[AMDGPU] Add builtin functions readlane ds_permute mov_dpp

[yaxunl]

Make some arguments constant.

[arsenm]

lib/CodeGen/CGBuiltin.cpp

8391–8398

Since these don't have mangling, you could add GCCBuiltin to the intrinsic definition and then you wouldn't need these to be handled here

[yaxunl]

lib/CodeGen/CGBuiltin.cpp

8391–8398

If I remove these, there will be error:

cannot compile this builtin function yet

[b-sumner]

mov_dpp should be under the VI+ comment

[arsenm]

lib/CodeGen/CGBuiltin.cpp

8391–8398

Yes, the GCCBuiltin isn't on the intrinsic declaration in llvm. If you fix that you shouldn't see that

[arsenm]

In D30551#691149, @b-sumner wrote:

mov_dpp should be under the VI+ comment

Also should be made to use TARGET_BUILTIN and dependent on dpp feature

[yaxunl]

remove redundant code.
make mov_dpp target builtin.

[arsenm]

LGTM

include/clang/Basic/BuiltinsAMDGPU.def

 BUILTIN(__builtin_amdgcn_s_decperflevel, "vIi", "n")
 BUILTIN(__builtin_amdgcn_uicmp, "LUiUiUiIi", "nc")
 BUILTIN(__builtin_amdgcn_uicmpl, "LUiLUiLUiIi", "nc")
 BUILTIN(__builtin_amdgcn_sicmp, "LUiiiIi", "nc")
 BUILTIN(__builtin_amdgcn_sicmpl, "LUiLiLiIi", "nc")
 BUILTIN(__builtin_amdgcn_fcmp, "LUiddIi", "nc")
 BUILTIN(__builtin_amdgcn_fcmpf, "LUiffIi", "nc")
 BUILTIN(__builtin_amdgcn_ds_swizzle, "iiIi", "nc")
+BUILTIN(__builtin_amdgcn_ds_permute, "iii", "nc")
+BUILTIN(__builtin_amdgcn_ds_bpermute, "iii", "nc")
+BUILTIN(__builtin_amdgcn_readfirstlane, "ii", "nc")
+BUILTIN(__builtin_amdgcn_readlane, "iii", "nc")
+BUILTIN(__builtin_amdgcn_mov_dpp, "iiIiIiIiIb", "nc")
 BUILTIN(__builtin_amdgcn_fmed3f, "ffff", "nc")
 
 //===----------------------------------------------------------------------===//
 // VI+ only builtins.
 //===----------------------------------------------------------------------===//
 
 TARGET_BUILTIN(__builtin_amdgcn_div_fixuph, "hhhh", "nc", "16-bit-insts")
 TARGET_BUILTIN(__builtin_amdgcn_rcph, "hh", "nc", "16-bit-insts")

lib/CodeGen/CGBuiltin.cpp

     llvm::Value *F = CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas,
                                       Src0->getType());
     llvm::Value *Src3ToBool = Builder.CreateIsNotNull(Src3);
     return Builder.CreateCall(F, {Src0, Src1, Src2, Src3ToBool});
   }
 
   case AMDGPU::BI__builtin_amdgcn_ds_swizzle:
     return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_ds_swizzle);
+  case AMDGPU::BI__builtin_amdgcn_ds_permute:
+    return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_ds_permute);
+  case AMDGPU::BI__builtin_amdgcn_ds_bpermute:
+    return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_ds_bpermute);
+  case AMDGPU::BI__builtin_amdgcn_readfirstlane:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_readfirstlane);
+  case AMDGPU::BI__builtin_amdgcn_readlane:
+    return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_readlane);
+  case AMDGPU::BI__builtin_amdgcn_mov_dpp: {
+    llvm::SmallVector<llvm::Value *, 5> Args;
+    for (unsigned I = 0; I != 5; ++I)
+      Args.push_back(EmitScalarExpr(E->getArg(I)));
+    Value *F = CGM.getIntrinsic(Intrinsic::amdgcn_mov_dpp,
+                                    Args[0]->getType());
+    return Builder.CreateCall(F, Args);
+  }
   case AMDGPU::BI__builtin_amdgcn_div_fixup:
   case AMDGPU::BI__builtin_amdgcn_div_fixupf:
   case AMDGPU::BI__builtin_amdgcn_div_fixuph:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_div_fixup);
   case AMDGPU::BI__builtin_amdgcn_trig_preop:
   case AMDGPU::BI__builtin_amdgcn_trig_preopf:
     return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_trig_preop);
   case AMDGPU::BI__builtin_amdgcn_rcp:

test/CodeGenOpenCL/builtins-amdgcn.cl

 
 // CHECK-LABEL: @test_ds_swizzle
 // CHECK: call i32 @llvm.amdgcn.ds.swizzle(i32 %a, i32 32)
 void test_ds_swizzle(global int* out, int a)
 {
   *out = __builtin_amdgcn_ds_swizzle(a, 32);
 }
 
+// CHECK-LABEL: @test_ds_permute
+// CHECK: call i32 @llvm.amdgcn.ds.permute.i32(i32 %a, i32 %b)
+void test_ds_permute(global int* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ds_permute(a, b);
+}
+
+// CHECK-LABEL: @test_ds_bpermute
+// CHECK: call i32 @llvm.amdgcn.ds.bpermute.i32(i32 %a, i32 %b)
+void test_ds_bpermute(global int* out, int a, int b)
+{
+  *out = __builtin_amdgcn_ds_bpermute(a, b);
+}
+
+// CHECK-LABEL: @test_readfirstlane
+// CHECK: call i32 @llvm.amdgcn.readfirstlane(i32 %a)
+void test_readfirstlane(global int* out, int a)
+{
+  *out = __builtin_amdgcn_readfirstlane(a);
+}
+
+// CHECK-LABEL: @test_readlane
+// CHECK: call i32 @llvm.amdgcn.readlane(i32 %a, i32 %b)
+void test_readlane(global int* out, int a, int b)
+{
+  *out = __builtin_amdgcn_readlane(a, b);
+}
+
+// CHECK-LABEL: @test_mov_dpp
+// CHECK: call i32 @llvm.amdgcn.mov.dpp.i32(i32 %src, i32 0, i32 0, i32 0, i1 false)
+void test_mov_dpp(global int* out, int src)
+{
+  *out = __builtin_amdgcn_mov_dpp(src, 0, 0, 0, false);
+}
+
 // CHECK-LABEL: @test_fcmp_f32
 // CHECK: call i64 @llvm.amdgcn.fcmp.f32(float %a, float %b, i32 5)
 void test_fcmp_f32(global ulong* out, float a, float b)
 {
   *out = __builtin_amdgcn_fcmpf(a, b, 5);
 }
 
 // CHECK-LABEL: @test_fcmp_f64

test/SemaOpenCL/builtins-amdgcn-error.cl

 {
   *out = __builtin_amdgcn_ds_swizzle(a, b); // expected-error {{argument to '__builtin_amdgcn_ds_swizzle' must be a constant integer}}
 }
 
 void test_s_getreg(global int* out, int a)
 {
   *out = __builtin_amdgcn_s_getreg(a); // expected-error {{argument to '__builtin_amdgcn_s_getreg' must be a constant integer}}
 }
+
+void test_mov_dpp(global int* out, int a, int b, int c, int d, bool e)
+{
+  *out = __builtin_amdgcn_mov_dpp(a, b, 0, 0, false); // expected-error {{argument to '__builtin_amdgcn_mov_dpp' must be a constant integer}}
+  *out = __builtin_amdgcn_mov_dpp(a, 0, c, 0, false); // expected-error {{argument to '__builtin_amdgcn_mov_dpp' must be a constant integer}}
+  *out = __builtin_amdgcn_mov_dpp(a, 0, 0, d, false); // expected-error {{argument to '__builtin_amdgcn_mov_dpp' must be a constant integer}}
+  *out = __builtin_amdgcn_mov_dpp(a, 0, 0, 0, e); // expected-error {{argument to '__builtin_amdgcn_mov_dpp' must be a constant integer}}
+}