|
Mantis - Quercus
|
|||||
| Viewing Issue Advanced Details | |||||
|
|
|||||
| ID: | Category: | Severity: | Reproducibility: | Date Submitted: | Last Update: |
| 2913 | minor | always | 09-08-08 12:36 | 01-04-09 11:24 | |
|
|
|||||
| Reporter: | ferg | Platform: | |||
| Assigned To: | ferg | OS: | |||
| Priority: | normal | OS Version: | |||
| Status: | closed | Product Version: | 3.2.0 | ||
| Product Build: | Resolution: | fixed | |||
| Projection: | none | ||||
| ETA: | none | Fixed in Version: | 4.0.0 | ||
|
|
|||||
| Summary: | 0002913: quercus precompile load | ||||
| Description: |
(rep by Steven Grimm) When I fire up Resin to work with Quercus, the first thing I do is hit a dummy page that is nothing but an include of our standard PHP library suite. Our libraries are pretty huge -- over a thousand files -- and this initial dummy page takes around three minutes to execute the first time as it loads all the .php files I compiled on the previous run. The debug output looks like you'd expect: [12:16:21.510] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/qrt/qrt.php] loading precompiled page [12:16:21.524] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/qrt/constants.php] loading precompiled page [12:16:21.716] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/qrt/utils.php] loading precompiled page [12:16:21.724] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/qrt/versions.php] loading precompiled page [12:16:21.815] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/display/feed/platform.php] loading precompiled page [12:16:22.287] {http--8080-6} Quercus[/data/users/sgrimm/www/lib/display/feed/broadcast.php] loading precompiled page It would be great if I could speed that up. I don't keep Resin running all the time and it's a bit of a pain to have to wait that long to do anything when I have a little time to spend on Quercus testing. Is there a way to speed things up as is by tweaking any configuration settings? If not, a couple of possible approaches: - Load the precompiled pages in the background, the same way compilation is kicked off as a background task, and use interpreted pages in the meantime. Maybe only do that once the number of precompiled pages loaded in a single request exceeds some threshold, so small/simple pages don't have to run any interpreted code. - Parallelize the loading of compiled pages. I'm running on a 4-core machine but only one of those cores gets used when I'm running this page. Probably the previous item is a prerequisite to this. Parallelizing would be great for compiling, too, by the way -- if there were multiple compiler threads it would be much faster to arrive at the desired "everything is compiled" state after blowing away the Quercus work directory, which I sometimes need to do after applying a Quercus bugfix. Thanks! I will probably have some more bug reports for you guys this week (there are still a couple things going wrong in ways I don't understand) but it's looking much better than it used to now. |
||||
| Steps To Reproduce: | |||||
| Additional Information: | |||||
| Relationships | |||||
| Attached Files: | |||||
| Notes | |||||
|
|
|||||
|
|
||||
|
|
|||||
|
|
||||