azonenberg_work has quit [Ping timeout: 240 seconds]
genii has quit [Quit: Welcome home, Mitch]
tms_ has quit [Ping timeout: 244 seconds]
freemint has quit [Ping timeout: 245 seconds]
indefini[m] has quit [Read error: Connection reset by peer]
scream has quit [Read error: Connection reset by peer]
swedishhat[m] has quit [Read error: Connection reset by peer]
henriknj has quit [Read error: Connection reset by peer]
thehurley3[m] has quit [Read error: Connection reset by peer]
xobs has quit [Read error: Connection reset by peer]
jfng has quit [Remote host closed the connection]
nrossi has quit [Read error: Connection reset by peer]
nrossi has joined ##openfpga
Bike has quit [Quit: Lost terminal]
nrossi has quit [*.net *.split]
flaviusb has quit [*.net *.split]
kmehall has quit [*.net *.split]
_whitenotifier has quit [*.net *.split]
pinoaffe has quit [*.net *.split]
jhol has quit [*.net *.split]
pointfree has quit [*.net *.split]
awygle has quit [*.net *.split]
sensille has quit [*.net *.split]
nrossi has joined ##openfpga
Jybz has joined ##openfpga
azonenberg_work has joined ##openfpga
indefini[m] has joined ##openfpga
jfng has joined ##openfpga
thehurley3[m] has joined ##openfpga
xobs has joined ##openfpga
henriknj has joined ##openfpga
scream has joined ##openfpga
swedishhat[m] has joined ##openfpga
flaviusb has joined ##openfpga
sensille has joined ##openfpga
jhol has joined ##openfpga
awygle has joined ##openfpga
pointfree has joined ##openfpga
kmehall has joined ##openfpga
_whitenotifier has joined ##openfpga
pinoaffe has joined ##openfpga
azonenberg has joined ##openfpga
tms_ has joined ##openfpga
bwidawsk has quit [Quit: Always remember, and never forget; I'll be back.]
emeb has quit [Quit: Leaving.]
bwidawsk has joined ##openfpga
rektide has quit [Remote host closed the connection]
Jybz has quit [Quit: Konversation terminated!]
emeb_mac has quit [Ping timeout: 240 seconds]
tlwoerner has quit [Ping timeout: 268 seconds]
tlwoerner has joined ##openfpga
rohitksingh has quit [Ping timeout: 245 seconds]
OmniMancer has joined ##openfpga
ym has joined ##openfpga
Zorix has quit [Quit: Leaving]
tlwoerner has quit [Ping timeout: 265 seconds]
genii has joined ##openfpga
tlwoerner has joined ##openfpga
carl0s has joined ##openfpga
freemint has joined ##openfpga
pie_ has quit [Ping timeout: 265 seconds]
emeb has joined ##openfpga
Asu has joined ##openfpga
OmniMancer has quit [Quit: Leaving.]
rohitksingh has joined ##openfpga
mumptai has joined ##openfpga
rohitksingh has quit [Ping timeout: 264 seconds]
rohitksingh has joined ##openfpga
rohitksingh has quit [Ping timeout: 245 seconds]
rohitksingh has joined ##openfpga
<ZirconiumX>
For all the faults Quartus has, it *does* produce some interesting statistics
<ZirconiumX>
My homebrew Cyclone V synthesis flow is ~1.5x worse than Quartus
<ZirconiumX>
Which isn't too bad for something I cooked up in a week
<mwk>
in what manner? area? timing?
craigjb has quit [Ping timeout: 245 seconds]
dh73 has joined ##openfpga
<ZirconiumX>
Cells at the moment. I can try area/timing, but it will take a little while
<ZirconiumX>
Because I'll have to rename all the primitives
Jybz has joined ##openfpga
<ZirconiumX>
"Whenever the processor discovers a mispredicted branch, it aborts all speculatively-executed instructions and restores the processor's state to the state it held before the branch. However, the cache state is not restored (see the section titled "Side Effects of Speculative Execution")"
<ZirconiumX>
Does this mean the MIPS R10K from '96 is vulnerable to Heartbleed?
<ZirconiumX>
....
<ZirconiumX>
That's *not* the vulnerability I was thinking of.
<ZirconiumX>
Meltdown. That one.
<sorear>
spectre, and yes
<sorear>
people blame intel but they were copying a standard design
<sorear>
all of the early ooo chips (ppro, 21264, r10k, pa-8000) function similarly here
<tnt>
Heh ... only if that speculative execution can depend on private data.
<tnt>
No idea if that's the case on a MIPS R10k
<sorear>
it can. dunno if anyone’s demonstrated it
<ZirconiumX>
I have a lot of weird manuals
<sorear>
also there’s working spectre attack code for BOOM, because apparently when you’re an academic “this will obviously work and nothing will be learned from doing it” doesn’t stop people
<ZirconiumX>
Are there any plans for BOOM to mitigate it, or just leave it to software?
<ZirconiumX>
Because hardware mitigations for Spectre *would* give something people to learn from
<sorear>
there are plans but I don’t think they really grapple with the problem scope
<ZirconiumX>
mwk: my homebrew flow does better than synth_intel at least
<ZirconiumX>
Not that anybody's particularly sentimental about that
<ZirconiumX>
~~because synth_intel breaks while compiling my testbench~~
* mwk
read that as boom, the improved doom engine, and wondered how the hell you invoke spectre from within a FPS game
<davidc__>
I'm not really sure what market segment they are targetting. They don't appear to be as featureful as iCE40, and pricing I could find doesn't seem to be as good
<gruetzkopf>
tnt: want a mipsr10k machine to test that on?
<gruetzkopf>
(ssh shell, at least)
<gruetzkopf>
(or was it r12k?)
rohitksingh has quit [Ping timeout: 245 seconds]
<tnt>
gruetzkopf: nope, tx :) I'm busy enough as it is :p
carl0s has quit [Remote host closed the connection]
freemint has quit [Remote host closed the connection]
freemint has joined ##openfpga
Shiz has quit [Ping timeout: 258 seconds]
rohitksingh has joined ##openfpga
<somlo>
speaking of spectre, what do people generally mean by "mitigated by software"? I mean, beyond just somehow avoiding speculative execution for "sensitive" parts of (generally kernel/os) code, trading in the extra speed/performance that speculation would have brought to those code sections?
freemint has quit [Ping timeout: 245 seconds]
<mwk>
more or less just that
<somlo>
also, what would be a sensible "hardware mitigation"? OK, checking permissions *before* speculations in Meltdown's special case, but the general Spectre problem...
<somlo>
un-touching tha cache would be crazy; perhaps avoiding the cache altogether during speculative execution?
<mwk>
well, that's a hard question
<emily>
there's stuff like retpoline
<emily>
which makes all branches in the kernel too indirect to be speculated iirc
<somlo>
emily: but wouldn't that fall under "software mitigation"?
<emily>
yes, sorry, that was an example of a software mitigation
<somlo>
since the problem is "cache side effects can be measured later by unprivileged code"
<somlo>
we'd want some way of speculating without leaving side effects in the cache; I can think of either going crazy trying to "un-touch" the cache, which I think is bat-shit insane, and my head hurts trying to think about it
<somlo>
or avoiding touching the cache to begin with
<somlo>
maybe some "safe speculation" mode, where aforementioned "sensitive" sections could still speculate, but not touch the cache
<somlo>
now, would that be worth having, performance wise? would it even be better than just losing speculation speed-up during those sensitive sections?
<sorear>
somlo: “un-touch the cache” is what the boom people are doing, it’s blatantly inadequate due to multi-core attacks but they don’t understand this
<sorear>
speculation in core 1 can be observed by core 2 measuring bus contention
<sorear>
if you have _any_ underprovisioned resources, you will have side channels
<somlo>
ok, so then it's either "just don't do it" (i.e., software mitigation), or something like "speculate without caching", which may be stupid but I can't tell right now :)
<somlo>
unless I'm missing some other hardware mitigation option :)
freemint has joined ##openfpga
rohitksingh has quit [Ping timeout: 258 seconds]
Maya-sama has joined ##openfpga
hackkitten has quit [Ping timeout: 276 seconds]
<sorear>
somlo: the “software mitigations” are all making assumptions about undocumented pipeline bits and might work on single-vendor ISAs but not here
<sorear>
fine-grained multitasking, confidentiality, high single thread performance: pick 2
<sorear>
I think the future is to have schedulers that are aware of confidentiality domains, and then don’t schedule threads from different domains simultaneously or sequentially without a machine clear and cache flush, on perf cores
<sorear>
tangentially I’m wondering how long it will take for the security community to notice Falkor value prediction
Maya-sama is now known as hackkitten
rohitksingh has joined ##openfpga
<emily>
Falkor value prediction?
<sorear>
you know how caches leak all memory addresses? value predictors are like that but for data loaded from memory
<sorear>
jonmasters has been heavily implying that Falkor has one
<sorear>
my interpretation of public tweets
<emily>
ah *nod*
dh73 has quit [Remote host closed the connection]
dh73 has joined ##openfpga
rohitksingh has quit [Ping timeout: 276 seconds]
rohitksingh has joined ##openfpga
rohitksingh has quit [Ping timeout: 245 seconds]
genii has quit [Quit: Welcome home, Mitch]
Asu has quit [Ping timeout: 268 seconds]
emeb_mac has joined ##openfpga
Asu` has joined ##openfpga
Asu` has quit [Read error: Connection reset by peer]
Asu` has joined ##openfpga
freeemint has joined ##openfpga
freemint has quit [Read error: Connection reset by peer]