Scott Todd

Results 78 issues of Scott Todd

Tested on a few programs, seems to still work: simple programs (no push constants) compile to WGSL that looks reasonable and complex programs (push constants) fail with errors like this:...

platform/web ๐ŸŒ

We can view executable sources for dispatch functions in Tracy thanks to https://github.com/openxla/iree/pull/9994, https://github.com/openxla/iree/pull/13333, and https://github.com/openxla/iree/pull/13500. * The definition of "executable sources" in a multi-level compiler can have multiple values....

performance โšก
platform/android ๐Ÿค–
quality of life ๐Ÿ˜Š

https://github.com/openxla/iree/pull/16622 fixes an issue that should have been caught by presubmit tests. We have a test but it has a soft failure mode if `onnx` is not installed: https://github.com/openxla/iree/blob/4c97ab6800820392589ce440f5b6b98f9bd855da/compiler/bindings/python/test/tools/import_onnx_test.py#L51-L56 I...

infrastructure
integrations/pytorch
integrations/onnx

## Overview Let's add more test coverage for full programs, individual ops/layers, and other system-level behavior coming from frameworks like PyTorch, JAX, TF/TFLite, ONNX, etc. ## "SHARK-TestSuite/e2eshark" https://github.com/nod-ai/SHARK-TestSuite/tree/main/e2eshark has a...

enhancement โž•
infrastructure
integrations

When diagnosing performance regressions it can be useful to run a bisect process to identify culprit commits. Git has a built in bisect tool: https://git-scm.com/docs/git-bisect, but bisecting across commits requires...

enhancement โž•
infrastructure
quality of life ๐Ÿ˜Š

Following [this discussion on Discord](https://discord.com/channels/689900678990135345/689957613152239638/1207827386058547220). These jobs in [the `ci.yml` workflow](https://github.com/openxla/iree/blob/main/.github/workflows/ci.yml) are increasing infrastructure complexity and are a bottleneck on total presubmit time: * `build_benchmark_tools` * `build_e2e_test_artifacts` * `test_benchmark_suites` *...

infrastructure
cleanup ๐Ÿงน
infrastructure/benchmark

This also refactors the job to use a GitHub-managed runner with less random GitHub Actions boilerplate for caching, env vars, etc. since the 'tracing' job is only building the runtime...

infrastructure

This was trickier that I wanted it to be... adding some screenshots and notes about settings to apply for myself/others. Progress on https://github.com/openxla/iree/issues/16372 (tidying up `regression_suite` while I evaluating extending...

documentation
infrastructure
bindings/python

## Current tasks ```[tasklist] ### Tasks - [x] Refactor CMake project so all "integration tests", even those using `lit`, can be run using packages instead of source builds (`-DIREE_BUILD_COMPILER`) -...

infrastructure

I spotted some low hanging fruit here: https://discord.com/channels/689900678990135345/1166024193599615006/1207085419959947294 and here: https://groups.google.com/g/iree-discuss/c/uy0L4Vdl3hs/m/YLe0iLCGAAAJ. * The `process_benchmark_results` step takes around 2 minutes to download a mysterious `iree-oss/benchmark-report` Dockerfile. The generation scripts only need...

infrastructure
cleanup ๐Ÿงน
infrastructure/benchmark