[PATCH] D61202: [AMDGPU] gfx1010 VOP3 and VOP3P implementation

Konstantin Zhuravlyov via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Apr 26 10:50:28 PDT 2019


kzhuravl added inline comments.


================
Comment at: lib/Target/AMDGPU/VOP3Instructions.td:446-450
 def V_FMA_F16_gfx9 : VOP3Inst <"v_fma_f16_gfx9", VOP3_Profile<VOP_F16_F16_F16_F16, VOP3_OPSEL>, fma> {
   let renamedInGFX9 = 1;
   let Predicates = [Has16BitInsts, isGFX9Plus];
   let FPDPRounding = 1;
 }
----------------
Yes, it is fixed.


================
Comment at: lib/Target/AMDGPU/VOP3Instructions.td:771
+  VOP3OpSel_Real_gfx10_with_name<0x340, "V_MAD_U16_gfx9", "v_mad_u16">;
+// FIXME-GFX10-V_FMA_F16: GFX10's V_FMA_F16 should be really matched to
+// V_FMA_F16_gfx9 because it has opsel. Currently if we match to V_FMA_F16_gfx9,
----------------
kzhuravl wrote:
> IIRC it is already fixed.
Comment should either be removed before submission or in a follow up change. Thanks.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D61202/new/

https://reviews.llvm.org/D61202





More information about the llvm-commits mailing list