Thanks for the report!

As I recently introduced an automated way of building simd version, let's explore the possibility that the build process is incorrect.

I wasn't aware of prefix simd opcode, this would be a great addition to CI to make sure this doesn't happen again, thanks!

incorrect test, in spoiler to avoid spreading misinformation.

The following was an incorrect test, corrected by the following reply #314 (comment).

First things first, let's cover the basis:

After a local build (and a cargo clean), I end up with this result:

wasm-objdump -d ./builds/rapier2d-simd/pkg/rapier_wasm2d_bg.wasm | grep 0xfd

10a67e: fd 0c 40 81 4e fd af 0a 8a | v128.const 0xfd4e8140 0x628a0aaf 0x34d03041 0x2cc06684

This is a bit underwhelming, but does that mean I have a simd instruction in there ?

Other than that, I realize that running ./build_rust.sh in rapier2d-simd then in rapier2d ends up in an error, I suspect the culprit is cargo reusing previously compiled crates without simd instructions. This is a risky error to have, a short term solution would be to cargo clean at the beginning of build_rust.sh scripts. I think I'd prefer abandoning the top level cargo workspace (if we can't trust rustup flags are invalidating cache).

rapier2d-simd-compat doesn't really use simd128 instruction #314

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions