Bug 811467. Increase image decode chunk size. r=joe
Telemetry shows us doing image decoding at around 10000 KB/s. This means we
can do about 10240 bytes per ms. Since we have 5ms to do the decoding that
means we should be able to 51200 bytes in 5ms. I'd suggest we change to 16384
as that gives a healthy margin.
It turns out tenfourfox has also increased the size for performance reasons.
--- a/modules/libpref/src/init/all.js
+++ b/modules/libpref/src/init/all.js
@@ -3609,17 +3609,17 @@ pref("image.mem.decodeondraw", true);
// value and twice this value.
//
// This used to be 120 seconds, but having it that high causes our working
// set to grow very large. Switching it back to 10 seconds will hopefully
// be better.
pref("image.mem.min_discard_timeout_ms", 10000);
// Chunk size for calls to the image decoders
-pref("image.mem.decode_bytes_at_a_time", 4096);
+pref("image.mem.decode_bytes_at_a_time", 16384);
// The longest time we can spend in an iteration of an async decode
pref("image.mem.max_ms_before_yield", 5);
// The maximum amount of decoded image data we'll willingly keep around (we
// might keep around more than this, but we'll try to get down to this value).
pref("image.mem.max_decoded_image_kb", 51200);
@@ -3835,9 +3835,9 @@ pref("network.activity.blipIntervalMilli
// to take effect.
pref("jsloader.reuseGlobal", false);
// When we're asked to take a screenshot, don't wait more than 2000ms for the
// event loop to become idle before actually taking the screenshot.
pref("dom.browserElement.maxScreenshotDelayMS", 2000);
// Whether we should show the placeholder when the element is focused but empty.
-pref("dom.placeholder.show_on_focus", true);
\ No newline at end of file
+pref("dom.placeholder.show_on_focus", true);