[Mesa-dev] [PATCH v3 1/3] glsl: teach copy_propagation_elements to deal with whole variables

Thomas Helland thomashelland90 at gmail.com
Thu Jul 26 19:32:43 UTC 2018


2018-07-25 3:03 GMT+02:00 Caio Marcelo de Oliveira Filho
<caio.oliveira at intel.com>:
> Keep information in acp_entry whether the entry is full or not, and
> use the ACP in more nodes when visiting the instructions:
>
> - add_copy: write whole variables to the ACP state (regardless the
>   type).
>
> - visit(ir_dereference_variable *): perform the propagation here if we have a
>   full candidate. Element-wise here doesn't apply because the mask
>   isn't available at this point.
>
> - visit_leave(ir_assignment *): process beyond scalar and vector, as
>   the full variables might have other types.
>
> Also import an improvement from opt_copy_propagation.cpp: if ir_call
> is an intrinsic, we know the variables affected, so keep going.
>
> v2: (all from Eric Anholt)
>     Describe how acp_entry attributes are used.
>     Don't do book-keeping to avoid adding repeated element to
>     the dsts in write_elements().
>
> Reviewed-by: Eric Anholt <eric at anholt.net>
> ---
>  .../glsl/opt_copy_propagation_elements.cpp    | 147 ++++++++++++++----
>  1 file changed, 118 insertions(+), 29 deletions(-)
>
> diff --git a/src/compiler/glsl/opt_copy_propagation_elements.cpp b/src/compiler/glsl/opt_copy_propagation_elements.cpp
> index 4c6ca790394..cae6d3c0707 100644
> --- a/src/compiler/glsl/opt_copy_propagation_elements.cpp
> +++ b/src/compiler/glsl/opt_copy_propagation_elements.cpp
> @@ -27,18 +27,9 @@
>   * Replaces usage of recently-copied components of variables with the
>   * previous copy of the variable.
>   *
> - * This pass can be compared with opt_copy_propagation, which operands
> - * on arbitrary whole-variable copies.  However, in order to handle
> - * the copy propagation of swizzled variables or writemasked writes,
> - * we want to track things on a channel-wise basis.  I found that
> - * trying to mix the swizzled/writemasked support here with the
> - * whole-variable stuff in opt_copy_propagation.cpp just made a mess,
> - * so this is separate despite the ACP handling being somewhat
> - * similar.
> - *
>   * This should reduce the number of MOV instructions in the generated
> - * programs unless copy propagation is also done on the LIR, and may
> - * help anyway by triggering other optimizations that live in the HIR.
> + * programs and help triggering other optimizations that live in GLSL
> + * level.
>   */
>
>  #include "ir.h"
> @@ -58,9 +49,22 @@ class acp_entry
>  public:
>     DECLARE_LINEAR_ZALLOC_CXX_OPERATORS(acp_entry)
>
> +   /* If set, rhs_full indicates that this ACP entry represents a
> +    * whole-variable copy.  The rhs_element[] array will still be filled,
> +    * to allow the swizzling from its components in case the variable
> +    * was a vector (and to simplify some of the erase() and write_vector()
> +    * logic).
> +    */
> +
> +   ir_variable *rhs_full;
>     ir_variable *rhs_element[4];
>     unsigned rhs_channel[4];
>
> +   /* Set of variables that use the variable associated with this acp_entry as
> +    * RHS.  This has the "reverse references" of rhs_full/rhs_element.  It is
> +    * used to speed up invalidating those references when the acp_entry
> +    * changes.
> +    */
>     set *dsts;
>  };
>
> @@ -91,6 +95,7 @@ public:
>     void erase(ir_variable *var, unsigned write_mask)
>     {
>        acp_entry *entry = pull_acp(var);
> +      entry->rhs_full = NULL;
>
>        for (int i = 0; i < 4; i++) {
>           if (!entry->rhs_element[i])
> @@ -114,6 +119,8 @@ public:
>              if (dst_entry->rhs_element[i] == var)
>                 dst_entry->rhs_element[i] = NULL;
>           }
> +         if (dst_entry->rhs_full == var)
> +            dst_entry->rhs_full = NULL;
>           _mesa_set_remove(entry->dsts, set_entry);
>        }
>     }
> @@ -128,9 +135,10 @@ public:
>        return NULL;
>     }
>
> -   void write(ir_variable *lhs, ir_variable *rhs, unsigned write_mask, int swizzle[4])
> +   void write_elements(ir_variable *lhs, ir_variable *rhs, unsigned write_mask, int swizzle[4])
>     {
>        acp_entry *lhs_entry = pull_acp(lhs);
> +      lhs_entry->rhs_full = NULL;
>
>        for (int i = 0; i < 4; i++) {
>           if ((write_mask & (1 << i)) == 0)
> @@ -146,6 +154,33 @@ public:
>        _mesa_set_add(rhs_entry->dsts, lhs);
>     }
>
> +   void write_full(ir_variable *lhs, ir_variable *rhs)
> +   {
> +      acp_entry *lhs_entry = pull_acp(lhs);
> +      if (lhs_entry->rhs_full == rhs)
> +         return;
> +
> +      if (lhs_entry->rhs_full) {
> +         remove_from_dsts(lhs_entry->rhs_full, lhs);
> +      } else if (lhs->type->is_vector()) {
> +         for (int i = 0; i < 4; i++) {
> +            if (lhs_entry->rhs_element[i])
> +               remove_from_dsts(lhs_entry->rhs_element[i], lhs);
> +         }
> +      }
> +
> +      lhs_entry->rhs_full = rhs;
> +      acp_entry *rhs_entry = pull_acp(rhs);
> +      _mesa_set_add(rhs_entry->dsts, lhs);
> +
> +      if (lhs->type->is_vector()) {
> +         for (int i = 0; i < 4; i++) {
> +            lhs_entry->rhs_element[i] = rhs;
> +            lhs_entry->rhs_channel[i] = i;
> +         }
> +      }
> +   }
> +
>     void remove_unused_var_from_dsts(acp_entry *lhs_entry, ir_variable *lhs, ir_variable *var)
>     {
>        if (!var)
> @@ -204,6 +239,14 @@ private:
>        return entry;
>     }
>
> +   void
> +   remove_from_dsts(ir_variable *var, ir_variable *to_remove)
> +   {
> +      acp_entry *entry = pull_acp(var);
> +      assert(entry);
> +      _mesa_set_remove(entry->dsts, _mesa_set_search(entry->dsts, to_remove));
> +   }
> +

Use the newly added _mesa_set_remove_key?
Apart from that, this looks good to me


Reviewed-by: Thomas Helland<thomashelland90 at gmail.com>


>     /** Available Copy to Propagate table, from variable to the entry
>      *  containing the current sources that can be used. */
>     hash_table *acp;
> @@ -247,6 +290,8 @@ public:
>        ralloc_free(mem_ctx);
>     }
>
> +   virtual ir_visitor_status visit(ir_dereference_variable *);
> +
>     void handle_loop(ir_loop *, bool keep_acp);
>     virtual ir_visitor_status visit_enter(class ir_loop *);
>     virtual ir_visitor_status visit_enter(class ir_function_signature *);
> @@ -282,6 +327,21 @@ public:
>
>  } /* unnamed namespace */
>
> +ir_visitor_status
> +ir_copy_propagation_elements_visitor::visit(ir_dereference_variable *ir)
> +{
> +   if (this->in_assignee)
> +      return visit_continue;
> +
> +   const acp_entry *entry = state->read(ir->var);
> +   if (entry && entry->rhs_full) {
> +      ir->var = (ir_variable *) entry->rhs_full;
> +      progress = true;
> +   }
> +
> +   return visit_continue;
> +}
> +
>  ir_visitor_status
>  ir_copy_propagation_elements_visitor::visit_enter(ir_function_signature *ir)
>  {
> @@ -316,16 +376,14 @@ ir_copy_propagation_elements_visitor::visit_leave(ir_assignment *ir)
>     ir_dereference_variable *lhs = ir->lhs->as_dereference_variable();
>     ir_variable *var = ir->lhs->variable_referenced();
>
> -   if (var->type->is_scalar() || var->type->is_vector()) {
> -      kill_entry *k;
> +   kill_entry *k;
>
> -      if (lhs)
> -        k = new(this->lin_ctx) kill_entry(var, ir->write_mask);
> -      else
> -        k = new(this->lin_ctx) kill_entry(var, ~0);
> +   if (lhs && var->type->is_vector())
> +      k = new(this->lin_ctx) kill_entry(var, ir->write_mask);
> +   else
> +      k = new(this->lin_ctx) kill_entry(var, ~0);
>
> -      kill(k);
> -   }
> +   kill(k);
>
>     add_copy(ir);
>
> @@ -460,11 +518,25 @@ ir_copy_propagation_elements_visitor::visit_enter(ir_call *ir)
>        }
>     }
>
> -   /* Since we're unlinked, we don't (necessarily) know the side effects of
> -    * this call.  So kill all copies.
> -    */
> -   this->state->erase_all();
> -   this->killed_all = true;
> +   if (!ir->callee->is_intrinsic()) {
> +      state->erase_all();
> +      this->killed_all = true;
> +   } else {
> +      if (ir->return_deref) {
> +         kill(new(this->lin_ctx) kill_entry(ir->return_deref->var, ~0));
> +      }
> +
> +      foreach_two_lists(formal_node, &ir->callee->parameters,
> +                        actual_node, &ir->actual_parameters) {
> +         ir_variable *sig_param = (ir_variable *) formal_node;
> +         if (sig_param->data.mode == ir_var_function_out ||
> +             sig_param->data.mode == ir_var_function_inout) {
> +            ir_rvalue *ir = (ir_rvalue *) actual_node;
> +            ir_variable *var = ir->variable_referenced();
> +            kill(new(this->lin_ctx) kill_entry(var, ~0));
> +         }
> +      }
> +   }
>
>     return visit_continue_with_parent;
>  }
> @@ -585,12 +657,29 @@ ir_copy_propagation_elements_visitor::kill(kill_entry *k)
>  void
>  ir_copy_propagation_elements_visitor::add_copy(ir_assignment *ir)
>  {
> -   int orig_swizzle[4] = {0, 1, 2, 3};
> -   int swizzle[4];
> -
>     if (ir->condition)
>        return;
>
> +   {
> +      ir_variable *lhs_var = ir->whole_variable_written();
> +      ir_dereference_variable *rhs = ir->rhs->as_dereference_variable();
> +
> +      if (lhs_var != NULL && rhs && rhs->var != NULL && lhs_var != rhs->var) {
> +         if (lhs_var->data.mode == ir_var_shader_storage ||
> +             lhs_var->data.mode == ir_var_shader_shared ||
> +             rhs->var->data.mode == ir_var_shader_storage ||
> +             rhs->var->data.mode == ir_var_shader_shared ||
> +             lhs_var->data.precise != rhs->var->data.precise) {
> +            return;
> +         }
> +         state->write_full(lhs_var, rhs->var);
> +         return;
> +      }
> +   }
> +
> +   int orig_swizzle[4] = {0, 1, 2, 3};
> +   int swizzle[4];
> +
>     ir_dereference_variable *lhs = ir->lhs->as_dereference_variable();
>     if (!lhs || !(lhs->type->is_scalar() || lhs->type->is_vector()))
>        return;
> @@ -645,7 +734,7 @@ ir_copy_propagation_elements_visitor::add_copy(ir_assignment *ir)
>     if (lhs->var->data.precise != rhs->var->data.precise)
>        return;
>
> -   state->write(lhs->var, rhs->var, write_mask, swizzle);
> +   state->write_elements(lhs->var, rhs->var, write_mask, swizzle);
>  }
>
>  bool
> --
> 2.18.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev


More information about the mesa-dev mailing list