From: Michel Danzer <michel.daenzer@amd.com>
Date: Fri, 16 Aug 2013 16:19:31 +0000 (+0000)
Subject: R600/SI: Add pattern for xor of i1
X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=b89f197713737e8c93045eb4dbd9c1d7ab41acd4;p=oota-llvm.git

R600/SI: Add pattern for xor of i1

Fixes two recent piglit regressions with radeonsi.

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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188559 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index 4eb3566c011..436a2cdcee6 100644
--- a/lib/Target/R600/SIInstructions.td
+++ b/lib/Target/R600/SIInstructions.td
@@ -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", []>;
diff --git a/test/CodeGen/R600/xor.ll b/test/CodeGen/R600/xor.ll
index f52729d4ae0..84d4cd49e93 100644
--- a/test/CodeGen/R600/xor.ll
+++ b/test/CodeGen/R600/xor.ll
@@ -37,3 +37,20 @@ define void @xor_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in
   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
+}