[PATCH] D93305: [AMDGPU][GlobalISel] Transform (fadd (fmul x, y), z) -> (fma x, y, z)

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/include/llvm/CodeGen/TargetLowering.h:796
+  virtual bool enableAggressiveFMAFusion(LLT Ty) const {
+    switch (Ty.getSizeInBits()) {
+    case 16:
----------------
This will break for vectors


================
Comment at: llvm/include/llvm/CodeGen/TargetLowering.h:1116
+                                bool LegalOnly = false) const {
+    switch (Ty.getSizeInBits()) {
+    case 16:
----------------
This will break for vectors. Also, this should not be built on top of the DAG legalizer rules. This should be something in LegalizerInfo


================
Comment at: llvm/include/llvm/CodeGen/TargetLowering.h:2744
+           "unexpected node in FMAD forming combine");
+    switch (Ty.getSizeInBits()) {
+    case 16:
----------------
Broken for vectors


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93305/new/

https://reviews.llvm.org/D93305