AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Apply .clangformat to all repo source files

Browse Source 
Change-Id: I7e79c6058f0303f9a98911e3b7dd2e8596079344


[ROCm/clr commit: 9e47fccc89]
This commit is contained in:
Maneesh Gupta
2018-03-12 11:29:03 +05:30
parent ecbb701440
commit 46ddefedee
293 changed files with 43980 additions and 45830 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/clr/hipamd/tests/src/deviceLib/hip_ballot.cpp
+47 -39
View File
@@ -28,60 +28,68 @@ THE SOFTWARE.
#include <hip/hip_runtime.h>
#include <hip/device_functions.h>
#define HIP_ASSERT(x) (assert((x)==hipSuccess))
#define HIP_ASSERT(x) (assert((x) == hipSuccess))
__global__ void
gpu_ballot(hipLaunchParm lp, unsigned int* device_ballot, int Num_Warps_per_Block,int pshift)
{
int tid = threadIdx.x + blockIdx.x * blockDim.x;
const unsigned int warp_num = threadIdx.x >> pshift;
__global__ void gpu_ballot(hipLaunchParm lp, unsigned int* device_ballot, int Num_Warps_per_Block,
int pshift) {
int tid = threadIdx.x + blockIdx.x * blockDim.x;
const unsigned int warp_num = threadIdx.x >> pshift;
#ifdef __HIP_PLATFORM_HCC__
atomicAdd(&device_ballot[warp_num+blockIdx.x*Num_Warps_per_Block],__popcll(__ballot(tid - 245)));
atomicAdd(&device_ballot[warp_num + blockIdx.x * Num_Warps_per_Block],
__popcll(__ballot(tid - 245)));
#else
atomicAdd(&device_ballot[warp_num+blockIdx.x*Num_Warps_per_Block],__popc(__ballot(tid - 245)));
atomicAdd(&device_ballot[warp_num + blockIdx.x * Num_Warps_per_Block],
__popc(__ballot(tid - 245)));
#endif
}
int main(int argc, char *argv[])
{ int warpSize, pshift;
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
int main(int argc, char* argv[]) {
int warpSize, pshift;
hipDeviceProp_t devProp;
hipGetDeviceProperties(&devProp, 0);
warpSize = devProp.warpSize;
warpSize = devProp.warpSize;
int w = warpSize;
pshift = 0;
while (w >>= 1) ++pshift;
int w = warpSize;
pshift = 0;
while (w >>= 1) ++pshift;
unsigned int Num_Threads_per_Block = 512;
unsigned int Num_Blocks_per_Grid = 1;
unsigned int Num_Warps_per_Block = Num_Threads_per_Block/warpSize;
unsigned int Num_Warps_per_Grid = (Num_Threads_per_Block*Num_Blocks_per_Grid)/warpSize;
unsigned int* host_ballot = (unsigned int*)malloc(Num_Warps_per_Grid*sizeof(unsigned int));
unsigned int* device_ballot;
HIP_ASSERT(hipMalloc((void**)&device_ballot, Num_Warps_per_Grid*sizeof(unsigned int)));
int divergent_count =0;
for (int i=0; i<Num_Warps_per_Grid; i++) host_ballot[i] = 0;
unsigned int Num_Threads_per_Block = 512;
unsigned int Num_Blocks_per_Grid = 1;
unsigned int Num_Warps_per_Block = Num_Threads_per_Block / warpSize;
unsigned int Num_Warps_per_Grid = (Num_Threads_per_Block * Num_Blocks_per_Grid) / warpSize;
unsigned int* host_ballot = (unsigned int*)malloc(Num_Warps_per_Grid * sizeof(unsigned int));
unsigned int* device_ballot;
HIP_ASSERT(hipMalloc((void**)&device_ballot, Num_Warps_per_Grid * sizeof(unsigned int)));
int divergent_count = 0;
for (int i = 0; i < Num_Warps_per_Grid; i++) host_ballot[i] = 0;
HIP_ASSERT(hipMemcpy(device_ballot, host_ballot, Num_Warps_per_Grid*sizeof(unsigned int), hipMemcpyHostToDevice));
HIP_ASSERT(hipMemcpy(device_ballot, host_ballot, Num_Warps_per_Grid * sizeof(unsigned int),
hipMemcpyHostToDevice));
hipLaunchKernel(gpu_ballot, dim3(Num_Blocks_per_Grid),dim3(Num_Threads_per_Block),0,0, device_ballot,Num_Warps_per_Block,pshift);
hipLaunchKernel(gpu_ballot, dim3(Num_Blocks_per_Grid), dim3(Num_Threads_per_Block), 0, 0,
device_ballot, Num_Warps_per_Block, pshift);
HIP_ASSERT(hipMemcpy(host_ballot, device_ballot, Num_Warps_per_Grid*sizeof(unsigned int), hipMemcpyDeviceToHost));
for (int i=0; i<Num_Warps_per_Grid; i++) {
HIP_ASSERT(hipMemcpy(host_ballot, device_ballot, Num_Warps_per_Grid * sizeof(unsigned int),
hipMemcpyDeviceToHost));
for (int i = 0; i < Num_Warps_per_Grid; i++) {
if ((host_ballot[i] == 0) || (host_ballot[i] / warpSize == warpSize))
std::cout << "Warp " << i << " IS convergent- Predicate true for "
<< host_ballot[i] / warpSize << " threads\n";
if ((host_ballot[i] == 0)||(host_ballot[i]/warpSize == warpSize)) std::cout << "Warp " << i << " IS convergent- Predicate true for " << host_ballot[i]/warpSize << " threads\n";
else {std::cout << " Warp " << i << " IS divergent - Predicate true for " << host_ballot[i]/warpSize<< " threads\n";
divergent_count++;}
}
if (divergent_count==1) printf("PASSED\n"); else printf("FAILED\n");
return EXIT_SUCCESS;
else {
std::cout << " Warp " << i << " IS divergent - Predicate true for "
<< host_ballot[i] / warpSize << " threads\n";
divergent_count++;
}
}
if (divergent_count == 1)
printf("PASSED\n");
else
printf("FAILED\n");
return EXIT_SUCCESS;
}