This repository has been archived by the owner on Dec 24, 2024. It is now read-only.
Fixes to work with latest driver and userspace BO sync #36
Conversation
This file contains 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
* Instead of using handles in the packet and in the configure hardware context sideband, we are instead passing virtual addresses. This removes some sidebands to get the handle from a virtual address and simplfies the application. The ROCr runtime has access to both so can always do the translation. * This is modifying the ioctl interface to xdna to work with the latest driver, which includes changing the heap size, passing the size of a buffer into the sync, and some small misc changes. This commit leaves one TODO which is having the user specify the sizes of operands so we know the size. This is an intermediate step to sync the BOs in userspace to avoid the syscall.
fifield
reviewed
Oct 31, 2024
| uint64_t base_addr = reinterpret_cast<uint64_t>(base) + offset; | ||
| for (int i = 0; i < len / cacheline_size; i++) { | ||
| uint64_t cur_addr = base_addr + cacheline_size * i; | ||
| _mm_clflush(reinterpret_cast<void *>(cur_addr)); |
There was a problem hiding this comment.
This failed for me w/ gcc13 with _mm_clflush undeclared. Is it in x86intrin.h?
There was a problem hiding this comment.
This failed for me w/ gcc13 with
_mm_clflushundeclared. Is it inx86intrin.h?
It built fine when I added my own declaration.
The correct fix on my system was to move #include <x86intrin.h> outside of the namespace rocr block in runtime/hsa-runtime/core/util/utils.h
| __u64 vaddr; | ||
| __u64 size; | ||
| __u32 type; |
There was a problem hiding this comment.
This struct field ordering change breaks other code:
/work/acdc/ROCR-Runtime/runtime/hsa-runtime/core/driver/xdna/amd_xdna_driver.cpp:300:61: error: designator order for field ‘amdxdna_drm_create_bo::vaddr’ does not match declaration order in ‘amdxdna_drm_create_bo’
300 | .size = dev_heap_size};
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
This PR includes:
get_handle_from_vaddr()sideband and instead only pass addresses from the application.