[PATCH] D48569: [AMDGPU] Convert rcp to rcp_iflag

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 26 00:05:22 PDT 2018


rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:6641
+                         N0.getOpcode() == ISD::SINT_TO_FP)) {
+    return DCI.DAG.getNode(AMDGPUISD::RCP_IFLAG, SDLoc(N), VT, N0);
+  }
----------------
arsenm wrote:
> Preserve fast math flags?
Makes sense.


================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:6644
+
+  return AMDGPUTargetLowering::PerformDAGCombine(N, DCI);
+}
----------------
arsenm wrote:
> This can just return SDValue(). The base PerformDAGCombine isn't going to do anything here
In fact not, there is combining code in parent and having no this call results in tests regression.


https://reviews.llvm.org/D48569





More information about the llvm-commits mailing list