markhahn

41 ms·

571.
by markhahn 11y ago
no content in your complaint, just your various dislikes.
572.
by markhahn 11y ago
Often they model how the warhead ages, rather than how it explodes.
573.
by markhahn 11y ago
not 20 real (64b) gflops/watt. and no, not much power is used on interconnect, and certainly not a PUE of 2.0.
574.
by markhahn 11y ago
Flash has had a drop in price because it gained mainstream utility, and therefore serious volume production. It's entirely unclear whether this will continue. The physics of flash cells is not all that promising: there's certainl
575.
by markhahn 11y ago
do geeks watch WWDC? appleheads, sure, but not real geeks...
576.
by markhahn 11y ago
ls's plethora of flags doesn't stop "echo *" from working.
577.
by markhahn 11y ago
jeez. worrying about books is pretty much antipodal to DH as I see it practiced...
578.
by markhahn 11y ago
Wake me up when there's a Docker-independent hypervisor engine.
579.
by markhahn 11y ago
because small, slow computers are easy to make power-efficient. the truth is that the processor isn't that interesting, if you're taking a lots-of-wimpy-nodes approach (like BGQ).
580.
by markhahn 11y ago
and they do a good job on products - well worth prefering their stuff...
581.
by markhahn 12y ago
a lesson for the author xor the reader.
582.
by markhahn 12y ago
"horrid for interactive" is ironic, since interactive and HPC are pretty much disjoint. (well, viz...) From an HPC perspective, Fortran IO should be performed by HDF5...
583.
by markhahn 12y ago
GPUs have a large advantage in a very narrow niche: regular and very intensive ops on regular and compact data. ML is not completely ideal (because it's not that compute-intensive), but MC simulations often are. Most things are not
584.
by markhahn 12y ago
blades were never more than a marketing trick: the offer nothing that can't be achieved in a standard chassis. there were a few multi-chassis SMP/NUMA machines that had cache coherency over external interfaces, but that was neith
585.
by markhahn 12y ago
RDMA doesn't really provide a flat memory model - all it's really doing is minimizing copies when you send a message. more like "put this 100K string into that node at <address>".
586.
by markhahn 12y ago
that's silly: HPC has been pinching pennies before big data was a thing. and the computer industry is biz: you get what you pay for. if you can live with Gb performance, you can drop around $2k (IB card, cables, switches) off your pr
587.
by markhahn 12y ago
the point is to avoid highly processed/prepared foods.
588.
by markhahn 12y ago
the basic premise is that remapping is necessarily expensive. unfortunately, neither the thesis nor its citations really support that. and we have plenty of examples where remapping is pervasive and apparently efficient (MMUs, for instanc
589.
by markhahn 12y ago
Fair enough. But this guy is talking about his takehome, not the fully-encumbered price for his position.
590.
by markhahn 12y ago
No, I don't believe incoming grad students would be told an outlandish lie like that. People in academia are extremely aware of how low the rate of "job creation" is. Profs are quite honest with grad students, since the sup
591.
by markhahn 12y ago
I'm salaried tech staff at a major .ca univ, and have done sessional teaching. I think the pay rate is approximately appropriate. yes, it's not great for a career, but neither is flipping burgers. sure, it's not the same t
592.
by markhahn 12y ago
$70k janitors, what a hoot! yeah, not so much.
593.
by markhahn 12y ago
cheesy old-left blather. it seems to offend some people that value is defined by whatever the market will bear, rather than something inherent.
594.
by markhahn 12y ago
Reading stuff from ESR considered a waste of time.
595.
by markhahn 12y ago
1. it's healthier, cheaper, fun and usually tastier. 2. no plan, unless "to use up the X" counts. 3. food. real food begins with, and consists mostly of vegetables. meat, cheese, eggs, grains, seeds, etc are great, but are
596.
by markhahn 12y ago
Perfectly clean of any Linux taint - is this a message?
597.
by markhahn 12y ago
wow, wrong on almost all the settings. seriously: maximize bufferbloat, cause huge intrusive IO pauses, keep useless pages in memory, etc? the only good ones there are kernel.panic = 300 kernel.sysrq = 1
598.
by markhahn 12y ago
because animations add so much value to communication via, eg, powerpoint...
599.
by markhahn 13y ago
192 kepler cores: don't bother. this is just a desperate attempt to get someone to take TK1 chips off their hands. coincidentally, I'm sure, but AMD is shipping their AM1 platform.
600.
by markhahn 13y ago
Stallman is wrong that proprietary software hurts users; that's the essence of his critique. It's sad that so much software has sailed under the GPL flag of convenience, simply because it existed and could avoid license bikeshedd

More ›