[llvm-commits] [llvm] r141498 - /llvm/trunk/test/CodeGen/X86/norex-subreg.ll

[Jakob Stoklund Olesen]

Author: stoklund
Date: Sat Oct  8 13:06:54 2011
New Revision: 141498

URL: http://llvm.org/viewvc/llvm-project?rev=141498&view=rev
Log:
Add missing test case for r141410.

Added:
    llvm/trunk/test/CodeGen/X86/norex-subreg.ll

Added: llvm/trunk/test/CodeGen/X86/norex-subreg.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/norex-subreg.ll?rev=141498&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/norex-subreg.ll (added)
+++ llvm/trunk/test/CodeGen/X86/norex-subreg.ll Sat Oct  8 13:06:54 2011
@@ -0,0 +1,39 @@
+; RUN: llc -O0 < %s
+target triple = "x86_64-apple-macosx10.7"
+
+; This test case extracts a sub_8bit_hi sub-register:
+;
+;	%R8B<def> = COPY %BH, %EBX<imp-use,kill>
+;	%ESI<def> = MOVZX32_NOREXrr8 %R8B<kill>
+;
+; The register allocation above is invalid, %BH can only be encoded without an
+; REX prefix, so the destination register must be GR8_NOREX.  The code above
+; triggers an assertion in copyPhysReg.
+;
+; <rdar://problem/10248099>
+
+define void @f() nounwind uwtable ssp {
+entry:
+  %0 = load i32* undef, align 4
+  %add = add i32 0, %0
+  %conv1 = trunc i32 %add to i16
+  %bf.value = and i16 %conv1, 255
+  %1 = and i16 %bf.value, 255
+  %2 = shl i16 %1, 8
+  %3 = load i16* undef, align 1
+  %4 = and i16 %3, 255
+  %5 = or i16 %4, %2
+  store i16 %5, i16* undef, align 1
+  %6 = load i16* undef, align 1
+  %7 = lshr i16 %6, 8
+  %bf.clear2 = and i16 %7, 255
+  %conv3 = zext i16 %bf.clear2 to i32
+  %rem = srem i32 %conv3, 15
+  %conv4 = trunc i32 %rem to i16
+  %bf.value5 = and i16 %conv4, 255
+  %8 = and i16 %bf.value5, 255
+  %9 = shl i16 %8, 8
+  %10 = or i16 undef, %9
+  store i16 %10, i16* undef, align 1
+  ret void
+}