AMDGPU: Set v2i32 any_extend to expand

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@314993 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/SIISelLowering.cpp b/lib/Target/AMDGPU/SIISelLowering.cpp
index 3a125c2e7e4689a965ce37807a4516ca94919f68..fac2b03fa2fac84f106cf284cc8b322ef7bffed7 100644 (file)
--- a/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -468,6 +468,7 @@ SITargetLowering::SITargetLowering(const TargetMachine &TM,
     setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2i16, Custom);
     setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::v2f16, Custom);
 
+    setOperationAction(ISD::ANY_EXTEND, MVT::v2i32, Expand);
     setOperationAction(ISD::ZERO_EXTEND, MVT::v2i32, Expand);
     setOperationAction(ISD::SIGN_EXTEND, MVT::v2i32, Expand);
     setOperationAction(ISD::FP_EXTEND, MVT::v2f32, Expand);

diff --git a/test/CodeGen/AMDGPU/anyext.ll b/test/CodeGen/AMDGPU/anyext.ll
index 3f220c40841298383e91f125897271b0c296cfe3..cabf0578aaf0c48d853e37e10e615711bab87f89 100644 (file)
--- a/test/CodeGen/AMDGPU/anyext.ll
+++ b/test/CodeGen/AMDGPU/anyext.ll
@@ -1,12 +1,13 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI %s
+; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,GFX89 %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9,GFX89 %s
 
 declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
 declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone
 
 ; GCN-LABEL: {{^}}anyext_i1_i32:
 ; GCN: v_cndmask_b32_e64
-define amdgpu_kernel void @anyext_i1_i32(i32 addrspace(1)* %out, i32 %cond) {
+define amdgpu_kernel void @anyext_i1_i32(i32 addrspace(1)* %out, i32 %cond) #0 {
 entry:
   %tmp = icmp eq i32 %cond, 0
   %tmp1 = zext i1 %tmp to i8
@@ -18,11 +19,11 @@ entry:
 }
 
 ; GCN-LABEL: {{^}}s_anyext_i16_i32:
-; VI: v_add_u16_e32 [[ADD:v[0-9]+]],
-; VI: v_xor_b32_e32 [[XOR:v[0-9]+]], -1, [[ADD]]
-; VI: v_and_b32_e32 [[AND:v[0-9]+]], 1, [[XOR]]
-; VI: buffer_store_dword [[AND]]
-define amdgpu_kernel void @s_anyext_i16_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %a, i16 addrspace(1)* %b) {
+; GFX89: v_add_u16_e32 [[ADD:v[0-9]+]],
+; GFX89: v_xor_b32_e32 [[XOR:v[0-9]+]], -1, [[ADD]]
+; GFX89: v_and_b32_e32 [[AND:v[0-9]+]], 1, [[XOR]]
+; GFX89: buffer_store_dword [[AND]]
+define amdgpu_kernel void @s_anyext_i16_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %a, i16 addrspace(1)* %b) #0 {
 entry:
   %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
   %tid.y = call i32 @llvm.amdgcn.workitem.id.y()
@@ -38,3 +39,27 @@ entry:
   store i32 %tmp4, i32 addrspace(1)* %out
   ret void
 }
+
+; GCN-LABEL: {{^}}anyext_v2i16_to_v2i32:
+; GFX9: global_load_short_d16_hi
+; GFX9: v_and_b32_e32 v{{[0-9]+}}, 0x80008000
+; GFX9: v_bfi_b32 v{{[0-9]+}}, v{{[0-9]+}}, 0, v{{[0-9]+}}
+; GFX9: v_cmp_eq_f32_e32
+; GFX9: v_cndmask_b32_e64 v{{[0-9]+}}, 0, 1, vcc
+define amdgpu_kernel void @anyext_v2i16_to_v2i32() #0 {
+bb:
+  %tmp = load i16, i16 addrspace(1)* undef, align 2
+  %tmp2 = insertelement <2 x i16> undef, i16 %tmp, i32 1
+  %tmp4 = and <2 x i16> %tmp2, <i16 -32768, i16 -32768>
+  %tmp5 = zext <2 x i16> %tmp4 to <2 x i32>
+  %tmp6 = shl nuw <2 x i32> %tmp5, <i32 16, i32 16>
+  %tmp7 = or <2 x i32> zeroinitializer, %tmp6
+  %tmp8 = bitcast <2 x i32> %tmp7 to <2 x float>
+  %tmp10 = fcmp oeq <2 x float> %tmp8, zeroinitializer
+  %tmp11 = zext <2 x i1> %tmp10 to <2 x i8>
+  %tmp12 = extractelement <2 x i8> %tmp11, i32 1
+  store i8 %tmp12, i8 addrspace(1)* undef, align 1
+  ret void
+}
+
+attributes #0 = { nounwind }