-
sech1
-
hyc
morons...
-
hyc
so much for all their "BWT is ASIC-proof" nonsense
-
hyc
omg that conversation is so dumb.
-
sech1
-
sech1
They'll soon reinvent RandomX at this pace
-
hyc
weren't they the ones who already reimplemented randomx in go?
-
hyc
or thatwas someone else
-
hyc
-
hyc
still can't believe they claimed CFROUND was undocumented and mfrs could arbitrarily decide to break or delete it. as if IEEE754 spec isn't a thing.
-
hyc
total morons
-
hyc
their lack of random branch instructions is still a flaw
-
hyc
their use of AND will eat entropy
-
hyc
funny they're still calling this AstroBWT, even though its guts are completely different
-
sech1
"The 2-bit fprc register determines the rounding mode of all floating point operations according to Table 4.3.1. The four rounding modes are defined by the IEEE 754 standard."
-
sech1
"This is a hardware dependent implementation. This means, if Intel/AMD/ARM or others change/fix their implementation for whatever reason, all RandomX blockchains implementations will encounter issues"
-
sech1
well, if Intel/AMD/ARM decide to abandon IEEE 754, there must be a damn good reason :D
-
hyc
LOL
-
sech1
I don't know how old is the guy, but I bet IEEE 754 is older than him :D
-
sech1
-
nioc
-
nioc
332 blocks to go
-
hyc
664 minutes, 11 hours
-
tromp
sech1 i see some code fragment in the discussion, but where is the full code?
-
sech1
-
tromp
thanks
-
tromp
looks like they made something that's quite FPGA/ASIC friendly
-
tromp
what was the point of AstroBWT in the first place?
-
tromp
They write "Dero also built out a RandomX implementation in Golang that has been discarded for use"
-
hyc
they said randomx wasn't suitable because the behavior of CFROUND is architecture-dependent, so they discarded it
-
hyc
despite the fact that it's an intrinsic part of the IEEE754 spec
-
hyc
they literally have no f#cking idea what they're doing
-
hyc
-
hyc
-
hyc
it's not the one I had in mind but I think it hits the main points
-
tromp
yes, i saw that "doc" and it fails to explain why you'd want to put BWT in a PoW
-
tromp
i agree they don't know what they're doing. just like those that come up with the X11-17 nonsense
-
hyc
their basic claim is BWT has been around for 30 years and ASICs for it exist, but aren't more than 2x better than CPUs
-
hyc
I believe they're discovering that the motivation just didn't exist before, and their network is now being dominated by FPGAs
-
tromp
those statement seem contradictory.
-
tromp
ASICs should be over 2x better performance/watt than FPGAs
-
hyc
yeah, I can't speak to the accuracy of their assumptions
-
hyc
but it was clear from the start that they were headed for disappointment
-
tromp
oh, you mean the claim of ASICs <=2 *CPU is just bogus. ok, then