[SYCL][CUDA] Support host-device memcpy2D #8181
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
sycl/plugins/cuda/pi_cuda.cpp
Outdated
| ? reinterpret_cast<CUdeviceptr>(src_ptr) | ||
| : 0; | ||
| cpyDesc.srcHost = (src_type == CU_MEMORYTYPE_HOST) ? src_ptr : nullptr; | ||
| } |
There was a problem hiding this comment.
Since this behavior is becoming more complex, would it make sense to split it into a helper function so we can avoid repeating the code? I can see two ways of doing it:
- Implement it as a function returning a triple, i.e. something like:
template<typename PtrT>
static std::tuple<CUmemorytype, CUdeviceptr, PtrT> getUSMHostOrDevicePtr(PtrT usm_ptr) { ... }
...
std::tie(cpyDesc.srcMemoryType, cpyDesc.srcDevice, cpyDesc.srcHost) = getUSMHostOrDevicePtr(src_ptr);
...
std::tie(cpyDesc.dstMemoryType, cpyDesc.dstDevice, cpyDesc.dstHost) = getUSMHostOrDevicePtr(dst_ptr);- Do the same, but make it
voidand pass the members to set as pointers/references:
template<typename PtrT>
static void getUSMHostOrDevicePtr(PtrT usm_ptr, CUmemorytype *out_mem_type, CUdeviceptr *out_dev_ptr, PtrT *out_host_ptr) { ... }
...
getUSMHostOrDevicePtr(src_ptr, &cpyDesc.srcMemoryType, &cpyDesc.srcDevice, &cpyDesc.srcHost);
...
getUSMHostOrDevicePtr(dst_ptr, &cpyDesc.dstMemoryType, &cpyDesc.dstDevice, &cpyDesc.dstHost);Note I use pointers above because I like to think it makes the intention clearer here, but I do not have a strong preference for one or the other.
jchlanda
reviewed
Feb 2, 2023
sycl/plugins/cuda/pi_cuda.cpp
Outdated
| "ARRAY, UNIFIED types are not supported!"); | ||
|
|
||
| // pointer not known to the CUDA subsystem (possibly a system allocated ptr) | ||
| if (ret == CUDA_ERROR_INVALID_VALUE) { |
There was a problem hiding this comment.
Is it safe to check only for two possible error codes from cuPointerGetAttribute?
There was a problem hiding this comment.
Not really. I could classify them into CUDA_ERROR_INVALID_VALUE since that is the only way to check for host-ptrs, CUDA_SUCCESS or not that asserts through the existing macro. Good tip.
|
/verify with intel/llvm-test-suite#1597 |
abagusetty
requested review from
jchlanda
and removed request for
sergey-semenov
steffenlarsen
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Addresses to support host-device memcpy2D copies