channel logs for 2004 - 2010 are archived at http://tunes.org/~nef/logs/old/ ·· can't be searched
#osdev2 = #osdev @ Libera from 23may2021 to present
#osdev @ OPN/FreeNode from 3apr2001 to 23may2021
all other channels are on OPN/FreeNode from 2004 to present
http://bespin.org/~qz/search/?view=1&c=osdev2&y=21&m=10&d=18
00:20:00 <eryjus> sdfgsdfgs: the wiki has a good article and is always a good place to start: https://wiki.osdev.org/Memory_management
00:23:00 <zid> might wanna look up what an instruction set is too
00:23:00 <zid> because apparently you don't know
00:23:00 <gog> imstructiom set
00:24:00 <zid> blimstruction sebt
00:25:00 * gog dabs
00:26:00 <gog> apparently university of iceland has a similar CS curriculum as i do because one of my wife's friends posted a screenshot of their homework and it was this "bomb defuse" program where you had to do a different buffer overflow or stack smash to get to the next stage
00:27:00 <zid> cute
00:27:00 <gog> s/as i do/as my alma mater did/
00:27:00 <gog> they are cute
00:35:00 <Oli> I wonder if you convey alma to be the Spanish word for soul, and mater, as a comrade.
00:36:00 <gog> tovarisch
00:40:00 <Oli> Thank you for expanding me about, gog!
00:41:00 <gog> :|
00:41:00 * gog expands
05:05:00 <sdfgsdfgs> that's like.. thanks for stretching me out and filling me up
05:05:00 <sdfgsdfgs> (with information)
05:06:00 <sdfgsdfgs> oh well, 6 hours late for this joke but anyways
05:06:00 <sdfgsdfgs> thanks for feeling my soul, feeling me from the inside, deep inside
05:06:00 <sdfgsdfgs> inside the fountain of knowledge
06:21:00 <vin> Why does cpu (say x86) need to access in word size and not byte size?
06:24:00 <Oli> Data bus size is a word; registers that processor opcodes may use, are word-sized.
06:24:00 <Oli> By the way, hello, vin! I feel curious about: Do you also go by the named of VinsCool?
06:26:00 <vin> Oli: Making data bus size a word seems like an ancient decision. Shouldn't the cost of batch vs burst accesses be relooked with modern technology?
06:27:00 <kazinsal> Most actual accesses are done a cache line at a time for performance
06:27:00 <vin> Oli: no sorry I don't.
06:28:00 <kazinsal> Chances are if you're going to access something at 0x123400, you might be about to access something in the same cache line in the near future
06:29:00 <vin> Right kazinsal but it is weird to require structs to be algined to 4 bytes on x86 because the processor operates on words and not bytes
06:30:00 <vin> imagine all the wasted cycles and bytes that go into padding structs in an average application.
06:30:00 <kazinsal> If you're accessing something unaligned and it spills across one data bus word into another, you need to make two accesses
06:31:00 <Oli> I am yet unfamiliar with the nomenclatures of batch and burst accesses. Thank you vin, for bringing up about, and also replying to my inquiry!
06:31:00 <kazinsal> It's potentially slower to *not* pad structs, and wasting a few dozen bytes per struct is a non issue when you have dozens of gigs of memory on modern machines
06:34:00 <vin> A struct with 10B key and size_t val results in 24B because it needs to be 8B aligned, now I can only fit 2 structs in line unless I create a custom struct that pads the last 10B and allows me to store 3 keys and val.
06:34:00 <Mondenkind> vin: what do you mean, the processor operates on words and not bytes? The processor operates on units of all sizes, from a byte to a cache line
06:37:00 <vin> Mondenkind: Mondenkind I am talking about the min size a processor can operate on. At this point all access are 64B so word size is really redundant
06:38:00 <Mondenkind> if a four-byte access crosses a cache line, performance will be worse
06:38:00 <Mondenkind> aligning all four-byte quantities to a multiple of four bytes prevents this
07:05:00 <Affliction> x86 can do unaligned accesses, but they might take a performance hit for the reasons above.
07:06:00 <Affliction> Some architectures do not allow them, so they don't have to worry about the additional complexity to implement unaligned accesses
07:06:00 <Affliction> Which could be significant on a smaller microcontrollerish core, for example
07:11:00 <Affliction> at any rate, feel free to turn off alignment for that struct, if the wasted memory or reduced cache efficiency causes problems
10:01:00 <vai> :]]] https://www.osnews.com/story/1599/exposed-the-linux-developers-lifestyle/
18:11:00 <Bitweasil> Aw. :( Those new M1 Max and M1 Pro chips look good.
18:11:00 <Bitweasil> DOUBLE the DRAM bandwidth on the big one? Not like the M1 was a slacker in DRAM bandwidth...
18:11:00 <j`ey> 10core is cool, especially when it doubles the number of big cores
18:12:00 <Bitweasil> Yeah, I guess they got the data from the M1 that you can cut the efficiency cores back and still be fine.
18:12:00 <j`ey> and 32G+ ram
18:12:00 <Bitweasil> Yeah, 64 on the big one.
18:35:00 <jimbzy> Yeah, they are pretty awesome.
18:50:00 <zid> yea that was always going to be the problem scaling m1, memory bw.. just going sideways and adding more channels'll do it
18:52:00 <Bitweasil> Sure seems like it did!
18:52:00 <Bitweasil> I'll be interested to see the benchmarks.
18:52:00 * Bitweasil soldiers on with... idk, what's a Pi4 got, a couple gig a second?
18:52:00 <Bitweasil> *runs mbw*
18:52:00 <Bitweasil> Heh. If that.
18:52:00 <Bitweasil> AVG Method: MEMCPY Elapsed: 0.33662 MiB: 512.00000 Copy: 1520.993 MiB/s
18:53:00 <jason1234> hello
18:53:00 <Bitweasil> AVG Method: DUMB Elapsed: 0.28403 MiB: 512.00000 Copy: 1802.642 MiB/s
18:53:00 <Bitweasil> AVG Method: MCBLOCK Elapsed: 0.19544 MiB: 512.00000 Copy: 2619.695 MiB/s
18:53:00 <Bitweasil> Hello!
18:53:00 <jason1234> is it possibel to boot grub on my USB sdb from the /dev/sda (small size) ?
18:53:00 <jason1234> I have about 1000 (count) on size on mbr / sda available, after > 1000 bad blocks
18:54:00 <jason1234> currently I use lilo (very small).
18:54:00 <jason1234> grub wont fit, it is too bit
18:54:00 <jason1234> lilo is smaller.
18:54:00 <jason1234> but I would like a bit of like mikeos or alike to learn boot in assembly.
18:55:00 <Bitweasil> The ODroid N2+ is better, 3092/3025/5454 on those tests.
18:55:00 <Bitweasil> erm.
18:55:00 <jason1234> it is a amd64
18:56:00 <Bitweasil> If you have a bunch of bad blocks, just replace the USB stick.
18:56:00 <Bitweasil> But booting from USB should be doable.
18:59:00 <jason1234> it is bad blaock on hte harddisk
18:59:00 <jason1234> i screewed hte harddisk due to systemd
18:59:00 <jason1234> i got crazy to wait for systemd !
18:59:00 <jason1234> I press and hold 0/1
19:01:00 <Bitweasil> No idea what that means. Yes, you can put a bootloader on a USB stick and boot from the hard drive, if you want, but if you have bad sectors on the drive, replace the drive.
19:02:00 <jason1234> i want to boot the boot loaeder on sda to boot on the usb then after
19:02:00 <jason1234> is there a little code for that?
19:02:00 <jason1234> amd64/ibm
19:03:00 <Bitweasil> Sure, standard x86 bootloaders should do it.
19:03:00 <jason1234> the asm code <= 512, shall call the USB driver (asm), and then init hte boot on the USB pendreive (on pendrive that will be live netbsd "https://gitlab.com/openbsd98324/netbsd-live/-/raw/master/live/9.1/amd64/v5/netbsd-live-image-2000mb-v5.img.gz" with grub on it)
19:04:00 <jason1234> do we have a little very little "x86 bootloaders" asm code or sthg available?
19:05:00 <Bitweasil> I mean, I don't, personally.
19:07:00 <jason1234> there are likely many many, but we havent a database where to look for.
19:07:00 <jason1234> only mikeos, has a USB handlers into the code.
19:53:00 <jason1234> any ideas?
19:53:00 <jason1234> no #mikeos :(
20:00:00 <Bitweasil> Sorry, no ideas.
20:00:00 <Bitweasil> I don't think you need a USB handler for what you want to do, if the BIOS can present the USB drive as a hard drive.
20:02:00 <jason1234> no worries
20:02:00 <jason1234> thank you kanynhow