cwpearson/astaroth
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fixed an out-of-bounds error with auto-optimization (introduced in the last few commits)

Browse Source
This commit is contained in:
jpekkila
2019-12-03 16:04:44 +02:00
parent 7e4212ddd9
commit 316d44b843
8 changed files with 84 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/core/CMakeLists.txt
View File

@@ -47,7 +47,7 @@ endif ()
## Create and link the library
#set(CMAKE_POSITION_INDEPENDENT_CODE ON) # fpic for shared libraries
cuda_add_library(astaroth_core STATIC astaroth.cc device.cc node.cc kernels/boundconds.cu kernels/integration.cu kernels/reductions.cu)
cuda_add_library(astaroth_core STATIC astaroth.cc node.cc device.cc kernels/boundconds.cu kernels/integration.cu kernels/reductions.cu)
target_include_directories(astaroth_core PRIVATE . ${CUDA_INCLUDE_DIRS})
target_link_libraries(astaroth_core ${CUDA_LIBRARIES} m)
target_compile_definitions(astaroth_core PRIVATE AC_USE_CUDA_RUNTIME_API)

27
src/core/device.cc
View File

@@ -43,6 +43,7 @@ struct device_s {
AcReal* reduce_scratchpad;
AcReal* reduce_result;
/*
#if AC_MPI_ENABLED
// Declare memory for buffers needed for packed data transfers here
AcReal* inner[2];
@@ -51,6 +52,7 @@ struct device_s {
AcReal* inner_host[2];
AcReal* outer_host[2];
#endif
*/
};
#include "kernels/boundconds.cuh"
@@ -109,12 +111,12 @@ acDeviceCreate(const int id, const AcMeshInfo device_config, Device* device_hand
acVertexBufferCompdomainSizeBytes(device_config)));
ERRCHK_CUDA_ALWAYS(cudaMalloc((void**)&device->reduce_result, sizeof(AcReal)));
/*
#if AC_MPI_ENABLED
// Allocate data required for packed transfers here (cudaMalloc)
const size_t block_size_bytes = device_config.int_params[AC_mx] *
device_config.int_params[AC_my] * NGHOST * NUM_VTXBUF_HANDLES *
sizeof(AcReal);
for (int i = 0; i < 2; ++i) {
device_config.int_params[AC_my] * NGHOST *
NUM_VTXBUF_HANDLES * sizeof(AcReal); for (int i = 0; i < 2; ++i) {
ERRCHK_CUDA_ALWAYS(cudaMalloc((void**)&device->inner[i], block_size_bytes));
ERRCHK_CUDA_ALWAYS(cudaMalloc((void**)&device->outer[i], block_size_bytes));
@@ -122,6 +124,7 @@ acDeviceCreate(const int id, const AcMeshInfo device_config, Device* device_hand
ERRCHK_CUDA_ALWAYS(cudaMallocHost(&device->outer_host[i], block_size_bytes));
}
#endif
*/
// Device constants
acDeviceLoadMeshInfo(device, STREAM_DEFAULT, device_config);
@@ -156,6 +159,7 @@ acDeviceDestroy(Device device)
cudaFree(device->reduce_scratchpad);
cudaFree(device->reduce_result);
/*
#if AC_MPI_ENABLED
// Free data required for packed tranfers here (cudaFree)
for (int i = 0; i < 2; ++i) {
@@ -166,6 +170,7 @@ acDeviceDestroy(Device device)
cudaFreeHost(device->outer_host[i]);
}
#endif
*/
// Concurrency
for (int i = 0; i < NUM_STREAMS; ++i) {
@@ -241,10 +246,16 @@ AcResult
acDeviceAutoOptimize(const Device device)
{
cudaSetDevice(device->id);
const int3 start = (int3){NGHOST, NGHOST, NGHOST};
const int3 end = (int3){device->local_config.int_params[AC_mx], //
device->local_config.int_params[AC_my], //
device->local_config.int_params[AC_mz]};
const int3 start = (int3){
device->local_config.int_params[AC_nx_min],
device->local_config.int_params[AC_ny_min],
device->local_config.int_params[AC_nz_min],
};
const int3 end = (int3){
device->local_config.int_params[AC_nx_max],
device->local_config.int_params[AC_ny_max],
device->local_config.int_params[AC_nz_max],
};
return acKernelAutoOptimizeIntegration(start, end, device->vba);
}
@@ -592,6 +603,8 @@ acDeviceReduceVec(const Device device, const Stream stream, const ReductionType
*/
#include <mpi.h>
#include <assert.h>
static int
mod(const int a, const int b)
{

3
src/core/kernels/boundconds.cu
View File

@@ -24,6 +24,7 @@
* Detailed info.
*
*/
#include "boundconds.cuh"
#include "common.cuh"
#include "src/core/errchk.h"
@@ -76,7 +77,7 @@ kernel_periodic_boundconds(const int3 start, const int3 end, AcReal* vtxbuf)
}
AcResult
acKernelPeriodicBoundconds(const cudaStream_t stream, const int3& start, const int3& end,
acKernelPeriodicBoundconds(const cudaStream_t stream, const int3 start, const int3 end,
AcReal* vtxbuf)
{
const dim3 tpb(8, 2, 8);

10
src/core/kernels/boundconds.cuh
View File

@@ -28,5 +28,13 @@
#include "astaroth.h"
AcResult acKernelPeriodicBoundconds(const cudaStream_t stream, const int3& start, const int3& end,
#ifdef __cplusplus
extern "C" {
#endif
AcResult acKernelPeriodicBoundconds(const cudaStream_t stream, const int3 start, const int3 end,
AcReal* vtxbuf);
#ifdef __cplusplus
} // extern "C"
#endif

3
src/core/kernels/integration.cu
View File

@@ -25,6 +25,7 @@
*
*/
#include "common.cuh"
#include "integration.cuh"
#include "src/core/errchk.h"
#include "src/core/math_utils.h"
@@ -283,7 +284,7 @@ dummy_kernel(void)
a* a;
}
AcReal
AcResult
acKernelDummy(void)
{
dummy_kernel<<<1, 1>>>();

8
src/core/kernels/integration.cuh
View File

@@ -26,9 +26,17 @@
*/
#pragma once
#ifdef __cplusplus
extern "C" {
#endif
AcResult acKernelDummy(void);
AcResult acKernelAutoOptimizeIntegration(const int3 start, const int3 end, VertexBufferArray vba);
AcResult acKernelIntegrateSubstep(const cudaStream_t stream, const int step_number,
const int3 start, const int3 end, VertexBufferArray vba);
#ifdef __cplusplus
} // extern "C"
#endif

9
src/core/kernels/reductions.cu
View File

@@ -25,6 +25,7 @@
*
*/
#include "common.cuh"
#include "reductions.cuh"
#include "src/core/errchk.h"
#include "src/core/math_utils.h" // is_power_of_two
@@ -173,8 +174,8 @@ kernel_reduce_block(const __restrict__ AcReal* scratchpad, const int num_blocks,
}
AcReal
acKernelReduceScal(const cudaStream_t stream, const ReductionType rtype, const int3& start,
const int3& end, const AcReal* vtxbuf, AcReal* scratchpad, AcReal* reduce_result)
acKernelReduceScal(const cudaStream_t stream, const ReductionType rtype, const int3 start,
const int3 end, const AcReal* vtxbuf, AcReal* scratchpad, AcReal* reduce_result)
{
const unsigned nx = end.x - start.x;
const unsigned ny = end.y - start.y;
@@ -227,8 +228,8 @@ acKernelReduceScal(const cudaStream_t stream, const ReductionType rtype, const i
}
AcReal
acKernelReduceVec(const cudaStream_t stream, const ReductionType rtype, const int3& start,
const int3& end, const AcReal* vtxbuf0, const AcReal* vtxbuf1,
acKernelReduceVec(const cudaStream_t stream, const ReductionType rtype, const int3 start,
const int3 end, const AcReal* vtxbuf0, const AcReal* vtxbuf1,
const AcReal* vtxbuf2, AcReal* scratchpad, AcReal* reduce_result)
{
const unsigned nx = end.x - start.x;

16
src/core/kernels/reductions.cuh
View File

@@ -27,10 +27,18 @@
#pragma once
#include <astaroth.h>
AcReal acKernelReduceScal(const cudaStream_t stream, const ReductionType rtype, const int3& start,
const int3& end, const AcReal* vtxbuf, AcReal* scratchpad,
#ifdef __cplusplus
extern "C" {
#endif
AcReal acKernelReduceScal(const cudaStream_t stream, const ReductionType rtype, const int3 start,
const int3 end, const AcReal* vtxbuf, AcReal* scratchpad,
AcReal* reduce_result);
AcReal acKernelReduceVec(const cudaStream_t stream, const ReductionType rtype, const int3& start,
const int3& end, const AcReal* vtxbuf0, const AcReal* vtxbuf1,
AcReal acKernelReduceVec(const cudaStream_t stream, const ReductionType rtype, const int3 start,
const int3 end, const AcReal* vtxbuf0, const AcReal* vtxbuf1,
const AcReal* vtxbuf2, AcReal* scratchpad, AcReal* reduce_result);
#ifdef __cplusplus
} // extern "C"
#endif