[Mesa-dev] [PATCH v2] radeonsi: emit_spi_map packets optimization

[Marek Olšák]

Thanks. I've pushed this and made some changes to it. See below for
the changes I made.

On Wed, Jul 18, 2018 at 5:48 PM, Sonny Jiang <sonny.jiang at amd.com> wrote:
> Signed-off-by: Sonny Jiang <sonny.jiang at amd.com>
> ---
>  src/gallium/drivers/radeonsi/si_build_pm4.h     | 24 ++++++++++++++++++++++++
>  src/gallium/drivers/radeonsi/si_gfx_cs.c        |  3 +++
>  src/gallium/drivers/radeonsi/si_state.h         |  1 +
>  src/gallium/drivers/radeonsi/si_state_shaders.c | 21 +++++++++++++--------
>  4 files changed, 41 insertions(+), 8 deletions(-)
>
> diff --git a/src/gallium/drivers/radeonsi/si_build_pm4.h b/src/gallium/drivers/radeonsi/si_build_pm4.h
> index b339cd5..8fc08f7 100644
> --- a/src/gallium/drivers/radeonsi/si_build_pm4.h
> +++ b/src/gallium/drivers/radeonsi/si_build_pm4.h
> @@ -214,4 +214,28 @@ static inline void radeon_opt_set_context_reg4(struct si_context *sctx, unsigned
>         }
>  }
>
> +/**
> + * Set consecutive registers if any registers value is different.
> + */
> +static inline void radeon_opt_set_context_regn(struct si_context *sctx, unsigned offset,
> +                                              unsigned * value, unsigned * saved_val,

I removed the space between * and variables, because * is a variable qualifier.
Example: int i, *a, *b, *c, x, **y;

> +                                              unsigned num)
> +{
> +       struct radeon_cmdbuf *cs = sctx->gfx_cs;
> +       int i, j;
> +
> +       for (i = 0; i < num; i++) {
> +               if (saved_val[i] != value[i]) {
> +                       radeon_set_context_reg_seq(cs, offset, num);
> +                       for (j = 0; j < num; j++) {
> +                               radeon_emit(cs, value[j]);
> +                       }

I removed { and } for the single-line block.


> +
> +                       memcpy(saved_val, value, sizeof(uint32_t) * num);
> +
> +                       break;

I removed the empty line.

> +               }
> +       }
> +}
> +
>  #endif
> diff --git a/src/gallium/drivers/radeonsi/si_gfx_cs.c b/src/gallium/drivers/radeonsi/si_gfx_cs.c
> index 628b6c5..16aa4f9 100644
> --- a/src/gallium/drivers/radeonsi/si_gfx_cs.c
> +++ b/src/gallium/drivers/radeonsi/si_gfx_cs.c
> @@ -353,4 +353,7 @@ void si_begin_new_gfx_cs(struct si_context *ctx)
>                 /* Set all saved registers state to unknown. */
>                 ctx->tracked_regs.reg_saved = 0;
>         }
> +
> +       /* 0xffffffff is a impossible value to register SPI_PS_INPUT_CNTL_n */
> +       memset(ctx->tracked_regs.reg_val_seq, 0xff, sizeof(uint32_t) * 32);
>  }
> diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h
> index 71056c7..c2d0287 100644
> --- a/src/gallium/drivers/radeonsi/si_state.h
> +++ b/src/gallium/drivers/radeonsi/si_state.h
> @@ -287,6 +287,7 @@ enum si_tracked_reg {
>  struct si_tracked_regs {
>         uint32_t                reg_saved;
>         uint32_t                reg_value[SI_NUM_TRACKED_REGS];
> +       uint32_t                reg_val_seq[32];

I renamed reg_val_seq to spi_ps_input_cntl, because the variable only
stores those registers and no others.

Marek