i965/fs: Define logical typed and untyped surface opcodes.

Each logical variant is largely equivalent to the original opcode but
instead of taking a single payload source it expects its arguments
separately as individual sources, like:

 typed_surface_write_logical null, coordinates, source, surface,
                                    num_coordinates, num_components

This patch defines the opcodes and usual instruction boilerplate,
including a placeholder lowering function provided mainly as
documentation for their source registers.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 373d6b8f8ab82cf0af5c583b6f1638d92d0096b0..f595366eeac3f7add34bb11ea222cf7e058b1c72 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -963,13 +963,33 @@ enum opcode {
 
    SHADER_OPCODE_SHADER_TIME_ADD,
 
+   /**
+    * Typed and untyped surface access opcodes.
+    *
+    * LOGICAL opcodes are eventually translated to the matching non-LOGICAL
+    * opcode but instead of taking a single payload blob they expect their
+    * arguments separately as individual sources:
+    *
+    * Source 0: [required] Surface coordinates.
+    * Source 1: [optional] Operation source.
+    * Source 2: [required] Surface index.
+    * Source 3: [required] Number of coordinate components (as UD immediate).
+    * Source 4: [required] Opcode-specific control immediate, same as source 2
+    *                      of the matching non-LOGICAL opcode.
+    */
    SHADER_OPCODE_UNTYPED_ATOMIC,
+   SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL,
    SHADER_OPCODE_UNTYPED_SURFACE_READ,
+   SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL,
    SHADER_OPCODE_UNTYPED_SURFACE_WRITE,
+   SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL,
 
    SHADER_OPCODE_TYPED_ATOMIC,
+   SHADER_OPCODE_TYPED_ATOMIC_LOGICAL,
    SHADER_OPCODE_TYPED_SURFACE_READ,
+   SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL,
    SHADER_OPCODE_TYPED_SURFACE_WRITE,
+   SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL,
 
    SHADER_OPCODE_MEMORY_FENCE,
 

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 289757f1e283c9c4352d29d4ec4ba6abefc386a3..2af7a892ca4b11b4348ebd279b3399df7a016207 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -711,6 +711,49 @@ fs_inst::components_read(unsigned i) const
       else
          return 1;
 
+   case SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL:
+   case SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL:
+      assert(src[3].file == IMM);
+      /* Surface coordinates. */
+      if (i == 0)
+         return src[3].fixed_hw_reg.dw1.ud;
+      /* Surface operation source (ignored for reads). */
+      else if (i == 1)
+         return 0;
+      else
+         return 1;
+
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL:
+   case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL:
+      assert(src[3].file == IMM &&
+             src[4].file == IMM);
+      /* Surface coordinates. */
+      if (i == 0)
+         return src[3].fixed_hw_reg.dw1.ud;
+      /* Surface operation source. */
+      else if (i == 1)
+         return src[4].fixed_hw_reg.dw1.ud;
+      else
+         return 1;
+
+   case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL:
+   case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL: {
+      assert(src[3].file == IMM &&
+             src[4].file == IMM);
+      const unsigned op = src[4].fixed_hw_reg.dw1.ud;
+      /* Surface coordinates. */
+      if (i == 0)
+         return src[3].fixed_hw_reg.dw1.ud;
+      /* Surface operation source. */
+      else if (i == 1 && op == BRW_AOP_CMPWR)
+         return 2;
+      else if (i == 1 && (op == BRW_AOP_INC || op == BRW_AOP_DEC ||
+                          op == BRW_AOP_PREDEC))
+         return 0;
+      else
+         return 1;
+   }
+
    default:
       return 1;
    }
@@ -3847,6 +3890,20 @@ lower_sampler_logical_send(const fs_builder &bld, fs_inst *inst, opcode op)
    }
 }
 
+static void
+lower_surface_logical_send(const fs_builder &bld, fs_inst *inst, opcode op,
+                           const fs_reg &sample_mask)
+{
+   /* Get the logical send arguments. */
+   const fs_reg &addr = inst->src[0];
+   const fs_reg &src = inst->src[1];
+   const fs_reg &surface = inst->src[2];
+   const fs_reg &dims = inst->src[3];
+   const fs_reg &arg = inst->src[4];
+
+   assert(!"Not implemented");
+}
+
 bool
 fs_visitor::lower_logical_sends()
 {
@@ -3914,6 +3971,42 @@ fs_visitor::lower_logical_sends()
          lower_sampler_logical_send(ibld, inst, SHADER_OPCODE_TG4_OFFSET);
          break;
 
+      case SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_UNTYPED_SURFACE_READ,
+                                    fs_reg());
+         break;
+
+      case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_UNTYPED_SURFACE_WRITE,
+                                    fs_reg());
+         break;
+
+      case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_UNTYPED_ATOMIC,
+                                    fs_reg());
+         break;
+
+      case SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_TYPED_SURFACE_READ,
+                                    fs_reg());
+         break;
+
+      case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_TYPED_SURFACE_WRITE,
+                                    fs_reg());
+         break;
+
+      case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL:
+         lower_surface_logical_send(ibld, inst,
+                                    SHADER_OPCODE_TYPED_ATOMIC,
+                                    fs_reg());
+         break;
+
       default:
          continue;
       }

diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index 9e3a9740d74e923d94f10d6dc4567dec4702fdc1..58587b24403880070e3911f9aa11cea7c2c51cc5 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -619,16 +619,28 @@ brw_instruction_name(enum opcode op)
 
    case SHADER_OPCODE_UNTYPED_ATOMIC:
       return "untyped_atomic";
+   case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL:
+      return "untyped_atomic_logical";
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
       return "untyped_surface_read";
+   case SHADER_OPCODE_UNTYPED_SURFACE_READ_LOGICAL:
+      return "untyped_surface_read_logical";
    case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
       return "untyped_surface_write";
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL:
+      return "untyped_surface_write_logical";
    case SHADER_OPCODE_TYPED_ATOMIC:
       return "typed_atomic";
+   case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL:
+      return "typed_atomic_logical";
    case SHADER_OPCODE_TYPED_SURFACE_READ:
       return "typed_surface_read";
+   case SHADER_OPCODE_TYPED_SURFACE_READ_LOGICAL:
+      return "typed_surface_read_logical";
    case SHADER_OPCODE_TYPED_SURFACE_WRITE:
       return "typed_surface_write";
+   case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL:
+      return "typed_surface_write_logical";
    case SHADER_OPCODE_MEMORY_FENCE:
       return "memory_fence";
 
@@ -1181,10 +1193,14 @@ backend_instruction::has_side_effects() const
 {
    switch (opcode) {
    case SHADER_OPCODE_UNTYPED_ATOMIC:
+   case SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL:
    case SHADER_OPCODE_GEN4_SCRATCH_WRITE:
    case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE_LOGICAL:
    case SHADER_OPCODE_TYPED_ATOMIC:
+   case SHADER_OPCODE_TYPED_ATOMIC_LOGICAL:
    case SHADER_OPCODE_TYPED_SURFACE_WRITE:
+   case SHADER_OPCODE_TYPED_SURFACE_WRITE_LOGICAL:
    case SHADER_OPCODE_MEMORY_FENCE:
    case SHADER_OPCODE_URB_WRITE_SIMD8:
    case FS_OPCODE_FB_WRITE: