Re: [casper] xst hangs
> a new workstation with 16 cores and 192GB of RAM memory. That sounds like a really beefy machine! Unfortunately, the Xilinx tools don't multi-thread very well. We've found it better to have fewer cores that run at higher clock rates. The modern trend for many cores clocked at lower speeds isn't helping our cause. Hopefully Vivado's a bit better. Anyone got any experience with large designs in Vivado for comparison? I played with it briefly last year and it seemed much faster. I think this is going to be important for ROACH-3. Jason > 2014-04-12 20:24 GMT+02:00 Matt Strader : >> Thanks, Glenn. Increasing the matlab java heap did the trick.And >> thanks, Ryan, for your tip. I may have to do that next. >> >> Matt >> >> >> On Fri, Apr 11, 2014 at 6:46 PM, Ryan Monroe >> wrote: >>> >>> If you're filling a roach 2, you'll have to use the black box trick (on a >>> phone so I can't find it). split the design into 2-4 equal parts on terms >>> of resources used! >>> >>> On Apr 11, 2014 6:34 PM, "G Jones" wrote: Did you increase your matlab java heap? Search the mailing list archives, sounds like it could be the culprit. Glenn On Apr 11, 2014 9:31 PM, "Matt Strader" wrote: > > Hi all, > > I have a rather large design I'm trying to compile for Roach2. > Compilation never gets past xst in System Generator, but doesn't give any > error. Progress seems to halt while trimming pins, stopping in the middle > of printing a warning, such as > "WARNING:Xst:1710 - FF/Latch Then cpu usage drops to near zero, and it just hangs there. It never > takes more than 16% of my RAM, so it's not lagging from lack of memory. > > If I cut a big chunk out of my design, it compiles successfully. > > Any suggestions? > > Thanks, > Matt Strader > UCSB Physics Dept. >> >> > > > > -- > Andrea Mattana > > I.N.A.F. - Istituto di Radioastronomia > Radiotelescopi di Medicina > Via Fiorentina, 3513 - 40059 Medicina (Bo) > Tel. 051/6965834 Fax. 051/6965810 >
Re: [casper] xst hangs
Hi Carspers, I have installed again the toolflow for the ROACH-1 in a new workstation with 16 cores and 192GB of RAM memory. I did my first compile test with a big project and the XST generation took more than 5 hours (rather than 1 on the old machine!) while the ISE/EDK stuff took only 40 minutes. I have read some Xilinx issue about permission of the cache directories. I will make other tests on Monday, it seems that the cached files have been generated on the right places, after reading to this mail I WILL DEFINITELY check the Matlab Java Heap ! Thanks a lot! Cheers, Andrea 2014-04-12 20:24 GMT+02:00 Matt Strader : > Thanks, Glenn. Increasing the matlab java heap did the trick.And > thanks, Ryan, for your tip. I may have to do that next. > > Matt > > > On Fri, Apr 11, 2014 at 6:46 PM, Ryan Monroe > wrote: >> >> If you're filling a roach 2, you'll have to use the black box trick (on a >> phone so I can't find it). split the design into 2-4 equal parts on terms >> of resources used! >> >> On Apr 11, 2014 6:34 PM, "G Jones" wrote: >>> >>> Did you increase your matlab java heap? Search the mailing list archives, >>> sounds like it could be the culprit. >>> >>> Glenn >>> >>> On Apr 11, 2014 9:31 PM, "Matt Strader" >>> wrote: Hi all, I have a rather large design I'm trying to compile for Roach2. Compilation never gets past xst in System Generator, but doesn't give any error. Progress seems to halt while trimming pins, stopping in the middle of printing a warning, such as "WARNING:Xst:1710 - FF/Latch >>> Then cpu usage drops to near zero, and it just hangs there. It never takes more than 16% of my RAM, so it's not lagging from lack of memory. If I cut a big chunk out of my design, it compiles successfully. Any suggestions? Thanks, Matt Strader UCSB Physics Dept. > > -- Andrea Mattana I.N.A.F. - Istituto di Radioastronomia Radiotelescopi di Medicina Via Fiorentina, 3513 - 40059 Medicina (Bo) Tel. 051/6965834 Fax. 051/6965810
Re: [casper] xst hangs
Thanks, Glenn. Increasing the matlab java heap did the trick.And thanks, Ryan, for your tip. I may have to do that next. Matt On Fri, Apr 11, 2014 at 6:46 PM, Ryan Monroe wrote: > If you're filling a roach 2, you'll have to use the black box trick (on a > phone so I can't find it). split the design into 2-4 equal parts on terms > of resources used! > On Apr 11, 2014 6:34 PM, "G Jones" wrote: > >> Did you increase your matlab java heap? Search the mailing list archives, >> sounds like it could be the culprit. >> >> Glenn >> On Apr 11, 2014 9:31 PM, "Matt Strader" >> wrote: >> >>> Hi all, >>> >>> I have a rather large design I'm trying to compile for Roach2. >>> Compilation never gets past xst in System Generator, but doesn't give any >>> error. Progress seems to halt while trimming pins, stopping in the middle >>> of printing a warning, such as >>> "WARNING:Xst:1710 - FF/Latch >> Then cpu usage drops to near zero, and it just hangs there. It never >>> takes more than 16% of my RAM, so it's not lagging from lack of memory. >>> >>> If I cut a big chunk out of my design, it compiles successfully. >>> >>> Any suggestions? >>> >>> Thanks, >>> Matt Strader >>> UCSB Physics Dept. >>> >>
Re: [casper] xst hangs
If you're filling a roach 2, you'll have to use the black box trick (on a phone so I can't find it). split the design into 2-4 equal parts on terms of resources used! On Apr 11, 2014 6:34 PM, "G Jones" wrote: > Did you increase your matlab java heap? Search the mailing list archives, > sounds like it could be the culprit. > > Glenn > On Apr 11, 2014 9:31 PM, "Matt Strader" wrote: > >> Hi all, >> >> I have a rather large design I'm trying to compile for Roach2. >> Compilation never gets past xst in System Generator, but doesn't give any >> error. Progress seems to halt while trimming pins, stopping in the middle >> of printing a warning, such as >> "WARNING:Xst:1710 - FF/Latch > Then cpu usage drops to near zero, and it just hangs there. It never >> takes more than 16% of my RAM, so it's not lagging from lack of memory. >> >> If I cut a big chunk out of my design, it compiles successfully. >> >> Any suggestions? >> >> Thanks, >> Matt Strader >> UCSB Physics Dept. >> >
Re: [casper] xst hangs
Did you increase your matlab java heap? Search the mailing list archives, sounds like it could be the culprit. Glenn On Apr 11, 2014 9:31 PM, "Matt Strader" wrote: > Hi all, > > I have a rather large design I'm trying to compile for Roach2. > Compilation never gets past xst in System Generator, but doesn't give any > error. Progress seems to halt while trimming pins, stopping in the middle > of printing a warning, such as > "WARNING:Xst:1710 - FF/Latch Then cpu usage drops to near zero, and it just hangs there. It never > takes more than 16% of my RAM, so it's not lagging from lack of memory. > > If I cut a big chunk out of my design, it compiles successfully. > > Any suggestions? > > Thanks, > Matt Strader > UCSB Physics Dept. >
[casper] xst hangs
Hi all, I have a rather large design I'm trying to compile for Roach2. Compilation never gets past xst in System Generator, but doesn't give any error. Progress seems to halt while trimming pins, stopping in the middle of printing a warning, such as "WARNING:Xst:1710 - FF/Latch
