32 --use_fast_math"