marcan changed the topic of #asahi-gpu to: Asahi Linux: porting Linux to Apple Silicon macs | GPU / 3D graphics stack black-box RE and development (NO binary reversing) | Keep things on topic | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Logs: https://alx.sh/l/asahi-gpu
<dougall> bloom: great solve! that all properly makes sense now :)
snalty has quit [Read error: Connection reset by peer]
snalty has joined #asahi-gpu
nickiminjaj has quit [Ping timeout: 246 seconds]
<bloom> :)
DragoonAethis has quit [Quit: hej-hej!]
DragoonAethis has joined #asahi-gpu
opticron has quit [Ping timeout: 265 seconds]
opticron has joined #asahi-gpu
odmir has quit [Remote host closed the connection]
odmir has joined #asahi-gpu
odmir has quit [Ping timeout: 240 seconds]
odmir has joined #asahi-gpu
phiologe has quit [Ping timeout: 250 seconds]
phiologe has joined #asahi-gpu
<bloom> and implemented in the compiler :)
<bloom> so now all the known float ops are handled
odmir has quit [Remote host closed the connection]
<bloom> dougall: "# okay, this is very lazy
<bloom> " wrote a proper C encoder
<bloom> will test tomorrow 😴
<dougall> yeah, that sounds much better :) g'night!
nickiminjaj has joined #asahi-gpu
nickiminjaj has quit [Ping timeout: 240 seconds]
nickiminjaj has joined #asahi-gpu
jix has quit [Quit: WeeChat 3.0]
jix has joined #asahi-gpu
nickiminjaj has quit [Ping timeout: 265 seconds]
nickiminjaj has joined #asahi-gpu
nickiminjaj has quit [Quit: leaving]
qyousef_ has quit [Ping timeout: 265 seconds]
qyousef_ has joined #asahi-gpu
Necrosporus has quit [Ping timeout: 265 seconds]
odmir has joined #asahi-gpu
linkmauve has quit [Ping timeout: 250 seconds]
Necrosporus has joined #asahi-gpu
odmir has quit [Remote host closed the connection]
odmir has joined #asahi-gpu
odmir has quit [Ping timeout: 240 seconds]
linkmauve has joined #asahi-gpu
odmir has joined #asahi-gpu
odmir has quit [Remote host closed the connection]
odmir has joined #asahi-gpu
odmir has quit []
<bloom> Interesting... Metal implements unary fneg/fabs with bitop (twiddling the sign bit) instead of fadd.
odmir has joined #asahi-gpu
bloom has quit [Ping timeout: 246 seconds]
bloom has joined #asahi-gpu
<bloom> note to self: `convert` is only for data type converts
<bloom> to change sizes, iadd/fadd are used.
<bloom> also, modifiers are unsupported (both float and int mods)
snalty has quit [Quit: ZNC 1.8.2 - https://znc.in]
<bloom> note: Metal compiler likes bitop_mov more than xor
<bloom> Dunno if that's deliberate.
<bloom> (was trying to find out if there's an xchg/swap instruction)
<bloom> The assembly prduced for precise::sin(x) is horrifying
<bloom> (~90 instructions)
<chrisf> ouch
<bloom> default in Metal is fast::sin(x) which is super cheap
<glibc> precise sin is expensive, there's no getting around that (either instruction, or LUT)
<bloom> that it is..
<bloom> what's more interesting to me is that they don't use the fast sine anywhere in it
<bloom> I would expect if you have fast imprecise transcendentals available, you'd use them and feed the estimate into Newton's method or something
<chrisf> what does the precise sin code look like?