[Mesa-stable] [Mesa-dev] [PATCH 1/2] glsl: Convert lower_vec_index_to_swizzle to a rvalue visitor.

[Ian Romanick]

Obligatory question: piglit tests?  This series looks like a good
solution for the problem in the bug, but I always worry that we've
caught N-1 possible cases.  That can be easier to detect from the set of
tests.

If there are more problem found later, we can fix them later.  This
series is

Reviewed-by: Ian Romanick <ian.d.romanick at intel.com>

On 04/28/2016 11:06 PM, Kenneth Graunke wrote:
> The old visitor missed some cases.  For example, it wouldn't handle
> an ir_dereference_array with a vector_extract as the index.
> 
> Rather than trying to add the missing cases, just rewrite it as an
> ir_rvalue_visitor.  This makes it easy to replace any expression,
> and is much less code.
> 
> Cc: mesa-stable at lists.freedesktop.org
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95164
> Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> ---
>  src/compiler/glsl/lower_vec_index_to_swizzle.cpp | 96 ++++--------------------
>  1 file changed, 13 insertions(+), 83 deletions(-)
> 
> diff --git a/src/compiler/glsl/lower_vec_index_to_swizzle.cpp b/src/compiler/glsl/lower_vec_index_to_swizzle.cpp
> index 8b18e95..b49255e 100644
> --- a/src/compiler/glsl/lower_vec_index_to_swizzle.cpp
> +++ b/src/compiler/glsl/lower_vec_index_to_swizzle.cpp
> @@ -30,18 +30,14 @@
>   */
>  
>  #include "ir.h"
> -#include "ir_visitor.h"
> +#include "ir_rvalue_visitor.h"
>  #include "ir_optimization.h"
>  #include "compiler/glsl_types.h"
>  #include "main/macros.h"
>  
> -/**
> - * Visitor class for replacing expressions with ir_constant values.
> - */
> -
>  namespace {
>  
> -class ir_vec_index_to_swizzle_visitor : public ir_hierarchical_visitor {
> +class ir_vec_index_to_swizzle_visitor : public ir_rvalue_visitor {
>  public:
>     ir_vec_index_to_swizzle_visitor()
>     {
> @@ -50,30 +46,28 @@ public:
>  
>     ir_rvalue *convert_vector_extract_to_swizzle(ir_rvalue *val);
>  
> -   virtual ir_visitor_status visit_enter(ir_expression *);
> -   virtual ir_visitor_status visit_enter(ir_swizzle *);
> -   virtual ir_visitor_status visit_enter(ir_assignment *);
> -   virtual ir_visitor_status visit_enter(ir_return *);
> -   virtual ir_visitor_status visit_enter(ir_call *);
> -   virtual ir_visitor_status visit_enter(ir_if *);
> +   virtual void handle_rvalue(ir_rvalue **);
>  
>     bool progress;
>  };
>  
>  } /* anonymous namespace */
>  
> -ir_rvalue *
> -ir_vec_index_to_swizzle_visitor::convert_vector_extract_to_swizzle(ir_rvalue *ir)
> +void
> +ir_vec_index_to_swizzle_visitor::handle_rvalue(ir_rvalue **rv)
>  {
> -   ir_expression *const expr = ir->as_expression();
> +   if (*rv == NULL)
> +      return;
> +
> +   ir_expression *const expr = (*rv)->as_expression();
>     if (expr == NULL || expr->operation != ir_binop_vector_extract)
> -      return ir;
> +      return;
>  
>     ir_constant *const idx = expr->operands[1]->constant_expression_value();
>     if (idx == NULL)
> -      return ir;
> +      return;
>  
> -   void *ctx = ralloc_parent(ir);
> +   void *ctx = ralloc_parent(expr);
>     this->progress = true;
>  
>     /* Page 40 of the GLSL 1.20 spec says:
> @@ -93,71 +87,7 @@ ir_vec_index_to_swizzle_visitor::convert_vector_extract_to_swizzle(ir_rvalue *ir
>     const int i = CLAMP(idx->value.i[0], 0,
>                         (int) expr->operands[0]->type->vector_elements - 1);
>  
> -   return new(ctx) ir_swizzle(expr->operands[0], i, 0, 0, 0, 1);
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_expression *ir)
> -{
> -   unsigned int i;
> -
> -   for (i = 0; i < ir->get_num_operands(); i++) {
> -      ir->operands[i] = convert_vector_extract_to_swizzle(ir->operands[i]);
> -   }
> -
> -   return visit_continue;
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_swizzle *ir)
> -{
> -   /* Can't be hit from normal GLSL, since you can't swizzle a scalar (which
> -    * the result of indexing a vector is.  But maybe at some point we'll end up
> -    * using swizzling of scalars for vector construction.
> -    */
> -   ir->val = convert_vector_extract_to_swizzle(ir->val);
> -
> -   return visit_continue;
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_assignment *ir)
> -{
> -   ir->rhs = convert_vector_extract_to_swizzle(ir->rhs);
> -
> -   return visit_continue;
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_call *ir)
> -{
> -   foreach_in_list_safe(ir_rvalue, param, &ir->actual_parameters) {
> -      ir_rvalue *new_param = convert_vector_extract_to_swizzle(param);
> -
> -      if (new_param != param) {
> -	 param->replace_with(new_param);
> -      }
> -   }
> -
> -   return visit_continue;
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_return *ir)
> -{
> -   if (ir->value) {
> -      ir->value = convert_vector_extract_to_swizzle(ir->value);
> -   }
> -
> -   return visit_continue;
> -}
> -
> -ir_visitor_status
> -ir_vec_index_to_swizzle_visitor::visit_enter(ir_if *ir)
> -{
> -   ir->condition = convert_vector_extract_to_swizzle(ir->condition);
> -
> -   return visit_continue;
> +   *rv = new(ctx) ir_swizzle(expr->operands[0], i, 0, 0, 0, 1);
>  }
>  
>  bool
>