R600 intrinsics patch

[Ronie Salgado]

Hello,

This patch is to be able to use some of the R600/AMDGPU intrinsics from
Clang, and also to use amdgpu instead of r600. Part of the discussions
concerning this patch is here:
https://bugs.freedesktop.org/show_bug.cgi?id=86326

Greetings,
Ronie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150317/e7d20769/attachment.html>
-------------- next part --------------
diff --git a/include/clang/Basic/BuiltinsNVPTX.def b/include/clang/Basic/BuiltinsNVPTX.def
index 9c920dc..a72db13 100644
--- a/include/clang/Basic/BuiltinsNVPTX.def
+++ b/include/clang/Basic/BuiltinsNVPTX.def
@@ -59,6 +59,11 @@ BUILTIN(__builtin_ptx_read_pm3, "i", "n")
 
 BUILTIN(__builtin_ptx_bar_sync, "vi", "n")
 
+BUILTIN(__builtin_ptx_read_workdim, "i", "nc")
+
+BUILTIN(__builtin_ptx_read_global_offset_x, "i", "nc")
+BUILTIN(__builtin_ptx_read_global_offset_y, "i", "nc")
+BUILTIN(__builtin_ptx_read_global_offset_z, "i", "nc")
 
 // Builtins exposed as part of NVVM
 // MISC
diff --git a/include/clang/Basic/BuiltinsR600.def b/include/clang/Basic/BuiltinsR600.def
index 84fc4fa..ad4361b 100644
--- a/include/clang/Basic/BuiltinsR600.def
+++ b/include/clang/Basic/BuiltinsR600.def
@@ -14,6 +14,26 @@
 
 // The format of this database matches clang/Basic/Builtins.def.
 
+BUILTIN(__builtin_amdgpu_read_global_size_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_global_size_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_global_size_z, "i", "nc")
+
+BUILTIN(__builtin_amdgpu_read_local_size_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_local_size_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_local_size_z, "i", "nc")
+
+BUILTIN(__builtin_amdgpu_read_ngroups_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_ngroups_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_ngroups_z, "i", "nc")
+
+BUILTIN(__builtin_amdgpu_read_tgid_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_tgid_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_tgid_z, "i", "nc")
+
+BUILTIN(__builtin_amdgpu_read_tidig_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_tidig_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_tidig_z, "i", "nc")
+
 BUILTIN(__builtin_amdgpu_div_scale, "dddbb*", "n")
 BUILTIN(__builtin_amdgpu_div_scalef, "fffbb*", "n")
 BUILTIN(__builtin_amdgpu_div_fmas, "ddddb", "nc")
@@ -33,4 +53,9 @@ BUILTIN(__builtin_amdgpu_ldexpf, "ffi", "nc")
 BUILTIN(__builtin_amdgpu_class, "bdi", "nc")
 BUILTIN(__builtin_amdgpu_classf, "bfi", "nc")
 
+BUILTIN(__builtin_amdgpu_read_workdim, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_global_offset_x, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_global_offset_y, "i", "nc")
+BUILTIN(__builtin_amdgpu_read_global_offset_z, "i", "nc")
+
 #undef BUILTIN
diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp
index 4c79a34..6d58e04 100644
--- a/lib/CodeGen/CGBuiltin.cpp
+++ b/lib/CodeGen/CGBuiltin.cpp
@@ -6428,7 +6428,7 @@ Value *CodeGenFunction::EmitR600BuiltinExpr(unsigned BuiltinID,
     llvm::Value *Y = EmitScalarExpr(E->getArg(1));
     llvm::Value *Z = EmitScalarExpr(E->getArg(2));
 
-    llvm::Value *Callee = CGM.getIntrinsic(Intrinsic::AMDGPU_div_scale,
+    llvm::Value *Callee = CGM.getIntrinsic(Intrinsic::amdgpu_div_scale,
                                            X->getType());
 
     llvm::Value *Tmp = Builder.CreateCall3(Callee, X, Y, Z);
@@ -6451,32 +6451,32 @@ Value *CodeGenFunction::EmitR600BuiltinExpr(unsigned BuiltinID,
     llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
     llvm::Value *Src3 = EmitScalarExpr(E->getArg(3));
 
-    llvm::Value *F = CGM.getIntrinsic(Intrinsic::AMDGPU_div_fmas,
+    llvm::Value *F = CGM.getIntrinsic(Intrinsic::amdgpu_div_fmas,
                                       Src0->getType());
     llvm::Value *Src3ToBool = Builder.CreateIsNotNull(Src3);
     return Builder.CreateCall4(F, Src0, Src1, Src2, Src3ToBool);
   }
   case R600::BI__builtin_amdgpu_div_fixup:
   case R600::BI__builtin_amdgpu_div_fixupf:
-    return emitTernaryFPBuiltin(*this, E, Intrinsic::AMDGPU_div_fixup);
+    return emitTernaryFPBuiltin(*this, E, Intrinsic::amdgpu_div_fixup);
   case R600::BI__builtin_amdgpu_trig_preop:
   case R600::BI__builtin_amdgpu_trig_preopf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_trig_preop);
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgpu_trig_preop);
   case R600::BI__builtin_amdgpu_rcp:
   case R600::BI__builtin_amdgpu_rcpf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rcp);
+    return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgpu_rcp);
   case R600::BI__builtin_amdgpu_rsq:
   case R600::BI__builtin_amdgpu_rsqf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rsq);
+    return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgpu_rsq);
   case R600::BI__builtin_amdgpu_rsq_clamped:
   case R600::BI__builtin_amdgpu_rsq_clampedf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rsq_clamped);
+    return emitUnaryFPBuiltin(*this, E, Intrinsic::amdgpu_rsq_clamped);
   case R600::BI__builtin_amdgpu_ldexp:
   case R600::BI__builtin_amdgpu_ldexpf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_ldexp);
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgpu_ldexp);
   case R600::BI__builtin_amdgpu_class:
   case R600::BI__builtin_amdgpu_classf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_class);
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgpu_class);
    default:
     return nullptr;
   }