From 5523502ff917803166051c8947f5dd3b23c6fcf8 Mon Sep 17 00:00:00 2001
From: Tom Stellard <thomas.stellard@amd.com>
Date: Wed, 23 May 2012 12:25:04 -0400
Subject: [PATCH] radeon/llvm: Use -1 as true value for SET* integer
 instructions

---
 .../drivers/radeon/AMDGPUISelLowering.cpp     | 36 +++++++++----------
 .../drivers/radeon/AMDGPUISelLowering.h       |  4 +--
 .../drivers/radeon/R600Instructions.td        | 20 +++++------
 3 files changed, 28 insertions(+), 32 deletions(-)

diff --git a/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp b/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
index a6827e5a24a..e90d71a5c27 100644
--- a/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
+++ b/src/gallium/drivers/radeon/AMDGPUISelLowering.cpp
@@ -119,10 +119,6 @@ SDValue AMDGPUTargetLowering::LowerSELECT_CC(SDValue Op,
   ISD::CondCode CCOpcode = cast<CondCodeSDNode>(CC)->get();
   SDValue Temp;
 
- //cmovlog = src0 != 0.0f ? src1 : src2
- //cmovlog = src0 == 0.0f ? src2 : src1
- //cnde = src0 == 0.0f ? src1 : src2
-
   // LHS and RHS are guaranteed to be the same value type
   EVT CompareVT = LHS.getValueType();
 
@@ -151,16 +147,16 @@ SDValue AMDGPUTargetLowering::LowerSELECT_CC(SDValue Op,
     RHS = DAG.getNode(ConversionOp, DL, VT, RHS);
   }
 
-  // If true is 1 and false is 0 or vice-versa we can handle this with a native
-  // instruction (SET* instructions).
-  if ((isOne(True) && isZero(False))) {
+  // If True is a hardware TRUE value and False is a hardware FALSE value or
+  // vice-versa we can handle this with a native instruction (SET* instructions).
+  if ((isHWTrueValue(True) && isHWFalseValue(False))) {
     return DAG.getNode(ISD::SELECT_CC, DL, VT, LHS, RHS, True, False, CC);
   }
 
-  // XXX If true is 0 and 1 is false, we can handle this with a native
-  // instruction, but we need to swap true and false and change the
-  // conditional.
-  if (isOne(False) && isZero(True)) {
+  // XXX If True is a hardware TRUE value and False is a hardware FALSE value,
+  // we can handle this with a native instruction, but we need to swap true
+  // and false and change the conditional.
+  if (isHWTrueValue(False) && isHWFalseValue(True)) {
   }
 
   // XXX Check if we can lower this to a SELECT or if it is supported by a native
@@ -196,14 +192,14 @@ SDValue AMDGPUTargetLowering::LowerSELECT_CC(SDValue Op,
 
   // If we make it this for it means we have no native instructions to handle
   // this SELECT_CC, so we must lower it.
-  SDValue One, Zero;
+  SDValue HWTrue, HWFalse;
 
   if (VT == MVT::f32) {
-    One = DAG.getConstantFP(1.0f, VT);
-    Zero = DAG.getConstantFP(0.0f, VT);
+    HWTrue = DAG.getConstantFP(1.0f, VT);
+    HWFalse = DAG.getConstantFP(0.0f, VT);
   } else if (VT == MVT::i32) {
-    One = DAG.getConstant(1, VT);
-    Zero = DAG.getConstant(0, VT);
+    HWTrue = DAG.getConstant(-1, VT);
+    HWFalse = DAG.getConstant(0, VT);
   }
   else {
     assert(!"Unhandled value type in LowerSELECT_CC");
@@ -211,7 +207,7 @@ SDValue AMDGPUTargetLowering::LowerSELECT_CC(SDValue Op,
 
   // Lower this unsupported SELECT_CC into a combination of two supported
   // SELECT_CC operations.
-  SDValue Cond = DAG.getNode(ISD::SELECT_CC, DL, VT, LHS, RHS, One, Zero, CC);
+  SDValue Cond = DAG.getNode(ISD::SELECT_CC, DL, VT, LHS, RHS, HWTrue, HWFalse, CC);
 
   return DAG.getNode(ISD::SELECT, DL, VT, Cond, True, False);
 }
@@ -220,18 +216,18 @@ SDValue AMDGPUTargetLowering::LowerSELECT_CC(SDValue Op,
 // Helper functions
 //===----------------------------------------------------------------------===//
 
-bool AMDGPUTargetLowering::isOne(SDValue Op) const
+bool AMDGPUTargetLowering::isHWTrueValue(SDValue Op) const
 {
   if (ConstantFPSDNode * CFP = dyn_cast<ConstantFPSDNode>(Op)) {
     return CFP->isExactlyValue(1.0);
   }
   if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(Op)) {
-    return C->isOne();
+    return C->isAllOnesValue();
   }
   return false;
 }
 
-bool AMDGPUTargetLowering::isZero(SDValue Op) const
+bool AMDGPUTargetLowering::isHWFalseValue(SDValue Op) const
 {
   if (ConstantFPSDNode * CFP = dyn_cast<ConstantFPSDNode>(Op)) {
     return CFP->getValueAPF().isZero();
diff --git a/src/gallium/drivers/radeon/AMDGPUISelLowering.h b/src/gallium/drivers/radeon/AMDGPUISelLowering.h
index d37473974bb..f1544b86920 100644
--- a/src/gallium/drivers/radeon/AMDGPUISelLowering.h
+++ b/src/gallium/drivers/radeon/AMDGPUISelLowering.h
@@ -37,8 +37,8 @@ protected:
                  MachineRegisterInfo & MRI, const TargetInstrInfo * TII,
 		 unsigned reg) const;
 
-  bool isOne(SDValue Op) const;
-  bool isZero(SDValue Op) const;
+  bool isHWTrueValue(SDValue Op) const;
+  bool isHWFalseValue(SDValue Op) const;
 
 public:
   AMDGPUTargetLowering(TargetMachine &TM);
diff --git a/src/gallium/drivers/radeon/R600Instructions.td b/src/gallium/drivers/radeon/R600Instructions.td
index 1f2c2d0f1e8..978ccecd339 100644
--- a/src/gallium/drivers/radeon/R600Instructions.td
+++ b/src/gallium/drivers/radeon/R600Instructions.td
@@ -446,7 +446,7 @@ def MIN_UINT : R600_2OP <
 def SETE_INT : R600_2OP <
   0x3A, "SETE_INT",
   [(set (i32 R600_Reg32:$dst),
-   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETEQ))]
+   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETEQ))]
 >;
 
 //  let AMDILOp = AMDILInst.IEQ;
@@ -454,13 +454,13 @@ def SETE_INT : R600_2OP <
 def SETGT_INT : R600_2OP <
   0x3B, "SGT_INT",
   [(set (i32 R600_Reg32:$dst),
-   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETGT))]
+   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETGT))]
 >;
 
 def SETGE_INT : R600_2OP <
 	0x3C, "SETGE_INT",
 	[(set (i32 R600_Reg32:$dst),
-   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETGE))]
+   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETGE))]
 >;
 //  let AMDILOp = AMDILInst.IGE;
 
@@ -468,7 +468,7 @@ def SETGE_INT : R600_2OP <
 def SETNE_INT : R600_2OP <
   0x3D, "SETNE_INT",
   [(set (i32 R600_Reg32:$dst),
-   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETNE))]
+   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETNE))]
 >;
 //let AMDILOp = AMDILInst.INE;
 
@@ -476,7 +476,7 @@ def SETNE_INT : R600_2OP <
 def SETGT_UINT : R600_2OP <
   0x3E, "SETGT_UINT",
   [(set (i32 R600_Reg32:$dst),
-   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETUGT))]
+   (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETUGT))]
 >;
 
 //  let AMDILOp = AMDILInst.UGT;
@@ -484,7 +484,7 @@ def SETGT_UINT : R600_2OP <
 def SETGE_UINT : R600_2OP <
   0x3F, "SETGE_UINT",
   [(set (i32 R600_Reg32:$dst),
-    (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETUGE))]
+    (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETUGE))]
 >;
 //  let AMDILOp = AMDILInst.UGE;
 
@@ -1134,25 +1134,25 @@ def : Pat <
 
 // SETGT_INT reverse args
 def : Pat <
-  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETLT),
+  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETLT),
   (SETGT_INT R600_Reg32:$src1, R600_Reg32:$src0)
 >;
 
 // SETGE_INT reverse args
 def : Pat <
-  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETLE),
+  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETLE),
   (SETGE_INT R600_Reg32:$src1, R600_Reg32:$src0)
 >;
 
 // SETGT_UINT reverse args
 def : Pat <
-  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETULT),
+  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETULT),
   (SETGT_UINT R600_Reg32:$src1, R600_Reg32:$src0)
 >;
 
 // SETGE_UINT reverse args
 def : Pat <
-  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, 1, 0, SETULE),
+  (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETULE),
   (SETGE_UINT R600_Reg32:$src0, R600_Reg32:$src1)
 >;
 
-- 
2.30.2