[PATCH] D75392: [AMDGPU] Fix the gfx10 scheduling model for f32 conversions

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/VOP1Instructions.td:199
 let FPDPRounding = 1 in {
 defm V_CVT_F16_F32 : VOP1Inst <"v_cvt_f16_f32", VOP_F16_F32, fpround>;
 } // End FPDPRounding = 1
----------------
As far as I can tell this is quarter rate.


================
Comment at: llvm/lib/Target/AMDGPU/VOP1Instructions.td:201
 } // End FPDPRounding = 1
 defm V_CVT_F32_F16 : VOP1Inst <"v_cvt_f32_f16", VOP_F32_F16, fpextend>;
 defm V_CVT_RPI_I32_F32 : VOP1Inst <"v_cvt_rpi_i32_f32", VOP_I32_F32, cvt_rpi_i32_f32>;
----------------
And this is full rate, as well as 3 opcodes below.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75392/new/

https://reviews.llvm.org/D75392