00:01
vagrantc has quit [Quit: leaving]
00:29
<
Abstract-Wizard >
just found out about the milk-v titan and im kind of excited
01:12
Abstract-Wizard has quit [Ping timeout: 248 seconds]
02:19
jn_ has joined #riscv
02:19
jn_ has quit [Changing host]
02:19
jn_ has joined #riscv
02:19
jn has quit [Ping timeout: 276 seconds]
02:36
_whitelogger has joined #riscv
03:18
_whitelogger has joined #riscv
03:34
<
shadows >
"Open Source in the Soul" from the vendor that did faff-all to upstream all products to date.
03:41
TMM has joined #riscv
04:06
BootLayer has quit [Quit: Leaving]
04:06
<
ganboing >
They are board integrator, not even SoC vendor, so don't put too much hope on the upstreaming effort.
05:10
r6fej has joined #riscv
05:14
lucmann has quit [Remote host closed the connection]
05:21
coldfeet has joined #riscv
05:22
dramforever[m]1 has joined #riscv
05:27
<
xypron >
Was there any discussion at DebConf25 whether the next Debian release (14, Forky) should move to RVA23?
05:29
<
dramforever[m]1 >
wasn't there, but i would be very surprised
05:30
<
dramforever[m]1 >
that would be unlike how debian has worked up til now at all
05:30
eremitah has quit [Remote host closed the connection]
05:40
agentcasey has joined #riscv
05:52
<
jrtc27 >
it's way too soon for it to make sense to move to RVA23
05:52
<
jrtc27 >
once there's actual hardware, sure, then have the conversation
05:53
<
jrtc27 >
but a baseline that nothing can run is a great way to get people to not use your distro
06:24
<
dramforever[m]1 >
it is not a hot take that nobody else is moving on to RVA23
06:24
<
dramforever[m]1 >
in a few years, possibly. definitely not anything recent
06:38
khem has quit [Quit: WeeChat 4.6.3]
06:43
khem has joined #riscv
06:43
lucmann has joined #riscv
06:55
r6fej has quit [Ping timeout: 260 seconds]
07:03
jobol has joined #riscv
07:40
<
ganboing >
drmpeg: sifive recently added their PMU driver for private L1/L2 and shared cache:
07:41
<
ganboing >
I did a perf using that with your code, and with the new prefetcher CSR values
07:41
<
ganboing >
The sifive_pl2_pmu/inner_prefetch_read (L1 cache prefetcher I guess), is only at 5% compared to the previous version
07:42
<
ganboing >
It looks like the prefetcher wasn't even working.
07:44
ganboing has quit [Quit: Leaving]
07:45
alperak has joined #riscv
07:45
ganboing has joined #riscv
07:53
<
drmpeg >
Have you looked at the STREAM benchmark performance in comparison?
07:54
<
ganboing >
Let me check. roughly, STREAM has a ~5x boost.
08:05
ldevulder has joined #riscv
08:08
<
ganboing >
With the new prefetcher config, sifive_pl2_pmu/inner_prefetch_read is ~9x with STREAM benchmark
08:10
<
drmpeg >
Woah. I wonder if there's some middle ground here.
08:10
<
ganboing >
I'm reading some Sifive cache docs, and trying to make sense of this. But likely Sifive/ESWIN needs to really come up with a somewhat workload agnostic setting
08:25
<
drmpeg >
That LDPC code has an interesting history. My DVB-T2 transmitter code was used as the basis for the 2017 Rohde and Schwarz engineering competition. The winning team came up with an optimization that made the LDPC lookup table more cache friendly.
08:39
pabs3 has quit [Read error: Connection reset by peer]
08:41
pabs3 has joined #riscv
08:41
coldfeet has quit [Ping timeout: 240 seconds]
08:58
prabhakalad has joined #riscv
09:02
<
ganboing >
Did some experiments and it seems STREAM performance can get on par with the new settings by just changing maxL1PFDist 0->2
09:02
<
ganboing >
It won't have very noticeable impact on your LDPC workload
09:27
ln5 has quit [Quit: .]
09:29
<
drmpeg >
Very nice.
09:48
ln5 has joined #riscv
09:48
hightower2 has joined #riscv
11:19
jjido has joined #riscv
11:36
Andre_Z has joined #riscv
11:56
hightower3 has joined #riscv
11:58
hightower2 has quit [Ping timeout: 276 seconds]
12:24
pabs3 has quit [Read error: Connection reset by peer]
12:24
coldfeet has joined #riscv
12:33
Andre_Z has quit [Quit: Leaving.]
12:45
coldfeet has quit [Quit: Lost terminal]
12:45
pabs3 has joined #riscv
12:51
tgamblin has joined #riscv
12:58
nexR has quit [Ping timeout: 276 seconds]
13:03
jjido has quit [Quit: My laptop has gone to sleep. ZZZzzz…]
13:04
pabs3 has quit [Quit: Don't rest until all the world is paved in moss and greenery.]
13:07
pabs3 has joined #riscv
13:09
Andre_Z has joined #riscv
13:31
jjido has joined #riscv
13:32
jjido has quit [Client Quit]
14:59
BootLayer has joined #riscv
15:26
jjido has joined #riscv
15:32
jjido has quit [Quit: My laptop has gone to sleep. ZZZzzz…]
15:57
jjido has joined #riscv
16:00
ldevulder has quit [Remote host closed the connection]
16:00
ldevulder has joined #riscv
16:02
cousteau has joined #riscv
16:03
jjido has quit [Quit: My laptop has gone to sleep. ZZZzzz…]
16:16
psydroid2 has joined #riscv
16:25
Trifton has quit [Ping timeout: 252 seconds]
16:28
Trifton has joined #riscv
16:33
Trifton has quit [Ping timeout: 252 seconds]
16:46
katzie has joined #riscv
16:52
ldevulder has quit [Ping timeout: 240 seconds]
17:19
prabhakalad has quit [Ping timeout: 240 seconds]
17:20
prabhakalad has joined #riscv
17:25
khem has quit [Quit: WeeChat 4.7.0]
17:26
khem has joined #riscv
17:47
jjido has joined #riscv
18:04
BootLayer has quit [Quit: Leaving]
18:05
TMM has joined #riscv
18:05
coldfeet has joined #riscv
18:08
cousteau has quit [Quit: Client closed]
18:32
haritz has joined #riscv
18:32
haritz has quit [Changing host]
18:32
haritz has joined #riscv
18:33
Trifton has joined #riscv
18:33
Trifton has quit [Remote host closed the connection]
18:33
tgamblin has quit [Ping timeout: 240 seconds]
18:34
tgamblin has joined #riscv
18:34
Trifton has joined #riscv
18:50
<
shadows >
if an SDIO card (AP6256 WiFi+BT module) needs power to be detected, from configuring a GPIO pin to output high), where in devicetree is the correct place for that?
18:52
<
shadows >
at the moment I have it as one of the pinctrl groups listed for the mmc1 node, but I wonder if it should be power sequenced
18:53
<
dramforever[m]1 >
shadows: do you have any specifics? right now i'm guessing mmc-pwrseq-simple
18:53
<
shadows >
or is power sequence not early enough for powering up an SDIO card?
18:54
<
shadows >
mmc-pwrseq-simple yes
18:54
<
dramforever[m]1 >
it seems weird that a gpio is for power, are you sure it isn't reset
18:55
<
dramforever[m]1 >
what board is this again?
18:56
<
shadows >
any of the JH7110 boards with AP6256 on SDIO, here it is Milk-V Mars CM Lite with WiFi option
18:58
<
shadows >
AP6256 is detailed on one of the last pages
19:00
<
dramforever[m]1 >
i see it, i think it should should just be mmc-pwrseq-simple
19:02
<
dramforever[m]1 >
it seems reasonable to treat this active-high "wifi_reg_on" as an active-low reset for power sequence purposes
19:02
<
shadows >
when I try instead from mmc-pwrseq-simple the mmc SDIO card is not detected
19:02
<
dramforever[m]1 >
what properties do you have?
19:03
<
dramforever[m]1 >
i'm thinking pinctrl-{names,0} for the pin, reset-gpios for the, well, gpio
19:05
<
dramforever[m]1 >
sdio_pwrseq from arch/arm64/boot/dts/rockchip/rk3308-bpi-p2-pro.dts looks good
19:09
<
shadows >
how would I add a delay to test if it just needs some time to power on?
19:10
<
dramforever[m]1 >
post-power-on-delay-ms possibly?
19:10
<
dramforever[m]1 >
Documentation/devicetree/bindings/mmc/mmc-pwrseq-simple.yaml
19:10
<
shadows >
btw am not near my dev workflow until +6hrs I will try then and share details
19:11
<
dramforever[m]1 >
sure, i literally do not have the hardware and just babbling :P
19:15
alperak has quit [Quit: Connection closed for inactivity]
19:26
sgerhold7 has quit [Quit: Ping timeout (120 seconds)]
19:26
sgerhold3 has joined #riscv
19:30
katzia has joined #riscv
19:31
katzie has quit [Read error: Connection reset by peer]
19:41
jjido has quit [Quit: My laptop has gone to sleep. ZZZzzz…]
20:13
Trifton has quit [Remote host closed the connection]
20:13
Trifton has joined #riscv
20:14
prabhakalad has quit [Quit: Konversation terminated!]
20:33
itrsea has joined #riscv
20:55
coldfeet has quit [Remote host closed the connection]
21:01
ZipCPU has joined #riscv
21:16
fossdd has quit [Ping timeout: 268 seconds]
21:21
indy has quit [Ping timeout: 276 seconds]
21:23
smaeul has quit [Read error: Connection reset by peer]
21:26
<
shadows >
(disregast line)
21:27
smaeul has joined #riscv
22:09
jobol has quit [Quit: Leaving]
22:09
indy has joined #riscv
22:45
guerby has quit [Ping timeout: 248 seconds]
22:45
guerby_ has joined #riscv
22:50
hightower3 has quit [Ping timeout: 248 seconds]
22:57
matrixbrain has joined #riscv
23:06
Andre_Z has quit [Quit: Leaving.]