[Mesa-dev] [PATCH 09/12] nv50/ir: add lock/unlock subops for load/store

[Samuel Pitoiset]

Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
---
 src/gallium/drivers/nouveau/codegen/nv50_ir.h            |  2 ++
 .../drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp        | 16 ++++++++++++++--
 src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp    | 10 ++++++++++
 3 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir.h b/src/gallium/drivers/nouveau/codegen/nv50_ir.h
index 9d7becf..97ebed4 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir.h
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir.h
@@ -232,6 +232,8 @@ enum operation
 #define NV50_IR_SUBOP_SHFL_UP   1
 #define NV50_IR_SUBOP_SHFL_DOWN 2
 #define NV50_IR_SUBOP_SHFL_BFLY 3
+#define NV50_IR_SUBOP_LOAD_LOCKED    1
+#define NV50_IR_SUBOP_STORE_UNLOCKED 2
 #define NV50_IR_SUBOP_MADSP_SD     0xffff
 // Yes, we could represent those with DataType.
 // Or put the type into operation and have a couple 1000 values in that enum.
diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp
index 8637db9..c7239b3 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp
@@ -1773,7 +1773,13 @@ CodeEmitterNVC0::emitSTORE(const Instruction *i)
    switch (i->src(0).getFile()) {
    case FILE_MEMORY_GLOBAL: opc = 0x90000000; break;
    case FILE_MEMORY_LOCAL:  opc = 0xc8000000; break;
-   case FILE_MEMORY_SHARED: opc = 0xc9000000; break;
+   case FILE_MEMORY_SHARED:
+      opc = 0xc8000000;
+      if (i->subOp == NV50_IR_SUBOP_STORE_UNLOCKED)
+         opc |= (1 << 26);
+      else
+         opc |= (1 << 24);
+      break;
    default:
       assert(!"invalid memory file");
       opc = 0;
@@ -1804,7 +1810,13 @@ CodeEmitterNVC0::emitLOAD(const Instruction *i)
    switch (i->src(0).getFile()) {
    case FILE_MEMORY_GLOBAL: opc = 0x80000000; break;
    case FILE_MEMORY_LOCAL:  opc = 0xc0000000; break;
-   case FILE_MEMORY_SHARED: opc = 0xc1000000; break;
+   case FILE_MEMORY_SHARED:
+      opc = 0xc0000000;
+      if (i->subOp == NV50_IR_SUBOP_LOAD_LOCKED)
+         opc |= (1 << 26);
+      else
+         opc |= (1 << 24);
+      break;
    case FILE_MEMORY_CONST:
       if (!i->src(0).isIndirect(0) && typeSizeof(i->dType) == 4) {
          emitMOV(i); // not sure if this is any better
diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
index 47285a2..85f7704 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
@@ -198,6 +198,11 @@ static const char *atomSubOpStr[] =
    "add", "min", "max", "inc", "dec", "and", "or", "xor", "cas", "exch"
 };
 
+static const char *ldstSubOpStr[] =
+{
+   "", "lock", "unlock"
+};
+
 static const char *DataTypeStr[] =
 {
    "-",
@@ -537,6 +542,11 @@ void Instruction::print() const
          if (subOp < Elements(atomSubOpStr))
             PRINT("%s ", atomSubOpStr[subOp]);
          break;
+      case OP_LOAD:
+      case OP_STORE:
+         if (subOp < Elements(ldstSubOpStr))
+            PRINT("%s ", ldstSubOpStr[subOp]);
+         break;
       default:
          if (subOp)
             PRINT("(SUBOP:%u) ", subOp);
-- 
2.6.4