#FastMath
LightNews LightNews
oguzcc.bsky.social
Three items, one mental total. Can you do shop math without a calculator?

📲 iPhone: https://apps.apple.com/app/id6450151568
📲 Android: https://play.google.com/store/apps/details?id=com.oguzcc.nerdy

🔊 AI-generated voiceover.

#mentalmath #reallifemath #mathhacks #fastmath #nerdyapp
November 4, 2026 at 10:05 AM
oguzcc.bsky.social
Coach Max has a 5-second dare for you. Can you solve it before the clock runs out?

📲 iPhone: https://apps.apple.com/app/id6792285717
📲 Android: https://play.google.com/store/apps/details?id=com.oguzcc.mathlon

🔊 AI-generated voiceover.

#mathgames #mathquiz #satmath #fastmath #brainteaser
November 3, 2026 at 6:16 PM
ustinoff.eurosky.social
2 Arrays mit je 1 Mio Zahlen werden addiert. Jetzt bin ich sogar schneller als numpy + (und das war schon 15 mal schneller als Standard-Python +).

NumPy Standard (+): 0.0037 Sekunden
Numba JIT Kogge-Stone (Parallel): 0.0027 Sekunden
Faktor : 0.72x
# 1. Numba JIT-kompilierter Kogge-Stone Addierer @njit(fastmath=True, parallel=True) def numba_kogge_stone_array(a_arr, b_arr): n = a_arr.shape[0] out = np.empty(n, dtype=np.uint32) for i in prange(n): a = a_arr[i] b = b_arr[i] g = a & b p = a ^ b g = g | (p & (g << 1)) p = p & (p << 1) g = g | (p & (g << 2)) p = p & (p << 2) g = g | (p & (g << 4)) p = p & (p << 4) g = g | (p & (g << 8)) p = p & (p << 8) g = g | (p & (g << 16)) p = p & (p << 16) carry = g << 1 out[i] = (a ^ b ^ carry) & 0xFFFFFFFF return out # Testdaten vorbereiten: Kleine Zahlen (max. 1.000.000), kein Überlauf möglich N = 1_000_000 arr_a = np.random.randint(0, 100_000_000, size=N, dtype=np.uint32) arr_b = np.random.randint(0, 100_000_000, size=N, dtype=np.uint32) # Warm-up-Call für Numba _ = numba_kogge_stone_array(arr_a[:10], arr_b[:10]) # --- Test 1: Standard NumPy Addition (ohne Cast & ohne Maske) --- start = time.perf_counter() res_standard = arr_a + arr_b time_standard = time.perf_counter() - start # --- Test 2: Numba JIT Kogge-Stone (Parallel) --- start = time.perf_counter() res_numba = numba_kogge_stone_array(arr_a, arr_b) time_numba = time.perf_counter() - start
October 27, 2026 at 8:58 PM
rwruedii.bsky.social
This looks like a fastmath error with the float, not a typical floating point error.
August 23, 2026 at 8:23 AM
de9uch1.bsky.social
今回の実装、速度厨なのでまだ速くしたい感ある。
numpyのufuncを作って関数をベクトル化、数値演算の制約をfastmath最適化で回避しつつJITコンパイルすれば速くなる可能性はある。ただ素のnumpyでのバッチ化matmulも相当速いので、どっちが速いかは謎。（結局0.05秒しか変わらないとかだと悲しい。）
September 7, 2024 at 8:15 AM
hardy-seifert.bsky.social
Eine gute App für das tägliche Üben.
www.fastmath.app
#BlueLZ #KopfGPT
FastMath: Mental Math Trainer
Become a mental math genius in no time. FastMath makes mental math easy and fun - try it today!
www.fastmath.app
May 31, 2026 at 7:33 AM
newsyc200.bsky.social
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/ (http://news.ycombinator.com/item?id=44142472)
May 31, 2025 at 3:23 PM
rmathew4bs.bsky.social
“How Does A Computer/Calculator Compute Logarithms?”, Zach Artrand (zachartrand.github.io/SoME-3-Living/).

Via HN: news.ycombinator.com/item?id=4074... (which provides important addenda)

#Calculator #Mathematics #Maths #Logarithms #TaylorSeries #NumericalMethods #NumericalAnalysis #FastMath
SoME-3-Living
Summer of Math Exposition 3 - How does a computer/calculator compute logarithms?
zachartrand.github.io
June 22, 2024 at 12:10 PM
daslu.bsky.social
#scicloj is planning a public meetup about Signal Processing in #Clojure.

The precise agenda and date (tentative: March 2026) will be determined based on your feedback.

clojureverse.org/t/upcoming-m...
Upcoming meetup - Signal processing in Clojure - please reach out
In recent months, there has been some encouraging progress in Clojure’s story for signal processing. A lot of that is work-in-progress and will be made more visible soon: updates to the Fastmath 3 li...
clojureverse.org
January 27, 2026 at 12:56 PM
users.rust-lang.org.web.brid.gy
Rust autovectorization issues
thanks for the suggestion. TIL that -Ofast enables `-ffast-math` flag. I removed `-ffast-math` and set -O3: g++ -Wall -Wextra -O3 -march=native -mtune=native -std=c++23 select_elements.cpp the elapsed time is `0.58802 s ± 0.01281 s`, so it is slower than Rust version now. Interesting. But I still wonder why is numba then produces quite fast code (even if I explicitly turn off fast math)? In [3]: @jit(fastmath=False) ...: def sum_selected(v: np.ndarray): ...: res_sum = np.float64(0.0) ...: for i in v: ...: if i > 0.5: ...: res_sum += i ...: return res_sum ...: In [4]: v = np.random.rand(100000000) In [5]: %timeit sum_selected(v) 163 ms ± 888 μs per loop (mean ± std. dev. of 7 runs, 1 loop each)
users.rust-lang.org
February 28, 2025 at 6:30 PM
hnfeed.bsky.social
Beware of Fast-Math
https://simonbyrne.github.io/notes/fastmath/
[comments] [280 points]
June 1, 2025 at 12:03 AM
newsyc250.bsky.social
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/ (http://news.ycombinator.com/item?id=44142472)
May 31, 2025 at 7:33 PM
sbseltzer.net
At the very least I think this covers all mainstream desktop platforms. Might be some odd behavior if any Unreal targets don't use IEEE standard NaN (e.g. -fastmath and the like).
July 9, 2025 at 9:35 PM
betterhn50.e-work.xyz
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/ (https://news.ycombinator.com/item?id=44142472)
Beware of Fast-Math
simonbyrne.github.io
May 31, 2025 at 9:13 AM
hnws.bsky.social
Beware of Fast-Math
L: https://simonbyrne.github.io/notes/fastmath/
C: https://news.ycombinator.com/item?id=44142472
posted on 2025.05.31 at 03:05:57 (c=0, p=4)
May 31, 2025 at 7:37 AM
betterhn300.e-work.xyz
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/ (https://news.ycombinator.com/item?id=44142472)
Beware of Fast-Math
simonbyrne.github.io
June 1, 2025 at 9:24 AM
ponder.ooo
makes sense to use fast math when ur just like deciding where to put the triangle on screen, doing super duper basic shit where everything stays bounded in nice predictable ways and don't rly care about precision or reproducibility at all but for a general-purpose API always-enabled fastmath is NUTS
July 30, 2025 at 7:55 PM
hn100.atproto.rocks
Beware of Fast-Math
https://simonbyrne.github.io/notes/fastmath/

https://news.ycombinator.com/item?id=44142472
Beware of Fast-Math
Beware of Fast-Math
simonbyrne.github.io
May 31, 2025 at 11:30 AM
hnbot.gsuscs.xyz
Beware of Fast-Math

https://simonbyrne.github.io/notes/fastmath/
May 31, 2025 at 8:00 AM
fclc.bsky.social
… reductions if fastmath/ieee754 conformance is not active
January 13, 2025 at 4:11 PM
fclc.bsky.social
I suspect it was an honest mistake for how that got in. Some compilers (correctly) only allow for unordered reductions if fastmath/ieee754 conformance is active
January 13, 2025 at 4:08 PM
betterhn20.e-work.xyz
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/ (https://news.ycombinator.com/item?id=44142472)
Beware of Fast-Math
simonbyrne.github.io
May 31, 2025 at 8:13 AM
be aware, if you are targetting cross platform and not just directx/hlsl then some shader compilers may act counterintuitively with NaN/Inf. Specifically, in Metal with fastmath enabled the shader compiler can treat NaN or INF as undefined behaviour. It may delete whole chunks of code
May 26, 2025 at 10:16 AM
alastair.mastodon.me.uk.ap.brid.gy
Testing some Python Vibe coding like the cool kids #claude #python #fractal

https://blog.0x32.co.uk/posts/fractalpy/
Vibe Coding fractalPy with Claude
Vibe coding a Mandelbrot renderer Spent today “vibe coding” with Claude, just describing what I wanted and letting it drive, to see how far that gets you on a small self-contained project. The result is fractalPy, a small Numba-accelerated Mandelbrot set renderer. It started as a single script — render the set to a PNG, colour it, done — using @njit(parallel=True, fastmath=True, cache=True) to JIT-compile the escape-time kernel so it’s fast enough to iterate on interactively:
blog.0x32.co.uk
August 2, 2026 at 3:01 PM
mtxvpb.bsky.social
Beware of Fast-Math https://simonbyrne.github.io/notes/fastmath/
July 6, 2025 at 12:01 PM