1 files changed, 36 insertions, 10 deletions
diff --git a/contrib/llvm/lib/Target/R600/AMDGPUCallingConv.td b/contrib/llvm/lib/Target/R600/AMDGPUCallingConv.td
index 9c30515..65cdb24 100644
--- a/contrib/llvm/lib/Target/R600/AMDGPUCallingConv.td
+++ b/contrib/llvm/lib/Target/R600/AMDGPUCallingConv.td
@@ -19,12 +19,13 @@ def CC_SI : CallingConv<[
 
   CCIfInReg<CCIfType<[f32, i32] , CCAssignToReg<[
     SGPR0, SGPR1, SGPR2, SGPR3, SGPR4, SGPR5, SGPR6, SGPR7,
-    SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15
+    SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
+    SGPR16
   ]>>>,
 
   CCIfInReg<CCIfType<[i64] , CCAssignToRegWithShadow<
     [ SGPR0, SGPR2, SGPR4, SGPR6, SGPR8, SGPR10, SGPR12, SGPR14 ],
-    [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR12, SGPR15 ]
+    [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR13, SGPR15 ]
   >>>,
 
   CCIfNotInReg<CCIfType<[f32, i32] , CCAssignToReg<[
@@ -34,15 +35,40 @@ def CC_SI : CallingConv<[
     VGPR24, VGPR25, VGPR26, VGPR27, VGPR28, VGPR29, VGPR30, VGPR31
   ]>>>,
 
-  // This is the default for i64 values.
-  // XXX: We should change this once clang understands the CC_AMDGPU.
-  CCIfType<[i64], CCAssignToRegWithShadow<
-   [ SGPR0, SGPR2, SGPR4, SGPR6, SGPR8, SGPR10, SGPR12, SGPR14 ],
-   [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR13, SGPR15 ]
-  >>
+  CCIfByVal<CCIfType<[i64] , CCAssignToRegWithShadow<
+    [ SGPR0, SGPR2, SGPR4, SGPR6, SGPR8, SGPR10, SGPR12, SGPR14 ],
+    [ SGPR1, SGPR3, SGPR5, SGPR7, SGPR9, SGPR11, SGPR13, SGPR15 ]
+  >>>
+
+]>;
+
+// Calling convention for R600
+def CC_R600 : CallingConv<[
+  CCIfInReg<CCIfType<[v4f32, v4i32] , CCAssignToReg<[
+    T0_XYZW, T1_XYZW, T2_XYZW, T3_XYZW, T4_XYZW, T5_XYZW, T6_XYZW, T7_XYZW,
+    T8_XYZW, T9_XYZW, T10_XYZW, T11_XYZW, T12_XYZW, T13_XYZW, T14_XYZW, T15_XYZW,
+    T16_XYZW, T17_XYZW, T18_XYZW, T19_XYZW, T20_XYZW, T21_XYZW, T22_XYZW,
+    T23_XYZW, T24_XYZW, T25_XYZW, T26_XYZW, T27_XYZW, T28_XYZW, T29_XYZW,
+    T30_XYZW, T31_XYZW, T32_XYZW
+  ]>>>
+]>;
+
+// Calling convention for compute kernels
+def CC_AMDGPU_Kernel : CallingConv<[
+  CCCustom<"allocateStack">
 ]>;
 
 def CC_AMDGPU : CallingConv<[
-  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>().device()"#
-       "->getGeneration() == AMDGPUDeviceInfo::HD7XXX", CCDelegateTo<CC_SI>>
+  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>().getGeneration() >= "
+       "AMDGPUSubtarget::SOUTHERN_ISLANDS && "
+       "State.getMachineFunction().getInfo<SIMachineFunctionInfo>()->"#
+       "ShaderType == ShaderType::COMPUTE", CCDelegateTo<CC_AMDGPU_Kernel>>,
+  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>().getGeneration() < "
+       "AMDGPUSubtarget::SOUTHERN_ISLANDS && "
+       "State.getMachineFunction().getInfo<R600MachineFunctionInfo>()->"
+       "ShaderType == ShaderType::COMPUTE", CCDelegateTo<CC_AMDGPU_Kernel>>,
+  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>()"#
+       ".getGeneration() >= AMDGPUSubtarget::SOUTHERN_ISLANDS", CCDelegateTo<CC_SI>>,
+  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>()"#
+       ".getGeneration() < AMDGPUSubtarget::SOUTHERN_ISLANDS", CCDelegateTo<CC_R600>>
 ]>;