nv50/ir: enable early fragment test with explicit user control

[mesa.git] / src / gallium / drivers / nouveau / codegen / nv50_ir_emit_gm107.cpp

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp
index 465319af5cec370fee8cb8dbafa58a8eeeb710a0..35930e3da4f7276b310b76ea01de7049e5b1fc62 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp
@@ -126,6 +126,7 @@ private:
    void emitF2I();
    void emitI2F();
    void emitI2I();
+   void emitSEL();
    void emitSHFL();
 
    void emitDADD();
@@ -177,6 +178,7 @@ private:
    void emitAL2P();
    void emitIPA();
    void emitATOM();
+   void emitATOMS();
    void emitCCTL();
 
    void emitPIXLD();
@@ -193,6 +195,11 @@ private:
    void emitNOP();
    void emitKIL();
    void emitOUT();
+
+   void emitBAR();
+   void emitMEMBAR();
+
+   void emitVOTE();
 };
 
 /*******************************************************************************
@@ -248,6 +255,8 @@ CodeEmitterGM107::emitSYS(int pos, const Value *val)
    case SV_INVOCATION_ID  : id = 0x11; break;
    case SV_THREAD_KILL    : id = 0x13; break;
    case SV_INVOCATION_INFO: id = 0x1d; break;
+   case SV_TID            : id = 0x21 + val->reg.data.sv.index; break;
+   case SV_CTAID          : id = 0x25 + val->reg.data.sv.index; break;
    default:
       assert(!"invalid system value");
       id = 0;
@@ -673,7 +682,12 @@ CodeEmitterGM107::emitMOV()
        (insn->sType != TYPE_F32 && !longIMMD(insn->src(0)))) {
       switch (insn->src(0).getFile()) {
       case FILE_GPR:
-         emitInsn(0x5c980000);
+         if (insn->def(0).getFile() == FILE_PREDICATE) {
+            emitInsn(0x5b6a0000);
+            emitGPR (0x08);
+         } else {
+            emitInsn(0x5c980000);
+         }
          emitGPR (0x14, insn->src(0));
          break;
       case FILE_MEMORY_CONST:
@@ -684,18 +698,32 @@ CodeEmitterGM107::emitMOV()
          emitInsn(0x38980000);
          emitIMMD(0x14, 19, insn->src(0));
          break;
+      case FILE_PREDICATE:
+         emitInsn(0x50880000);
+         emitPRED(0x0c, insn->src(0));
+         emitPRED(0x1d);
+         emitPRED(0x27);
+         break;
       default:
          assert(!"bad src file");
          break;
       }
-      emitField(0x27, 4, insn->lanes);
+      if (insn->def(0).getFile() != FILE_PREDICATE &&
+          insn->src(0).getFile() != FILE_PREDICATE)
+         emitField(0x27, 4, insn->lanes);
    } else {
       emitInsn (0x01000000);
       emitIMMD (0x14, 32, insn->src(0));
       emitField(0x0c, 4, insn->lanes);
    }
 
-   emitGPR(0x00, insn->def(0));
+   if (insn->def(0).getFile() == FILE_PREDICATE) {
+      emitPRED(0x27);
+      emitPRED(0x03, insn->def(0));
+      emitPRED(0x00);
+   } else {
+      emitGPR(0x00, insn->def(0));
+   }
 }
 
 void
@@ -866,6 +894,32 @@ CodeEmitterGM107::emitI2I()
    emitGPR  (0x00, insn->def(0));
 }
 
+void
+CodeEmitterGM107::emitSEL()
+{
+   switch (insn->src(1).getFile()) {
+   case FILE_GPR:
+      emitInsn(0x5ca00000);
+      emitGPR (0x14, insn->src(1));
+      break;
+   case FILE_MEMORY_CONST:
+      emitInsn(0x4ca00000);
+      emitCBUF(0x22, -1, 0x14, 16, 2, insn->src(1));
+      break;
+   case FILE_IMMEDIATE:
+      emitInsn(0x38a00000);
+      emitIMMD(0x14, 19, insn->src(1));
+      break;
+   default:
+      assert(!"bad src1 file");
+      break;
+   }
+
+   emitPRED(0x27, insn->src(2));
+   emitGPR (0x08, insn->src(0));
+   emitGPR (0x00, insn->def(0));
+}
+
 void
 CodeEmitterGM107::emitSHFL()
 {
@@ -1512,7 +1566,10 @@ CodeEmitterGM107::emitFSWZADD()
    emitRND  (0x27);
    emitField(0x26, 1, insn->lanes); /* abused for .ndv */
    emitField(0x1c, 8, insn->subOp);
-   emitGPR  (0x14, insn->src(1));
+   if (insn->predSrc != 1)
+      emitGPR  (0x14, insn->src(1));
+   else
+      emitGPR  (0x14);
    emitGPR  (0x08, insn->src(0));
    emitGPR  (0x00, insn->def(0));
 }
@@ -1554,11 +1611,13 @@ CodeEmitterGM107::emitLOP()
          break;
       }
       emitPRED (0x30);
+      emitX    (0x2b);
       emitField(0x29, 2, lop);
       emitINV  (0x28, insn->src(1));
       emitINV  (0x27, insn->src(0));
    } else {
       emitInsn (0x04000000);
+      emitX    (0x39);
       emitINV  (0x38, insn->src(1));
       emitINV  (0x37, insn->src(0));
       emitField(0x35, 2, lop);
@@ -1626,9 +1685,11 @@ CodeEmitterGM107::emitIADD()
       emitNEG(0x31, insn->src(0));
       emitNEG(0x30, insn->src(1));
       emitCC (0x2f);
+      emitX  (0x2b);
    } else {
       emitInsn(0x1c000000);
       emitSAT (0x36);
+      emitX   (0x35);
       emitCC  (0x34);
       emitIMMD(0x14, 32, insn->src(1));
    }
@@ -2304,22 +2365,34 @@ void
 CodeEmitterGM107::emitATOM()
 {
    unsigned dType, subOp;
-   switch (insn->dType) {
-   case TYPE_U32: dType = 0; break;
-   case TYPE_S32: dType = 1; break;
-   case TYPE_U64: dType = 2; break;
-   case TYPE_F32: dType = 3; break;
-   case TYPE_B128: dType = 4; break;
-   case TYPE_S64: dType = 5; break;
-   default: assert(!"unexpected dType"); dType = 0; break;
-   }
-   if (insn->subOp == NV50_IR_SUBOP_ATOM_EXCH)
-      subOp = 8;
-   else
-      subOp = insn->subOp;
-   assert(insn->subOp != NV50_IR_SUBOP_ATOM_CAS); /* XXX */
 
-   emitInsn (0xed000000);
+   if (insn->subOp == NV50_IR_SUBOP_ATOM_CAS) {
+      switch (insn->dType) {
+      case TYPE_U32: dType = 0; break;
+      case TYPE_U64: dType = 1; break;
+      default: assert(!"unexpected dType"); dType = 0; break;
+      }
+      subOp = 15;
+
+      emitInsn (0xee000000);
+   } else {
+      switch (insn->dType) {
+      case TYPE_U32: dType = 0; break;
+      case TYPE_S32: dType = 1; break;
+      case TYPE_U64: dType = 2; break;
+      case TYPE_F32: dType = 3; break;
+      case TYPE_B128: dType = 4; break;
+      case TYPE_S64: dType = 5; break;
+      default: assert(!"unexpected dType"); dType = 0; break;
+      }
+      if (insn->subOp == NV50_IR_SUBOP_ATOM_EXCH)
+         subOp = 8;
+      else
+         subOp = insn->subOp;
+
+      emitInsn (0xed000000);
+   }
+
    emitField(0x34, 4, subOp);
    emitField(0x31, 3, dType);
    emitField(0x30, 1, insn->src(0).getIndirect(0)->getSize() == 8);
@@ -2328,6 +2401,45 @@ CodeEmitterGM107::emitATOM()
    emitGPR  (0x00, insn->def(0));
 }
 
+void
+CodeEmitterGM107::emitATOMS()
+{
+   unsigned dType, subOp;
+
+   if (insn->subOp == NV50_IR_SUBOP_ATOM_CAS) {
+      switch (insn->dType) {
+      case TYPE_U32: dType = 0; break;
+      case TYPE_U64: dType = 1; break;
+      default: assert(!"unexpected dType"); dType = 0; break;
+      }
+      subOp = 4;
+
+      emitInsn (0xee000000);
+      emitField(0x34, 1, dType);
+   } else {
+      switch (insn->dType) {
+      case TYPE_U32: dType = 0; break;
+      case TYPE_S32: dType = 1; break;
+      case TYPE_U64: dType = 2; break;
+      case TYPE_S64: dType = 3; break;
+      default: assert(!"unexpected dType"); dType = 0; break;
+      }
+
+      if (insn->subOp == NV50_IR_SUBOP_ATOM_EXCH)
+         subOp = 8;
+      else
+         subOp = insn->subOp;
+
+      emitInsn (0xec000000);
+      emitField(0x1c, 3, dType);
+   }
+
+   emitField(0x34, 4, subOp);
+   emitGPR  (0x14, insn->src(1));
+   emitADDR (0x08, 0x12, 22, 0, insn->src(0));
+   emitGPR  (0x00, insn->def(0));
+}
+
 void
 CodeEmitterGM107::emitCCTL()
 {
@@ -2604,6 +2716,85 @@ CodeEmitterGM107::emitOUT()
    emitGPR  (0x00, insn->def(0));
 }
 
+void
+CodeEmitterGM107::emitBAR()
+{
+   uint8_t subop;
+
+   emitInsn (0xf0a80000);
+
+   switch (insn->subOp) {
+   case NV50_IR_SUBOP_BAR_RED_POPC: subop = 0x02; break;
+   case NV50_IR_SUBOP_BAR_RED_AND:  subop = 0x0a; break;
+   case NV50_IR_SUBOP_BAR_RED_OR:   subop = 0x12; break;
+   case NV50_IR_SUBOP_BAR_ARRIVE:   subop = 0x81; break;
+   default:
+      subop = 0x80;
+      assert(insn->subOp == NV50_IR_SUBOP_BAR_SYNC);
+      break;
+   }
+
+   emitField(0x20, 8, subop);
+
+   // barrier id
+   if (insn->src(0).getFile() == FILE_GPR) {
+      emitGPR(0x08, insn->src(0));
+   } else {
+      ImmediateValue *imm = insn->getSrc(0)->asImm();
+      assert(imm);
+      emitField(0x08, 8, imm->reg.data.u32);
+      emitField(0x2b, 1, 1);
+   }
+
+   // thread count
+   if (insn->src(1).getFile() == FILE_GPR) {
+      emitGPR(0x14, insn->src(1));
+   } else {
+      ImmediateValue *imm = insn->getSrc(0)->asImm();
+      assert(imm);
+      emitField(0x14, 12, imm->reg.data.u32);
+      emitField(0x2c, 1, 1);
+   }
+
+   if (insn->srcExists(2) && (insn->predSrc != 2)) {
+      emitPRED (0x27, insn->src(2));
+      emitField(0x2a, 1, insn->src(2).mod == Modifier(NV50_IR_MOD_NOT));
+   } else {
+      emitField(0x27, 3, 7);
+   }
+}
+
+void
+CodeEmitterGM107::emitMEMBAR()
+{
+   emitInsn (0xef980000);
+   emitField(0x08, 2, insn->subOp >> 2);
+}
+
+void
+CodeEmitterGM107::emitVOTE()
+{
+   int subOp;
+
+   assert(insn->src(0).getFile() == FILE_PREDICATE &&
+          insn->def(1).getFile() == FILE_PREDICATE);
+
+   switch (insn->subOp) {
+   case NV50_IR_SUBOP_VOTE_ANY: subOp = 1; break;
+   default:
+      assert(insn->subOp == NV50_IR_SUBOP_VOTE_ALL);
+      subOp = 0;
+      break;
+   }
+
+   emitInsn (0x50d80000);
+   emitField(0x30, 2, subOp);
+   emitGPR  (0x00, insn->def(0));
+   emitPRED (0x2d, insn->def(1));
+   emitField(0x2a, 1, insn->src(0).mod == Modifier(NV50_IR_MOD_NOT));
+   emitPRED (0x27, insn->src(0));
+}
+
 /*******************************************************************************
  * assembler front-end
  ******************************************************************************/
@@ -2680,11 +2871,7 @@ CodeEmitterGM107::emitInstruction(Instruction *i)
       emitRAM();
       break;
    case OP_MOV:
-      if (insn->def(0).getFile() == FILE_GPR &&
-          insn->src(0).getFile() != FILE_PREDICATE)
-         emitMOV();
-      else
-         assert(!"R2P/P2R");
+      emitMOV();
       break;
    case OP_RDSV:
       emitS2R();
@@ -2696,7 +2883,10 @@ CodeEmitterGM107::emitInstruction(Instruction *i)
    case OP_CEIL:
    case OP_TRUNC:
    case OP_CVT:
-      if (isFloatType(insn->dType)) {
+      if (insn->op == OP_CVT && (insn->def(0).getFile() == FILE_PREDICATE ||
+                                 insn->src(0).getFile() == FILE_PREDICATE)) {
+         emitMOV();
+      } else if (isFloatType(insn->dType)) {
          if (isFloatType(insn->sType))
             emitF2F();
          else
@@ -2800,6 +2990,9 @@ CodeEmitterGM107::emitInstruction(Instruction *i)
             emitISETP();
       }
       break;
+   case OP_SELP:
+      emitSEL();
+      break;
    case OP_PRESIN:
    case OP_PREEX2:
       emitRRO();
@@ -2838,13 +3031,16 @@ CodeEmitterGM107::emitInstruction(Instruction *i)
       case FILE_MEMORY_SHARED: emitSTS(); break;
       case FILE_MEMORY_GLOBAL: emitST(); break;
       default:
-         assert(!"invalid load");
+         assert(!"invalid store");
          emitNOP();
          break;
       }
       break;
    case OP_ATOM:
-      emitATOM();
+      if (insn->src(0).getFile() == FILE_MEMORY_SHARED)
+         emitATOMS();
+      else
+         emitATOM();
       break;
    case OP_CCTL:
       emitCCTL();
@@ -2904,6 +3100,15 @@ CodeEmitterGM107::emitInstruction(Instruction *i)
    case OP_RESTART:
       emitOUT();
       break;
+   case OP_BAR:
+      emitBAR();
+      break;
+   case OP_MEMBAR:
+      emitMEMBAR();
+      break;
+   case OP_VOTE:
+      emitVOTE();
+      break;
    default:
       assert(!"invalid opcode");
       emitNOP();