aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

+1

Browse Source
This commit is contained in:
Evghenii
2014-01-05 11:04:10 +01:00
parent 2f01af0595
commit c429e930b2
2 changed files with 11 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples_ptx/common_gpu.mk
View File

@@ -5,7 +5,7 @@ CXX=g++
CXXFLAGS=-O3 -I$(CUDATK)/include -Iobjs_gpu/ -D_CUDA_ CXXFLAGS=-O3 -I$(CUDATK)/include -Iobjs_gpu/ -D_CUDA_
# #
NVCC=nvcc NVCC=nvcc
NVCC_FLAGS=-O3 -arch=sm_35 -D_CUDA_ NVCC_FLAGS=-O3 -arch=sm_35 -D_CUDA_ -I../
ifdef PTXCC_REGMAX ifdef PTXCC_REGMAX
NVCC_FLAGS += --maxrregcount=$(PTXCC_REGMAX) NVCC_FLAGS += --maxrregcount=$(PTXCC_REGMAX)
endif endif

10
examples_ptx/cuda_helpers.cuh Normal file
View File

@@ -0,0 +1,10 @@
#pragma once
#define programCount 32
#define programIndex (threadIdx.x & 31)
#define taskIndex0 (blockIdx.x*4 + (threadIdx.x >> 5))
#define taskCount0 (gridDim.x*4)
#define taskIndex1 (blockIdx.y)
#define taskCount1 (gridDim.y)
#define warpIdx (threadIdx.x >> 5)
#define launch(ntx,nty,ntz,func) if (programIndex==0) func<<<dim3(((ntx)+4-1)/4,nty,ntz),128>>>