4 files changed, 21 insertions, 30 deletions
diff --git a/src/gallium/drivers/radeon/AMDIL.h b/src/gallium/drivers/radeon/AMDIL.h
index 45638f64280..8bd024a4bd8 100644
--- a/src/gallium/drivers/radeon/AMDIL.h
+++ b/src/gallium/drivers/radeon/AMDIL.h
@@ -129,7 +129,8 @@ enum AddressSpaces {
   ADDRESS_NONE     = 5, // Address space for unknown memory.
   PARAM_D_ADDRESS  = 6, // Address space for direct addressible parameter memory (CONST0)
   PARAM_I_ADDRESS  = 7, // Address space for indirect addressible parameter memory (VTX1)
-  LAST_ADDRESS     = 8
+  USER_SGPR_ADDRESS = 8, // Address space for USER_SGPRS on SI
+  LAST_ADDRESS     = 9
 };
 
 // This union/struct combination is an easy way to read out the
diff --git a/src/gallium/drivers/radeon/SIISelLowering.cpp b/src/gallium/drivers/radeon/SIISelLowering.cpp
index a79aba9358b..1cdcd36c648 100644
--- a/src/gallium/drivers/radeon/SIISelLowering.cpp
+++ b/src/gallium/drivers/radeon/SIISelLowering.cpp
@@ -188,6 +188,9 @@ void SITargetLowering::lowerUSE_SGPR(MachineInstr *MI,
   unsigned dstReg = MI->getOperand(0).getReg();
   int64_t newIndex = MI->getOperand(1).getImm();
   const TargetRegisterClass * dstClass = MRI.getRegClass(dstReg);
+  unsigned DwordWidth = dstClass->getSize() / 4;
+  assert(newIndex % DwordWidth == 0 && "USER_SGPR not properly aligned");
+  newIndex = newIndex / DwordWidth;
 
   unsigned newReg = dstClass->getRegister(newIndex);
   addLiveIn(MI, MF, MRI, TII, newReg); 
diff --git a/src/gallium/drivers/radeon/SIInstructions.td b/src/gallium/drivers/radeon/SIInstructions.td
index a77b8bd7c11..fc8ec4a3394 100644
--- a/src/gallium/drivers/radeon/SIInstructions.td
+++ b/src/gallium/drivers/radeon/SIInstructions.td
@@ -7,6 +7,18 @@
 //
 //===----------------------------------------------------------------------===//
 
+def load_user_sgpr : PatFrag<(ops node:$ptr),
+  (load node:$ptr),
+  [{
+    const Value *Src = cast<LoadSDNode>(N)->getSrcValue();
+    if (Src) {
+      PointerType * PT = dyn_cast<PointerType>(Src->getType());
+      return PT && PT->getAddressSpace() == AMDILAS::USER_SGPR_ADDRESS;
+    }
+    return false;
+  }]
+>;
+
 
 def isSI : Predicate<"Subtarget.device()"
                             "->getGeneration() == AMDILDeviceInfo::HD7XXX">;
@@ -826,26 +838,19 @@ def SI_INTERP_CONST : InstSI <
                                                  imm:$attr, SReg_32:$params))]
 >;
 
-
 def USE_SGPR_32 : InstSI <
   (outs SReg_32:$dst),
   (ins i32imm:$src0),
   "USE_SGPR_32",
-  [(set SReg_32:$dst, (int_SI_use_sgpr imm:$src0))]
-
-> {
-  field bits<32> Inst = 0;
-}
+  [(set (i32 SReg_32:$dst), (load_user_sgpr imm:$src0))]
+>;
 
 def USE_SGPR_64 : InstSI <
   (outs SReg_64:$dst),
   (ins i32imm:$src0),
   "USE_SGPR_64",
-  [(set SReg_64:$dst, (int_SI_use_sgpr imm:$src0))]
-
-> {
-  field bits<32> Inst = 0;
-}
+  [(set (i64 SReg_64:$dst), (load_user_sgpr imm:$src0))]
+>;
 
 def VS_LOAD_BUFFER_INDEX : InstSI <
   (outs VReg_32:$dst),
@@ -869,19 +874,6 @@ def : Pat<
                            0, 0, (i32 SREG_LIT_0))
 >;
 
-def : Pat<
-  (int_SI_use_sgprptrcf32 imm:$src0),
-  (USE_SGPR_64 imm:$src0)
->;
-def : Pat<
-  (int_SI_use_sgprptrci128 imm:$src0),
-  (USE_SGPR_64 imm:$src0)
->;
-def : Pat<
-  (int_SI_use_sgprptrci256 imm:$src0),
-  (USE_SGPR_64 imm:$src0)
->;
-
 /* int_SI_export */
 def : Pat <
   (int_SI_export imm:$en, imm:$vm, imm:$done, imm:$tgt, imm:$compr,
diff --git a/src/gallium/drivers/radeon/SIIntrinsics.td b/src/gallium/drivers/radeon/SIIntrinsics.td
index d8bf4fab3db..95273a2a68f 100644
--- a/src/gallium/drivers/radeon/SIIntrinsics.td
+++ b/src/gallium/drivers/radeon/SIIntrinsics.td
@@ -21,11 +21,6 @@ let TargetPrefix = "SI", isTarget = 1 in {
   def int_SI_vs_load_input : Intrinsic <[llvm_v4f32_ty], [llvm_v4i32_ty, llvm_i16_ty, llvm_i32_ty], []> ;
 
   def int_SI_sample : Intrinsic <[llvm_v4f32_ty], [llvm_i32_ty, llvm_v4f32_ty, llvm_v8i32_ty, llvm_v4i32_ty]>;
-  def int_SI_use_sgpr : Intrinsic <[llvm_anyint_ty], [llvm_i32_ty], [IntrNoMem]>;
-  class int_SI_use_sgprptr : Intrinsic <[llvm_anyptr_ty], [llvm_i32_ty], []>;
-  def int_SI_use_sgprptrcf32 : int_SI_use_sgprptr;
-  def int_SI_use_sgprptrci128 : int_SI_use_sgprptr;
-  def int_SI_use_sgprptrci256 : int_SI_use_sgprptr;
 
   /* Interpolation Intrinsics */