R600: support f16 -> f64 conversion intrinsic.

Unfortunately, we don't seem to have a direct truncation, but the
extension can be legally split into two operations so we should
support that.

llvm-svn: 213357

diff --git a/llvm/lib/Target/R600/AMDGPUISelLowering.cpp b/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
index 89d5b08..42d2a13 100644 (file)
--- a/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
@@ -242,6 +242,8 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
     setOperationAction(ISD::FCOPYSIGN, MVT::f64, Expand);
   }
 
+  setOperationAction(ISD::FP16_TO_FP, MVT::f64, Expand);
+
   const MVT ScalarIntVTs[] = { MVT::i32, MVT::i64 };
   for (MVT VT : ScalarIntVTs) {
     setOperationAction(ISD::SREM, VT, Expand);

diff --git a/llvm/test/CodeGen/R600/fp32_to_fp16.ll b/llvm/test/CodeGen/R600/fp16_to_fp.ll
similarity index 50%
rename from llvm/test/CodeGen/R600/fp32_to_fp16.ll
rename to llvm/test/CodeGen/R600/fp16_to_fp.ll
index 3a051f8..777eadc 100644 (file)
--- a/llvm/test/CodeGen/R600/fp32_to_fp16.ll
+++ b/llvm/test/CodeGen/R600/fp16_to_fp.ll
@@ -1,6 +1,7 @@
 ; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
 
 declare float @llvm.convert.from.fp16.f32(i16) nounwind readnone
+declare double @llvm.convert.from.fp16.f64(i16) nounwind readnone
 
 ; SI-LABEL: @test_convert_fp16_to_fp32:
 ; SI: BUFFER_LOAD_USHORT [[VAL:v[0-9]+]]
@@ -12,3 +13,16 @@ define void @test_convert_fp16_to_fp32(float addrspace(1)* noalias %out, i16 add
   store float %cvt, float addrspace(1)* %out, align 4
   ret void
 }
+
+
+; SI-LABEL: @test_convert_fp16_to_fp64:
+; SI: BUFFER_LOAD_USHORT [[VAL:v[0-9]+]]
+; SI: V_CVT_F32_F16_e32 [[RESULT32:v[0-9]+]], [[VAL]]
+; SI: V_CVT_F64_F32_e32 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[RESULT32]]
+; SI: BUFFER_STORE_DWORDX2 [[RESULT]]
+define void @test_convert_fp16_to_fp64(double addrspace(1)* noalias %out, i16 addrspace(1)* noalias %in) nounwind {
+  %val = load i16 addrspace(1)* %in, align 2
+  %cvt = call double @llvm.convert.from.fp16.f64(i16 %val) nounwind readnone
+  store double %cvt, double addrspace(1)* %out, align 4
+  ret void
+}