[Mesa-dev] [PATCH v3 16/30] nvir/nir: implement nir_intrinsic_store_(per_vertex_)output

[Karol Herbst]

v3: add workaround for RA issues
    indirects have to be multiplied by 0x10
    fix indirect access

Signed-off-by: Karol Herbst <kherbst at redhat.com>
---
 .../drivers/nouveau/codegen/nv50_ir_from_nir.cpp   | 54 ++++++++++++++++++++++
 1 file changed, 54 insertions(+)

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_nir.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_nir.cpp
index 75d74a6379..74edec0c97 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_nir.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_nir.cpp
@@ -1093,6 +1093,11 @@ Converter::visit(nir_function *function)
 
    setPosition(entry, true);
 
+   if (info->io.genUserClip > 0) {
+      for (int c = 0; c < 4; ++c)
+         clipVtx[c] = getScratch();
+   }
+
    switch (prog->getType()) {
    case Program::TYPE_TESSELLATION_CONTROL:
       outBase = mkOp2v(
@@ -1119,6 +1124,8 @@ Converter::visit(nir_function *function)
    bb->cfg.attach(&exit->cfg, Graph::Edge::TREE);
    setPosition(exit, true);
 
+   if (info->io.genUserClip > 0)
+      handleUserClipPlanes();
    // TODO: for non main function this needs to be a OP_RETURN
    mkOp(OP_EXIT, TYPE_NONE, NULL)->terminator = 1;
    return true;
@@ -1339,6 +1346,53 @@ Converter::visit(nir_intrinsic_instr *insn)
       }
       break;
    }
+   case nir_intrinsic_store_output:
+   case nir_intrinsic_store_per_vertex_output: {
+      Value *indirect;
+      auto idx = nir_intrinsic_base(insn) + getIndirect(&insn->src[op == nir_intrinsic_store_output ? 1 : 2], 0, &indirect);
+      uint8_t offset = insn->const_index[2];
+
+      if (indirect)
+         // we have to multiply with 16
+         mkOp2(OP_MUL, TYPE_U32, indirect, indirect, loadImm(getScratch(), 16));
+
+      for (auto i = 0u; i < insn->num_components; ++i) {
+         if (!((1u << i) & nir_intrinsic_write_mask(insn)))
+            continue;
+
+         Value *src = getSrc(&insn->src[0], i);
+         switch (prog->getType()) {
+         case Program::TYPE_FRAGMENT: {
+            if (info->out[idx].sn == TGSI_SEMANTIC_POSITION) {
+               // TGSI uses a different interface than NIR, TGSI stores that value in the z component, NIR in X
+               offset += 2;
+               src = mkOp1v(OP_SAT, TYPE_F32, getScratch(), src);
+            }
+            break;
+         }
+         case Program::TYPE_VERTEX: {
+            if (info->io.genUserClip > 0) {
+               mkMov(clipVtx[i], src);
+               src = clipVtx[i];
+            }
+            break;
+         }
+         default:
+            break;
+         }
+
+         assert(i + offset < 4);
+         uint32_t address = info->out[idx].slot[i + offset];
+
+         // TODO: RA doesn't like exorts without moving the sources...
+         mkStore(OP_EXPORT,
+                 TYPE_F32,
+                 mkSymbol(FILE_SHADER_OUTPUT, 0, TYPE_U32, address * 4),
+                 indirect,
+                 mkMov(getSSA(), src)->getDef(0))->perPatch = info->out[idx].patch;
+      }
+      break;
+   }
    default:
       ERROR("unknown nir_intrinsic_op %s\n", nir_intrinsic_infos[op].name);
       return false;
-- 
2.14.3