Add avoiding_rmw blog entry

[cworth.org] / src / exa / mozilla_i965_profiles.mdwn

diff --git a/src/exa/mozilla_i965_profiles.mdwn b/src/exa/mozilla_i965_profiles.mdwn
index bb066ca85b8e4b33a60cd8fc3254c849e12a0e56..0c4b220bdf422349b3bba4cb70dc69b0d07fa372 100644 (file)
--- a/src/exa/mozilla_i965_profiles.mdwn
+++ b/src/exa/mozilla_i965_profiles.mdwn
@@ -1,3 +1,8 @@
+[[meta title="Profiles for Mozilla Trender on i965"]]
+
+[[tag exa performance xorg]]
+
+
 I recently [[posted|mozilla_trender]] results showing EXA (and XAA)
 performing quite badly on the Mozilla Trender benchmarks. As a
 reminder, here is the chart showing the results on an i965 card:
@@ -77,7 +82,7 @@ reason?
 And here with EXA we see some good, and some really bad. The good news
 is that the pixman functions doing software compositing have
 disappeared from the top of the profile, leaving only software
-rasterization. But what's with this new i965_prepare_composite
+rasterization. But what's with this new `i965_prepare_composite`
 function that's taking even more time than all of libxul? That seems
 like rather excessive overhead.
 
@@ -87,3 +92,9 @@ A quick glimpse at the
 assignment statements, and then a "long sequence of commands needed to
 set up the 3D rendering pipe". Is any of that setup redundant and
 could it be easily eliminated?
+
+I noticed two calls to i830WaitSync which seemed to have "slowdown"
+written all over them. But I ran with these two calls removed and
+didn't notice any change in the performance, (and if they were causing
+a problem, shouldn't they have shown up separately in the profile
+anyway?).