On Wed, 2016-09-07 at 18:48 -0700, Francisco Jerez wrote:
> The previous regs_read value can be recovered by rewriting each
> reference of regs_read() like 'x = i.regs_read(j)' to 'x =
> DIV_ROUND_UP(i.size_read(j), reg_unit)'.
> 
> For the same reason as in the previous patches, this doesn't attempt
> to be particularly clever about simplifying the result in the
> interest
> of keeping the rather lengthy patch as obvious as possible.  I'll
> come
> back later to clean up any ugliness introduced here.
> ---
>  src/mesa/drivers/dri/i965/brw_ir_vec4.h            |  6 +++--
>  src/mesa/drivers/dri/i965/brw_vec4.cpp             | 30
> ++++++++++++++--------
>  .../drivers/dri/i965/brw_vec4_copy_propagation.cpp |  2 +-
>  3 files changed, 25 insertions(+), 13 deletions(-)
> 
> diff --git a/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> index 5a79062..2fd5441 100644
> --- a/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> +++ b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> @@ -167,7 +167,7 @@ public:
>     unsigned sol_vertex; /**< gen6: used for setting dst index in SVB
> header */
>  
>     bool is_send_from_grf();
> -   unsigned regs_read(unsigned arg) const;
> +   unsigned size_read(unsigned arg) const;
>     bool can_reswizzle(const struct gen_device_info *devinfo, int
> dst_writemask,
>                        int swizzle, int swizzle_mask);
>     void reswizzle(int dst_writemask, int swizzle);
> @@ -278,7 +278,9 @@ inline unsigned
>  regs_read(const vec4_instruction *inst, unsigned i)
>  {
>     /* XXX - Take into account register-misaligned offsets correctly.
> */
> -   return inst->regs_read(i);
> +   const unsigned reg_size =
> +      inst->src[i].file == UNIFORM || inst->src[i].file == IMM ? 16
> : REG_SIZE;
> +   return DIV_ROUND_UP(inst->size_read(i), reg_size);
>  }
>  
>  } /* namespace brw */
> diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp
> b/src/mesa/drivers/dri/i965/brw_vec4.cpp
> index bdd6e59..561170c 100644
> --- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
> @@ -199,11 +199,8 @@
> vec4_instruction::has_source_and_destination_hazard() const
>  }
>  
>  unsigned
> -vec4_instruction::regs_read(unsigned arg) const
> +vec4_instruction::size_read(unsigned arg) const
>  {
> -   if (src[arg].file == BAD_FILE)
> -      return 0;
> -
>     switch (opcode) {
>     case SHADER_OPCODE_SHADER_TIME_ADD:
>     case SHADER_OPCODE_UNTYPED_ATOMIC:
> @@ -213,13 +210,26 @@ vec4_instruction::regs_read(unsigned arg) const
>     case SHADER_OPCODE_TYPED_SURFACE_READ:
>     case SHADER_OPCODE_TYPED_SURFACE_WRITE:
>     case TCS_OPCODE_URB_WRITE:
> -      return arg == 0 ? mlen : 1;
> -
> +      if (arg == 0)
> +         return mlen * REG_SIZE;
> +      break;
>     case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
> -      return arg == 1 ? mlen : 1;
> +      if (arg == 1)
> +         return mlen * REG_SIZE;
> +      break;
> +   default:
> +      break;
> +   }
>  
> +   switch (src[arg].file) {
> +   case BAD_FILE:
> +      return 0;
> +   case IMM:
> +   case UNIFORM:
> +      return 4 * type_sz(src[arg].type);
>     default:
> -      return 1;
> +      /* XXX - Represent actual execution size and vertical stride.
> */
> +      return 8 * type_sz(src[arg].type);
>     }
>  }
>  
> @@ -1188,7 +1198,7 @@ vec4_visitor::opt_register_coalesce()
>        bool interfered = false;
>        for (int i = 0; i < 3; i++) {
>              if (inst->src[0].in_range(scan_inst->src[i],
> -                                      scan_inst->regs_read(i)))
> +                                      DIV_ROUND_UP(scan_inst-
> >size_read(i), REG_SIZE)))

Why not just swap scan_inst->regs_read(i) with regs_read(scan_inst, i)
instead?

>              interfered = true;
>        }
>        if (interfered)
> @@ -1214,7 +1224,7 @@ vec4_visitor::opt_register_coalesce()
>           } else {
>              for (int i = 0; i < 3; i++) {
>                 if (inst->dst.in_range(scan_inst->src[i],
> -                                      scan_inst->regs_read(i)))
> +                                      DIV_ROUND_UP(scan_inst-
> >size_read(i), REG_SIZE)))

Same here.

>                    interfered = true;
>              }
>              if (interfered)
> diff --git a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
> b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
> index f98c7ac..777d252 100644
> --- a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp
> @@ -437,7 +437,7 @@ vec4_visitor::opt_copy_propagation(bool
> do_constant_prop)
>           continue;
>  
>           /* We only handle single-register copies. */
> -         if (inst->regs_read(i) != 1)
> +         if (inst->size_read(i) != REG_SIZE)
>              continue;
>  
>           const unsigned reg = (alloc.offsets[inst->src[i].nr] +
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to