[Mesa-dev] [PATCH 1/6] i965/vec4: Move can_do_writemask to vec4_instruction

[Michael Schellenberger Costa]

Hi Jason,

Am 06.04.2016 um 06:11 schrieb Jason Ekstrand:
> ---
>  src/mesa/drivers/dri/i965/brw_ir_vec4.h            |  1 +
>  src/mesa/drivers/dri/i965/brw_vec4.cpp             | 28 +++++++++++++++++++
>  .../dri/i965/brw_vec4_dead_code_eliminate.cpp      | 31 +---------------------
>  3 files changed, 30 insertions(+), 30 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_ir_vec4.h b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> index 2b6872e..81b6a13 100644
> --- a/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> +++ b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
> @@ -172,6 +172,7 @@ public:
>                        int swizzle, int swizzle_mask);
>     void reswizzle(int dst_writemask, int swizzle);
>     bool can_do_source_mods(const struct brw_device_info *devinfo);
> +   bool can_do_writemask(const struct brw_device_info *devinfo);
>     bool can_change_types() const;
>     bool has_source_and_destination_hazard() const;
>  
> diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
> index 0025343..4d0efa8 100644
> --- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
> @@ -239,6 +239,34 @@ vec4_instruction::can_do_source_mods(const struct brw_device_info *devinfo)
>  }
>  
>  bool
> +vec4_instruction::can_do_writemask(const struct brw_device_info *devinfo)
> +{
> +   switch (opcode) {
> +   case SHADER_OPCODE_GEN4_SCRATCH_READ:
> +   case VS_OPCODE_PULL_CONSTANT_LOAD:
> +   case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
> +   case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9:
Can you order the stages alphabetically here, that improves readability
a lot?
Michael
> +   case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
> +   case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
> +   case TES_OPCODE_CREATE_INPUT_READ_HEADER:
> +   case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
> +   case VEC4_OPCODE_URB_READ:
> +      return false;
> +   default:
> +      /* The MATH instruction on Gen6 only executes in align1 mode, which does
> +       * not support writemasking.
> +       */
> +      if (devinfo->gen == 6 && is_math())
> +         return false;
> +
> +      if (is_tex())
> +         return false;
> +
> +      return true;
> +   }
> +}
> +
> +bool
>  vec4_instruction::can_change_types() const
>  {
>     return dst.type == src[0].type &&
> diff --git a/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp b/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp
> index 166bc17..c643212 100644
> --- a/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp
> @@ -36,35 +36,6 @@
>  
>  using namespace brw;
>  
> -static bool
> -can_do_writemask(const struct brw_device_info *devinfo,
> -                 const vec4_instruction *inst)
> -{
> -   switch (inst->opcode) {
> -   case SHADER_OPCODE_GEN4_SCRATCH_READ:
> -   case VS_OPCODE_PULL_CONSTANT_LOAD:
> -   case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
> -   case VS_OPCODE_SET_SIMD4X2_HEADER_GEN9:
> -   case TCS_OPCODE_SET_INPUT_URB_OFFSETS:
> -   case TCS_OPCODE_SET_OUTPUT_URB_OFFSETS:
> -   case TES_OPCODE_CREATE_INPUT_READ_HEADER:
> -   case TES_OPCODE_ADD_INDIRECT_URB_OFFSET:
> -   case VEC4_OPCODE_URB_READ:
> -      return false;
> -   default:
> -      /* The MATH instruction on Gen6 only executes in align1 mode, which does
> -       * not support writemasking.
> -       */
> -      if (devinfo->gen == 6 && inst->is_math())
> -         return false;
> -
> -      if (inst->is_tex())
> -         return false;
> -
> -      return true;
> -   }
> -}
> -
>  bool
>  vec4_visitor::dead_code_eliminate()
>  {
> @@ -101,7 +72,7 @@ vec4_visitor::dead_code_eliminate()
>              /* If the instruction can't do writemasking, then it's all or
>               * nothing.
>               */
> -            if (!can_do_writemask(devinfo, inst)) {
> +            if (!inst->can_do_writemask(devinfo)) {
>                 bool result = result_live[0] | result_live[1] |
>                               result_live[2] | result_live[3];
>                 result_live[0] = result;