Rohanjames1997 comments

Repositories
Issues
Comments

Results 24 comments of


                                            Rohanjames1997

[Inductor] Add support for NEON ISA in the Inductor C++ backend

To witness the perf gains of this PR, follow these simple steps: 1. Spawn an aarch64 instance (I tested mostly on Graviton3 instances). 2. Checkout PyTorch to the main branch...

[Inductor] Add support for NEON ISA in the Inductor C++ backend

@pytorchmergebot merge -i

[Inductor] Add support for NEON ISA in the Inductor C++ backend

Refer to #123584

Stackcollapse issues with output from bcc's mallocstacks.py

Has anyone found a workaround for this? Edit: It seems like adding a print statement "fixes" the issue, and stackcollapse now recognizes the output format https://github.com/iovisor/bcc/blob/472f9b3bd6bcc63e6e959148320eb76184c51870/examples/tracing/mallocstacks.py#L68-L72 ➡️ ``` for k,...