[llvm] r222942 - R600/SI: Fix assertion on sign extend of 3 vectors
Matt Arsenault
Matthew.Arsenault at amd.com
Fri Nov 28 14:51:38 PST 2014
Author: arsenm
Date: Fri Nov 28 16:51:38 2014
New Revision: 222942
URL: http://llvm.org/viewvc/llvm-project?rev=222942&view=rev
Log:
R600/SI: Fix assertion on sign extend of 3 vectors
This was trying to create an MVT with 3x vectors which
created an invalid EVT
Modified:
llvm/trunk/lib/Target/R600/SIISelLowering.cpp
llvm/trunk/test/CodeGen/R600/setcc.ll
Modified: llvm/trunk/lib/Target/R600/SIISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/SIISelLowering.cpp?rev=222942&r1=222941&r2=222942&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/SIISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/R600/SIISelLowering.cpp Fri Nov 28 16:51:38 2014
@@ -640,11 +640,11 @@ MachineBasicBlock * SITargetLowering::Em
return BB;
}
-EVT SITargetLowering::getSetCCResultType(LLVMContext &, EVT VT) const {
+EVT SITargetLowering::getSetCCResultType(LLVMContext &Ctx, EVT VT) const {
if (!VT.isVector()) {
return MVT::i1;
}
- return MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
+ return EVT::getVectorVT(Ctx, MVT::i1, VT.getVectorNumElements());
}
MVT SITargetLowering::getScalarShiftAmountTy(EVT VT) const {
Modified: llvm/trunk/test/CodeGen/R600/setcc.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/R600/setcc.ll?rev=222942&r1=222941&r2=222942&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/R600/setcc.ll (original)
+++ llvm/trunk/test/CodeGen/R600/setcc.ll Fri Nov 28 16:51:38 2014
@@ -1,5 +1,7 @@
-;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=R600 --check-prefix=FUNC %s
-;RUN: llc < %s -march=r600 -mcpu=SI -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+
+declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; FUNC-LABEL: {{^}}setcc_v2i32:
; R600-DAG: SETE_INT * T{{[0-9]+\.[XYZW]}}, KC0[3].X, KC0[3].Z
@@ -343,3 +345,46 @@ entry:
store i32 %1, i32 addrspace(1)* %out
ret void
}
+
+; FIXME: This does 4 compares
+; FUNC-LABEL: {{^}}v3i32_eq:
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI: s_endpgm
+define void @v3i32_eq(<3 x i32> addrspace(1)* %out, <3 x i32> addrspace(1)* %ptra, <3 x i32> addrspace(1)* %ptrb) {
+ %tid = call i32 @llvm.r600.read.tidig.x() nounwind readnone
+ %gep.a = getelementptr <3 x i32> addrspace(1)* %ptra, i32 %tid
+ %gep.b = getelementptr <3 x i32> addrspace(1)* %ptrb, i32 %tid
+ %gep.out = getelementptr <3 x i32> addrspace(1)* %out, i32 %tid
+ %a = load <3 x i32> addrspace(1)* %gep.a
+ %b = load <3 x i32> addrspace(1)* %gep.b
+ %cmp = icmp eq <3 x i32> %a, %b
+ %ext = sext <3 x i1> %cmp to <3 x i32>
+ store <3 x i32> %ext, <3 x i32> addrspace(1)* %gep.out
+ ret void
+}
+
+; FUNC-LABEL: {{^}}v3i8_eq:
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI-DAG: v_cmp_eq_i32
+; SI-DAG: v_cndmask_b32_e64 {{v[0-9]+}}, 0, -1,
+; SI: s_endpgm
+define void @v3i8_eq(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %ptra, <3 x i8> addrspace(1)* %ptrb) {
+ %tid = call i32 @llvm.r600.read.tidig.x() nounwind readnone
+ %gep.a = getelementptr <3 x i8> addrspace(1)* %ptra, i32 %tid
+ %gep.b = getelementptr <3 x i8> addrspace(1)* %ptrb, i32 %tid
+ %gep.out = getelementptr <3 x i8> addrspace(1)* %out, i32 %tid
+ %a = load <3 x i8> addrspace(1)* %gep.a
+ %b = load <3 x i8> addrspace(1)* %gep.b
+ %cmp = icmp eq <3 x i8> %a, %b
+ %ext = sext <3 x i1> %cmp to <3 x i8>
+ store <3 x i8> %ext, <3 x i8> addrspace(1)* %gep.out
+ ret void
+}
More information about the llvm-commits
mailing list