From: Boris Brezillon Date: Fri, 2 Aug 2019 17:18:44 +0000 (+0200) Subject: panfrost: Bail out early when doing a wallpaper blit X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=95507a3dd4536a5abacba3f8f302066f96fe2a5e;p=mesa.git panfrost: Bail out early when doing a wallpaper blit The wallpaper blit is a bit special in that the operation is targetting the current FB, but the u_blitter logic creates a new surface for it which makes util_framebuffer_state_equal() return false. In that case we don't want a new FB descriptor to be emitted/attached, so let's just copy the new state into ctx->pipe_framebuffer and exit the function. Signed-off-by: Boris Brezillon Reviewed-by: Alyssa Rosenzweig --- diff --git a/src/gallium/drivers/panfrost/pan_context.c b/src/gallium/drivers/panfrost/pan_context.c index 65e6824a9b0..a6412de7646 100644 --- a/src/gallium/drivers/panfrost/pan_context.c +++ b/src/gallium/drivers/panfrost/pan_context.c @@ -2369,10 +2369,22 @@ panfrost_set_framebuffer_state(struct pipe_context *pctx, if (util_framebuffer_state_equal(&ctx->pipe_framebuffer, fb)) return; - if (!ctx->wallpaper_batch && (!is_scanout || has_draws)) { - panfrost_flush(pctx, NULL, PIPE_FLUSH_END_OF_FRAME); + /* The wallpaper logic sets a new FB state before doing the blit and + * restore the old one when it's done. Those FB states are reported to + * be different because the surface they are pointing to are different, + * but those surfaces actually point to the same cbufs/zbufs. In that + * case we definitely don't want new FB descs to be emitted/attached + * since the job is expected to be flushed just after the blit is done, + * so let's just copy the new state and return here. + */ + if (ctx->wallpaper_batch) { + util_copy_framebuffer_state(&ctx->pipe_framebuffer, fb); + return; } + if (!is_scanout || has_draws) + panfrost_flush(pctx, NULL, PIPE_FLUSH_END_OF_FRAME); + /* Invalidate the FBO job cache since we've just been assigned a new * FB state. */