commit b72888647baee2093c2cbd7a0c3da6dc8b870f20
author Matt Arsenault <Matthew.Arsenault@amd.com> Mon Jan 28 23:59:18 2019 +0000
committer Matt Arsenault <Matthew.Arsenault@amd.com> Mon Jan 28 23:59:18 2019 +0000
tree 46b3c066b9e0f03ef065f854908f0d96ba14ef61
parent 33f87b8aef87cd365ec59f0961ad8de8bd848bc0

AMDGPU: Add ds append/consume builtins

llvm-svn: 352443

clang/include/clang/Basic/BuiltinsAMDGPU.def

diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def
index 5478704..fa3db61 100644
--- a/clang/include/clang/Basic/BuiltinsAMDGPU.def
+++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def
@@ -98,6 +98,8 @@
 BUILTIN(__builtin_amdgcn_ds_faddf, "ff*3fIiIiIb", "n")
 BUILTIN(__builtin_amdgcn_ds_fminf, "ff*3fIiIiIb", "n")
 BUILTIN(__builtin_amdgcn_ds_fmaxf, "ff*3fIiIiIb", "n")
+BUILTIN(__builtin_amdgcn_ds_append, "ii*3", "n")
+BUILTIN(__builtin_amdgcn_ds_consume, "ii*3", "n")
 
 //===----------------------------------------------------------------------===//
 // CI+ only builtins.

clang/lib/CodeGen/CGBuiltin.cpp

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 195b0c0..fa4d512 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -12478,6 +12478,14 @@
   case AMDGPU::BI__builtin_amdgcn_fmed3f:
   case AMDGPU::BI__builtin_amdgcn_fmed3h:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_fmed3);
+  case AMDGPU::BI__builtin_amdgcn_ds_append:
+  case AMDGPU::BI__builtin_amdgcn_ds_consume: {
+    Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append ?
+      Intrinsic::amdgcn_ds_append : Intrinsic::amdgcn_ds_consume;
+    Value *Src0 = EmitScalarExpr(E->getArg(0));
+    Function *F = CGM.getIntrinsic(Intrin, { Src0->getType() });
+    return Builder.CreateCall(F, { Src0, Builder.getFalse() });
+  }
   case AMDGPU::BI__builtin_amdgcn_read_exec: {
     CallInst *CI = cast<CallInst>(
       EmitSpecialRegisterBuiltin(*this, E, Int64Ty, Int64Ty, true, "exec"));

clang/test/CodeGenOpenCL/builtins-amdgcn.cl

diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
index dc7f480..6b7ea52 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
@@ -536,6 +536,18 @@
   *out = __builtin_amdgcn_s_getpc();
 }
 
+// CHECK-LABEL: @test_ds_append_lds(
+// CHECK: call i32 @llvm.amdgcn.ds.append.p3i32(i32 addrspace(3)* %ptr, i1 false)
+kernel void test_ds_append_lds(global int* out, local int* ptr) {
+  *out = __builtin_amdgcn_ds_append(ptr);
+}
+
+// CHECK-LABEL: @test_ds_consume_lds(
+// CHECK: call i32 @llvm.amdgcn.ds.consume.p3i32(i32 addrspace(3)* %ptr, i1 false)
+kernel void test_ds_consume_lds(global int* out, local int* ptr) {
+  *out = __builtin_amdgcn_ds_consume(ptr);
+}
+
 // CHECK-DAG: [[$WI_RANGE]] = !{i32 0, i32 1024}
 // CHECK-DAG: attributes #[[$NOUNWIND_READONLY:[0-9]+]] = { nounwind readonly }
 // CHECK-DAG: attributes #[[$READ_EXEC_ATTRS]] = { convergent }