Skip to content

merge CUDA and ROCm codes in op #2847

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
wanghan-iapcm merged 25 commits into from
Sep 26, 2023
Merged

merge CUDA and ROCm codes in op #2847

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
25 commits
Select commit Hold shift + click to select a range
2fa777f
use the same macro between CUDA and ROCm
njzjz Sep 19, 2023
3f8d05d
checkpoint
njzjz Sep 19, 2023
6ee416e
gpuMemset
njzjz Sep 19, 2023
79b31ca
checkpoint
njzjz Sep 19, 2023
972919f
merge locate_xx_se_a and locate_xx_se_r: the same function
njzjz Sep 19, 2023
797dad2
merge cmake files
njzjz Sep 19, 2023
6ef697f
renaming
njzjz Sep 19, 2023
c32407c
Revert "merge locate_xx_se_a and locate_xx_se_r: the same function"
njzjz Sep 19, 2023
a1cdc20
fix typo
njzjz Sep 19, 2023
f7bfadb
fix trap
njzjz Sep 19, 2023
245964c
typo
njzjz Sep 19, 2023
15f5729
fix tabulate_fusion_se_a_fifth_order_polynomial for rocm
njzjz Sep 19, 2023
f36ecc6
use GPU_DYNAMIC_SHARED_MEM_DECL
njzjz Sep 19, 2023
074bbdc
clean never used shared memory decalare
njzjz Sep 19, 2023
4edce5f
fix submodule
njzjz Sep 19, 2023
30b61e7
merge CUDA and ROCm codes in op
njzjz Sep 20, 2023
28dc52d
address comments
njzjz Sep 20, 2023
1f2e049
add GpuSyncThreads
njzjz Sep 20, 2023
e3edda1
Merge branch 'devel' into merge-cuda-rocm
njzjz Sep 21, 2023
06c3eb6
fix offset in warp_reduce
njzjz Sep 21, 2023
2049ac4
Merge remote-tracking branch 'origin/devel' into merge-cuda-rocm-op
njzjz Sep 21, 2023
7313d73
cuda -> gpu
njzjz Sep 21, 2023
357417f
Merge branch 'merge-cuda-rocm' into merge-cuda-rocm-op
njzjz Sep 21, 2023
b2ee402
define gpuMemcpyDeviceToDevice
njzjz Sep 21, 2023
b627c86
Merge branch 'devel' into merge-cuda-rocm-op
njzjz Sep 22, 2023
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
1 change: 1 addition & 0 deletions source/lib/include/gpu_cuda.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
#define gpuMemcpy cudaMemcpy
#define gpuMemcpyDeviceToHost cudaMemcpyDeviceToHost
#define gpuMemcpyHostToDevice cudaMemcpyHostToDevice
#define gpuMemcpyDeviceToDevice cudaMemcpyDeviceToDevice
#define gpuMemset cudaMemset

#define GPU_MAX_NBOR_SIZE 4096
Expand Down
1 change: 1 addition & 0 deletions source/lib/include/gpu_rocm.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@
#define gpuMemcpy hipMemcpy
#define gpuMemcpyDeviceToHost hipMemcpyDeviceToHost
#define gpuMemcpyHostToDevice hipMemcpyHostToDevice
#define gpuMemcpyDeviceToDevice hipMemcpyDeviceToDevice
#define gpuMemset hipMemset

#define DPErrcheck(res) \
Expand Down
24 changes: 6 additions & 18 deletions source/op/gelu_multi_device.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,13 +64,9 @@ class GeluOp : public OpKernel {
const int_64 size = static_cast<int_64>(output_tensor->NumElements());

if (device == "GPU") {
#if GOOGLE_CUDA
deepmd::gelu_gpu(out, x, size);
#endif // GOOGLE_CUDA

#if TENSORFLOW_USE_ROCM
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
deepmd::gelu_gpu(out, x, size);
#endif // TENSORFLOW_USE_ROCM
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
} else if (device == "CPU") {
deepmd::gelu_cpu(out, x, size);
}
Expand Down Expand Up @@ -108,13 +104,9 @@ class GeluGradOp : public OpKernel {
const int_64 size = static_cast<int_64>(output_tensor->NumElements());

if (device == "GPU") {
#if GOOGLE_CUDA
deepmd::gelu_grad_gpu(out, x, dy, size);
#endif // GOOGLE_CUDA

#if TENSORFLOW_USE_ROCM
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
deepmd::gelu_grad_gpu(out, x, dy, size);
#endif // TENSORFLOW_USE_ROCM
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
} else if (device == "CPU") {
deepmd::gelu_grad_cpu(out, x, dy, size);
}
Expand Down Expand Up @@ -154,13 +146,9 @@ class GeluGradGradOp : public OpKernel {
const int_64 size = static_cast<int_64>(output_tensor->NumElements());

if (device == "GPU") {
#if GOOGLE_CUDA
deepmd::gelu_grad_grad_gpu(out, x, dy, dy_2, size);
#endif // GOOGLE_CUDA

#if TENSORFLOW_USE_ROCM
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
deepmd::gelu_grad_grad_gpu(out, x, dy, dy_2, size);
#endif // TENSORFLOW_USE_ROCM
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
} else if (device == "CPU") {
deepmd::gelu_grad_grad_cpu(out, x, dy, dy_2, size);
}
Expand Down
Loading