UPDATE: The speed (for me on my slow Chromebook) when the website isn't changing is a stable 1.014 seconds!!! I have squeezed every ounce of performance out of this, using coding techniques I normally wouldn't do... It will be WAYY faster in a use-case scenario because the pixels will be similar in color to one another. Right now it is just making a random assortment of pixels. This project refreshes every 5 seconds. See inside to check out the TIMES list (note that I hid it for more optimization). This list shows you how long each frame took to render.
Many thanks to Chromecat_test for the "Every Block Timed v2" The best I've seen with this hunk of junk (on my slow ahhh Chromebook) was ~1.002 Optimizations I tried and failed: - Set pen transparency instead of penup/pen down (since setpencolorparam RTC is apparently faster than pen down/pen up. This, interestingly, produced slower but more consistent results. - Making a "blank" sprite. Probably only works for projects with a lot of assets. Optimizations that worked: - repeat until instead of repeat (because repeat's RTC is higher than repeat until's). This dropped it down from 1.7 average to 1.5 average. - Instead of repeat until i > length of list, I did ypos < -79. This is because length of list is EXPENSIVE (2.0 RTC) and the (variable) block is 2x more expensive than the x/ypos blocks. This showed WAY more 1 second times than before. - Using the old trick of "if <not <color = item i of frame>>" - It appears to me that the sweet spot for the sprite's costume size is exactly 4x4 pixels. - Unsurprisingly, hiding the TIMES list made the whole project go faster. Huh.
