i965: Add untyped surface write opcode.

Reviewed-by: Topi Pohjolainen <[email protected]>
Acked-by: Kenneth Graunke <[email protected]>

9 files changed, 81 insertions, 0 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 4a4ff1d26d7..37494f892f4 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -904,6 +904,7 @@ enum opcode {
 
    SHADER_OPCODE_UNTYPED_ATOMIC,
    SHADER_OPCODE_UNTYPED_SURFACE_READ,
+   SHADER_OPCODE_UNTYPED_SURFACE_WRITE,
 
    SHADER_OPCODE_GEN4_SCRATCH_READ,
    SHADER_OPCODE_GEN4_SCRATCH_WRITE,
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 1d0e5634027..a1f4c536bc2 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -417,6 +417,13 @@ brw_untyped_surface_read(struct brw_codegen *p,
                          unsigned num_channels);
 
 void
+brw_untyped_surface_write(struct brw_codegen *p,
+                          struct brw_reg payload,
+                          struct brw_reg surface,
+                          unsigned msg_length,
+                          unsigned num_channels);
+
+void
 brw_pixel_interpolator_query(struct brw_codegen *p,
                              struct brw_reg dest,
                              struct brw_reg mrf,
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 3ae1a1a56ac..cc45da17a93 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -2894,6 +2894,57 @@ brw_untyped_surface_read(struct brw_codegen *p,
       p, insn, num_channels);
 }
 
+static void
+brw_set_dp_untyped_surface_write_message(struct brw_codegen *p,
+                                         struct brw_inst *insn,
+                                         unsigned num_channels)
+{
+   const struct brw_device_info *devinfo = p->devinfo;
+   /* Set mask of 32-bit channels to drop. */
+   unsigned msg_control = 0xf & (0xf << num_channels);
+
+   if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
+      if (p->compressed)
+         msg_control |= 1 << 4; /* SIMD16 mode */
+      else
+         msg_control |= 2 << 4; /* SIMD8 mode */
+   } else {
+      if (devinfo->gen >= 8 || devinfo->is_haswell)
+         msg_control |= 0 << 4; /* SIMD4x2 mode */
+      else
+         msg_control |= 2 << 4; /* SIMD8 mode */
+   }
+
+   brw_inst_set_dp_msg_type(devinfo, insn,
+                            devinfo->gen >= 8 || devinfo->is_haswell ?
+                             HSW_DATAPORT_DC_PORT1_UNTYPED_SURFACE_WRITE :
+                             GEN7_DATAPORT_DC_UNTYPED_SURFACE_WRITE);
+   brw_inst_set_dp_msg_control(devinfo, insn, msg_control);
+}
+
+void
+brw_untyped_surface_write(struct brw_codegen *p,
+                          struct brw_reg payload,
+                          struct brw_reg surface,
+                          unsigned msg_length,
+                          unsigned num_channels)
+{
+   const struct brw_device_info *devinfo = p->devinfo;
+   const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
+                          HSW_SFID_DATAPORT_DATA_CACHE_1 :
+                          GEN7_SFID_DATAPORT_DATA_CACHE);
+   const bool align1 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1;
+   /* Mask out unused components -- See comment in brw_untyped_atomic(). */
+   const unsigned mask = devinfo->gen == 7 && !devinfo->is_haswell && !align1 ?
+                          WRITEMASK_X : WRITEMASK_XYZW;
+   struct brw_inst *insn = brw_send_indirect_surface_message(
+      p, sfid, brw_writemask(brw_null_reg(), mask),
+      payload, surface, msg_length, 0, align1);
+
+   brw_set_dp_untyped_surface_write_message(
+      p, insn, num_channels);
+}
+
 void
 brw_pixel_interpolator_query(struct brw_codegen *p,
                              struct brw_reg dest,
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index c0abff1ba0e..ec4c1a87fd2 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -502,6 +502,7 @@ fs_inst::is_send_from_grf() const
    case FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET:
    case SHADER_OPCODE_UNTYPED_ATOMIC:
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
    case SHADER_OPCODE_URB_WRITE_SIMD8:
       return true;
    case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD:
@@ -956,6 +957,8 @@ fs_inst::regs_read(int arg) const
       return mlen;
    } else if (opcode == SHADER_OPCODE_UNTYPED_SURFACE_READ && arg == 0) {
       return mlen;
+   } else if (opcode == SHADER_OPCODE_UNTYPED_SURFACE_WRITE && arg == 0) {
+      return mlen;
    } else if (opcode == FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET && arg == 0) {
       return mlen;
    } else if (opcode == FS_OPCODE_LINTERP && arg == 0) {
@@ -1048,6 +1051,7 @@ fs_visitor::implied_mrf_writes(fs_inst *inst)
       return 2;
    case SHADER_OPCODE_UNTYPED_ATOMIC:
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
    case SHADER_OPCODE_URB_WRITE_SIMD8:
    case FS_OPCODE_INTERPOLATE_AT_CENTROID:
    case FS_OPCODE_INTERPOLATE_AT_SAMPLE:
diff --git a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
index efe05e34ae4..7fe1afa6b32 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_generator.cpp
@@ -2030,6 +2030,12 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width)
          brw_mark_surface_used(prog_data, src[1].dw1.ud);
          break;
 
+      case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
+         assert(src[2].file == BRW_IMMEDIATE_VALUE);
+         brw_untyped_surface_write(p, src[0], src[1],
+                                   inst->mlen, src[2].dw1.ud);
+         break;
+
       case FS_OPCODE_SET_SIMD4X2_OFFSET:
          generate_set_simd4x2_offset(inst, dst, src[0]);
          break;
diff --git a/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp b/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp
index a8bd93d7caa..a439399fe38 100644
--- a/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp
+++ b/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp
@@ -358,6 +358,7 @@ schedule_node::set_latency_gen7(bool is_haswell)
       break;
 
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
       /* Test code:
        *   mov(8)    g112<1>UD       0x00000000UD       { align1 WE_all 1Q };
        *   mov(1)    g112.7<1>UD     g1.7<0,1,0>UD      { align1 WE_all };
diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index 4fbd8d063e3..df09c226547 100644
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -494,6 +494,8 @@ brw_instruction_name(enum opcode op)
       return "untyped_atomic";
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
       return "untyped_surface_read";
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
+      return "untyped_surface_write";
 
    case SHADER_OPCODE_LOAD_PAYLOAD:
       return "load_payload";
@@ -1038,6 +1040,7 @@ backend_instruction::has_side_effects() const
    switch (opcode) {
    case SHADER_OPCODE_UNTYPED_ATOMIC:
    case SHADER_OPCODE_GEN4_SCRATCH_WRITE:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
    case SHADER_OPCODE_URB_WRITE_SIMD8:
    case FS_OPCODE_FB_WRITE:
       return true;
diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
index 1160fdfdb54..31d8af305ff 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
@@ -216,6 +216,7 @@ vec4_instruction::is_send_from_grf()
    case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
    case SHADER_OPCODE_UNTYPED_ATOMIC:
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
       return true;
    default:
       return false;
@@ -232,6 +233,7 @@ vec4_instruction::regs_read(unsigned arg) const
    case SHADER_OPCODE_SHADER_TIME_ADD:
    case SHADER_OPCODE_UNTYPED_ATOMIC:
    case SHADER_OPCODE_UNTYPED_SURFACE_READ:
+   case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
       return arg == 0 ? mlen : 1;
 
    case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
index c6d98d584e8..f1fce1735c7 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
@@ -1484,6 +1484,12 @@ vec4_generator::generate_code(const cfg_t *cfg)
          brw_mark_surface_used(&prog_data->base, src[1].dw1.ud);
          break;
 
+      case SHADER_OPCODE_UNTYPED_SURFACE_WRITE:
+         assert(src[2].file == BRW_IMMEDIATE_VALUE);
+         brw_untyped_surface_write(p, src[0], src[1], inst->mlen,
+                                   src[2].dw1.ud);
+         break;
+
       case VS_OPCODE_UNPACK_FLAGS_SIMD4X2:
          generate_unpack_flags(dst);
          break;