Add CUDA Examples/Improve CUDA robustness#202
Open
SousaTrashBin wants to merge 104 commits into
Open
Conversation
alcides
force-pushed
the
add_some_cuda_examples
branch
from
152e24b to
54a8846
Compare
… normal/core ast and gpu ast (not sure if this is the smartest way of doing this)
… before finishing this)
…e GPU AST and llvm instructions
…hings are more isolated and less likely to break
…and LLVM IR generation
… normal/core ast and gpu ast (not sure if this is the smartest way of doing this)
… before finishing this)
…e GPU AST and llvm instructions
…nagement on vectors
…rations (still didn't test if this works on gpu)
…werer When _extract_call_info returns prev_args from a partial application, eff_ty only contains the remaining parameter types. Using eff_ty with offset=len(prev_args) caused incorrect type expectations and premature full-application detection, dropping trailing arguments. This fix uses target.type (the full function type) when prev_args exist, matching what _lower_builtin_call already does. Fixes histogram.ae producing all-zero word counts due to the wordcount call being silently dropped during lowering. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…nodes - Remove stale gpu/llvm imports from decorators/__init__.py - Add missing LLVMVectorGet, LLVMVectorSet, LLVMVectorSize classes to llvm_ast.py - Add find_calls() method to LLVMTerm and all compound term subclasses - Fix target_machine attribute error in CPULLVMExecutionEngine - Update old import "X.ae" syntax to open X in all LLVM examples Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
alcides
force-pushed
the
add_some_cuda_examples
branch
from
6345117 to
2dbc106
Compare
The LLVM test examples use Vector_size which is handled natively by the LLVM backend but was missing from the Vector library, causing KeyError when the evaluator falls back. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…g LLVMVectorSize alias - Remove leftover `print(f"DEBUG: ...")` in lowerer.py - Upgrade silent fallback/disable log messages from debug to warning level - Add `LLVMVectorSize = Any` to runtime else-branch in core.py for consistency Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CPULLVMPipeline was never imported anywhere. MultiBackendPipeline in aeon/llvm/pipeline.py handles CPU-only as its default and is the one used by the driver. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
alcides
requested changes
May 1, 2026
Owner
alcides
left a comment
alcides
left a comment
There was a problem hiding this comment.
A bit of code cleaning would make this good to be merged.
| if ir_hash not in self._module_cache: | ||
| ptx = self._compile_to_ptx(llvm_ir) | ||
| self._module_cache[ir_hash] = self._load_module(ptx) | ||
| with open("debug.ll", "w") as f: |
Owner
There was a problem hiding this comment.
Isto pode ir para uma função auxiliar e estar atrás de uma flag?
Collaborator
Author
There was a problem hiding this comment.
só a parte de escrever para um ficheiro? ou toda a parte de compilação do ptx?
| finally: | ||
| for d_ptr in device_ptrs: | ||
| self.libcuda.cuMemFree_v2(d_ptr) | ||
| if isinstance(ret_type, LLVMPointerType) or "Vector" in str(ret_type): |
Owner
There was a problem hiding this comment.
Em vez de Vector in str(...), devíamos ter uma resolução de nomes para colocar tudo qualified, e comparar com o qualified name.
Collaborator
Author
There was a problem hiding this comment.
qualified ou neste caso unqualified? assumo que a parte do "qualified" seja à esquerda do '.' e à direita seja a parte unqualified, adiciono dois fields aos LLVMTypes, um para a qualified e outro para a unqualified? (e vejo se qualified == "Vector" e unqualified == "Vector")
Collaborator
Author
…er by extracting multiple implementation steps to simpler individual functions
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
the only example that is not working properly is histogram.ae, I'm pretty sure it's cause I have a kernel inside another kernel. LLVM/CPU was not tested and needs to be verified posteriorly.