glsl: Move lower_vector_insert to GLSL-to-NIR.

We already have a nir_builder equivalent for generating this code, just
use that instead of doing it in GLSL.

No change on r300 shader-db.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21476>

src/compiler/glsl/glsl_parser_extras.cpp

@@ -2410,7 +2410,6 @@ do_common_optimization(exec_list *ir, bool linked,
    OPT(do_algebraic, ir, native_integers, options);
    OPT(do_lower_jumps, ir, true, true, options->EmitNoMainReturn,
        options->EmitNoCont);
-   OPT(lower_vector_insert, ir, false);
 
    /* If an optimization pass fails to preserve the invariant flag, calling
     * the pass only once earlier may result in incorrect code generation. Always call

src/compiler/glsl/glsl_to_nir.cpp

@@ -2381,9 +2381,13 @@ nir_visitor::visit(ir_expression *ir)
    case ir_binop_dot:
       result = nir_fdot(&b, srcs[0], srcs[1]);
       break;
+
    case ir_binop_vector_extract:
       result = nir_vector_extract(&b, srcs[0], srcs[1]);
       break;
+   case ir_triop_vector_insert:
+      result = nir_vector_insert(&b, srcs[0], srcs[1], srcs[2]);
+      break;
 
    case ir_binop_atan2:
       result = nir_atan2(&b, srcs[0], srcs[1]);

src/compiler/glsl/ir_optimization.h

@@ -75,7 +75,6 @@ bool lower_packing_builtins(exec_list *instructions,
                             bool has_shading_language_packing,
                             bool has_gpu_shader5,
                             bool has_half_float_packing);
-bool lower_vector_insert(exec_list *instructions, bool lower_nonconstant_index);
 bool lower_vector_derefs(gl_linked_shader *shader);
 void lower_named_interface_blocks(void *mem_ctx, gl_linked_shader *shader);
 void optimize_dead_builtin_variables(exec_list *instructions,

src/compiler/glsl/lower_vector_insert.cpp

@@ -1,178 +0,0 @@
-/*
- * Copyright © 2013 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
- * DEALINGS IN THE SOFTWARE.
- */
-#include "ir.h"
-#include "ir_builder.h"
-#include "ir_rvalue_visitor.h"
-#include "ir_optimization.h"
-
-using namespace ir_builder;
-
-namespace {
-
-class vector_insert_visitor : public ir_rvalue_visitor {
-public:
-   vector_insert_visitor(bool lower_nonconstant_index)
-      : progress(false), lower_nonconstant_index(lower_nonconstant_index),
-        remove_assignment(false)
-   {
-      factory.instructions = &factory_instructions;
-   }
-
-   virtual ~vector_insert_visitor()
-   {
-      assert(factory_instructions.is_empty());
-   }
-
-   virtual void handle_rvalue(ir_rvalue **rv);
-   virtual ir_visitor_status visit_leave(ir_assignment *expr);
-
-   ir_factory factory;
-   exec_list factory_instructions;
-   bool progress;
-   bool lower_nonconstant_index;
-   bool remove_assignment;
-};
-
-} /* anonymous namespace */
-
-void
-vector_insert_visitor::handle_rvalue(ir_rvalue **rv)
-{
-   if (*rv == NULL || (*rv)->ir_type != ir_type_expression)
-      return;
-
-   ir_expression *const expr = (ir_expression *) *rv;
-
-   if (likely(expr->operation != ir_triop_vector_insert))
-      return;
-
-   factory.mem_ctx = ralloc_parent(expr);
-
-   ir_constant *const idx =
-      expr->operands[2]->constant_expression_value(factory.mem_ctx);
-   if (idx != NULL) {
-      unsigned index = idx->value.u[0];
-
-      if (index >= expr->operands[0]->type->vector_elements) {
-         /* Section 5.11 (Out-of-Bounds Accesses) of the GLSL 4.60 spec says:
-          *
-          *  In the subsections described above for array, vector, matrix and
-          *  structure accesses, any out-of-bounds access produced undefined
-          *  behavior.... Out-of-bounds writes may be discarded or overwrite
-          *  other variables of the active program.
-          */
-         this->remove_assignment = true;
-         this->progress = true;
-         return;
-      }
-
-      /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
-       * a new temporary.  The new temporary gets assigned as
-       *
-       *     t = vec
-       *     t.mask = scalar
-       *
-       * where mask is the component selected by index.
-       */
-      ir_variable *const temp =
-         factory.make_temp(expr->operands[0]->type, "vec_tmp");
-
-      const int mask = 1 << idx->value.i[0];
-
-      factory.emit(assign(temp, expr->operands[0]));
-      factory.emit(assign(temp, expr->operands[1], mask));
-
-      this->progress = true;
-      *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
-   } else if (this->lower_nonconstant_index) {
-      /* Replace (vector_insert (vec) (scalar) (index)) with a dereference of
-       * a new temporary.  The new temporary gets assigned as
-       *
-       *     t = vec
-       *     if (index == 0)
-       *         t.x = scalar
-       *     if (index == 1)
-       *         t.y = scalar
-       *     if (index == 2)
-       *         t.z = scalar
-       *     if (index == 3)
-       *         t.w = scalar
-       */
-      ir_variable *const temp =
-         factory.make_temp(expr->operands[0]->type, "vec_tmp");
-
-      ir_variable *const src_temp =
-         factory.make_temp(expr->operands[1]->type, "src_temp");
-
-      factory.emit(assign(temp, expr->operands[0]));
-      factory.emit(assign(src_temp, expr->operands[1]));
-
-      assert(expr->operands[2]->type == glsl_type::int_type ||
-             expr->operands[2]->type == glsl_type::uint_type);
-
-      for (unsigned i = 0; i < expr->type->vector_elements; i++) {
-         ir_constant *const cmp_index =
-            ir_constant::zero(factory.mem_ctx, expr->operands[2]->type);
-         cmp_index->value.u[0] = i;
-
-         ir_variable *const cmp_result =
-            factory.make_temp(glsl_type::bool_type, "index_condition");
-
-         factory.emit(assign(cmp_result,
-                             equal(expr->operands[2]->clone(factory.mem_ctx,
-                                                            NULL),
-                                   cmp_index)));
-
-         factory.emit(if_tree(cmp_result,
-                              assign(temp, src_temp, WRITEMASK_X << i)));
-      }
-
-      this->progress = true;
-      *rv = new(factory.mem_ctx) ir_dereference_variable(temp);
-   }
-
-   base_ir->insert_before(factory.instructions);
-}
-
-ir_visitor_status
-vector_insert_visitor::visit_leave(ir_assignment *ir)
-{
-   ir_rvalue_visitor::visit_leave(ir);
-
-   if (this->remove_assignment) {
-      ir->remove();
-      this->remove_assignment = false;
-   }
-
-   return visit_continue;
-}
-
-bool
-lower_vector_insert(exec_list *instructions, bool lower_nonconstant_index)
-{
-   vector_insert_visitor v(lower_nonconstant_index);
-
-   visit_list_elements(&v, instructions);
-
-   return v.progress;
-}

src/compiler/glsl/meson.build

@@ -213,7 +213,6 @@ files_libglsl = files(
   'lower_tess_level.cpp',
   'lower_vec_index_to_cond_assign.cpp',
   'lower_vector_derefs.cpp',
-  'lower_vector_insert.cpp',
   'opt_algebraic.cpp',
   'opt_constant_folding.cpp',
   'opt_constant_propagation.cpp',

src/mesa/state_tracker/st_glsl_to_ir.cpp

@@ -86,7 +86,6 @@ link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
                          ctx->Extensions.ARB_gpu_shader5);
 
       do_vec_index_to_cond_assign(ir);
-      lower_vector_insert(ir, true);
       if (options->MaxIfDepth == 0) {
          lower_discard(ir);
       }