Alex Reinking
Alex Reinking
The failures in these workflows on CI match what I observed locally while testing #1969. * Opened #1972 for an s390x failure discovered by this testing. * Opened #1973 for...
When running the full test suite on i386, many floating-point-related failures occur. The test summary is reproduced below. ``` **** FAILED ****************************************************************** - test/spec/float_exprs.txt /ws/build/spectest-interp out/test/spec/float_exprs/float_exprs.json - test/spec/float_misc.txt /ws/build/spectest-interp out/test/spec/float_misc/float_misc.json...
When running the unittests on an s390x Docker image, the Rot13 test fails. ``` [ RUN ] InterpTest.Rot13 ../src/test-interp.cc:546: Failure Expected equality of these values: "Uryyb, JroNffrzoyl!" string_data Which is:...
This PR consists of two tightly related changes. It depends on #1968 ### 1. Fix x87 math detection `TARGET_ARCH` was only used to determine whether to add gcc-specific SSE math...
_tl;dr_ the obvious way to use Halide via CMake (`find_package(Halide)`) does not lead to easy cross-compiling builds. However, fixing this is almost certainly a breaking change. --- If you run...
The de-facto standard tool for documenting Python projects is Sphinx, which unfortunately expects to scan Python sources as its default workflow. However, Sphinx is _very_ powerful and can generate documentation...
This is in-progress, but I need an open PR to test the workflow. Look away!
WIP -- still to-do is to audit our packaging rules (and maybe add some GHA tests, hmm) and update the READMEs. This PR replaces all usage of `FetchContent` inside our...
Since I'm not at Google anymore, I have to work from an external fork... this is going to the internal branch to keep the PR number stable when it's eventually...
Needs to be rebased on main still.