I'll try to build an App with the old Visual Studio of 2003 (instead of VS2005). At least the /G7 optimization should work there. Let's see if it helps...
It can be found on the Power User's Apps page. This is definitely not a release candidate, just something to see in which direction to proceed.
BM
My first pure 4.26 workunit is in progress on my 3500+ machine. With about 21% of it done, it looks like it's going to show a huge decrease in required processing time. Also, as someone else has noted, it looks like my processor isn't running as hot as it was with the old apps.
@Bernd--This is probably another silly question, but. . .
With all of the problems you're encountering with the Microsoft compilers, would it be possible to try some other compiler, like for example, Borland or OpenWatcom? I'd be curious as to whether something else might produce more bug-free code.
Of course, I'm only a very novice programmer, so I don't have any solid information on whether this would really be the case.
First full 4.26 WU finished. Gave a speedup around 17 percent. But it is not valiedated yet, waiting for partner.
Edit wants to ask: Am I right, that there is no SSE-Code in this app? Would be incredible for this speedup...
Correct, no SSE code, so there's more to come in the future. The performance of the recent Linux power App should be a rough indication what to expect on the Windows platform as well.
Also, as someone else has noted, it looks like my processor isn't running as hot as it was with the old apps.
Actually, if anything it should run a tiny bit hotter :-). More instructions processed per real-time second ==> more head to dissipate. But with temp-regulated fans, the fan's average RPM would be a better indication of heat output than temperature of the die.
Quote:
Sooooooo, lets see. If I download and install this new app I will get a noticeable performance increase on my Pentium D then, yes?
With all of the problems you're encountering with the Microsoft compilers, would it be possible to try some other compiler, like for example, Borland or OpenWatcom? I'd be curious as to whether something else might produce more bug-free code.
Actually I only have seen one thing that I'd consider a serious bug. Difference in performance of the produced binaries between versions of compilers are common, I know quite some cases for gcc, too.
BOINC is currently bound to the M$ compiler, and mixed linking is painful and the results still have their drawbacks. I'd really like to use the Intel compiler for all x86-based platforms, but so far I could get neither BOINC nor our code compiled with these.
With all of the problems you're encountering with the Microsoft compilers, would it be possible to try some other compiler, like for example, Borland or OpenWatcom? I'd be curious as to whether something else might produce more bug-free code.
Actually I only have seen one thing that I'd consider a serious bug. Difference in performance of the produced binaries between versions of compilers are common, I know quite some cases for gcc, too.
BOINC is currently bound to the M$ compiler, and mixed linking is painful and the results still have their drawbacks. I'd really like to use the Intel compiler for all x86-based platforms, but so far I could get neither BOINC nor our code compiled with these.
Actually, if anything it should run a tiny bit hotter :-). More instructions processed per real-time second ==> more head to dissipate. But with temp-regulated fans, the fan's average RPM would be a better indication of heat output than temperature of the die.
Well, my fans have zero temperature regulation. So changes in reported core temperature are pretty directly related to changes in power consumption, if the environment is stable.
You perhaps have heard the management jargon term "work smarter, not harder". Not all code speedup tweaks just get more of the hardware working at once. Some find ways to avoid doing some of the computing. I had the impression that the linear trig function trick is meant to do just that.
RE: back and forth, back
)
Imho, it is nice to see all this activity on E@H, very exciting! :)
Sooooooo, lets see. If I
)
Sooooooo, lets see. If I download and install this new app I will get a noticeable performance increase on my Pentium D then, yes?
RE: From the 4.25 App
)
My first pure 4.26 workunit is in progress on my 3500+ machine. With about 21% of it done, it looks like it's going to show a huge decrease in required processing time. Also, as someone else has noted, it looks like my processor isn't running as hot as it was with the old apps.
@Bernd--This is probably another silly question, but. . .
With all of the problems you're encountering with the Microsoft compilers, would it be possible to try some other compiler, like for example, Borland or OpenWatcom? I'd be curious as to whether something else might produce more bug-free code.
Of course, I'm only a very novice programmer, so I don't have any solid information on whether this would really be the case.
First full 4.26 WU finished.
)
First full 4.26 WU finished. Gave a speedup around 17 percent. But it is not valiedated yet, waiting for partner.
Edit wants to ask: Am I right, that there is no SSE-Code in this app? Would be incredible for this speedup...
I loaded 4.26 last night and
)
I loaded 4.26 last night and have noticed a ~15% drop in the amount of CPU time my units take. Even better, it appears the results are successful. :)
Congrats on what seems to be a successfully working, faster App.
Marc
RE: First full 4.26 WU
)
Correct, no SSE code, so there's more to come in the future. The performance of the recent Linux power App should be a rough indication what to expect on the Windows platform as well.
Bikeman
RE: Also, as someone else
)
Actually, if anything it should run a tiny bit hotter :-). More instructions processed per real-time second ==> more head to dissipate. But with temp-regulated fans, the fan's average RPM would be a better indication of heat output than temperature of the die.
Yup!
Bikeman
RE: With all of the
)
Actually I only have seen one thing that I'd consider a serious bug. Difference in performance of the produced binaries between versions of compilers are common, I know quite some cases for gcc, too.
BOINC is currently bound to the M$ compiler, and mixed linking is painful and the results still have their drawbacks. I'd really like to use the Intel compiler for all x86-based platforms, but so far I could get neither BOINC nor our code compiled with these.
BM
BM
RE: RE: With all of the
)
Consider talking with Matthias...?
RE: Actually, if anything
)
Well, my fans have zero temperature regulation. So changes in reported core temperature are pretty directly related to changes in power consumption, if the environment is stable.
You perhaps have heard the management jargon term "work smarter, not harder". Not all code speedup tweaks just get more of the hardware working at once. Some find ways to avoid doing some of the computing. I had the impression that the linear trig function trick is meant to do just that.