nv50/ir: fix bb positions after exit instructions

[android-x86/external-mesa.git] / src / gallium / drivers / nouveau / codegen / nv50_ir_emit_nv50.cpp

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nv50.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nv50.cpp
index 216e119..ca10848 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nv50.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nv50.cpp
@@ -527,7 +527,8 @@ CodeEmitterNV50::emitForm_ADD(const Instruction *i)
 
    setSrcFileBits(i, NV50_OP_ENC_LONG_ALT);
    setSrc(i, 0, 0);
-   setSrc(i, 1, 2);
+   if (i->predSrc != 1)
+      setSrc(i, 1, 2);
 
    if (i->getIndirect(0, 0)) {
       assert(!i->getIndirect(1, 0));
@@ -840,7 +841,7 @@ CodeEmitterNV50::emitQUADOP(const Instruction *i, uint8_t lane, uint8_t quOp)
 
    emitForm_ADD(i);
 
-   if (!i->srcExists(1))
+   if (!i->srcExists(1) || i->predSrc == 1)
       srcId(i->src(0), 32 + 14);
 }
 
@@ -881,8 +882,7 @@ CodeEmitterNV50::emitPFETCH(const Instruction *i)
 }
 
 static void
-interpApply(const InterpEntry *entry, uint32_t *code,
-      bool force_persample_interp, bool flatshade)
+interpApply(const FixupEntry *entry, uint32_t *code, const FixupData& data)
 {
    int ipa = entry->ipa;
    int encSize = entry->reg;
@@ -890,7 +890,7 @@ interpApply(const InterpEntry *entry, uint32_t *code,
 
    if ((ipa & NV50_IR_INTERP_SAMPLE_MASK) == NV50_IR_INTERP_DEFAULT &&
        (ipa & NV50_IR_INTERP_MODE_MASK) != NV50_IR_INTERP_FLAT) {
-      if (force_persample_interp) {
+      if (data.force_persample_interp) {
          if (encSize == 8)
             code[loc + 1] |= 1 << 16;
          else
@@ -1124,12 +1124,15 @@ CodeEmitterNV50::emitIMUL(const Instruction *i)
 {
    code[0] = 0x40000000;
 
+   if (i->src(1).getFile() == FILE_IMMEDIATE) {
+      if (i->sType == TYPE_S16)
+         code[0] |= 0x8100;
+      code[1] = 0;
+      emitForm_IMM(i);
+   } else
    if (i->encSize == 8) {
       code[1] = (i->sType == TYPE_S16) ? (0x8000 | 0x4000) : 0x0000;
-      if (i->src(1).getFile() == FILE_IMMEDIATE)
-         emitForm_IMM(i);
-      else
-         emitForm_MAD(i);
+      emitForm_MAD(i);
    } else {
       if (i->sType == TYPE_S16)
          code[0] |= 0x8100;
@@ -1190,29 +1193,45 @@ CodeEmitterNV50::emitDMUL(const Instruction *i)
 void
 CodeEmitterNV50::emitIMAD(const Instruction *i)
 {
+   int mode;
    code[0] = 0x60000000;
-   if (isSignedType(i->sType))
-      code[1] = i->saturate ? 0x40000000 : 0x20000000;
-   else
-      code[1] = 0x00000000;
 
-   int neg1 = i->src(0).mod.neg() ^ i->src(1).mod.neg();
-   int neg2 = i->src(2).mod.neg();
-
-   assert(!(neg1 & neg2));
-   code[1] |= neg1 << 27;
-   code[1] |= neg2 << 26;
+   assert(!i->src(0).mod && !i->src(1).mod && !i->src(2).mod);
+   if (!isSignedType(i->sType))
+      mode = 0;
+   else if (i->saturate)
+      mode = 2;
+   else
+      mode = 1;
 
-   if (i->src(1).getFile() == FILE_IMMEDIATE)
+   if (i->src(1).getFile() == FILE_IMMEDIATE) {
+      code[1] = 0;
       emitForm_IMM(i);
-   else
+      code[0] |= (mode & 1) << 8 | (mode & 2) << 14;
+      if (i->flagsSrc >= 0) {
+         assert(!(code[0] & 0x10400000));
+         assert(SDATA(i->src(i->flagsSrc)).id == 0);
+         code[0] |= 0x10400000;
+      }
+   } else
+   if (i->encSize == 4) {
+      emitForm_MUL(i);
+      code[0] |= (mode & 1) << 8 | (mode & 2) << 14;
+      if (i->flagsSrc >= 0) {
+         assert(!(code[0] & 0x10400000));
+         assert(SDATA(i->src(i->flagsSrc)).id == 0);
+         code[0] |= 0x10400000;
+      }
+   } else {
+      code[1] = mode << 29;
       emitForm_MAD(i);
 
-   if (i->flagsSrc >= 0) {
-      // add with carry from $cX
-      assert(!(code[1] & 0x0c000000) && !i->getPredicate());
-      code[1] |= 0xc << 24;
-      srcId(i->src(i->flagsSrc), 32 + 12);
+      if (i->flagsSrc >= 0) {
+         // add with carry from $cX
+         assert(!(code[1] & 0x0c000000) && !i->getPredicate());
+         code[1] |= 0xc << 24;
+         srcId(i->src(i->flagsSrc), 32 + 12);
+      }
    }
 }
 
@@ -1444,6 +1463,7 @@ CodeEmitterNV50::emitSFnOp(const Instruction *i, uint8_t subOp)
 
    if (i->encSize == 4) {
       assert(i->op == OP_RCP);
+      assert(!i->saturate);
       code[0] |= i->src(0).mod.abs() << 15;
       code[0] |= i->src(0).mod.neg() << 22;
       emitForm_MUL(i);
@@ -1451,6 +1471,10 @@ CodeEmitterNV50::emitSFnOp(const Instruction *i, uint8_t subOp)
       code[1] = subOp << 29;
       code[1] |= i->src(0).mod.abs() << 20;
       code[1] |= i->src(0).mod.neg() << 26;
+      if (i->saturate) {
+         assert(subOp == 6 && i->op == OP_EX2);
+         code[1] |= 1 << 27;
+      }
       emitForm_MAD(i);
    }
 }
@@ -1609,7 +1633,9 @@ CodeEmitterNV50::emitTEX(const TexInstruction *i)
    code[1] |= (i->tex.mask & 0xc) << 12;
 
    if (i->tex.liveOnly)
-      code[1] |= 4;
+      code[1] |= 1 << 2;
+   if (i->tex.derivAll)
+      code[1] |= 1 << 3;
 
    defId(i->def(0), 2);
 
@@ -2054,8 +2080,9 @@ CodeEmitterNV50::getMinEncodingSize(const Instruction *i) const
 
    // check constraints on short MAD
    if (info.srcNr >= 2 && i->srcExists(2)) {
-      if (!i->defExists(0) || !isFloatType(i->dType) ||
-          i->def(0).rep()->reg.data.id != i->src(2).rep()->reg.data.id)
+      if (!i->defExists(0) ||
+          (i->flagsSrc >= 0 && SDATA(i->src(i->flagsSrc)).id > 0) ||
+          DDATA(i->def(0)).id != SDATA(i->src(2)).id)
          return 8;
    }
 
@@ -2085,7 +2112,7 @@ makeInstructionLong(Instruction *insn)
    insn->encSize = 8;
 
    for (int i = fn->bbCount - 1; i >= 0 && fn->bbArray[i] != insn->bb; --i) {
-      fn->bbArray[i]->binPos += 4;
+      fn->bbArray[i]->binPos += adj;
    }
    fn->binSize += adj;
    insn->bb->binSize += adj;
@@ -2137,9 +2164,16 @@ replaceExitWithModifier(Function *func)
             return;
       }
    }
-   epilogue->binSize -= 8;
-   func->binSize -= 8;
+
+   int adj = epilogue->getExit()->encSize;
+   epilogue->binSize -= adj;
+   func->binSize -= adj;
    delete_Instruction(func->getProgram(), epilogue->getExit());
+
+   // There may be BB's that are laid out after the exit block
+   for (int i = func->bbCount - 1; i >= 0 && func->bbArray[i] != epilogue; --i) {
+      func->bbArray[i]->binPos -= adj;
+   }
 }
 
 void