Skip to content

HIP update #643

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jeffhammond merged 2 commits into from
Oct 29, 2024
Merged

HIP update #643

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2325705
sync the HIP version with the CUDA version
jeffhammond Oct 29, 2024
2db9cce
fix
jeffhammond Oct 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 1 addition & 3 deletions Cxx11/nstream-managed-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,7 @@
///
/// Copyright (c) 2017, Intel Corporation
/// Copyright (c) 2021, NVIDIA
/// Copyright (c) 2024, NVIDIA
///

// Copyright (c) 2021, NVIDIA
/// Redistribution and use in source and binary forms, with or without
/// modification, are permitted provided that the following conditions
/// are met:
Expand Down
48 changes: 36 additions & 12 deletions Cxx11/nstream-managed-hip.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
///
/// Copyright (c) 2017, Intel Corporation
/// Copyright (c) 2024, NVIDIA
///
/// Redistribution and use in source and binary forms, with or without
/// modification, are permitted provided that the following conditions
Expand Down Expand Up @@ -104,10 +105,10 @@ int main(int argc, char * argv[])

int iterations;
int length;
bool system_memory;
bool system_memory, grid_stride, ordered_fault, prefetch;
try {
if (argc < 3) {
throw "Usage: <# iterations> <vector length> [<use_system_memory>]";
throw "Usage: <# iterations> <vector length> [<use_system_memory> <grid_stride> <ordered_fault> <prefetch>]";
}

iterations = std::atoi(argv[1]);
Expand All @@ -121,6 +122,9 @@ int main(int argc, char * argv[])
}

system_memory = (argc>3) ? prk::parse_boolean(std::string(argv[3])) : false;
grid_stride = (argc>4) ? prk::parse_boolean(std::string(argv[4])) : false;
ordered_fault = (argc>5) ? prk::parse_boolean(std::string(argv[5])) : false;
prefetch = (argc>6) ? prk::parse_boolean(std::string(argv[6])) : false;
}
catch (const char * e) {
std::cout << e << std::endl;
Expand All @@ -130,6 +134,9 @@ int main(int argc, char * argv[])
std::cout << "Number of iterations = " << iterations << std::endl;
std::cout << "Vector length = " << length << std::endl;
std::cout << "Memory allocator = " << (system_memory ? "system (malloc)" : "hipMallocManaged") << std::endl;
std::cout << "Grid stride = " << (grid_stride ? "yes" : "no") << std::endl;
std::cout << "Ordered fault = " << (ordered_fault ? "yes" : "no") << std::endl;
std::cout << "Prefetch = " << (prefetch ? "yes" : "no") << std::endl;

const int blockSize = 256;
dim3 dimBlock(blockSize, 1, 1);
Expand All @@ -147,20 +154,18 @@ int main(int argc, char * argv[])
prk_float * B;
prk_float * C;

const size_t bytes = length * sizeof(prk_float);
if (system_memory) {
A = new double[length];
B = new double[length];
C = new double[length];
} else {

int managed_memory = 0;
prk::HIP::check( hipDeviceGetAttribute(&managed_memory, hipDeviceAttributeManagedMemory, 0) );
std::cout << "hipDeviceGetAttribute(..hipDeviceAttributeManagedMemory..) => " << managed_memory << std::endl;

prk::HIP::check( hipMallocManaged((void**)&A, bytes) );
prk::HIP::check( hipMallocManaged((void**)&B, bytes) );
prk::HIP::check( hipMallocManaged((void**)&C, bytes) );
A = prk::HIP::malloc_managed<double>(length);
B = prk::HIP::malloc_managed<double>(length);
C = prk::HIP::malloc_managed<double>(length);
}

// initialize on CPU to ensure pages are faulted there
Expand All @@ -170,12 +175,31 @@ int main(int argc, char * argv[])
C[i] = static_cast<prk_float>(2);
}

if (ordered_fault) {
fault_pages<<<1,1>>>(static_cast<unsigned>(length), A, B, C);
prk::HIP::sync();
}

if (prefetch) {
prk::HIP::prefetch(A, length);
prk::HIP::prefetch(B, length);
prk::HIP::prefetch(C, length);
}

prk_float scalar(3);
{
for (int iter = 0; iter<=iterations; iter++) {

if (iter==1) nstream_time = prk::wtime();
hipLaunchKernelGGL(nstream, dim3(dimGrid), dim3(dimBlock), 0, 0, static_cast<unsigned>(length), scalar, A, B, C);
if (iter==1) {
prk::HIP::sync();
nstream_time = prk::wtime();
}

if (grid_stride) {
hipLaunchKernelGGL(nstream2, dim3(dimGrid), dim3(dimBlock), 0, 0, static_cast<unsigned>(length), scalar, A, B, C);
} else {
hipLaunchKernelGGL(nstream, dim3(dimGrid), dim3(dimBlock), 0, 0, static_cast<unsigned>(length), scalar, A, B, C);
}
prk::HIP::check( hipDeviceSynchronize() );
}
nstream_time = prk::wtime() - nstream_time;
Expand Down Expand Up @@ -203,9 +227,9 @@ int main(int argc, char * argv[])
free(B);
free(C);
} else {
prk::HIP::check( hipFree(A) );
prk::HIP::check( hipFree(B) );
prk::HIP::check( hipFree(C) );
prk::HIP::free(A);
prk::HIP::free(B);
prk::HIP::free(C);
}

double epsilon=1.e-8;
Expand Down