ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-18 03:49:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

Code clean up (#20)

Browse Source 
* tuning para,

* testing on v100

* add fp16

* remove deprecated tensor descriptor

* sync with miopen

* update build script

Co-authored-by: Jing Zhang <jizhan@amd.com>
This commit is contained in:
Chao Liu
2020-06-23 20:31:27 -05:00
committed by GitHub
parent 7d09790a0a
commit 5c7cec1115
94 changed files with 9477 additions and 13782 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
driver/src/device.cpp
View File

@@ -6,7 +6,7 @@ DeviceMem::DeviceMem(std::size_t mem_size) : mMemSize(mem_size)
#if CK_DEVICE_BACKEND_AMD
hipGetErrorString(hipMalloc(static_cast<void**>(&mpDeviceBuf), mMemSize));
#elif CK_DEVICE_BACKEND_NVIDIA
checkCudaErrors(cudaMalloc(static_cast<void**>(&mpDeviceBuf), mMemSize));
cudaMalloc(static_cast<void**>(&mpDeviceBuf), mMemSize);
#endif
}
@@ -18,8 +18,7 @@ void DeviceMem::ToDevice(const void* p)
hipGetErrorString(
hipMemcpy(mpDeviceBuf, const_cast<void*>(p), mMemSize, hipMemcpyHostToDevice));
#elif CK_DEVICE_BACKEND_NVIDIA
checkCudaErrors(
cudaMemcpy(mpDeviceBuf, const_cast<void*>(p), mMemSize, cudaMemcpyHostToDevice));
cudaMemcpy(mpDeviceBuf, const_cast<void*>(p), mMemSize, cudaMemcpyHostToDevice);
#endif
}
@@ -28,7 +27,7 @@ void DeviceMem::FromDevice(void* p)
#if CK_DEVICE_BACKEND_AMD
hipGetErrorString(hipMemcpy(p, mpDeviceBuf, mMemSize, hipMemcpyDeviceToHost));
#elif CK_DEVICE_BACKEND_NVIDIA
checkCudaErrors(cudaMemcpy(p, mpDeviceBuf, mMemSize, cudaMemcpyDeviceToHost));
cudaMemcpy(p, mpDeviceBuf, mMemSize, cudaMemcpyDeviceToHost);
#endif
}
@@ -37,7 +36,7 @@ DeviceMem::~DeviceMem()
#if CK_DEVICE_BACKEND_AMD
hipGetErrorString(hipFree(mpDeviceBuf));
#elif CK_DEVICE_BACKEND_NVIDIA
checkCudaErrors(cudaFree(mpDeviceBuf));
cudaFree(mpDeviceBuf);
#endif
}
@@ -68,8 +67,10 @@ struct KernelTimerImpl
void Start()
{
#if CK_DEVICE_BACKEND_AMD
hipDeviceSynchronize();
hipEventRecord(mStart, 0);
#elif CK_DEVICE_BACKEND_NVIDIA
cudaDeviceSynchronize();
cudaEventRecord(mStart, 0);
#endif
}