i965: add support for textureSamples function

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
[v2: kayden-supplied code in fs_nir replacing need for logical opcode]
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index a8594afa48614ac7c771c95ab41a08d42bded773..d18f7ca4f73ed577f8a583e23264df2d02d9cd6d 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -979,6 +979,7 @@ enum opcode {
    SHADER_OPCODE_TG4_LOGICAL,
    SHADER_OPCODE_TG4_OFFSET,
    SHADER_OPCODE_TG4_OFFSET_LOGICAL,
+   SHADER_OPCODE_SAMPLEINFO,
 
    /**
     * Combines multiple sources of size 1 into a larger virtual GRF.
@@ -1511,6 +1512,7 @@ enum brw_message_target {
 #define GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4      8
 #define GEN5_SAMPLER_MESSAGE_LOD                 9
 #define GEN5_SAMPLER_MESSAGE_SAMPLE_RESINFO      10
+#define GEN6_SAMPLER_MESSAGE_SAMPLE_SAMPLEINFO   11
 #define GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_C    16
 #define GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO   17
 #define GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO_C 18

diff --git a/src/mesa/drivers/dri/i965/brw_disasm.c b/src/mesa/drivers/dri/i965/brw_disasm.c
index 7401e325638f4dd511b49951aa7113115a14ade7..db23a187a93a2f350a687c9850476aba0613e3a8 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_disasm.c
+++ b/src/mesa/drivers/dri/i965/brw_disasm.c
@@ -617,6 +617,7 @@ static const char *const gen5_sampler_msg_type[] = {
    [GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4]      = "gather4",
    [GEN5_SAMPLER_MESSAGE_LOD]                 = "lod",
    [GEN5_SAMPLER_MESSAGE_SAMPLE_RESINFO]      = "resinfo",
+   [GEN6_SAMPLER_MESSAGE_SAMPLE_SAMPLEINFO]   = "sampleinfo",
    [GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_C]    = "gather4_c",
    [GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO]   = "gather4_po",
    [GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO_C] = "gather4_po_c",

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index d240371767b1c18ba8c704ce288c4a4bf43c6d53..17cbdf46582c510f98bbdf7c9be1bd26ac4b330f 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -877,6 +877,7 @@ fs_visitor::implied_mrf_writes(fs_inst *inst)
    case SHADER_OPCODE_TXL:
    case SHADER_OPCODE_TXS:
    case SHADER_OPCODE_LOD:
+   case SHADER_OPCODE_SAMPLEINFO:
       return 1;
    case FS_OPCODE_FB_WRITE:
       return 2;

diff --git a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
index c86ca043b63246f675b1b25c021f3391efd389a6..90805e45ad7af476fcd26a68e8c6e44065f3f191 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
@@ -646,6 +646,9 @@ fs_generator::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src
             msg_type = GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO;
          }
          break;
+      case SHADER_OPCODE_SAMPLEINFO:
+         msg_type = GEN6_SAMPLER_MESSAGE_SAMPLE_SAMPLEINFO;
+         break;
       default:
         unreachable("not reached");
       }
@@ -1920,6 +1923,7 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width)
       case SHADER_OPCODE_LOD:
       case SHADER_OPCODE_TG4:
       case SHADER_OPCODE_TG4_OFFSET:
+      case SHADER_OPCODE_SAMPLEINFO:
         generate_tex(inst, dst, src[0], src[1]);
         break;
       case FS_OPCODE_DDX_COARSE:

diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
index a6c6a2fa8db29c0c6e3c50ff1f8838f76cf9a2f8..0902e1c396e0df84b4be556f0c021f8adcd92532 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
@@ -1842,6 +1842,16 @@ fs_visitor::nir_emit_texture(const fs_builder &bld, nir_tex_instr *instr)
    case nir_texop_txf_ms: op = ir_txf_ms; break;
    case nir_texop_txl: op = ir_txl; break;
    case nir_texop_txs: op = ir_txs; break;
+   case nir_texop_texture_samples: {
+      fs_reg dst = retype(get_nir_dest(instr->dest), BRW_REGISTER_TYPE_D);
+      fs_inst *inst = bld.emit(SHADER_OPCODE_SAMPLEINFO, dst,
+                               bld.vgrf(BRW_REGISTER_TYPE_D, 1),
+                               sampler_reg);
+      inst->mlen = 1;
+      inst->header_size = 1;
+      inst->base_mrf = -1;
+      return;
+   }
    default:
       unreachable("unknown texture opcode");
    }

diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index de1a7fe81a8f672c2ed75a8ef401073ba07be1b3..d353591c356a0407743af05663b7cdb00a8fd1a8 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -618,6 +618,8 @@ brw_instruction_name(enum opcode op)
       return "tg4_offset";
    case SHADER_OPCODE_TG4_OFFSET_LOGICAL:
       return "tg4_offset_logical";
+   case SHADER_OPCODE_SAMPLEINFO:
+      return "sampleinfo";
 
    case SHADER_OPCODE_SHADER_TIME_ADD:
       return "shader_time_add";

diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
index 85dc37238d1db0ab33ac122e561795c434395e68..893ff356afa3eb0a3b9381e789c45bb1a18b1571 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
@@ -331,6 +331,7 @@ vec4_visitor::implied_mrf_writes(vec4_instruction *inst)
    case SHADER_OPCODE_TXS:
    case SHADER_OPCODE_TG4:
    case SHADER_OPCODE_TG4_OFFSET:
+   case SHADER_OPCODE_SAMPLEINFO:
       return inst->header_size;
    default:
       unreachable("not reached");

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
index 92050b94d339249c4dbc4fc597bb4531370c61cc..195033358fb87bd6f41e1e25ed463cf925bd7ce3 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
@@ -286,6 +286,9 @@ vec4_generator::generate_tex(vec4_instruction *inst,
             msg_type = GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4_PO;
          }
          break;
+      case SHADER_OPCODE_SAMPLEINFO:
+         msg_type = GEN6_SAMPLER_MESSAGE_SAMPLE_SAMPLEINFO;
+         break;
       default:
         unreachable("should not get here: invalid vec4 texture opcode");
       }
@@ -1374,6 +1377,7 @@ vec4_generator::generate_code(const cfg_t *cfg)
       case SHADER_OPCODE_TXS:
       case SHADER_OPCODE_TG4:
       case SHADER_OPCODE_TG4_OFFSET:
+      case SHADER_OPCODE_SAMPLEINFO:
          generate_tex(inst, dst, src[0], src[1]);
          break;
 

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp b/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
index 751ec73f709b5cc83f2111b42fc5965aea456706..c21fd02bf68aa320852a2be5854471c4b1562050 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
@@ -1337,6 +1337,7 @@ ir_texture_opcode_for_nir_texop(nir_texop texop)
    switch (texop) {
    case nir_texop_lod: op = ir_lod; break;
    case nir_texop_query_levels: op = ir_query_levels; break;
+   case nir_texop_texture_samples: op = ir_texture_samples; break;
    case nir_texop_tex: op = ir_tex; break;
    case nir_texop_tg4: op = ir_tg4; break;
    case nir_texop_txb: op = ir_txb; break;

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
index f6e59ce41969ebf1c26c374a4ee0e5a459764e9e..0465770440553c4ceebbf6bb245336fad3eb5984 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp
@@ -2550,6 +2550,7 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
    case ir_tg4: opcode = offset_value.file != BAD_FILE
                          ? SHADER_OPCODE_TG4_OFFSET : SHADER_OPCODE_TG4; break;
    case ir_query_levels: opcode = SHADER_OPCODE_TXS; break;
+   case ir_texture_samples: opcode = SHADER_OPCODE_SAMPLEINFO; break;
    case ir_txb:
       unreachable("TXB is not valid for vertex shaders.");
    case ir_lod:
@@ -2569,13 +2570,15 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
     * - Texel offsets
     * - Gather channel selection
     * - Sampler indices too large to fit in a 4-bit value.
+    * - Sampleinfo message - takes no parameters, but mlen = 0 is illegal
     */
    inst->header_size =
       (devinfo->gen < 5 || devinfo->gen >= 9 ||
        inst->offset != 0 || op == ir_tg4 ||
+       op == ir_texture_samples ||
        is_high_sampler(sampler_reg)) ? 1 : 0;
    inst->base_mrf = 2;
-   inst->mlen = inst->header_size + 1; /* always at least one */
+   inst->mlen = inst->header_size;
    inst->dst.writemask = WRITEMASK_XYZW;
    inst->shadow_compare = shadow_comparitor.file != BAD_FILE;
 
@@ -2587,6 +2590,9 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
    if (op == ir_txs || op == ir_query_levels) {
       int writemask = devinfo->gen == 4 ? WRITEMASK_W : WRITEMASK_X;
       emit(MOV(dst_reg(MRF, param_base, lod.type, writemask), lod));
+      inst->mlen++;
+   } else if (op == ir_texture_samples) {
+      inst->dst.writemask = WRITEMASK_X;
    } else {
       /* Load the coordinate */
       /* FINISHME: gl_clamp_mask and saturate */
@@ -2595,6 +2601,7 @@ vec4_visitor::emit_texture(ir_texture_opcode op,
 
       emit(MOV(dst_reg(MRF, param_base, coordinate.type, coord_mask),
                coordinate));
+      inst->mlen++;
 
       if (zero_mask != 0) {
          emit(MOV(dst_reg(MRF, param_base, coordinate.type, zero_mask),
@@ -2817,6 +2824,7 @@ vec4_visitor::visit(ir_texture *ir)
    case ir_txb:
    case ir_lod:
    case ir_tg4:
+   case ir_texture_samples:
       break;
    }