Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[llvm] Allocate the buffers of the real function in the entry block (#…
…8206) Issue: fix #8203 ### Brief Summary <!-- copilot:summary --> ### <samp>🤖 Generated by Copilot at 4f313de</samp> Refactor the LLVM backend to use a better memory allocation strategy with `create_entry_block_alloca`. This improves performance and correctness of the generated code. ### Walkthrough <!-- copilot:walkthrough --> ### <samp>🤖 Generated by Copilot at 4f313de</samp> * Use `create_entry_block_alloca` function to allocate memory for LLVM variables ([link](https://github.com/taichi-dev/taichi/pull/8206/files?diff=unified&w=0#diff-3c663c78745adcd3f6a7ac81fe99e628decc3040f292ea1e20ecd4b85a7f4313L2784-R2788), [link](https://github.com/taichi-dev/taichi/pull/8206/files?diff=unified&w=0#diff-3c663c78745adcd3f6a7ac81fe99e628decc3040f292ea1e20ecd4b85a7f4313L2795-R2795)). This avoids potential stack size and alignment issues in the LLVM backend. The variables are `new_ctx`, `buffer`, and `result_buffer` in `codegen_llvm.cpp`.
- Loading branch information