Attachment 101697 Details for Bug 80500 – Experimental patch that appears to work around flickering shadows issue.

Experimental patch that appears to work around flickering shadows issue.

0001-Experimental-hack-to-help-isolate-flickering-shadows.patch (text/plain), 1.79 KB, created by Courtney Goeltzenleuchter on 2014-06-24 21:07:32 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Courtney Goeltzenleuchter

Created: 2014-06-24 21:07:32 UTC

Size: 1.79 KB

patch

obsolete

>From fca741c06e58af27cea55fae9c1ba09968c3a45b Mon Sep 17 00:00:00 2001
>From: Cody Northrop <cody@lunarg.com>
>Date: Mon, 23 Jun 2014 16:33:50 -0600
>Subject: [PATCH] Experimental hack to help isolate flickering shadows
>
>---
> src/mesa/drivers/dri/i965/intel_buffer_objects.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
>diff --git a/src/mesa/drivers/dri/i965/intel_buffer_objects.c b/src/mesa/drivers/dri/i965/intel_buffer_objects.c
>index 96dacde..8895b7b 100644
>--- a/src/mesa/drivers/dri/i965/intel_buffer_objects.c
>+++ b/src/mesa/drivers/dri/i965/intel_buffer_objects.c
>@@ -240,6 +240,10 @@ intel_bufferobj_subdata(struct gl_context * ctx,
> 
>    assert(intel_obj);
> 
>+   // DO NOT COMMIT!
>+   // This is just an interesting hack that circumvents the real problem
>+   bool isUniformBufferUpdate = (ctx->UniformBuffer == obj);
>+
>    /* See if we can unsynchronized write the data into the user's BO. This
>     * avoids GPU stalls in unfortunately common user patterns (uploading
>     * sequentially into a BO, with draw calls in between each upload).
>@@ -289,6 +293,12 @@ intel_bufferobj_subdata(struct gl_context * ctx,
> 				size);
> 
> 	 drm_intel_bo_unreference(temp_bo);
>+
>+	 // DO NOT COMMIT!
>+	 // This is just an interesting hack that circumvents the real problem
>+	 if (isUniformBufferUpdate)
>+	    intel_batchbuffer_flush(brw);
>+
>          return;
>       } else {
>          perf_debug("Stalling on glBufferSubData(%ld, %ld) (%ldkb) to a busy "
>@@ -303,6 +313,11 @@ intel_bufferobj_subdata(struct gl_context * ctx,
> 
>    drm_intel_bo_subdata(intel_obj->buffer, offset, size, data);
>    intel_bufferobj_mark_inactive(intel_obj);
>+
>+   // DO NOT COMMIT!
>+   // This is just an interesting hack that circumvents the real problem
>+   if (isUniformBufferUpdate)
>+      intel_batchbuffer_flush(brw);
> }
> 
> 
>-- 
>1.8.3.2
>

Actions: View

Attachments on bug 80500: 101697 | 116847

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.