PPC QPX requires a 32-byte aligned stack

On systems which support the QPX vector instructions, the stack must be
32-byte aligned.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@173993 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/PowerPC/PPCFrameLowering.h b/lib/Target/PowerPC/PPCFrameLowering.h
index 3517d8c086229526ed16c9f63dcc9b668fe3cd8e..b1d63ab53714123ff50ae0a0a1c27dbbf9574e6c 100644 (file)
--- a/lib/Target/PowerPC/PPCFrameLowering.h
+++ b/lib/Target/PowerPC/PPCFrameLowering.h
@@ -27,7 +27,8 @@ class PPCFrameLowering: public TargetFrameLowering {
 
 public:
   PPCFrameLowering(const PPCSubtarget &sti)
-    : TargetFrameLowering(TargetFrameLowering::StackGrowsDown, 16, 0),
+    : TargetFrameLowering(TargetFrameLowering::StackGrowsDown,
+        (sti.hasQPX() || sti.isBGQ()) ? 32 : 16, 0),
       Subtarget(sti) {
   }
 

diff --git a/lib/Target/PowerPC/PPCSubtarget.cpp b/lib/Target/PowerPC/PPCSubtarget.cpp
index 91beeaee680c52ecd893ceac9eca4106bcb72824..18e4c07942b966fef69ab938f06dfb749c819af1 100644 (file)
--- a/lib/Target/PowerPC/PPCSubtarget.cpp
+++ b/lib/Target/PowerPC/PPCSubtarget.cpp
@@ -83,6 +83,12 @@ PPCSubtarget::PPCSubtarget(const std::string &TT, const std::string &CPU,
   // Set up darwin-specific properties.
   if (isDarwin())
     HasLazyResolverStubs = true;
+
+  // QPX requires a 32-byte aligned stack. Note that we need to do this if
+  // we're compiling for a BG/Q system regardless of whether or not QPX
+  // is enabled because external functions will assume this alignment.
+  if (hasQPX() || isBGQ())
+    StackAlignment = 32;
 }
 
 /// SetJITMode - This is called to inform the subtarget info that we are

diff --git a/test/CodeGen/PowerPC/a2q-stackalign.ll b/test/CodeGen/PowerPC/a2q-stackalign.ll
new file mode 100644 (file)
index 0000000..00c3291
--- /dev/null
+++ b/test/CodeGen/PowerPC/a2q-stackalign.ll
@@ -0,0 +1,23 @@
+; RUN: llc < %s -march=ppc64 -mcpu=a2 | FileCheck -check-prefix=CHECK-A2 %s
+; RUN: llc < %s -march=ppc64 -mcpu=a2q | FileCheck -check-prefix=CHECK-A2Q %s
+; RUN: llc < %s -march=ppc64 -mtriple=powerpc64-bgq-linux -mcpu=a2 | FileCheck -check-prefix=CHECK-BGQ %s
+target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f128:128:128-v128:128:128-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+declare i32 @bar(i8* %a) nounwind;
+define i32 @foo() nounwind {
+  %p = alloca i8, i8 115
+  store i8 0, i8* %p
+  %r = call i32 @bar(i8* %p)
+  ret i32 %r
+}
+
+; Without QPX, the allocated stack frame is 240 bytes, but with QPX
+; (because we require 32-byte alignment), it is 256 bytes.
+; CHECK-A2: @foo
+; CHECK-A2: stdu 1, -240(1)
+; CHECK-A2Q: @foo
+; CHECK-A2Q: stdu 1, -256(1)
+; CHECK-BGQ: @foo
+; CHECK-BGQ: stdu 1, -256(1)
+