[PATCH] D118356: [DAGCombiner] Limit scaling of repeated FP divisor by splat factor

[David Sherwood via Phabricator via llvm-commits]

david-arm added inline comments.


================
Comment at: llvm/test/CodeGen/AArch64/fdiv-combine.ll:199
-; CHECK-NEXT:    fmul z0.d, z1.d, z3.d
-; CHECK-NEXT:    fmul z1.d, z2.d, z3.d
 ; CHECK-NEXT:    b foo_2_nxv2f64
----------------
To be honest, the original code looks faster to me even with the extra fmul and fmov. The latency of a fmul is a lot lower than fdiv and the throughput for fdiv is terrible, whereas it's pretty good for fmul. 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118356/new/

https://reviews.llvm.org/D118356