[Mesa-dev] [PATCH v02 27/37] i965: Port gen8+ 3DSTATE_PS_EXTRA to genxml.

Kenneth Graunke kenneth at whitecape.org
Thu Apr 27 06:26:15 UTC 2017


On Monday, April 24, 2017 3:19:22 PM PDT Rafael Antognolli wrote:
[snip]
> diff --git a/src/mesa/drivers/dri/i965/genX_state_upload.c b/src/mesa/drivers/dri/i965/genX_state_upload.c
> index 0f7a222..7ed79b2 100644
> --- a/src/mesa/drivers/dri/i965/genX_state_upload.c
> +++ b/src/mesa/drivers/dri/i965/genX_state_upload.c
> @@ -1668,6 +1668,99 @@ static const struct brw_tracked_state genX(raster_state) = {
>     },
>     .emit = genX(upload_raster),
>  };
> +
> +/* ---------------------------------------------------------------------- */
> +
> +static void
> +genX(upload_ps_extra)(struct brw_context *brw)
> +{
> +   const struct brw_wm_prog_data *prog_data =
> +      brw_wm_prog_data(brw->wm.base.prog_data);
> +
> +   brw_batch_emit(brw, GENX(3DSTATE_PS_EXTRA), pse) {

Could we call this psx?  That's the abbreviation I used in the old code.

> +      pse.PixelShaderValid = true;
> +      pse.PixelShaderComputedDepthMode = prog_data->computed_depth_mode;
> +      pse.PixelShaderKillsPixel = prog_data->uses_kill;
> +      pse.AttributeEnable = prog_data->num_varying_inputs != 0;
> +      pse.PixelShaderUsesSourceDepth = prog_data->uses_src_depth;
> +      pse.PixelShaderUsesSourceW = prog_data->uses_src_w;
> +      pse.PixelShaderIsPerSample = prog_data->persample_dispatch;
> +
> +      /* _NEW_MULTISAMPLE | BRW_NEW_CONSERVATIVE_RASTERIZATION */
> +      if (prog_data->uses_sample_mask) {
> +#if GEN_GEN >= 9
> +         struct gl_context *ctx = &brw->ctx;

Let's just move ctx to the top and mark it UNUSED to shut up warnings.
(That way, if we introduce something else that uses ctx, it'll already
be in place...plus, that's just where it normally is...)

> +
> +         if (prog_data->post_depth_coverage)
> +            pse.InputCoverageMaskState = ICMS_DEPTH_COVERAGE;
> +         else if (prog_data->inner_coverage && ctx->IntelConservativeRasterization)
> +            pse.InputCoverageMaskState = ICMS_INNER_CONSERVATIVE;
> +         else
> +            pse.InputCoverageMaskState = ICMS_NORMAL;
> +#else
> +         pse.PixelShaderUsesInputCoverageMask = true;
> +#endif
> +      }
> +
> +      pse.oMaskPresenttoRenderTarget = prog_data->uses_omask;
> +#if GEN_GEN >= 9
> +      pse.PixelShaderPullsBary = prog_data->pulls_bary;

Let's move the computed stencil stuff here to combine blocks:

      psx.PixelShaderComputesStencil = prog_data->computed_stencil;

That's 1 line instead of 6.

(I'm not worried about losing the assert - if anything, we could
assert gen >= 9 in the compiler when setting the prog_data field.)

> +#endif
> +
> +      /* The stricter cross-primitive coherency guarantees that the hardware
> +       * gives us with the "Accesses UAV" bit set for at least one shader stage
> +       * and the "UAV coherency required" bit set on the 3DPRIMITIVE command
> +       * are redundant within the current image, atomic counter and SSBO GL
> +       * APIs, which all have very loose ordering and coherency requirements
> +       * and generally rely on the application to insert explicit barriers when
> +       * a shader invocation is expected to see the memory writes performed by
> +       * the invocations of some previous primitive.  Regardless of the value
> +       * of "UAV coherency required", the "Accesses UAV" bits will implicitly
> +       * cause an in most cases useless DC flush when the lowermost stage with
> +       * the bit set finishes execution.
> +       *
> +       * It would be nice to disable it, but in some cases we can't because on
> +       * Gen8+ it also has an influence on rasterization via the PS UAV-only
> +       * signal (which could be set independently from the coherency mechanism
> +       * in the 3DSTATE_WM command on Gen7), and because in some cases it will
> +       * determine whether the hardware skips execution of the fragment shader
> +       * or not via the ThreadDispatchEnable signal.  However if we know that
> +       * GEN8_PS_BLEND_HAS_WRITEABLE_RT is going to be set and
> +       * GEN8_PSX_PIXEL_SHADER_NO_RT_WRITE is not set it shouldn't make any
> +       * difference so we may just disable it here.
> +       *
> +       * Gen8 hardware tries to compute ThreadDispatchEnable for us but doesn't
> +       * take into account KillPixels when no depth or stencil writes are
> +       * enabled.  In order for occlusion queries to work correctly with no
> +       * attachments, we need to force-enable here.
> +       *
> +       * BRW_NEW_FS_PROG_DATA | BRW_NEW_FRAGMENT_PROGRAM | _NEW_BUFFERS |
> +       * _NEW_COLOR
> +       */
> +      if ((prog_data->has_side_effects || prog_data->uses_kill) &&
> +          !brw_color_buffer_write_enabled(brw))
> +         pse.PixelShaderHasUAV = true;;

One too many semicolons here

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

> +
> +      if (prog_data->computed_stencil) {
> +         assert(brw->gen >= 9);
> +#if GEN_GEN >= 9
> +         pse.PixelShaderComputesStencil = true;
> +#endif
> +      }
> +   }
> +}
> +
> +const struct brw_tracked_state genX(ps_extra) = {
> +   .dirty = {
> +      .mesa  = _NEW_BUFFERS | _NEW_COLOR,
> +      .brw   = BRW_NEW_BLORP |
> +               BRW_NEW_CONTEXT |
> +               BRW_NEW_FRAGMENT_PROGRAM |
> +               BRW_NEW_FS_PROG_DATA |
> +               BRW_NEW_CONSERVATIVE_RASTERIZATION,
> +   },
> +   .emit = genX(upload_ps_extra),
> +};
>  #endif
>  
>  /* ---------------------------------------------------------------------- */
> @@ -1957,7 +2050,7 @@ genX(init_atoms)(struct brw_context *brw)
>        &genX(sbe_state),
>        &genX(sf_state),
>        &gen8_ps_blend,
> -      &gen8_ps_extra,
> +      &genX(ps_extra),
>        &genX(ps_state),
>        &genX(depth_stencil_state),
>        &genX(wm_state),
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20170426/7965f282/attachment.sig>


More information about the mesa-dev mailing list