Improve code caching |
||
Issue descriptionDetails & design doc TBD. We want to improve code caching [1], particularly in two aspects: - Currently, code caching triggers only rarely, meaning that some of its potential is wastes. - Currently, code caching makes all decisions at initial compile time, which forces its implementation to be rather conservative. The idea, roughly, is to change the code caching implementation to allow incremental updates to the code cache. That way, we could do less caching up-front (which is currently a costly operation, and its costliness is one of the reasons why we trigger so conservatively), and instead update the cache after the initial compile, maybe function-by-function. [1] https://blog.chromium.org/2015/03/new-javascript-techniques-for-rapid.html
,
Feb 3 2017
,
Feb 3 2017
One thing to keep in mind, currently our code looks something like this:
__d("Module1", ["Dep1","Dep2"], function(Dep1, Dep2) { module.exports = ... });
__d("Module2", ["Dep1","Dep2"], function(Dep1, Dep2) { module.exports = ... });
So any incremental updates we did would likely consist of giving a new version of one of these modules (and the function we pass to the __d function)
,
Feb 6 2017
Design doc (work in progress): https://docs.google.com/document/d/1-HhAgqRpHs9YQF9Zs7pTovUw8EE06tR3_zMIxaeCyOw/view
,
Feb 6 2017
#3: This issue is primarily about improving the existing code caching mechanism. It's not about extending the reach of the feature, e.g. mixing content from different (versions of) resources, or exposing control of the cache to web developers via ServiceWorkers (or some-such). (The mechanism would likely be beneficial for doing any of those things, however, in case we do want to go down that path.) |
||
►
Sign in to add a comment |
||
Comment 1 by marja@chromium.org
, Jan 26 2017