"hardware accelerated" is such a weird term
=> More informations about this toot | View the thread
The latency is measured using iPhone’s 240 Hz camera and Is It Snappy; so there’s ~4ms error inherent in measurements plus measurement using physical keyboard has some wiggle room because the key travel is somewhat ambiguous - I try to use full pressing point as a start but it’s somewhat error prone.
=> More informations about this toot | View the thread
In my endless pursuit of typing latency I’m happy to report that the keyboard can have a material impact. Switching from Durgod K320 to (pictured) NuPhy Halo65 HE cuts end to end typing latency from 60ms to 40ms.
I also built a small fake “keyboard” with Arduino that allows me to measure the entire stack after the keyboard sends data over USB which is about 20ms. Using Wayland with 144Hz monitor for this plus Zed or builtin terminal.
=> View attached media | View attached media
=> More informations about this toot | View the thread
So this was a funny coincidence...
re:
https://mastodon.gamedev.place/@zeux/113766835964360251
Of course I ended up accidentally missing one vreinterpret and didn't test the final snippet on gcc/arm.
Someone filed a bug today, which I fixed.
During breakfast walk I was thinking, it's such a shame that GitHub doesn't have Linux ARM CI, so I can't test this easily in the future.
And lo and behold, as of today, they do! https://github.blog/changelog/2025-01-16-linux-arm64-hosted-runners-now-available-for-free-in-public-repositories-public-preview/
So I'm now able to actually test this in CI :)
=> More informations about this toot | View the thread
Using C++03 apparently can break your STL vector implementation completely, nice.
=> More informations about this toot | View the thread
Glad I stopped using WordPress long before whatever the heck is happening to it now.
My blog started on Blogspot in 2007, migrated from Blogspot to Wordpress in 2010 but then switched to a static site generator in 2014. Feels good to not care about any of that. In a sense it’s yet another form of independence which turns out has immense value.
=> More informations about this toot | View the thread
(obviously pugixml does not require C++17. in fact it should still compile on MSVC 2003! not sure about VC6)
=> More informations about this toot | View the thread
[...] Since C++17 standardized string_view and became the default in many compilers in the last few years, this felt like the best route forward that maintains ABI/API compatibility but makes pugixml easier to use.
Naturally if you don't want to have anything to do with STL you can still use PUGIXML_NO_STL and ignore all that :)
=> More informations about this toot | View the thread
pugixml 1.15 is live! In addition to regular maintenance (warning fixes, build system improvements, etc.), many APIs now have std::string_view overloads (and, as a byproduct, transparently support std::string).
This was a long time coming; I've wanted to do something like this before C++17 existed but there were various small issues associated with this so it never happened. [...]
https://github.com/zeux/pugixml/releases/tag/v1.15
=> More informations about this toot | View the thread
Yup - this is exactly why Luau went with method JIT instead of tracing. In addition to unlocking AOT, predictable easy to reason about performance was just way more important. It helped that gradual typing that was being established provided a clear way to keep the generated code lean when the source was type-annotated.
From https://www.pypy.org/posts/2025/01/musings-tracing.html
=> More informations about this toot | View the thread
and this is part of their portable SIMD stuff as well of course!
=> More informations about this toot | View the thread
Both Windows and Android have come a long way.
It is now substantially faster to setup an Android phone and Android Studio development environment compared to setting up Windows with Visual Studio.
=> More informations about this toot | View the thread
Got a Snapdragon X laptop, 3.72 GB/s => 4.54 GB/s with the same optimization, nice :) Not so Apple specific after all!
(also huh Qualcomm does know how to make CPUs now)
=> More informations about this toot | View the thread
What's people's current experience with filing MSVC codegen perf bugs?
I used to not want to file bugs because experience was bad (circa 2018?), then there was a period of "I can file bugs and bugs get fixed" (circa 2021?), but then bugs stopped getting fixed and experience got bad again so I lost interest. Not sure if I should bother creating minimal reproducers etc.
=> More informations about this toot | View the thread
Every time I write NEON code I wish the strongly typed vectors did not exist in GCC. vreinterpret_* stuff is useless and should just die.
It's sad to see Rust repeating this mistake. Burst doesn't.
=> More informations about this toot | View the thread
3.79 GB/s => 4.85 GB/s decode on Apple M2, noice
=> More informations about this toot | View the thread
New blog post, "Year of independence"! https://zeux.io/2024/12/31/year-of-independence/
=> More informations about this toot | View the thread
The trend continues which is kinda funny.
=> More informations about this toot | View the thread
high refresh heartrate
=> More informations about this toot | View the thread
@mbr In http://marc-b-reynolds.github.io/quaternions/2017/05/02/QuatQuantPart1.html, is the Cayley transform correct? The inverse seems to produce the conjugate quaternion; vec4(-s*v.xyz, 1.0 - s) seems to produce correct reconstruction.
=> More informations about this toot | View the thread
=> This profile with reblog | Go to zeux@mastodon.gamedev.place account This content has been proxied by September (ba2dc).Proxy Information
text/gemini
