AI/rocm-systems
2
0
Tükrözés 0
Kód Hibajegyek Egyesítési kérések Actions Packages Projektek Kiadások Wiki Tevékenység

Added overflow check in kernel launch (#1770)

Forráskód böngészése 
[ROCm/hip commit: 13316f724f]
This commit is contained in:
Sarbojit2019
2020-02-04 09:02:16 +05:30
committed by GitHub
szülő 7ee73c0b5b
commit 91d9cfd64d
2 fájl változott, egészen pontosan 41 új sor hozzáadva és 22 régi sor törölve
Download Patch File Download Diff File Expand all files Collapse all files
projects/hip/docs/markdown/hip_faq.md
+3
Fájl megtekintése
@@ -237,3 +237,6 @@ To disable it and use the legancy grid launch method, please either change the d
$HIP/include/hip/hcc_detail/hip_runtime_api.h $HIP/include/hip/hcc_detail/hip_runtime_api.h
$HIP/include/hip/hcc_detail/host_defines.h $HIP/include/hip/hcc_detail/host_defines.h
Or pass "-DGENERIC_GRID_LAUNCH=0" to hipcc at application compilation time. Or pass "-DGENERIC_GRID_LAUNCH=0" to hipcc at application compilation time.
### What is maximum limit of Generic Grid Launch parameters (grid and block)?
Product of (grid.x and block.x), (grid.y and block.y) or (grid.z and block.z) should always be less than UINT_MAX.
projects/hip/src/hip_module.cpp
+38 -22
Fájl megtekintése
@@ -134,10 +134,10 @@ extern hipError_t ihipGetDeviceProperties(hipDeviceProp_t* props, int device);
return ihipLogStatus(hipStatus); \ return ihipLogStatus(hipStatus); \
} }
hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t globalWorkSizeX, hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t gridSizeX,
uint32_t globalWorkSizeY, uint32_t globalWorkSizeZ, uint32_t gridSizeY, uint32_t gridSizeZ,
uint32_t localWorkSizeX, uint32_t localWorkSizeY, uint32_t blockSizeX, uint32_t blockSizeY,
uint32_t localWorkSizeZ, size_t sharedMemBytes, uint32_t blockSizeZ, size_t sharedMemBytes,
hipStream_t hStream, void** kernelParams, void** extra, hipStream_t hStream, void** kernelParams, void** extra,
hipEvent_t startEvent, hipEvent_t stopEvent, uint32_t flags, bool isStreamLocked = 0, hipEvent_t startEvent, hipEvent_t stopEvent, uint32_t flags, bool isStreamLocked = 0,
void** impCoopParams = 0) { void** impCoopParams = 0) {
@@ -146,6 +146,14 @@ hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t global
auto ctx = ihipGetTlsDefaultCtx(); auto ctx = ihipGetTlsDefaultCtx();
hipError_t ret = hipSuccess; hipError_t ret = hipSuccess;
size_t globalWorkSizeX = (size_t)gridSizeX * (size_t)blockSizeX;
size_t globalWorkSizeY = (size_t)gridSizeY * (size_t)blockSizeY;
size_t globalWorkSizeZ = (size_t)gridSizeZ * (size_t)blockSizeZ;
if(globalWorkSizeX > UINT32_MAX || globalWorkSizeY > UINT32_MAX || globalWorkSizeZ > UINT32_MAX)
{
return hipErrorInvalidConfiguration;
}
if (ctx == nullptr) { if (ctx == nullptr) {
ret = hipErrorInvalidDevice; ret = hipErrorInvalidDevice;
@@ -202,8 +210,8 @@ hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t global
lp.dynamic_group_mem_bytes = lp.dynamic_group_mem_bytes =
sharedMemBytes; // TODO - this should be part of preLaunchKernel. sharedMemBytes; // TODO - this should be part of preLaunchKernel.
hStream = ihipPreLaunchKernel( hStream = ihipPreLaunchKernel(
hStream, dim3(globalWorkSizeX/localWorkSizeX, globalWorkSizeY/localWorkSizeY, globalWorkSizeZ/localWorkSizeZ), hStream, dim3(globalWorkSizeX/blockSizeX, globalWorkSizeY/blockSizeY, globalWorkSizeZ/blockSizeZ),
dim3(localWorkSizeX, localWorkSizeY, localWorkSizeZ), &lp, f->_name.c_str(), isStreamLocked); dim3(blockSizeX, blockSizeY, blockSizeZ), &lp, f->_name.c_str(), isStreamLocked);
hsa_kernel_dispatch_packet_t aql; hsa_kernel_dispatch_packet_t aql;
@@ -212,9 +220,9 @@ hipError_t ihipModuleLaunchKernel(TlsData *tls, hipFunction_t f, uint32_t global
// aql.completion_signal._handle = 0; // aql.completion_signal._handle = 0;
// aql.kernarg_address = 0; // aql.kernarg_address = 0;
aql.workgroup_size_x = localWorkSizeX; aql.workgroup_size_x = blockSizeX;
aql.workgroup_size_y = localWorkSizeY; aql.workgroup_size_y = blockSizeY;
aql.workgroup_size_z = localWorkSizeZ; aql.workgroup_size_z = blockSizeZ;
aql.grid_size_x = globalWorkSizeX; aql.grid_size_x = globalWorkSizeX;
aql.grid_size_y = globalWorkSizeY; aql.grid_size_y = globalWorkSizeY;
aql.grid_size_z = globalWorkSizeZ; aql.grid_size_z = globalWorkSizeZ;
@@ -275,7 +283,7 @@ hipError_t hipModuleLaunchKernel(hipFunction_t f, uint32_t gridDimX, uint32_t gr
HIP_INIT_API(hipModuleLaunchKernel, f, gridDimX, gridDimY, gridDimZ, blockDimX, blockDimY, blockDimZ, sharedMemBytes, HIP_INIT_API(hipModuleLaunchKernel, f, gridDimX, gridDimY, gridDimZ, blockDimX, blockDimY, blockDimZ, sharedMemBytes,
hStream, kernelParams, extra); hStream, kernelParams, extra);
return ihipLogStatus(ihipModuleLaunchKernel(tls, return ihipLogStatus(ihipModuleLaunchKernel(tls,
f, blockDimX * gridDimX, blockDimY * gridDimY, gridDimZ * blockDimZ, blockDimX, blockDimY, f, gridDimX, gridDimY, gridDimZ, blockDimX, blockDimY,
blockDimZ, sharedMemBytes, hStream, kernelParams, extra, nullptr, nullptr, 0)); blockDimZ, sharedMemBytes, hStream, kernelParams, extra, nullptr, nullptr, 0));
} }
@@ -287,8 +295,12 @@ hipError_t hipExtModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
hipEvent_t startEvent, hipEvent_t stopEvent, uint32_t flags) { hipEvent_t startEvent, hipEvent_t stopEvent, uint32_t flags) {
HIP_INIT_API(hipExtModuleLaunchKernel, f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX, HIP_INIT_API(hipExtModuleLaunchKernel, f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX,
localWorkSizeY, localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra); localWorkSizeY, localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra);
if(localWorkSizeX == 0 || localWorkSizeY == 0 || localWorkSizeZ == 0)
return hipErrorInvalidValue;
return ihipLogStatus(ihipModuleLaunchKernel(tls, return ihipLogStatus(ihipModuleLaunchKernel(tls,
f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX, localWorkSizeY, f, globalWorkSizeX/localWorkSizeX, globalWorkSizeY/localWorkSizeY, globalWorkSizeZ/localWorkSizeZ, localWorkSizeX, localWorkSizeY,
localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra, startEvent, stopEvent, flags)); localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra, startEvent, stopEvent, flags));
} }
@@ -300,8 +312,12 @@ hipError_t hipHccModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
hipEvent_t startEvent, hipEvent_t stopEvent) { hipEvent_t startEvent, hipEvent_t stopEvent) {
HIP_INIT_API(hipHccModuleLaunchKernel, f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX, HIP_INIT_API(hipHccModuleLaunchKernel, f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX,
localWorkSizeY, localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra); localWorkSizeY, localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra);
if(localWorkSizeX == 0 || localWorkSizeY == 0 || localWorkSizeZ == 0)
return hipErrorInvalidValue;
return ihipLogStatus(ihipModuleLaunchKernel(tls, return ihipLogStatus(ihipModuleLaunchKernel(tls,
f, globalWorkSizeX, globalWorkSizeY, globalWorkSizeZ, localWorkSizeX, localWorkSizeY, f, globalWorkSizeX/localWorkSizeX, globalWorkSizeY/localWorkSizeY, globalWorkSizeZ/localWorkSizeZ, localWorkSizeX, localWorkSizeY,
localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra, startEvent, stopEvent, 0)); localWorkSizeZ, sharedMemBytes, hStream, kernelParams, extra, startEvent, stopEvent, 0));
} }
@@ -352,9 +368,9 @@ hipError_t ihipExtLaunchMultiKernelMultiDevice(hipLaunchParams* launchParamsList
const hipLaunchParams& lp = launchParamsList[i]; const hipLaunchParams& lp = launchParamsList[i];
result = ihipModuleLaunchKernel(tls, kds[i], result = ihipModuleLaunchKernel(tls, kds[i],
lp.gridDim.x * lp.blockDim.x, lp.gridDim.x,
lp.gridDim.y * lp.blockDim.y, lp.gridDim.y,
lp.gridDim.z * lp.blockDim.z, lp.gridDim.z,
lp.blockDim.x, lp.blockDim.y, lp.blockDim.x, lp.blockDim.y,
lp.blockDim.z, lp.sharedMem, lp.blockDim.z, lp.sharedMem,
lp.stream, lp.args, nullptr, nullptr, nullptr, 0, lp.stream, lp.args, nullptr, nullptr, nullptr, 0,
@@ -458,9 +474,9 @@ hipError_t ihipLaunchCooperativeKernel(const void* f, dim3 gridDim,
// launch the main kernel // launch the main kernel
result = ihipModuleLaunchKernel(tls, kd, result = ihipModuleLaunchKernel(tls, kd,
gridDim.x * blockDimX.x, gridDim.x,
gridDim.y * blockDimX.y, gridDim.y,
gridDim.z * blockDimX.z, gridDim.z,
blockDimX.x, blockDimX.y, blockDimX.z, blockDimX.x, blockDimX.y, blockDimX.z,
sharedMemBytes, stream, kernelParams, nullptr, nullptr, sharedMemBytes, stream, kernelParams, nullptr, nullptr,
nullptr, 0, true, impCoopParams); nullptr, 0, true, impCoopParams);
@@ -612,9 +628,9 @@ hipError_t ihipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsL
impCoopParams[0] = &mg_info_ptr[i]; impCoopParams[0] = &mg_info_ptr[i];
result = ihipModuleLaunchKernel(tls, kds[i], result = ihipModuleLaunchKernel(tls, kds[i],
lp.gridDim.x * lp.blockDim.x, lp.gridDim.x,
lp.gridDim.y * lp.blockDim.y, lp.gridDim.y,
lp.gridDim.z * lp.blockDim.z, lp.gridDim.z,
lp.blockDim.x, lp.blockDim.y, lp.blockDim.x, lp.blockDim.y,
lp.blockDim.z, lp.sharedMem, lp.blockDim.z, lp.sharedMem,
lp.stream, lp.args, nullptr, nullptr, nullptr, 0, lp.stream, lp.args, nullptr, nullptr, nullptr, 0,
@@ -1469,6 +1485,6 @@ hipError_t hipLaunchKernel(
&szKernArg, &szKernArg,
HIP_LAUNCH_PARAM_END}; HIP_LAUNCH_PARAM_END};
return ihipLogStatus(ihipModuleLaunchKernel(tls, kd, numBlocks.x * dimBlocks.x, numBlocks.y * dimBlocks.y, numBlocks.z * dimBlocks.z, return ihipLogStatus(ihipModuleLaunchKernel(tls, kd, numBlocks.x, numBlocks.y, numBlocks.z,
dimBlocks.x, dimBlocks.y, dimBlocks.z, sharedMemBytes, stream, nullptr, (void**)&config, nullptr, nullptr, 0)); dimBlocks.x, dimBlocks.y, dimBlocks.z, sharedMemBytes, stream, nullptr, (void**)&config, nullptr, nullptr, 0));
} }