I think it was per axle weight, but since Lady P’s FAB 1 is not on the road, I should stop being a pedant 😝

I found my success rate with rose cuttings was (maybe) only 10%, but given those were left over bits from the end of Autumn/Winter pruning that otherwise were going on the compost heap, it’s still a win 😀

I recently remembered we were told about it in 'CS101'.

Maybe bits have become more accurate in the intervening 40+ years. 😝

FACTCHECK UPDATE: No, the UK can't "be more like Norway" in the North Sea

🇬🇧 extracted most of its oil & gas after privatisation
🇳🇴 has much more left due to state stewardship
🤔Lawson spent UK windfall on tax cuts; Norway got a sovereign wealth fund

www.carbonbrief.org/...

Just got back from Revision demoparty—my 64KB intro took 1st place. Huge thanks to @alkama.bsky.social for the audio work that made this possible. It’s a fully procedural 3:40 animation packed into a 58KB executable. I had a blast building it, and the audience reaction was incredible.

PVRIC was primarily a lossless buffer (eg RGBA, Z, YUV) compression scheme, getting say 40-60% reduction (YYMV) in bandwidth, but not memory footprint since it has to cope with non-compressible blocks.

It was extended to have a 'guaranteed' 50% mode that gets used if the lossless doesn't fit.

Hang on... didn't you complain about Storm Dave in an earlier post? Surely 'he' was partially responsible 😀

FWIW the SRM submissions website seems to be 'down'

My kids' high school Computer Science text book said (paraphrasing) "Use floating point type for money".

I could cry.

Angry Cockatoo Destroys Anti-Bird Spikes 😂 YouTube video by Pawsome Pets

A sulphur crested cockatoo took a different approach …

m.youtube.com/watch?v=bzjP...

And remove the padding page stuck on the front?

A butternut pumpkin that looks like a posterior (childish snigger)

In that case (though technically a fruit and not a vegetable) …

Got to the section where the piano spectrum is analysed and was surprised there was a significant 7th harmonic. I thought the hammers in pianos were arranged to strike the strings on a node of the 7th harmonic so that it was lessened because the 7th was meant to be less pleasant to the human ear.

Rupert the Bear and Bill Badger comic book characters

How can you claim such a thing!
I’m sure there’s evidence UK badgers are equal to bears!

images.fineartamerica.com/images/artwo...

Reminds me when I arrived in the UK from Oz and discovered magpies were entirely different species on the two continents.
I was thinking the European ones were harmless (compared to Aus) until I saw one kill all the chicks in a blackbird nest. 😳
At least the magpies in Aus have a nice song.

This site (www.electrickettlesguide.com/best-electri...) suggests 1.5kW is common in the US.
I’m certainly more used to ≈2.3

If America hadn’t got a cocked-up maths dictionary (mathworld.wolfram.com/Trapezium.html) maybe trapezium would have been easier to rhyme 😝.

/me goes and looks up Otsu’s algorithm…

Hmm, yes, that’s probably the same concept.

FWIW you might find Delp & Mitchell’s BTC of interest - it was a very early, block-based image compression scheme.

I can’t recall what was in the paper, but in the DC VQ compressor, during the initial phase, the split of a given set is done to minimise its local error by sliding the partition point. That can be done in linear time.

6/n
The second phase is to do 'some number' of global 'k-means' passes to improve the reps.
That is, for each original vector, try all of the N reps to see which is closest, and once done, update the reps to be the average of all that map to it. It should converge after a while.

5/n
Once you have your "N" subsets, generate your N initial reps by taking the average of each subset.

4/n
5) Put the left and right subsets into a list of vector sets.
6) stop when the # of subsets == the number of colours you want (eg N=256)
otherwise...
7) from the list of, remove the one with the biggest error and go to back to step 2

3/n ...
2) Find the principal axis through that vector set.
3) Take the dot product of each vector VS the axis, and sort the vectors based on dot prod.
4) Find optimal partition point that splits the set into "left" and "right" subsets so that the local error sum is minimised.

How does hardware texture compression work? That it compresses the data compared to the pixel array is obvious. But what makes it different from from normal compression (like png, jpeg)?

2/n ... so effectively 12 or 16D (rather than 3 for just palettes), but that shouldn't really matter. (FWIW there's an example in computergraphics.stackexchange.com/a/1551/209)

The approach I took was done in 2 phases.
Phase 1 was...
1) create a set of vectors from the image.

Oh gosh! Now I need to remember exactly what I did and how it may have differed from Wu's paper! (It was back in the late 90s so it's not *exactly* fresh in my memory)

First a disclaimer. This for the Dreamcast VQ compressor which was doing 2x2 pixel vectors (or sometimes 4x2) of RGB or RGBA ...

Four weeks to the EGSR papers deadline! (I'm so excited to be on the Program Committee this year - cannot wait to see all the amazing papers submitted!) So - submit away!

Important dates:

Abstracts deadline: 4/8

Papers deadline: 4/15

Submit your work, and join in Bordeaux!

egsr2026.inria.fr

High-Performance Graphics 2026 CFP High-Performance Graphics 2026 Call for Participation

The papers deadline for High-Performance Graphics 2026 is 3 weeks from now and the abstract deadline is 2 weeks from now (April 1 AoE).
If you want to submit, but have concerns about traveling to LA, please contact the paper chairs for options.
www.highperformancegraphics.org/2026/call-fo...

“Color image quantization for frame buffer display” by Heckbert

Would you know how that compares to Heckbert’s Median cut approach (history.siggraph.org/learning/col...)?

FWIW I think I used a variant of Wu’s “Color Quantization by Dynamic Programming and Principal Analysis” as the basis for the Dreamcast texture compressor. Might be worth a look.

If you can learn dot products and homogeneous vectors then you’re 95% the way there.

FWIW I use DDD for debugging C (and occasionally C++) on Linux.