07:58 <landorfini> rellla: have you even seen daddy the libv lately, i really i think i was a cause of his bad moods? Cause i did not behave very well with him, as i later understood, for me everything started to take shape, with huge simplifications to my methods.

07:59 <landorfini> that is the only developer with whom conflicts i regret, cause he had given support and momentum for me earlier.

07:59 <landorfini> added to that dutch, belgian and german people i have always consistently liked too.

08:05 <landorfini> you see rellla: i got injured three times starting from 2001 as a really young man with one of the lowsiest methods ever, i got damaged deliberately and i allready have skeletons long times in my closets, i can not tolerate it, but you see, my brain never underwent any serious injury though, i have been due to physical injury which i never entirely got treated locked to computer technology

08:06 <landorfini> and i do it really well

08:06 <landorfini> but this was more like not the first cards i had or first plans how to live my life, it is a backup plan

08:08 <landorfini> belive me if you were in my shoes, they committed so upsetting worlds lowsiest terror at me, you would be even more unstable and pissed off in my shoes, the anger towards them is totally justified.

08:10 <landorfini> it is very mentally stressful inconvineient and painful to see, that such raw terrorists and fellons kind agents to boot off native estonians are supported by internationals or other europeans, this is huge pain for me really.

08:14 <landorfini> I tend to think about justice my own, and it is worth rather to charge badly serious fellons like this, rather than endorse them :(

13:35 <rellla> ok, i tried to do some benchmarking with https://gitlab.freedesktop.org/rellla/gfx/-/blob/master/gbm-indexed-draw/main.c

13:36 <rellla> mesa already does rebuild the index buffer, if there is a gap between the indices. afaik this is done here: https://elixir.bootlin.com/mesa/mesa-19.2.0-rc1/source/src/gallium/auxiliary/util/u_vbuf.c#L1372

13:38 <rellla> so i tested both, with and without this optimization. i also tried to execute glDrawElements 1 and 3 times. in the latter case this should trigger lima cache

13:38 <rellla> and i did this with different index buffer ranges. the result is here:

13:38 <rellla> http://imkreisrum.de/tmp/indexed_draw.jpg

13:40 <rellla> what i wonder is, that we only benefit from the indices_unroll in one case out of four with mesa master ...

13:41 <rellla> though difference is 2~20ms

13:45 <rellla> and i wonder why [0,1,2,997,998,999] seems to be faster than [0,1,2,97,98,99] ...

13:46 <rellla> anyway, i'll leave analysis up to you for today :)