R600: Add mul24 intrinsics

llvm-svn: 208604

5 files changed, 35 insertions, 4 deletions

diff --git a/llvm/lib/Target/R600/AMDGPUISelLowering.cpp b/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
index 285ab2d6702..2d2f4f4e6d4 100644
--- a/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
@@ -699,6 +699,14 @@ SDValue AMDGPUTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
       return DAG.getNode(AMDGPUISD::UMIN, DL, VT, Op.getOperand(1),
                                                   Op.getOperand(2));
 
+    case AMDGPUIntrinsic::AMDGPU_umul24:
+      return DAG.getNode(AMDGPUISD::MUL_U24, DL, VT,
+                         Op.getOperand(1), Op.getOperand(2));
+
+    case AMDGPUIntrinsic::AMDGPU_imul24:
+      return DAG.getNode(AMDGPUISD::MUL_I24, DL, VT,
+                         Op.getOperand(1), Op.getOperand(2));
+
     case AMDGPUIntrinsic::AMDGPU_bfe_i32:
       return DAG.getNode(AMDGPUISD::BFE_I32, DL, VT,
                          Op.getOperand(1),
diff --git a/llvm/lib/Target/R600/AMDGPUIntrinsics.td b/llvm/lib/Target/R600/AMDGPUIntrinsics.td
index c6521d07cfb..9f30bd8f1c9 100644
--- a/llvm/lib/Target/R600/AMDGPUIntrinsics.td
+++ b/llvm/lib/Target/R600/AMDGPUIntrinsics.td
@@ -49,6 +49,8 @@ let TargetPrefix = "AMDGPU", isTarget = 1 in {
   def int_AMDGPU_imin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_umax : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_umin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_AMDGPU_umul24 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_AMDGPU_imul24 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_cube : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
   def int_AMDGPU_bfi : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_AMDGPU_bfe_i32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
diff --git a/llvm/lib/Target/R600/AMDILIntrinsics.td b/llvm/lib/Target/R600/AMDILIntrinsics.td
index 658deb5bc01..4a3e02e202b 100644
--- a/llvm/lib/Target/R600/AMDILIntrinsics.td
+++ b/llvm/lib/Target/R600/AMDILIntrinsics.td
@@ -92,10 +92,6 @@ let TargetPrefix = "AMDIL", isTarget = 1 in {
           BinaryIntInt;
   def int_AMDIL_mulhi_u32 : GCCBuiltin<"__amdil_umul_high">,
           BinaryIntInt;
-  def int_AMDIL_mul24_i32 : GCCBuiltin<"__amdil_imul24">,
-          BinaryIntInt;
-  def int_AMDIL_mul24_u32 : GCCBuiltin<"__amdil_umul24">,
-          BinaryIntInt;
   def int_AMDIL_mulhi24_i32 : GCCBuiltin<"__amdil_imul24_high">,
           BinaryIntInt;
   def int_AMDIL_mulhi24_u32 : GCCBuiltin<"__amdil_umul24_high">,
diff --git a/llvm/test/CodeGen/R600/llvm.AMDGPU.imul24.ll b/llvm/test/CodeGen/R600/llvm.AMDGPU.imul24.ll
new file mode 100644
index 00000000000..33a1b8204e2
--- /dev/null
+++ b/llvm/test/CodeGen/R600/llvm.AMDGPU.imul24.ll
@@ -0,0 +1,14 @@
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=FUNC %s
+
+declare i32 @llvm.AMDGPU.imul24(i32, i32) nounwind readnone
+
+; FUNC-LABEL: @test_imul24
+; SI: V_MUL_I32_I24
+; CM: MUL_INT24
+define void @test_imul24(i32 addrspace(1)* %out, i32 %src0, i32 %src1) nounwind {
+  %mul = call i32 @llvm.AMDGPU.imul24(i32 %src0, i32 %src1) nounwind readnone
+  store i32 %mul, i32 addrspace(1)* %out, align 4
+  ret void
+}
+
diff --git a/llvm/test/CodeGen/R600/llvm.AMDGPU.umul24.ll b/llvm/test/CodeGen/R600/llvm.AMDGPU.umul24.ll
new file mode 100644
index 00000000000..21f824a65fc
--- /dev/null
+++ b/llvm/test/CodeGen/R600/llvm.AMDGPU.umul24.ll
@@ -0,0 +1,11 @@
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+
+declare i32 @llvm.AMDGPU.umul24(i32, i32) nounwind readnone
+
+; SI-LABEL: @test_umul24
+define void @test_umul24(i32 addrspace(1)* %out, i32 %src0, i32 %src1) nounwind {
+  %mul = call i32 @llvm.AMDGPU.umul24(i32 %src0, i32 %src1) nounwind readnone
+  store i32 %mul, i32 addrspace(1)* %out, align 4
+  ret void
+}
+