[PATCH] R600/SI: Add pattern for xor of i1
Michel Dänzer
michel at daenzer.net
Fri Aug 16 07:04:37 PDT 2013
From: Michel Dänzer <michel.daenzer at amd.com>
Fixes two recent piglit regressions with radeonsi.
Signed-off-by: Michel Dänzer <michel.daenzer at amd.com>
---
lib/Target/R600/SIInstructions.td | 4 +++-
test/CodeGen/R600/xor.ll | 17 +++++++++++++++++
2 files changed, 20 insertions(+), 1 deletion(-)
diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index 4eb3566..436a2cd 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 f52729d..84d4cd4 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
+}
--
1.8.4.rc2
More information about the llvm-commits
mailing list