aseipp

34 ms·

391.
by aseipp 3y ago
I also have a bridge I can sell you. And trust me: you've never seen one like this, at rates like I'm offering! Buy now!
392.
by aseipp 3y ago
Everyone who uses NVMe-over-network-transport simply does redundancy at the client layer. The networking gear is very robust, and it is easier to optimize the "data plane" path this way (map storage queues <-> network queues
393.
by aseipp 3y ago
Yeah! Did/do you work on it? The original publications were good timing; I was working as a consultant on an FPGA-based ML accelerator at the time the original stuff was talked about, and I really enjoyed reading everything I could abo
394.
by aseipp 3y ago
Microsoft absolutely runs their own models on their own hardware, at scale, and they have done so for years just like every other hyperscaler -- Project Brainwave was first publicly talked about as far back as 2018. The generative LLM craze
395.
by aseipp 3y ago
Hacker News puts `noreferrer` on the link (because apparently Internet Points and Arguments are more important than the courtesy of respecting the wishes of the author.)
396.
by aseipp 3y ago
No, it's not enough. One of the most important features missing is that kernel developers will want the equivalent of `git range-diff` (i.e. the ability to compare two versions of a single commit which was submitted as part of a batch
397.
by aseipp 3y ago
Facebook uses a heavily modified fork of Mercurial which they have (as of last year) released publicly, and now refer to it by a new name, "Sapling" (though the command may still be called 'hg' internally). And Sapling a
398.
by aseipp 3y ago
Racket today and for several years now has already been using Chez Scheme, so any of the latest versions you download will be built that way. On all platforms. So, just download away, and you're good to go. Rather, this announcement is
399.
by aseipp 3y ago
They even switched to mandatory yearly renewals a while back when Libero SoC launched. Buy a copy of Libero Gold for $999? You lose access after 365 days unless you renew. Every license tier used to be perpetual upon purchase, but now a sin
400.
by aseipp 3y ago
My rule of thumb is that I don't listen to anyone who can't write exploits for modern software stacks (OS kernels, browsers.) And even then, you can still ignore 50% of those people too, as they are cranks. In contrast, most of th
401.
by aseipp 3y ago
That's a lot of words to say "I don't know."
402.
by aseipp 3y ago
Using a TPM for remote attestation of device state is basically worthless in the generic PC UEFI world because there is pretty much unlimited surface area for holes to leverage, from motherboards that falsely report secure boot states to fi
403.
by aseipp 3y ago
From a pragmatic viewpoint of saving some space, it might make sense. But from a design viewpoint it's not really a "minimal necessary feature" to do something you can't do already, so specifying it as part of the standa
404.
by aseipp 3y ago
Diffs are purely a UI mechanic of Git. Computed on demand, but they don't really exist at the data layer, where blobs are stored. That's the real spot where all the problems are. Binaries can't realistically be packed and com
405.
by aseipp 3y ago
It's a reference to a famous Twitter post: https://twitter.com/dril/status/107911000199671808
406.
by aseipp 3y ago
The kernel will work fine, but at minimum EL2 runs the Qualcomm Hypervisor (Gunyah) which prevents native KVM virtualization from taking place. This is true of all Snapdragon platforms. Windows supports virtualization on the 8 Gen 3 only be
407.
by aseipp 3y ago
There's nothing risky about developing ARM cores. Well, other than the fact that doing it wrong is very costly. Millions of them ship every year, and there have been many third-party designs, shipped in millions of units, by many diffe
408.
by aseipp 3y ago
> How does this manage to scale? "Okay" to "very poorly", depending on your view. And for the record, many times the drivers go untested, frankly. I've absolutely found completely broken drivers in the kernel (as
409.
by aseipp 3y ago
Windows does actually have a universal C runtime library now, but it's pretty recent; only Windows 10+ https://learn.microsoft.com/en-us/cpp/porting/upgrade-your-c... A big motivation for it was security
410.
by aseipp 3y ago
If by "outgrow" you mean you want to improve concurrent read/write performance, availability, and throughput via replication, then I would -- personally, for peace of mind and ease -- look at just buying a managed RDBMS from
411.
by aseipp 3y ago
You can spec out the prices here, which are the same boxes used in this article. The entry price is about $2k USD for the 32-core SKU: https://www.ipi.wiki/products/com-hpc-ampere-altra# The Dev Kit is probably what mo
412.
by aseipp 3y ago
To be fair the Neoverse N-series was designed for datacenter style usage where lower clock frequencies and going wide is more common even on x86. For client uses some of the things the Apple M-series does for single core performance may or
413.
by aseipp 3y ago
On that note, Wasmtime has an option to enable NaN canonicalization, which should patch that hole, too: https://docs.rs/wasmtime/latest/wasmtime/struct.Config.html#...
414.
by aseipp 3y ago
I've been using it for years and pretty happy, but: have you tried just replacing the modem and router with your own kit? Last I remember, IPv6 is slightly fiddly, and you need some VLAN tagging set up, but you can just use any DOCSIS
415.
by aseipp 3y ago
MPTCP has been in use at massive scale on iOS for years now for core Apple services and any iOS application that uses Network.framework can leverage it too.
416.
by aseipp 3y ago
Weekly? Now, this is nowhere near SF levels, but I live in South Austin, and I probably see at least 1-2 autonomous cars every single day I go driving in town.
417.
by aseipp 3y ago
Good points; though I admit, I was thinking more Linux-capable/"Linux-class" cores in the mid-range FPGAs which is where I spend a bunch of my time. For that, the paltry cache sizes due to limited amounts of BRAM are more not
418.
by aseipp 3y ago
Nope, I was wrong! I was curious about a reference and went digging, and I was misremembering the details of a particular Linux-class system that didn't implement C (Shakti), so they needed an entirely new set of binary packages from
419.
by aseipp 3y ago
I think an example is something like opcodes crossing I-cache lines, re: fetch and decode complication; instructions are 16-bit aligned when C is present, so you can have a 32-byte instruction cross cache lines easily. At minimum it will de
420.
by aseipp 3y ago
Well, the limit is much higher now, I can tell you that, but that's only because I just never missed a payment for a few years. They expanded my limits pretty quickly. I mostly didn't use credit cards before that when I made less

More ›