appujee
appujee
Pipeline cost model for vector instructions: [D149495](https://reviews.llvm.org/D149495) posted by [michaelmaitland](https://reviews.llvm.org/p/michaelmaitland/). We can use `-mcpu=sifive-x280` to try it out.
yeah, that's a good idea. if lld could incorporate these features, it'll make working with orderfile a lot easier. Currently there is no standard way to generate orderfile. the instrumentation...
aosp testing in progress: https://android-review.googlesource.com/c/platform/build/soong/+/2734373
i think it is okay to keep it open as the currently used (backend) flags will be updated (to -munaligned-access) with the new toolchain.
yes, `-munaligned-access` flag works with the latest clang update.
I'll replace the previous workaround with -munaligned-access
The change landed few weeks back, https://android-review.googlesource.com/c/platform/build/soong/+/2977952 I forgot to close this bug.
Interestingly `-munaligned-access` was replaced by `-mno-strict-align` in llvm upstream a few days back. Both flags were aliases. Pirama has change to that effect https://android-review.googlesource.com/c/platform/build/soong/+/3002035
I'll enable the flag by default for Android. https://github.com/llvm/llvm-project/pull/85704 WIP.