The scalar zero + splat yields more intermediate code than the direct
dense zero constant, and ultimately is lowered to exactly the same
LLVM IR operations, so no point wasting the intermediate code.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
No. I was a bit surprised too I did not even have to change even one, but all tests looked at a later lowered stage, which is identical.