[PATCH] D22092: AMDGPU: Reduce the duration of whole-quad-mode

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Thu Sep 1 10:46:22 PDT 2016

arsenm accepted this revision.
arsenm added a comment.
This revision is now accepted and ready to land.


Comment at: lib/Target/AMDGPU/SIWholeQuadMode.cpp:232-235
@@ +231,6 @@
+        // Since we're in machine SSA, we do not need to track physical
+        // registers across basic blocks.
+        if (Value->isPHIDef())
+          continue;
I thought I had an example similar to this a long time ago. Is duplicating the s_cmp an optimization that started working relatively recently? I guess you can ignore this for now, maybe you can do something with inline asm but it seems to be pretty broken with scc right now

Comment at: lib/Target/AMDGPU/SIWholeQuadMode.cpp:475
@@ -424,1 +474,3 @@
+    DEBUG(dbgs() << "  " << MI);
This looks like a contextless instruction printing which probably isn't helpful


More information about the llvm-commits mailing list