alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
megi has quit [Quit: WeeChat 2.9]
megi has joined #panfrost
yann has quit [Ping timeout: 244 seconds]
stikonas has quit [Remote host closed the connection]
dstzd has quit [Quit: ZNC - https://znc.in]
dstzd has joined #panfrost
kaspter has joined #panfrost
_whitelogger has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
dstzd has joined #panfrost
kaspter has quit [Ping timeout: 260 seconds]
kaspter has joined #panfrost
<HdkR> Time to be concerned about how long until Mali is wiped away :/
davidlt has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 240 seconds]
camus1 is now known as kaspter
davidlt has quit [Ping timeout: 260 seconds]
_whitelogger has joined #panfrost
buzzmarshall has quit [Remote host closed the connection]
davidlt has joined #panfrost
kaspter has quit [Ping timeout: 258 seconds]
kaspter has joined #panfrost
nlhowell has quit [Quit: WeeChat 2.9]
kinkinkijkin has quit [Remote host closed the connection]
<icecream95> HdkR: Maybe
<icecream95> it will just be artificially slowed down with `for (int i = 0; i < 100000; ++i);` statements in the driver...
davidlt has quit [Ping timeout: 244 seconds]
<HdkR> That's just a waste of time. Nvidia will just kill it
kinkinkijkin has joined #panfrost
<HdkR> Still questionable if they will care about pushing their desktop chips back down to commodity performance levels that cell phones require.
<HdkR> Theoretically they could now hit perf/w numbers even if they are more inefficient on average
<HdkR> (Ampere now hits something like ~400GFLOP/SM)
cowsay_ has joined #panfrost
cowsay has quit [Ping timeout: 240 seconds]
guillaume_g has joined #panfrost
yann has joined #panfrost
raster has joined #panfrost
<daniels> HdkR: why would NV kill Mali when it brings in a great deal of revenue and doesn't actually compete with Tegra?
<daniels> just lop off the top end so you can't put it in datacentre or automotive
stikonas has joined #panfrost
robmur01 has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
<robmur01> HdkR: you mean that same Nvidia that gave up on mobile years ago because they couldn't compete on power? :P
nlhowell has joined #panfrost
davidlt has joined #panfrost
robclark has quit [Ping timeout: 240 seconds]
<robmur01> I'm no expert, but I'd imagine that trying to replace the market leader with something unproven (and tainted by the whiff of previous reputation) because of NIH would be a gift to Arm's direct competitors
ezequielg has quit [Ping timeout: 240 seconds]
robclark has joined #panfrost
ezequielg has joined #panfrost
<robmur01> But obviously I can't comment on the future. Not least because I know f all anyway... first I heard was on the radio news this morning, since I lack the ability to read my work email while asleep :(
raster has joined #panfrost
icecream95 has quit [Ping timeout: 240 seconds]
nlhowell has quit [Ping timeout: 256 seconds]
<urjaman> meanwhile i did a news search after going "the f yall talking about nvidia and mali..." :P
nlhowell has joined #panfrost
nlhowell has quit [Ping timeout: 244 seconds]
dstzd has quit [Quit: ZNC - https://znc.in]
<alyssa> F.
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
dstzd has joined #panfrost
<daniels> robmur01: right, and it's not like MediaTek/Allwinner need another excuse to move over from Arm ISA to RISC-V
vstehle has joined #panfrost
<robmur01> let's just hope it gets stuck at the regulatory stage, everybody gets bored and gives up... :)
raster has quit [Quit: Gettin' stinky!]
raster has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
dstzd has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
buzzmarshall has joined #panfrost
jernej has joined #panfrost
dstzd has joined #panfrost
unoccupied has quit [Ping timeout: 240 seconds]
unoccupied has joined #panfrost
vstehle has quit [Ping timeout: 256 seconds]
rak-zero has quit [Ping timeout: 240 seconds]
vstehle has joined #panfrost
rak-zero has joined #panfrost
<bbrezillon> a301084b
<daniels> bbrezillon: cryptic!
<HdkR> daniels: I'm highly pessimistic about the whole deal. Especially from a company that said things like "Our future is data center" and "We don't deal in commodity"
<daniels> I completely agree, their business models are violently opposed, which is why it took me a long time to believe the rumours
<daniels> but now they've committed, walking away from those would be vandalism - they wouldn't just set fire to an enormous pile of cash by binning Mali and vacating the smartphone/STB/etc markets, and also Cortex-M is a pretty pure definition of commodity
<daniels> if all they wanted was vertically-integrated DC SKUs, why would they spend $40b on an architecture license?
<HdkR> I don't believe it'll be an entire binning, but considering most things in the Mali family already conflict with Nvidia internal things, It'll more likely be Nvidia slowly eroding Mali's top end in to Nvidia instead. Where efficiency is "close enough".
<HdkR> They already tried licensing GPUs a few years ago and nobody took them up on it
<HdkR> In a post-Ampere, going in to a Hopper world where 2SM designs can be <5w. Easy enough to eat in to the higher end SoC space
<daniels> totally agree on that, but given an 18-month regulatory approval timeline, and how long it takes to develop & tape out credible LP NV IP, that's quite some distance away
<daniels> but given that Mali's first high-end design win is the automotive Valhall parts, it's not really much of a change to the industry status quo :P
<HdkR> hah, I guess so
nlhowell has joined #panfrost
<HdkR> I guesss if ARM gets a bigger R&D and pushes out more fun designs like the Cortex-X1 then it can't be all bad
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
<alyssa> bbrezillon: i don't know what a301084b means either tbf
dstzd has quit [Quit: ZNC - https://znc.in]
jernej has joined #panfrost
dstzd has joined #panfrost
remexre has joined #panfrost
<daniels> HdkR: presumably ultra-high-end cores like X1/N1 would be the first to go in favour of whatever they're calling Denver this week?
<HdkR> Depends on if Denver can even be competitive. Xavier's Carmel has some terrible performance characteristics
<HdkR> Worst case it ends up being quite a bit worse than Cortex-A57
<bbrezillon> daniels, alyssa: oops, wrong window
guillaume_g has quit [Quit: Konversation terminated!]
nlhowell has quit [Ping timeout: 240 seconds]
_whitelogger has joined #panfrost
davidlt has quit [Ping timeout: 256 seconds]
nerdboy has joined #panfrost
icecream95 has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
<icecream95> It turns out that AFBC is a lot simpler than I thought it was: https://gitlab.freedesktop.org/icecream95/afbc/-/blob/master/format.md#data
<icecream95> Now for the hard part, writing an encoder and decoder that aren't extremely slow compared to tiling...
<alyssa> icecream95: sounds like you're having fun :-)
jernej has quit [Remote host closed the connection]
dstzd has quit [Quit: ZNC - https://znc.in]
dstzd has joined #panfrost
jernej has joined #panfrost
jernej has quit [Client Quit]
jernej has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
jernej has quit [Client Quit]
jernej has joined #panfrost
dstzd has joined #panfrost
dstzd has quit [Client Quit]
dstzd has joined #panfrost
dstzd has quit [Client Quit]
jernej has quit [Client Quit]
dstzd has joined #panfrost
jernej has joined #panfrost
tomboy64 has joined #panfrost
austriancoder has quit [Ping timeout: 272 seconds]
austriancoder has joined #panfrost
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
jernej has quit [Client Quit]
jernej has joined #panfrost
yann has quit [Ping timeout: 256 seconds]