Programming Leftovers

-
Qt 6 Build System
Our beloved configure script still exists. It accepts most of the options you are used to from Qt 5.
At the moment the qmake build system is still in place. You can choose between buiding Qt with qmake or CMake. Pass -cmake to configure to use the CMake-based build.
For the final release, expect that the qmake build system will vanish together with the need to pass -cmake.
-
There and Back Again - fixing a 10-year-old bug in QNAM - KDAB -
A few days ago, a customer using our KD Soap library reported an interesting problem with it. The library worked perfectly to access the SOAP server, but there was some noticeable latency when doing a series of requests in short succession from the client. This latency should not have been there, as both the server and the client were in the same local network.
An investigation began, regarding whether the client code or KD Soap were doing something suboptimal, like reconnecting to the server all the time, or something like that. The actual reason turned out to be something much more profound: a 6-year-old bug in QNetworkAccessManager (at least, that’s when it was reported. I think the bug is at least 10 years old…)
-
OpenCV 4.5 Released With Support For Multiple OpenCL Contexts, OpenVINO Backend
OpenCV as the widely-used, real-time computer vision library is out this week with version 4.5 as a big feature release.
OpenCV 4.5 brings OpenCL support for multiple contexts, its DNN module now supports OpenVINO 2020.1 as a possible back-end for inference, and there are optimizations to the NVIDIA CUDA DNN code path. OpenVINO supports inference on both CPUs and Intel graphics hardware.
-
plocate 1.0.0 released
I've released version 1.0.0 of plocate, my faster locate(1)! (Actually, I'm now at 1.0.2, after some minor fixes and improvements.) It has a new build system, portability fixes, man pages, support for case-insensitive searches (still quite fast), basic and extended regex searches (as slow as mlocate) and a few other options. The latter two were mostly to increase mlocate compatibility, not because I think either is very widely used. That, and supporting case-insensitive searches was an interesting problem in its own right

It now also has a small home page with tarballs. And access() checking is also now asynchronous via io_uring via a small trick (assuming Linux 5.6 or newer, it can run an asynchronous statx() to prime the cache, all but guaranteeing that the access() call itself won't lead to I/O), speeding up certain searches on non-SSDs even more.
-
Performance Can Only Increase – Mike Blumenkrantz – Super. Good. Code.
Well, as I’m now caching descriptor sets across descriptor pools, it occurred to me that, assuming my descriptor state hashing mechanism is accurate, all the resources used in a given set must be identical. This means that all resources for a given type (e.g., UBO, SSBO, sampler, image) must be completely identical to previous uses across all shader stages. Extrapolating further, this also means that the way in which these resources are used must also identical, which means the pipeline barriers for access and image layouts must also be identical.
Which means they can be stored onto the struct zink_descriptor_set object and reused instead of being accumulated every time. This reuse completely eliminates add_transition() from using any CPU time (it’s the left-most block above update_sampler_descriptors() in the first graph), and it thus massively reduces overall time for descriptor updates.
This marks a notable landmark, as it’s the point at which update_descriptors() begins to use only ~50% of the total CPU time consumed in zink_draw_vbo(), with the other half going to the draw command where it should be.
-
San Diego Perl Mongers Meeting, Tuesday, October 13th
This is a quick reminder that the quarterly San Diego Perl Mongers meeting will be occurring on Tuesday evening, starting at our normal time of 7 PM PDT. As has been for the last several meetings, we’re going to meet online, as in-person meetings are discouraged, and online meetings seem to be a bit more popular.
-
Tau Station, The Internal Adventures: DBIx::Class::Row cache (avoid the rabbits!)
Tau Station, "the free-to-play narrative sci-fi MMORPG (TM)", has a nicely complex database. Currently, we have 190 tables with 740 relationships between those tables. DBIx::Class does an amazing job at managing that complexity, since each relationship is simply an accessor on the DBIx::Class::Row object.
However, there is a subtle issue when using those relationship accessors. Using a relationship accessor creates a new Row object and stores it in the calling object. This behavior can easily leads to duplicate DBIC Row objects for a single database row. At best, the duplicates cause wasted resources duplicating the Rows. At worst, they cause update anomalies, since updates done to one Row object are not seen by the duplicate objects.
With a highly-connected schema like we have in Tau Station, trying to handle the object duplication can pretty soon feel like we're trying to handle rabbits in Australia.
In order to avoid this Row duplication, we have developed a cache of DBIC Row objects that is shared within the application. In most cases, this allows us to ensure that we have one DBIC Row object per database row while processing an HTTP request, avoiding those subtle update anomalies.
-

- Login or register to post comments
Printer-friendly version- 2730 reads
PDF version
More in Tux Machines
- Highlights
- Front Page
- Latest Headlines
- Archive
- Recent comments
- All-Time Popular Stories
- Hot Topics
- New Members
digiKam 7.7.0 is released
After three months of active maintenance and another bug triage, the digiKam team is proud to present version 7.7.0 of its open source digital photo manager. See below the list of most important features coming with this release.
|
Dilution and Misuse of the "Linux" Brand
|
Samsung, Red Hat to Work on Linux Drivers for Future Tech
The metaverse is expected to uproot system design as we know it, and Samsung is one of many hardware vendors re-imagining data center infrastructure in preparation for a parallel 3D world.
Samsung is working on new memory technologies that provide faster bandwidth inside hardware for data to travel between CPUs, storage and other computing resources. The company also announced it was partnering with Red Hat to ensure these technologies have Linux compatibility.
|
today's howtos
|








.svg_.png)
Content (where original) is available under CC-BY-SA, copyrighted by original author/s.

Recent comments
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago
1 year 11 weeks ago