I gotta say....I don't know what I'm doing wrong with 5 of my hosts. They all have the same version of Linux x64 (Mandriva 2007 spring). A recent edition of boinc 5.10.x. I swear I've downloaded the 4.49 files, unzipped them into a folder called 4.49somthing, deleted the one with the _1 in the new folder. Downloaded the SSE2 app, uzipped it, renamed it by adding the _1 to the name. Copied the newly renamed app into the unzipped 4.49 folder. Then exited boinc, opened console, typed "killall -9 boinc", waited for it to terminate (checked with ksysguard). Then copied the contents of the 4.49 folder into the einstein directory. picked the "overwrite all" box when it asked about the app_info.xml file. When I restart boinc it gives all work a computation error. Then the transfer tab shows it trying unsuccessfully to download 4.49.
My two quads are working fine, but my three duals and two single cores are yeilding the errors.
any ideas?
tony
oh and yes, every computer I've got running is atleast SSE2 capable.
AMD64 2800
AMD64 3700 sandiego
AMD64 X2 4800 dual
AMD64 X2 5200 dual
AMD64 X2 6000 dual
AMD 9600 black edition quad
Intel Q6600 quad
First, the SSE2 executable file doesn't have executable permissions set. Use the "chmod a+x" command for that file to set them.
Second, the file is misnamed in two ways. Not only is the "_1" at the end missing, but it also has an extra "_1" in the middle of the filename.
Quote:
I gotta say....I don't know what I'm doing wrong with 5 of my hosts. They all have the same version of Linux x64 (Mandriva 2007 spring). A recent edition of boinc 5.10.x. I swear I've downloaded the 4.49 files, unzipped them into a folder called 4.49somthing, deleted the one with the _1 in the new folder. Downloaded the SSE2 app, uzipped it, renamed it by adding the _1 to the name. Copied the newly renamed app into the unzipped 4.49 folder. Then exited boinc, opened console, typed "killall -9 boinc", waited for it to terminate (checked with ksysguard). Then copied the contents of the 4.49 folder into the einstein directory. picked the "overwrite all" box when it asked about the app_info.xml file. When I restart boinc it gives all work a computation error. Then the transfer tab shows it trying unsuccessfully to download 4.49.
My two quads are working fine, but my three duals and two single cores are yeilding the errors.
any ideas?
tony
oh and yes, every computer I've got running is atleast SSE2 capable.
AMD64 2800
AMD64 3700 sandiego
AMD64 X2 4800 dual
AMD64 X2 5200 dual
AMD64 X2 6000 dual
AMD 9600 black edition quad
Intel Q6600 quad
Since I've switched my X9650 Extreme Quad-core machine over to the SSE2 app, I've been getting crunch times as short as just under three hours. It's also done some good on my AMD 6000+ machine, though probably to not as great an extent.
Just a remark: Intel Core2 has 128-bit wide SSE engine, but AMD 6000+ not.
So the AMD 6000+ offers the same speed on SSE2 and FPU instructions.
Core2 exectues SSE2 two times faster than FPU code.
Since I've switched my X9650 Extreme Quad-core machine over to the SSE2 app, I've been getting crunch times as short as just under three hours. It's also done some good on my AMD 6000+ machine, though probably to not as great an extent.
Just a remark: Intel Core2 has 128-bit wide SSE engine, but AMD 6000+ not.
So the AMD 6000+ offers the same speed on SSE2 and FPU instructions.
Core2 exectues SSE2 two times faster than FPU code.
But will the Intel 128-bit width benefit scalar SSE2 instructions as well?
The SSE2 instructions in the app should all be scalar (I don't think the gcc will auto-vectorize anything?) so a 128 bit data path should not be that much an advantage, right?
As far as I could see the main benefit from SSE2 in this app is caused by a better double to int conversion (still kind of slow with the old fashioned FPU control word setting method even on Core 2). Maybe it's not that big a difference on AMD K8?
But will the Intel 128-bit width benefit scalar SSE2 instructions as well?
The SSE2 instructions in the app should all be scalar (I don't think the gcc will auto-vectorize anything?) so a 128 bit data path should not be that much an advantage, right?
As far as I could see the main benefit from SSE2 in this app is caused by a better double to int conversion (still kind of slow with the old fashioned FPU control word setting method even on Core 2). Maybe it's not that big a difference on AMD K8?
I gotta say....I don't know
)
I gotta say....I don't know what I'm doing wrong with 5 of my hosts. They all have the same version of Linux x64 (Mandriva 2007 spring). A recent edition of boinc 5.10.x. I swear I've downloaded the 4.49 files, unzipped them into a folder called 4.49somthing, deleted the one with the _1 in the new folder. Downloaded the SSE2 app, uzipped it, renamed it by adding the _1 to the name. Copied the newly renamed app into the unzipped 4.49 folder. Then exited boinc, opened console, typed "killall -9 boinc", waited for it to terminate (checked with ksysguard). Then copied the contents of the 4.49 folder into the einstein directory. picked the "overwrite all" box when it asked about the app_info.xml file. When I restart boinc it gives all work a computation error. Then the transfer tab shows it trying unsuccessfully to download 4.49.
My two quads are working fine, but my three duals and two single cores are yeilding the errors.
any ideas?
tony
oh and yes, every computer I've got running is atleast SSE2 capable.
AMD64 2800
AMD64 3700 sandiego
AMD64 X2 4800 dual
AMD64 X2 5200 dual
AMD64 X2 6000 dual
AMD 9600 black edition quad
Intel Q6600 quad
Two reasons: First, the
)
Two reasons:
First, the SSE2 executable file doesn't have executable permissions set. Use the "chmod a+x" command for that file to set them.
Second, the file is misnamed in two ways. Not only is the "_1" at the end missing, but it also has an extra "_1" in the middle of the filename.
6 of the 7 are now working.
)
6 of the 7 are now working. I borked up the q6600 and ran out of quota, so I guess it'll run WCG for a day.
thanks
by the way, I didn't have to do any text based chmod. I right clicked, picked properties, permissions, advanced, and set them there.
RE: 6 of the 7 are now
)
Yeah, that works, too. I'm just used to doing things on the command-line.
RE: by the way, I didn't
)
wimp :-)
Real BOINC-ers use the CLI.
/me goes to install the package again, now that I know the executable needs to be made executable....
Kathryn :o)
Einstein@Home Moderator
RE: wimp :-) Real
)
What was that about the long walk off the short pier ... ;-).
Cheers,
Gary.
RE: RE: wimp :-) Real
)
Never said I was a real BOINC-er :)
I don't mind the CLI (at least not anymore). But boinc_cmd is just a pain.
Kathryn :o)
Einstein@Home Moderator
RE: Since I've switched my
)
Just a remark: Intel Core2 has 128-bit wide SSE engine, but AMD 6000+ not.
So the AMD 6000+ offers the same speed on SSE2 and FPU instructions.
Core2 exectues SSE2 two times faster than FPU code.
RE: RE: Since I've
)
But will the Intel 128-bit width benefit scalar SSE2 instructions as well?
The SSE2 instructions in the app should all be scalar (I don't think the gcc will auto-vectorize anything?) so a 128 bit data path should not be that much an advantage, right?
As far as I could see the main benefit from SSE2 in this app is caused by a better double to int conversion (still kind of slow with the old fashioned FPU control word setting method even on Core 2). Maybe it's not that big a difference on AMD K8?
CU
Bikeman
RE: But will the Intel
)
You are right, master...