From 9d548d0343774636e72713d678a078c8e808ed29 Mon Sep 17 00:00:00 2001
From: Jakob Stoklund Olesen <stoklund@2pi.dk>
Date: Thu, 1 Sep 2011 17:18:50 +0000
Subject: [PATCH] Prevent remat of partial register redefinitions.

An instruction that redefines only part of a larger register can never
be rematerialized since the virtual register value depends on the old
value in other parts of the register.

This was fixed for the inline spiller in r138794.  This patch fixes the
problem for all register allocators, and includes a small test case.

<rdar://problem/10032939>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@138944 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/CodeGen/TargetInstrInfoImpl.cpp |  9 +++++++++
 test/CodeGen/ARM/subreg-remat.ll    | 28 ++++++++++++++++++++++++++++
 2 files changed, 37 insertions(+)
 create mode 100644 test/CodeGen/ARM/subreg-remat.ll

diff --git a/lib/CodeGen/TargetInstrInfoImpl.cpp b/lib/CodeGen/TargetInstrInfoImpl.cpp
index ea0cfe2b940..1cd42ca2ae0 100644
--- a/lib/CodeGen/TargetInstrInfoImpl.cpp
+++ b/lib/CodeGen/TargetInstrInfoImpl.cpp
@@ -362,6 +362,15 @@ isReallyTriviallyReMaterializableGeneric(const MachineInstr *MI,
   const TargetInstrInfo &TII = *TM.getInstrInfo();
   const TargetRegisterInfo &TRI = *TM.getRegisterInfo();
 
+  // A sub-register definition can only be rematerialized if the instruction
+  // doesn't read the other parts of the register.  Otherwise it is really a
+  // read-modify-write operation on the full virtual register which cannot be
+  // moved safely.
+  unsigned Reg = MI->getOperand(0).getReg();
+  if (TargetRegisterInfo::isVirtualRegister(Reg) &&
+      MI->getOperand(0).getSubReg() && MI->readsVirtualRegister(Reg))
+    return false;
+
   // A load from a fixed stack slot can be rematerialized. This may be
   // redundant with subsequent checks, but it's target-independent,
   // simple, and a common case.
diff --git a/test/CodeGen/ARM/subreg-remat.ll b/test/CodeGen/ARM/subreg-remat.ll
new file mode 100644
index 00000000000..cf45c034514
--- /dev/null
+++ b/test/CodeGen/ARM/subreg-remat.ll
@@ -0,0 +1,28 @@
+; RUN: llc < %s -relocation-model=pic -disable-fp-elim -mcpu=cortex-a8 -pre-RA-sched=source | FileCheck %s
+target triple = "thumbv7-apple-ios"
+; <rdar://problem/10032939>
+;
+; The vector %v2 is built like this:
+;
+;   %vreg6:ssub_1<def> = VMOVSR %vreg0<kill>, pred:14, pred:%noreg, %vreg6<imp-def>; DPR_VFP2:%vreg6 GPR:%vreg0
+;   %vreg6:ssub_0<def> = VLDRS <cp#0>, 0, pred:14, pred:%noreg; mem:LD4[ConstantPool] DPR_VFP2:%vreg6
+;
+; When %vreg6 spills, the VLDRS constant pool load cannot be rematerialized
+; since it implicitly reads the ssub_1 sub-register.
+;
+; CHECK: f1
+; CHECK: vmov    s1, r0
+; CHECK: vldr.32 s0, LCPI
+; The vector must be spilled:
+; CHECK: vstr.64 d0,
+; CHECK: asm clobber d0
+; And reloaded after the asm:
+; CHECK: vldr.64 [[D16:d[0-9]+]],
+; CHECK: vstr.64 [[D16]], [r1]
+define void @f1(float %x, <2 x float>* %p) {
+  %v1 = insertelement <2 x float> undef, float %x, i32 1
+  %v2 = insertelement <2 x float> %v1, float 0x400921FB60000000, i32 0
+  %y = call double asm sideeffect "asm clobber $0", "=w,0,~{d1},~{d2},~{d3},~{d4},~{d5},~{d6},~{d7},~{d8},~{d9},~{d10},~{d11},~{d12},~{d13},~{d14},~{d15},~{d16},~{d17},~{d18},~{d19},~{d20},~{d21},~{d22},~{d23},~{d24},~{d25},~{d26},~{d27},~{d28},~{d29},~{d30},~{d31}"(<2 x float> %v2) nounwind
+  store <2 x float> %v2, <2 x float>* %p, align 8
+  ret void
+}