momchil-velikov wrote:

> Can you please add phase ordering test(s) that are representative for the 
> cases you _do_ want to address with this patch? That would make it clearer 
> what the motivation relative to the existing hoisting support is.

I don't have specific cases right now, this is a continuation of the previous 
GVN-based hosting of scalars and memory accesses patches, as generally useful.

I've seen about 2.5% improvement in SPEC2026 `708.sqlite_r` by using GVNHoist 
pass with `--gvn-hoist-max-bbs=1`, but I have not yet checked if it reproduces 
with this patch.

https://github.com/llvm/llvm-project/pull/210337
_______________________________________________
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

Reply via email to