[Mesa-dev] [PATCH 9/9] i965: Enable ARB_shader_atomic_counter_ops

[Ian Romanick]

From: Ian Romanick <ian.d.romanick at intel.com>

Signed-off-by: Ian Romanick <ian.d.romanick at intel.com>
Reviewed-by: Iago Toral Quiroga <itoral at igalia.com>
---
 src/mesa/drivers/dri/i965/brw_fs_nir.cpp     | 23 ++++++++++++++++++++---
 src/mesa/drivers/dri/i965/brw_shader.cpp     | 16 ++++++++++++++++
 src/mesa/drivers/dri/i965/brw_vec4_nir.cpp   | 16 +++++++++++++---
 src/mesa/drivers/dri/i965/intel_extensions.c |  1 +
 4 files changed, 50 insertions(+), 6 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
index b60e97a..16444a1 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
@@ -3798,23 +3798,40 @@ fs_visitor::nir_emit_intrinsic(const fs_builder &bld, nir_intrinsic_instr *instr
    switch (instr->intrinsic) {
    case nir_intrinsic_atomic_counter_inc:
    case nir_intrinsic_atomic_counter_dec:
-   case nir_intrinsic_atomic_counter_read: {
+   case nir_intrinsic_atomic_counter_read:
+   case nir_intrinsic_atomic_counter_add:
+   case nir_intrinsic_atomic_counter_min:
+   case nir_intrinsic_atomic_counter_max:
+   case nir_intrinsic_atomic_counter_and:
+   case nir_intrinsic_atomic_counter_or:
+   case nir_intrinsic_atomic_counter_xor:
+   case nir_intrinsic_atomic_counter_exchange:
+   case nir_intrinsic_atomic_counter_comp_swap: {
       if (stage == MESA_SHADER_FRAGMENT &&
           instr->intrinsic != nir_intrinsic_atomic_counter_read)
          ((struct brw_wm_prog_data *)prog_data)->has_side_effects = true;
 
+      /* Get some metadata from the image intrinsic. */
+      const nir_intrinsic_info *info = &nir_intrinsic_infos[instr->intrinsic];
+
       /* Get the arguments of the atomic intrinsic. */
       const fs_reg offset = get_nir_src(instr->src[0]);
       const unsigned surface = (stage_prog_data->binding_table.abo_start +
                                 instr->const_index[0]);
+      const fs_reg src0 = (info->num_srcs >= 2
+                           ? get_nir_src(instr->src[1]) : fs_reg());
+      const fs_reg src1 = (info->num_srcs >= 3
+                           ? get_nir_src(instr->src[2]) : fs_reg());
       fs_reg tmp;
 
+      assert(info->num_srcs <= 3);
+
       /* Emit a surface read or atomic op. */
       if (instr->intrinsic == nir_intrinsic_atomic_counter_read) {
          tmp = emit_untyped_read(bld, brw_imm_ud(surface), offset, 1, 1);
       } else {
-         tmp = emit_untyped_atomic(bld, brw_imm_ud(surface), offset, fs_reg(),
-                                   fs_reg(), 1, 1,
+         tmp = emit_untyped_atomic(bld, brw_imm_ud(surface), offset, src0,
+                                   src1, 1, 1,
                                    get_atomic_counter_op(instr->intrinsic));
       }
 
diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index 6de5f7b..1477676 100644
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -612,6 +612,22 @@ get_atomic_counter_op(nir_intrinsic_op op)
       return BRW_AOP_INC;
    case nir_intrinsic_atomic_counter_dec:
       return BRW_AOP_PREDEC;
+   case nir_intrinsic_atomic_counter_add:
+      return BRW_AOP_ADD;
+   case nir_intrinsic_atomic_counter_min:
+      return BRW_AOP_UMIN;
+   case nir_intrinsic_atomic_counter_max:
+      return BRW_AOP_UMAX;
+   case nir_intrinsic_atomic_counter_and:
+      return BRW_AOP_AND;
+   case nir_intrinsic_atomic_counter_or:
+      return BRW_AOP_OR;
+   case nir_intrinsic_atomic_counter_xor:
+      return BRW_AOP_XOR;
+   case nir_intrinsic_atomic_counter_exchange:
+      return BRW_AOP_MOV;
+   case nir_intrinsic_atomic_counter_comp_swap:
+      return BRW_AOP_CMPWR;
    default:
       unreachable("Not reachable.");
    }
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp b/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
index a06af47..ce11ab6 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_nir.cpp
@@ -739,11 +739,21 @@ vec4_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
    case nir_intrinsic_atomic_counter_dec: {
       unsigned surf_index = prog_data->base.binding_table.abo_start +
          (unsigned) instr->const_index[0];
+      const vec4_builder bld =
+         vec4_builder(this).at_end().annotate(current_annotation, base_ir);
+
+      /* Get some metadata from the image intrinsic. */
+      const nir_intrinsic_info *info = &nir_intrinsic_infos[instr->intrinsic];
+
+      /* Get the arguments of the atomic intrinsic. */
       src_reg offset = get_nir_src(instr->src[0], nir_type_int,
                                    instr->num_components);
       const src_reg surface = brw_imm_ud(surf_index);
-      const vec4_builder bld =
-         vec4_builder(this).at_end().annotate(current_annotation, base_ir);
+      const src_reg src0 = (info->num_srcs >= 2
+                           ? get_nir_src(instr->src[1]) : src_reg());
+      const src_reg src1 = (info->num_srcs >= 3
+                           ? get_nir_src(instr->src[2]) : src_reg());
+
       src_reg tmp;
 
       dest = get_nir_dest(instr->dest);
@@ -752,7 +762,7 @@ vec4_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
          tmp = emit_untyped_read(bld, surface, offset, 1, 1);
       } else {
          tmp = emit_untyped_atomic(bld, surface, offset,
-                                   src_reg(), src_reg(),
+                                   src0, src1,
                                    1, 1,
                                    get_atomic_counter_op(instr->intrinsic));
       }
diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/dri/i965/intel_extensions.c
index 0f28546..bb104d1 100644
--- a/src/mesa/drivers/dri/i965/intel_extensions.c
+++ b/src/mesa/drivers/dri/i965/intel_extensions.c
@@ -345,6 +345,7 @@ intelInitExtensions(struct gl_context *ctx)
       ctx->Extensions.ARB_framebuffer_no_attachments = true;
       ctx->Extensions.ARB_gpu_shader5 = true;
       ctx->Extensions.ARB_shader_atomic_counters = true;
+      ctx->Extensions.ARB_shader_atomic_counter_ops = true;
       ctx->Extensions.ARB_shader_clock = true;
       ctx->Extensions.ARB_shader_image_load_store = true;
       ctx->Extensions.ARB_shader_image_size = true;
-- 
2.5.5