[PATCH] D18573: [X86] Enable call frame optimization ("mov to push") not only for optsize (PR26325)

[Reid Kleckner via llvm-commits]

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

lgtm I think this is production ready. We *mostly* build chromium with -Os, and we've fixed some bugs in this code.


================
Comment at: lib/Target/X86/X86CallFrameOptimization.cpp:176-178
@@ -175,5 +175,5 @@
   ContextVector &CallSeqVector) {
   // This transformation is always a win when we do not expect to have
   // a reserved call frame. Under other circumstances, it may be either
   // a win or a loss, and requires a heuristic.
   bool CannotReserveFrame = MF.getFrameInfo()->hasVarSizedObjects();
----------------
Huh, so we're already doing this a lot due to inalloca.

================
Comment at: test/CodeGen/X86/win32-seh-nested-finally.ll:46
@@ -45,3 +45,3 @@
 ; CHECK: movl $1, -[[state]](%ebp)
-; CHECK: movl $1, (%esp)
+; CHECK: pushl $1
 ; CHECK: calll _f
----------------
Great, I was halfway through checking that before I got distracted...

================
Comment at: test/CodeGen/X86/zext-fold.ll:38-39
@@ -37,3 +37,4 @@
 ; CHECK: movzbl {{[0-9]+}}(%esp), [[REGISTER:%e[a-z]{2}]]
-; CHECK-NEXT: movl [[REGISTER]], 4(%esp)
+; CHECK: subl $8, %esp
+; CHECK-NEXT: pushl [[REGISTER]]
 ; CHECK-NEXT: andl $224, [[REGISTER]]
----------------
Actually, the sub probably won't be emitted on Mac and Windows, it will only appear on platforms with 16 byte stack alignment.


http://reviews.llvm.org/D18573