[llvm] r188559 - R600/SI: Add pattern for xor of i1

[Michel Danzer]

Author: daenzer
Date: Fri Aug 16 11:19:31 2013
New Revision: 188559

URL: http://llvm.org/viewvc/llvm-project?rev=188559&view=rev
Log:
R600/SI: Add pattern for xor of i1

Fixes two recent piglit regressions with radeonsi.

Reviewed-by: Tom Stellard <thomas.stellard at amd.com>

Modified:
    llvm/trunk/lib/Target/R600/SIInstructions.td
    llvm/trunk/test/CodeGen/R600/xor.ll

Modified: llvm/trunk/lib/Target/R600/SIInstructions.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/SIInstructions.td?rev=188559&r1=188558&r2=188559&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/SIInstructions.td (original)
+++ llvm/trunk/lib/Target/R600/SIInstructions.td Fri Aug 16 11:19:31 2013
@@ -1144,7 +1144,9 @@ def : Pat <
   (S_OR_B64 $src0, $src1)
 >;
 def S_XOR_B32 : SOP2_32 <0x00000012, "S_XOR_B32", []>;
-def S_XOR_B64 : SOP2_64 <0x00000013, "S_XOR_B64", []>;
+def S_XOR_B64 : SOP2_64 <0x00000013, "S_XOR_B64",
+  [(set i1:$dst, (xor i1:$src0, i1:$src1))]
+>;
 def S_ANDN2_B32 : SOP2_32 <0x00000014, "S_ANDN2_B32", []>;
 def S_ANDN2_B64 : SOP2_64 <0x00000015, "S_ANDN2_B64", []>;
 def S_ORN2_B32 : SOP2_32 <0x00000016, "S_ORN2_B32", []>;

Modified: llvm/trunk/test/CodeGen/R600/xor.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/R600/xor.ll?rev=188559&r1=188558&r2=188559&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/R600/xor.ll (original)
+++ llvm/trunk/test/CodeGen/R600/xor.ll Fri Aug 16 11:19:31 2013
@@ -37,3 +37,20 @@ define void @xor_v4i32(<4 x i32> addrspa
   store <4 x i32> %result, <4 x i32> addrspace(1)* %out
   ret void
 }
+
+;EG-CHECK: @xor_i1
+;EG-CHECK: XOR_INT {{\*? *}}T{{[0-9]+\.[XYZW], PV\.[XYZW], PV\.[XYZW]}}
+
+;SI-CHECK: @xor_i1
+;SI-CHECK: S_XOR_B64 {{SGPR[0-9]+_SGPR[0-9]+, SGPR[0-9]+_SGPR[0-9]+, SGPR[0-9]+_SGPR[0-9]+}}
+
+define void @xor_i1(float addrspace(1)* %out, float addrspace(1)* %in0, float addrspace(1)* %in1) {
+  %a = load float addrspace(1) * %in0
+  %b = load float addrspace(1) * %in1
+  %acmp = fcmp oge float %a, 0.000000e+00
+  %bcmp = fcmp oge float %b, 0.000000e+00
+  %xor = xor i1 %acmp, %bcmp
+  %result = select i1 %xor, float %a, float %b
+  store float %result, float addrspace(1)* %out
+  ret void
+}