Add riscv64 and Imagination Technologies GPUs support #232
+61
−17
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
vkutuev
force-pushed
the
vkutuev/riscv-img-povervr
branch
from
85f336b to
6621830
Compare
| m_vendor_name.find("img") != std::string::npos || | ||
| m_vendor_id == 0x1010) { | ||
| m_vendor_code = VENDOR_CODE_IMG; | ||
| m_default_wgs = 32; |
There was a problem hiding this comment.
Why you decided to set default wgs = 32, and not 64? It seems that 64 suits in most cases well
There was a problem hiding this comment.
RECOMMENDATION: If possible use a workgroup size of 32.
This is a recommendation from the official Imagination Technologies blog
gsvgit
reviewed
May 17, 2025
| @@ -138,7 +143,12 @@ namespace spla { | |||
| auto* cl_coo = s.template get<CLCooVec<T>>(); | |||
| auto* cpu_coo = s.template get<CpuCooVec<T>>(); | |||
| cpu_coo_vec_resize(cl_coo->values, *cpu_coo); | |||
| cl_coo_vec_read(cl_coo->values, cpu_coo->Ai.data(), cpu_coo->Ax.data(), *cl_coo, cl_acc->get_queue_default()); | |||
| if (!cl_acc->is_img()) { | |||
There was a problem hiding this comment.
I guess, comments with explanation and references to related documents are necessary here.
| @@ -146,7 +146,12 @@ namespace spla { | |||
| auto* cl_csr = s.template get<CLCsr<T>>(); | |||
| auto* cpu_csr = s.template get<CpuCsr<T>>(); | |||
| cpu_csr_resize(s.get_n_rows(), cl_csr->values, *cpu_csr); | |||
| cl_csr_read(s.get_n_rows(), cl_csr->values, cpu_csr->Ap.data(), cpu_csr->Aj.data(), cpu_csr->Ax.data(), *cl_csr, cl_acc->get_queue_default()); | |||
| if (!cl_acc->is_img()) { | |||
There was a problem hiding this comment.
I guess, comments with explanation and references to related documents are necessary here.
| @@ -126,7 +126,12 @@ namespace spla { | |||
| auto* cl_acc = get_acc_cl(); | |||
| auto* cl_dense = s.template get<CLDenseVec<T>>(); | |||
| auto* cpu_dense = s.template get<CpuDenseVec<T>>(); | |||
| cl_dense_vec_read(s.get_n_rows(), cpu_dense->Ax.data(), *cl_dense, cl_acc->get_queue_default()); | |||
| if (!cl_acc->is_img()) { | |||
There was a problem hiding this comment.
I guess, comments with explanation and references to related documents are necessary here.
gsvgit
approved these changes
May 19, 2025
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.
Closes #231
Problem
In src/opencl/cl_format_*.hpp, copying data to staging buffer created with
CL_MEM_READ_ONLYflag does not affect this buffer.Workaround
Remove this flag if device vendor is Imagination Technologies.
Testing
Solution was tested on
All tests were passed