[PATCH] D144010: [X86] AMD Znver4 (Genoa) Scheduler enablement

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/lib/Target/X86/X86ScheduleZnver4.td:905
+defm : Zn4WriteResYMMPair<WriteFAddY, [Zn4FPFAdd01], 3, [1], 1>; // Floating point add/sub (YMM).
+defm : Zn4WriteResZMMPair<WriteFAddZ, [Zn4FPFAdd01], 6, [1], 1>; // Floating point add/sub (ZMM).
+defm : Zn4WriteResXMMPair<WriteFAdd64, [Zn4FPFAdd01], 3, [1], 1>;  // Floating point double add/sub.
----------------
double the latency just for vaddps?


================
Comment at: llvm/lib/Target/X86/X86ScheduleZnver4.td:909
+defm : Zn4WriteResYMMPair<WriteFAdd64Y, [Zn4FPFAdd01], 3, [1], 1>; // Floating point double add/sub (YMM).
+defm : Zn4WriteResZMMPair<WriteFAdd64Z, [Zn4FPFAdd01], 3, [1], 1>; // Floating point double add/sub (ZMM).
+defm : Zn4WriteResXMMPair<WriteFCmp, [Zn4FPFMul01], 2, [2], 1>;  // Floating point compare.
----------------
Don't you need to account for double pumping for ZMM? In the Jaguar model we'd typically double the resource usage to [2] to simulate it - so uops stays at 1.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D144010/new/

https://reviews.llvm.org/D144010