Once you waste time trying to parse the timestamp output of the %(#&'ing boost C++ libraries, you'll know the pain. Why can't they at least output in one of the dozen formats that/bin/time understands?
For continuous speech, the best continuous voice recognition programs still have around 2-5% word error rates with *good* sound quality. Imagine the error rate with the lousy sound quality (and background noise) from most cell phones. Then consider a speech synthesis program where every 50th word is completely wrong.
I'm considering CMU Sphinx 3/4 as an example of state-of-the-art voice recognition application. It barely runs in real-time on a P4.
I'm also suspicious. The distinction between many sounds is the placement or movement of the tongue. For instance, I can whisper and be understandable without moving my vocal cords. They describe this device as something that "detects" speech by observing the vocal cords, not the tongue. How does it work?
Also, it sounds like the speech is recognized and converted into words in this system (as in Sphyinx or commercial voice recognition software?). The accuracy of even the best voice recognition software is still too poor to be used in general applications (and requires a fast P4 to do the recognition in real-time). It'll be a while before any cell phones carry this.
Re:mnb Re:Perhaps there is a reason...
on
DVHS on a Budget
·
· Score: 1
This is actually correct. The parent specifically said that CD _recorders_ look for this metadata. If you buy a consumer audio CD burner (i.e., not your PC's CD burner), they will generally only accept the "digital audio" or similar marking CD-Rs. These CDs will play just fine in any CD player or CD-ROM drive.
It makes a lot more sense to have separate clock domains for each core. You can take two cores that have already been designed (with their clock trees) and throw them on the die. Why duplicate work?
There are already many clock domains in single-core processors (external bus, sometimes L2/L3 cache, etc..), so this isn't a hard problem.
This increases the fabrication costs for the silicon die because the processes used to create high-performance CMOS logic and high-density DRAM are different. Because of the cost, it's not likely to happen for commodity microprocessors any time soon.
We have already seen that the investment bankers and analysts can't do basic math (c.f., the late '90s) . Take away their calculators and they might have to start thinking. Bring it on!
I have mod points, but I'd rather add to the dicussion. Condor does an excellent job of scheduling tasks with the resource constraints that you mentioned and it works across machines. I'd highly suggest looking into it. While it doesn't have a periodic submission feature (AFAIK, I wouldn't use that, even if it did), I don't see why you couldn't use cron to submit an initial condor job that starts everything else off every day.
The dag scripts handle dependencies (setting these up the first time might be a big hairy). Setting resource requirements is easy. Scheduling comes for free.
Slashdot Mirror
How about those URLs that span 3 lines? Linebreaks don't paste so well.
You realize that only people who have HTML-capable email clients will bother trying to click those two URLs, right?
It's a PITA to find, copy, and paste a long URL.
This ain't C. You don't have to end every statement with a semicolon.
Once you waste time trying to parse the timestamp output of the %(#&'ing boost C++ libraries, you'll know the pain. Why can't they at least output in one of the dozen formats that /bin/time understands?
Pretty good for a 50 Cent OS.
To push your analogy a bit further, a waveguide is the RF version of a ditch. A length of coax would also probably do.
Pffft. Had they run been running NTP, they would have had the time.
For continuous speech, the best continuous voice recognition programs still have around 2-5% word error rates with *good* sound quality. Imagine the error rate with the lousy sound quality (and background noise) from most cell phones. Then consider a speech synthesis program where every 50th word is completely wrong.
I'm considering CMU Sphinx 3/4 as an example of state-of-the-art voice recognition application. It barely runs in real-time on a P4.
I'm also suspicious. The distinction between many sounds is the placement or movement of the tongue. For instance, I can whisper and be understandable without moving my vocal cords. They describe this device as something that "detects" speech by observing the vocal cords, not the tongue. How does it work?
Also, it sounds like the speech is recognized and converted into words in this system (as in Sphyinx or commercial voice recognition software?). The accuracy of even the best voice recognition software is still too poor to be used in general applications (and requires a fast P4 to do the recognition in real-time). It'll be a while before any cell phones carry this.
Light from the sun is SO 8.3 minutes ago.
This is actually correct. The parent specifically said that CD _recorders_ look for this metadata. If you buy a consumer audio CD burner (i.e., not your PC's CD burner), they will generally only accept the "digital audio" or similar marking CD-Rs. These CDs will play just fine in any CD player or CD-ROM drive.
It makes a lot more sense to have separate clock domains for each core. You can take two cores that have already been designed (with their clock trees) and throw them on the die. Why duplicate work?
There are already many clock domains in single-core processors (external bus, sometimes L2/L3 cache, etc..), so this isn't a hard problem.
Puns aside, don't forget that the Mars Rover had serious space issues on their 32MB flash volume that caused repeated reboots.
This increases the fabrication costs for the silicon die because the processes used to create high-performance CMOS logic and high-density DRAM are different. Because of the cost, it's not likely to happen for commodity microprocessors any time soon.
>And let's all speculate aimlessly until we know which.
If we do this, then we are no better than the interceptor missiles, which also speculate and fly aimlessly.
When I was an IRCOp, .NO meant NO.
Intel did, in fact, spend a fair amount of chip real estate in the Itanium on executing x86 code. It works. It's just not very fast.
Now we can buy from PalmTwo?
You must not be new here.
You seem to know a lot about assurance companies. Tell me again, how do these assurance companies help you when you have a tree fall on your house?
What do you get for your money? Assurance that everything will be alright?
Looks like someone else found a way to make some money. Amazing, huh?
We have already seen that the investment bankers and analysts can't do basic math (c.f., the late '90s) . Take away their calculators and they might have to start thinking. Bring it on!
If you're learning from so-n-so's blog, you've got bigger problems than you think.
It's ultraviolet. You're not going to be able to see it. Trust me.
I have mod points, but I'd rather add to the dicussion. Condor does an excellent job of scheduling tasks with the resource constraints that you mentioned and it works across machines. I'd highly suggest looking into it. While it doesn't have a periodic submission feature (AFAIK, I wouldn't use that, even if it did), I don't see why you couldn't use cron to submit an initial condor job that starts everything else off every day.
The dag scripts handle dependencies (setting these up the first time might be a big hairy). Setting resource requirements is easy. Scheduling comes for free.