The issue is that the AVX circuitry is actually powered down when not in use to save power and run cooler, so as soon as an instruction is used, the CPU starts to power up the circuits but this takes ~70 microseconds (and this reduces the base clock of the chip pretty much immediately), and after ~700 microseconds of not using the instructions it powers down the circuits again.
So if you know you have a batch of milliseconds (or more) of intensive work to do (eg crypto, rendering), it may be worth it. But I work on a maths library and while I have written SSE2 and AVX vectorised version of key routines, calling the AVX ones is for us generally a net slowdown as it slows down all the logic that occurs around those routines (eg most of our matrix operations are around 20x20 to 200x200 size... these are worth speeding up when we do lots of them, but we rarely do "batches")
If you have a startup that is taking a proven idea and exploiting a niche ("twitter for left handers", "ebay for counterfeit goods", "airbnb for dyslexic vegetarians") or you're taking an existing business model and disrupting it via the internet (car rental, clothes swaps, food delivery) then you may well find seed money in the UK.
But if it's a new idea (new product/service) then in my experience even with a team with a proven track record, and a product built up and running to proof-of-engineering-principle, there aren't many people who can look at an idea and a first version and see beyond the fact that it's new, different and raw to see the opportunity to invest.
It would seem the early stage money is not tech-savvy, not excited by tech, and is looking for iteration over innovation. That is, it's risk averse and favours the reliability of a decent chance of incremental gains rather than lower chance of much bigger gains.
I haven't met everyone, by a long stretch, and it may well be that my product/pitch is wrong, but genuine innovation doesn't seem to come through the UK seed/VC channels, while such products start in the UK they generally find seed money from the US.
See http://www.pushing-pixels.org/2011/11/04/about-those-vector-... for a detailed exploration of precisely the issue with icons and real-life examples of how smaller versions are actually different images (eg a large 3d image with drop shadows gets gradually flattened to a simplified 2d version).
One thing you can do for patching SVG is to have multiple versions of images/shapes and use CSS rules to select just the most appropriate one to show depending on a class you add to some root or parent item - hiding the other until needed.
I haven't tested this, but wonder if the browser would be smart enough to prioritise loading of the different images depending on the visibility.
Otherwise you could generate those parts of the SVG as required, and use the load event to only swap the display to the more appropriate version of an image once loaded.. hence as you zoom the display, you first get the "core" image/shapes etc simply scaled by the browser, and then as background images are loaded, these scaled versions are replaced by more appropriate versions.
This would match the way that zooming works with tiles in apps like Google maps - when you first zoom in you simply get the base tile images scaled by the browser, and then replaced by the more detailed versions as they're loaded.
https://stackoverflow.com/questions/35663635/why-do-processo...
The issue is that the AVX circuitry is actually powered down when not in use to save power and run cooler, so as soon as an instruction is used, the CPU starts to power up the circuits but this takes ~70 microseconds (and this reduces the base clock of the chip pretty much immediately), and after ~700 microseconds of not using the instructions it powers down the circuits again.
So if you know you have a batch of milliseconds (or more) of intensive work to do (eg crypto, rendering), it may be worth it. But I work on a maths library and while I have written SSE2 and AVX vectorised version of key routines, calling the AVX ones is for us generally a net slowdown as it slows down all the logic that occurs around those routines (eg most of our matrix operations are around 20x20 to 200x200 size... these are worth speeding up when we do lots of them, but we rarely do "batches")