[llvm] r252088 - [x86] Teach the shrink-wrapping hooks to do the proper thing with Win64.
Quentin Colombet via llvm-commits
llvm-commits at lists.llvm.org
Wed Nov 4 14:37:28 PST 2015
Author: qcolombet
Date: Wed Nov 4 16:37:28 2015
New Revision: 252088
URL: http://llvm.org/viewvc/llvm-project?rev=252088&view=rev
Log:
[x86] Teach the shrink-wrapping hooks to do the proper thing with Win64.
Win64 has some strict requirements for the epilogue. As a result, we disable
shrink-wrapping for Win64 unless the block that gets the epilogue is already an
exit block.
Fixes PR24193.
Added:
llvm/trunk/test/CodeGen/X86/x86-win64-shrink-wrapping.ll
Modified:
llvm/trunk/lib/Target/X86/X86FrameLowering.cpp
Modified: llvm/trunk/lib/Target/X86/X86FrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FrameLowering.cpp?rev=252088&r1=252087&r2=252088&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FrameLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86FrameLowering.cpp Wed Nov 4 16:37:28 2015
@@ -2184,6 +2184,14 @@ eliminateCallFramePseudoInstr(MachineFun
bool X86FrameLowering::canUseAsEpilogue(const MachineBasicBlock &MBB) const {
assert(MBB.getParent() && "Block is not attached to a function!");
+ // Win64 has strict requirements in terms of epilogue and we are
+ // not taking a chance at messing with them.
+ // I.e., unless this block is already an exit block, we can't use
+ // it as an epilogue.
+ if (MBB.getParent()->getSubtarget<X86Subtarget>().isTargetWin64() &&
+ !MBB.succ_empty() && !MBB.isReturnBlock())
+ return false;
+
if (canUseLEAForSPInEpilogue(*MBB.getParent()))
return true;
Added: llvm/trunk/test/CodeGen/X86/x86-win64-shrink-wrapping.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-win64-shrink-wrapping.ll?rev=252088&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/x86-win64-shrink-wrapping.ll (added)
+++ llvm/trunk/test/CodeGen/X86/x86-win64-shrink-wrapping.ll Wed Nov 4 16:37:28 2015
@@ -0,0 +1,122 @@
+; RUN: llc %s -o - -enable-shrink-wrap=true | FileCheck %s --check-prefix=CHECK --check-prefix=ENABLE
+; RUN: llc %s -o - -enable-shrink-wrap=false | FileCheck %s --check-prefix=CHECK --check-prefix=DISABLE
+
+target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
+target triple = "x86_64--windows-gnu"
+
+; The output of this function with or without shrink-wrapping
+; shouldn't change.
+; Indeed, the epilogue block would have been if.else, meaning
+; after the pops, we will have additional instruction (jump, mov,
+; etc.) prior to the return and this is forbidden for Win64.
+; CHECK-LABEL: loopInfoSaveOutsideLoop:
+; CHECK: push
+; CHECK-NOT: popq
+; CHECK: popq
+; CHECK-NOT: popq
+; CHECK-NEXT: retq
+define i32 @loopInfoSaveOutsideLoop(i32 %cond, i32 %N) #0 {
+entry:
+ %tobool = icmp eq i32 %cond, 0
+ br i1 %tobool, label %if.else, label %for.preheader
+
+for.preheader: ; preds = %entry
+ tail call void asm "nop", ""()
+ br label %for.body
+
+for.body: ; preds = %for.body, %for.preheader
+ %i.05 = phi i32 [ %inc, %for.body ], [ 0, %for.preheader ]
+ %sum.04 = phi i32 [ %add, %for.body ], [ 0, %for.preheader ]
+ %call = tail call i32 asm "movl $$1, $0", "=r,~{ebx}"()
+ %add = add nsw i32 %call, %sum.04
+ %inc = add nuw nsw i32 %i.05, 1
+ %exitcond = icmp eq i32 %inc, 10
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body
+ tail call void asm "nop", "~{ebx}"()
+ %shl = shl i32 %add, 3
+ br label %if.end
+
+if.else: ; preds = %entry
+ %mul = shl nsw i32 %N, 1
+ br label %if.end
+
+if.end: ; preds = %if.else, %for.end
+ %sum.1 = phi i32 [ %shl, %for.end ], [ %mul, %if.else ]
+ ret i32 %sum.1
+}
+
+; When we can sink the epilogue of the function into an existing exit block,
+; this is Ok for shrink-wrapping to kicks in.
+; CHECK-LABEL: loopInfoSaveOutsideLoop2:
+; ENABLE: testl %ecx, %ecx
+; ENABLE-NEXT: je [[ELSE_LABEL:.LBB[0-9_]+]]
+;
+; Prologue code.
+; Make sure we save the CSR used in the inline asm: rbx.
+; CHECK: pushq %rbx
+;
+; DISABLE: testl %ecx, %ecx
+; DISABLE-NEXT: je [[ELSE_LABEL:.LBB[0-9_]+]]
+;
+; CHECK: nop
+; CHECK: xorl [[SUM:%eax]], [[SUM]]
+; CHECK-NEXT: movl $10, [[IV:%e[a-z]+]]
+;
+; CHECK: [[LOOP_LABEL:.LBB[0-9_]+]]: # %for.body
+; CHECK: movl $1, [[TMP:%e[a-z]+]]
+; CHECK: addl [[TMP]], [[SUM]]
+; CHECK-NEXT: decl [[IV]]
+; CHECK-NEXT: jne [[LOOP_LABEL]]
+; Next BB.
+; CHECK: nop
+; CHECK: shll $3, [[SUM]]
+;
+; DISABLE: jmp [[EPILOG_BB:.LBB[0-9_]+]]
+;
+; ENABLE-NEXT: popq %rbx
+; ENABLE-NEXT: retq
+;
+; CHECK: [[ELSE_LABEL]]: # %if.else
+; Shift second argument by one and store into returned register.
+; CHECK: addl %edx, %edx
+; CHECK: movl %edx, %eax
+;
+; DISABLE: [[EPILOG_BB]]: # %if.end
+; DISABLE-NEXT: popq %rbx
+;
+; CHECK: retq
+;
+define i32 @loopInfoSaveOutsideLoop2(i32 %cond, i32 %N) #0 {
+entry:
+ %tobool = icmp eq i32 %cond, 0
+ br i1 %tobool, label %if.else, label %for.preheader
+
+for.preheader: ; preds = %entry
+ tail call void asm "nop", ""()
+ br label %for.body
+
+for.body: ; preds = %for.body, %for.preheader
+ %i.05 = phi i32 [ %inc, %for.body ], [ 0, %for.preheader ]
+ %sum.04 = phi i32 [ %add, %for.body ], [ 0, %for.preheader ]
+ %call = tail call i32 asm "movl $$1, $0", "=r,~{ebx}"()
+ %add = add nsw i32 %call, %sum.04
+ %inc = add nuw nsw i32 %i.05, 1
+ %exitcond = icmp eq i32 %inc, 10
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body
+ tail call void asm "nop", "~{ebx}"()
+ %shl = shl i32 %add, 3
+ ret i32 %shl
+
+if.else: ; preds = %entry
+ %mul = shl nsw i32 %N, 1
+ br label %if.end
+
+if.end: ; preds = %if.else, %for.end
+ ret i32 %mul
+}
+
+attributes #0 = { uwtable }
More information about the llvm-commits
mailing list