first impressions of the Lego smart brick, before I do any actual tearing down: wow, I forgot how good they are at working with plastic.

[gsuberland@chaos.social]

@ldcd @whitequark @ftg almost certainly a hardware log2 instruction bolted on. I wonder if they're using it for FFT or something.

[azonenberg@ioc.exchange]

@ftg @ldcd @whitequark @gsuberland i used to be big on mips. I remember being very disappointed when i discovered BFC00000 was not a valid license plate in my state

[ldcd@social.treehouse.systems]

@whitequark @gsuberland @ftg cursed thought: pentium based ECU

[whitequark@social.treehouse.systems]

complete success

[gsuberland@chaos.social]

@ldcd @whitequark @ftg "why did you crash your car?" "F00F bug, your honour"

[whitequark@social.treehouse.systems]

contrary to everything i've seen online, this is a Winbond W25Q16JWBY part (but one person got the closest, they thought it is a W25Q16JVBY. the difference is major: one is 3.3V, the other is 1.8V)

[whitequark@social.treehouse.systems]

@gsuberland @ldcd @ftg well yes they call it a log2() accelerator but why

[whitequark@social.treehouse.systems]

NotImplementedError: quad enablement SFDPJEDECQuadEnableRequirements.Reg2Bit1_Read35h_Write05h not implemented yet

ah yes. i remember why i hate SPI flashes now

[whitequark@social.treehouse.systems]

tried using dual mode and the SFDP tells me to transmit half a byte

once i dump it this flash is definitely going into the naughty pile (of unit tests in glasgow)

[wren6991@types.pl]

@whitequark @gsuberland @ldcd @ftg ...are you sure it's not just CLZ?

[gsuberland@chaos.social]

@whitequark @ldcd @ftg hmm, seems to be hardware log2p0 and log2p1 intrinsics. weird.

[whitequark@social.treehouse.systems]

@gsuberland @ldcd @ftg what is the p0/p1?

[gsuberland@chaos.social]

@whitequark @ldcd @ftg log2p0 is plain integer log base 2 (I think with always round down behaviour?) and log2p1 is the same but add 1 after. the latter is used to calculate how many bits are needed to store a given number. I know the latter gets used in some AV compression algorithms, idk about the decompression algorithms though.

[whitequark@social.treehouse.systems]

here's the flash contents https://upload.whitequark.org/1775953651-lego_brick_00F2MZ_749DF5_W25Q16JWBY.bin

[ignaloidas@not.acu.lt]

@wren6991@types.pl @whitequark@social.treehouse.systems @gsuberland@chaos.social @ldcd@social.treehouse.systems @ftg@mastodon.radio ceil(log2(n+1)) is clz, and is somewhat often defined as such so wouldn't surprise me (I doubt it would actually have floating point log)

[whitequark@social.treehouse.systems]

@ignaloidas @gsuberland @ldcd @ftg @wren6991 i have literally never seen clz implemented that way

[ignaloidas@not.acu.lt]

@whitequark@social.treehouse.systems @gsuberland@chaos.social @ldcd@social.treehouse.systems @ftg@mastodon.radio @wren6991@types.pl jpeg xl uses the log2 way in it's spec (it took me a bit of thinking to reverse it)

(wait, now that I'm re-thinking, that's not clz but bitsize minus clz but basically same)

[wren6991@types.pl]

@ignaloidas @gsuberland @ldcd @ftg @whitequark (implements XOR instruction) Galois Field Addition Accelerator

[nicolas17@social.treehouse.systems]

@whitequark hey @zhuowei does this match anything you found in the LEGO Smart Brick app? Or does the app have a newer firmware version?

[maehw@chaos.social]

@nicolas17 @whitequark @zhuowei for everyone else who got invested now, I think you're referring to this thread: https://notnow.dev/objects/cccfc047-da57-444a-be6f-63a5a766bcf7