Boosted by @joe@f.duriansoftware.com
Replying to
@joe@f.duriansoftware.com compilers still benefit from human assistance with the transpose: https://godbolt.org/z/MoEsKMxnb
Remote
Alexander Monakov
@amonakov@mastodon.gamedev.place
mastodon 4.7.3293 Followers
198 Following
28 Posts
Joined January 15, 2018
codeberg:
Replying to
@joe@f.duriansoftware.com in Linux land, they kind of did (and didn't stop there):
Replying to
@joe one can do many wonderful things with just CSS: https://lyra.horse/
(have you played CSS Puzzle Box? https://suricrasia.online/puzzlebox/ )
Replying to
@mid_kid@fosstodon.org printf in shell iterates its argument list :)
printf '%d \n' $(seq 10)
Replying to
@pervognsen@mastodon.social @dougall@mastodon.social
He also took care of all pending nanoBench bugreports: https://github.com/andreas-abel/nanoBench/issues/33#issuecomment-4151237205
Replying to
@joe@f.duriansoftware.com it piles acorns?
Replying to
@lisyarus@mastodon.gamedev.place thanks for sharing this. What do you think about using double precision for intermediates when inputs are in single precision? For example, in your "vector length" example it's much cheaper to cast x, y, z to double instead of dividing by max(x, y, z) (in context of a CPU target, not GPU) — after the cast extreme exponent values do not pose an issue. This is exactly how hypotf in a modern C math library works — it's the double-precision hypot that needs to be fancier than that.
Replying to
@fatlimey@mastodon.gamedev.place err, not just on any gridline — no three points can belong to one (possibly oblique) straight line
the "not on one gridline" problem is not so difficult, because if you take a solution and permute rows/columns arbitrarily, it doesn't change occupancy of grid lines and so still satisfies the restriction
In CSS, is there a crime^Wway to unset max-height if the element was "focused" somehow? I'm looking at collapsing quote blocks in public-inbox. I can do
.q {
display: inline-block;
max-height: 9em;
overflow-y: auto;
resize: vertical;
}
.q[style*=height] {
max-height: unset;
}
which collapses blocks, and lets me resize them. But if I add
.q:active {
max-height: unset;
}
that's not useful, because the block is expanded only as long as the mouse button is pressed, and :focus doesn't work.
Replying to
@joe@f.duriansoftware.com my feed would additionally demand them to be 3D
Oh, so that's why PC HPET runs at that frequency
(from https://en.wikipedia.org/wiki/Color_burst)
Replying to
@never_released@mastodon.social also,
"silicon vendors including [...] Mesa"? 🧐
Replying to
@wolf480pl@mstdn.io it's not a ruling policy if it doesn't include rules for rulering the ruler
“The way they handled it was heartless.”
https://time.com/article/2026/04/30/oracle-layoffs-ai-tech-jobs/
Replying to
@wolf480pl not sure. does not look implausible that they wanted to be more conservative with boost clocks for SIMD/FP workloads for power delivery reasons
Replying to
@lisyarus@mastodon.gamedev.place yeah, mixing floats/doubles becomes quite inconvenient in SIMD context, but scalar code in math library functions really benefits from higher intermediate precision/range
Replying to
@Doomed_Daniel@mastodon.gamedev.place @zeux@mastodon.gamedev.place @lisyarus@mastodon.gamedev.place as far as I'm aware, C++ standards do not have anything analogous to C's Annex F that would say how the language binds to IEEE 754 / IEC 60559.
Replying to
@wolf480pl yeah, not my best code for sure. I've pushed a commit swapping around one if/elif to correct that and match my intention.
My take on HTML templating: https://codeberg.org/amonakov/fml.py
(my Python is cringe probably, please be gentle)
Replying to
@rygorous @kenshirriff re. first paragraph, I was wondering about that too, and turns out the article explains that better: 8x translates to incrementing the next three-bit group
Replying to
@wolf480pl hm but on the other hand instruction mix is not the whole story here
it would be for "max watts at fixed frequency"
are you keeping an eye at CPU's frequency/voltage? it might be it runs at higher turbo when SIMD/FP units are idle
Replying to
@harold "use strict" and "use asm" say hello
Replying to
@wolf480pl right, sorry about not covering this properly in the readme and the code yet.
The intention is that snippets are separated by empty lines or file boundaries.
The above is intended to be taken as two snippets, the first invoking :text2 as its last node.
Replying to
@wolf480pl like a superpower! :)
(also, sorry! I probably should have corrected that before posting here)
Replying to
@MasonRemaley@mastodon.gamedev.place I've made a similar presentation-latency demo recently: https://mastodon.gamedev.place/@amonakov/116073617904950047
It is somewhat more well-defined in that it synchronizes against the presentation event reported by the X server, so the presentation latency is properly controlled (never more than one frame, sleep amount controls the in-frame slop).
Another idea it showcases is submitting all input-independent rendering before the sleep/poll.
Replying to
@horenmar@mastodon.social seeing how PDF producer seems to be Safari, I'm quite curious about your workflow! can you tell how the slides were made?