[Mesa-dev] [PATCH v4 (part2) 32/59] nir: Implement __intrinsic_store_load

[Iago Toral Quiroga]

---
 src/glsl/nir/glsl_to_nir.cpp            | 67 ++++++++++++++++++++++++++++++++-
 src/glsl/nir/nir_intrinsics.h           |  2 +-
 src/glsl/nir/nir_lower_phis_to_scalar.c |  2 +
 3 files changed, 69 insertions(+), 2 deletions(-)

diff --git a/src/glsl/nir/glsl_to_nir.cpp b/src/glsl/nir/glsl_to_nir.cpp
index cbec2df..be4cd66 100644
--- a/src/glsl/nir/glsl_to_nir.cpp
+++ b/src/glsl/nir/glsl_to_nir.cpp
@@ -643,6 +643,8 @@ nir_visitor::visit(ir_call *ir)
          op = nir_intrinsic_memory_barrier;
       } else if (strcmp(ir->callee_name(), "__intrinsic_store_ssbo") == 0) {
          op = nir_intrinsic_store_ssbo;
+      } else if (strcmp(ir->callee_name(), "__intrinsic_load_ssbo") == 0) {
+         op = nir_intrinsic_load_ssbo;
       } else {
          unreachable("not reached");
       }
@@ -766,11 +768,74 @@ nir_visitor::visit(ir_call *ir)
          instr->src[1] = evaluate_rvalue(block);
          break;
       }
+      case nir_intrinsic_load_ssbo: {
+         exec_node *param = ir->actual_parameters.get_head();
+         ir_rvalue *block = ((ir_instruction *)param)->as_rvalue();
+
+         param = param->get_next();
+         ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
+
+         /* Check if we need the indirect version */
+         ir_constant *const_offset = offset->as_constant();
+         if (!const_offset) {
+            op = nir_intrinsic_load_ssbo_indirect;
+            ralloc_free(instr);
+            instr = nir_intrinsic_instr_create(shader, op);
+            instr->src[1] = evaluate_rvalue(offset);
+            instr->const_index[0] = 0;
+         } else {
+            instr->const_index[0] = const_offset->value.u[0];
+         }
+
+         instr->src[0] = evaluate_rvalue(block);
+
+         instr->const_index[1] = 1; /* number of vec4's */
+
+         const glsl_type *type = ir->return_deref->var->type;
+         instr->num_components = type->vector_elements;
+
+         /* Setup destination register */
+         nir_ssa_dest_init(&instr->instr, &instr->dest,
+                           type->vector_elements, NULL);
+
+         /* Insert the created nir instruction now since in the case of boolean
+          * result we will need to emit another instruction after it
+          */
+         nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+
+         /*
+          * In SSBO/UBO's, a true boolean value is any non-zero value, but we
+          * consider a true boolean to be ~0. Fix this up with a != 0
+          * comparison.
+          */
+         if (type->base_type == GLSL_TYPE_BOOL) {
+            nir_load_const_instr *const_zero =
+               nir_load_const_instr_create(shader, 1);
+            const_zero->value.u[0] = 0;
+            nir_instr_insert_after_cf_list(this->cf_node_list,
+                                           &const_zero->instr);
+
+            nir_alu_instr *compare = nir_alu_instr_create(shader, nir_op_ine);
+            compare->src[0].src.is_ssa = true;
+            compare->src[0].src.ssa = &instr->dest.ssa;
+            compare->src[1].src.is_ssa = true;
+            compare->src[1].src.ssa = &const_zero->def;
+            for (unsigned i = 0; i < type->vector_elements; i++)
+               compare->src[1].swizzle[i] = 0;
+            compare->dest.write_mask = (1 << type->vector_elements) - 1;
+
+            nir_instr_insert_after_cf_list(this->cf_node_list, &compare->instr);
+         }
+
+         break;
+      }
       default:
          unreachable("not reached");
       }
 
-      nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+      /* nir_intrinsic_load_ssbo{_indirect} were already emitted */
+      if (op != nir_intrinsic_load_ssbo && op != nir_intrinsic_load_ssbo_indirect)
+         nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
 
       if (ir->return_deref) {
          nir_intrinsic_instr *store_instr =
diff --git a/src/glsl/nir/nir_intrinsics.h b/src/glsl/nir/nir_intrinsics.h
index 83eeecd..9877ea4 100644
--- a/src/glsl/nir/nir_intrinsics.h
+++ b/src/glsl/nir/nir_intrinsics.h
@@ -168,7 +168,7 @@ SYSTEM_VALUE(invocation_id, 1)
 LOAD(uniform, 0, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 LOAD(ubo, 1, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 LOAD(input, 0, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
-/* LOAD(ssbo, 1, 0) */
+LOAD(ssbo, 1, 0)
 
 /*
  * Stores work the same way as loads, except now the first register input is
diff --git a/src/glsl/nir/nir_lower_phis_to_scalar.c b/src/glsl/nir/nir_lower_phis_to_scalar.c
index 739170d..bdb7e6a 100644
--- a/src/glsl/nir/nir_lower_phis_to_scalar.c
+++ b/src/glsl/nir/nir_lower_phis_to_scalar.c
@@ -94,6 +94,8 @@ is_phi_src_scalarizable(nir_phi_src *src,
       case nir_intrinsic_load_uniform_indirect:
       case nir_intrinsic_load_ubo:
       case nir_intrinsic_load_ubo_indirect:
+      case nir_intrinsic_load_ssbo:
+      case nir_intrinsic_load_ssbo_indirect:
       case nir_intrinsic_load_input:
       case nir_intrinsic_load_input_indirect:
          return true;
-- 
1.9.1