nvc0: enable FBFETCH with a special slot for color buffer 0

We don't need to support all the color buffers for advanced blend, just
cb0. For Fermi, we use the special binding slots so that we don't
overlap with user textures, while Kepler+ gets a dedicated position for
the fb handle in the driver constbuf.

This logic is only triggered when a FBFETCH is actually present so it
should be a no-op most of the time.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

1 files changed, 41 insertions, 0 deletions

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
index 6b38d5f0fa9..64bfd084326 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
@@ -1459,6 +1459,9 @@ bool Source::scanInstruction(const struct tgsi_full_instruction *inst)
    if (insn.getOpcode() == TGSI_OPCODE_BARRIER)
       info->numBarriers = 1;
 
+   if (insn.getOpcode() == TGSI_OPCODE_FBFETCH)
+      info->prop.fp.readsFramebuffer = true;
+
    if (insn.dstCount()) {
       Instruction::DstRegister dst = insn.getDst(0);
 
@@ -1574,6 +1577,7 @@ private:
    void handleTEX(Value *dst0[4], int R, int S, int L, int C, int Dx, int Dy);
    void handleTXF(Value *dst0[4], int R, int L_M);
    void handleTXQ(Value *dst0[4], enum TexQuery, int R);
+   void handleFBFETCH(Value *dst0[4]);
    void handleLIT(Value *dst0[4]);
    void handleUserClipPlanes();
 
@@ -2283,6 +2287,40 @@ Converter::handleTXF(Value *dst[4], int R, int L_M)
 }
 
 void
+Converter::handleFBFETCH(Value *dst[4])
+{
+   TexInstruction *texi = new_TexInstruction(func, OP_TXF);
+   unsigned int c, d;
+
+   texi->tex.target = TEX_TARGET_2D_MS_ARRAY;
+   texi->tex.levelZero = 1;
+   texi->tex.useOffsets = 0;
+
+   for (c = 0, d = 0; c < 4; ++c) {
+      if (dst[c]) {
+         texi->setDef(d++, dst[c]);
+         texi->tex.mask |= 1 << c;
+      }
+   }
+
+   Value *x = mkOp1v(OP_RDSV, TYPE_F32, getScratch(), mkSysVal(SV_POSITION, 0));
+   Value *y = mkOp1v(OP_RDSV, TYPE_F32, getScratch(), mkSysVal(SV_POSITION, 1));
+   Value *z = mkOp1v(OP_RDSV, TYPE_U32, getScratch(), mkSysVal(SV_LAYER, 0));
+   Value *ms = mkOp1v(OP_RDSV, TYPE_U32, getScratch(), mkSysVal(SV_SAMPLE_INDEX, 0));
+
+   mkCvt(OP_CVT, TYPE_U32, x, TYPE_F32, x)->rnd = ROUND_Z;
+   mkCvt(OP_CVT, TYPE_U32, y, TYPE_F32, y)->rnd = ROUND_Z;
+   texi->setSrc(0, x);
+   texi->setSrc(1, y);
+   texi->setSrc(2, z);
+   texi->setSrc(3, ms);
+
+   texi->tex.r = texi->tex.s = -1;
+
+   bb->insertTail(texi);
+}
+
+void
 Converter::handleLIT(Value *dst0[4])
 {
    Value *val0 = NULL;
@@ -3323,6 +3361,9 @@ Converter::handleInstruction(const struct tgsi_full_instruction *insn)
       handleTXQ(dst0, TXQ_TYPE, 0);
       std::swap(dst0[0], dst0[2]);
       break;
+   case TGSI_OPCODE_FBFETCH:
+      handleFBFETCH(dst0);
+      break;
    case TGSI_OPCODE_F2I:
    case TGSI_OPCODE_F2U:
       FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi)