Re: new upper limit with BIGMEM
Tell the Voyager 1; it's about trespassing even that limit xDDD El 05/04/2011 23:02, James A. Peltier escribiC3: - Original Message - |real mem = 137428045824 (131061MB) |avail mem = 133755703296 (127559MB) | |seems to work ok... | | But have you hit the limit? | | The sky is the limit, but his is not a flying machine. | | Miod Umm, we conquered the skies a while ago. Really the solar system is the limit currently.
Re: new upper limit with BIGMEM
On 04/05/11 01:31, Otto Moerbeek wrote: On Tue, Apr 05, 2011 at 12:57:28AM -0400, STeve Andre' wrote: On 04/05/11 00:52, Otto Moerbeek wrote: On Mon, Apr 04, 2011 at 09:23:48PM -0400, STeve Andre' wrote: On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre' As vague as the first question. Addressing memory is done from a context. -Otto Hmm. OK -- can amd64 then handle 256G of memory? Still missing context. There's the kernel and userland, physical and virtual memory to name a few things that tell a bit about context. MAXDSIZE is the maximum of virtual memory a process can allocate using malloc and similar mechanisms. MAXDSIZE has already been at 8GB for a while. That didn't change with bigmem, which talks about addressable physical memory from the kernel. The two limits have no immediate connection. So the answer is: no, enabling bigmem did not change MAXDSIZE. Processes cannot allocate more virtual memory from the heap with bigmem enabled. Of course now that the kernel can address more physical memory, it might consider backing more of the allocated virtual memory with physical pages at any point in time. -Otto All right, let me ask again, or a little differently. I understand what you are saying, but the 32 bit limit on amd64 has surprised a lot of people lately (I know, they didn't read up). How much physical memory can I stuff into an amd64 box now? Could I have a 64G machine and run ten 6G processes without swapping? --STeve Andre'
Re: new upper limit with BIGMEM
The new official amd64 limit is 1GB. Bryan [bra...@gmail.com] wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit?
Re: new upper limit with BIGMEM
On Tue, Apr 05, 2011 at 02:11:02AM -0400, STeve Andre' wrote: On 04/05/11 01:31, Otto Moerbeek wrote: On Tue, Apr 05, 2011 at 12:57:28AM -0400, STeve Andre' wrote: On 04/05/11 00:52, Otto Moerbeek wrote: On Mon, Apr 04, 2011 at 09:23:48PM -0400, STeve Andre' wrote: On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre' As vague as the first question. Addressing memory is done from a context. -Otto Hmm. OK -- can amd64 then handle 256G of memory? Still missing context. There's the kernel and userland, physical and virtual memory to name a few things that tell a bit about context. MAXDSIZE is the maximum of virtual memory a process can allocate using malloc and similar mechanisms. MAXDSIZE has already been at 8GB for a while. That didn't change with bigmem, which talks about addressable physical memory from the kernel. The two limits have no immediate connection. So the answer is: no, enabling bigmem did not change MAXDSIZE. Processes cannot allocate more virtual memory from the heap with bigmem enabled. Of course now that the kernel can address more physical memory, it might consider backing more of the allocated virtual memory with physical pages at any point in time. -Otto All right, let me ask again, or a little differently. I understand what you are saying, but the 32 bit limit on amd64 has surprised a lot of people lately (I know, they didn't read up). How much physical memory can I stuff into an amd64 box now? Could I have a 64G machine and run ten 6G processes without swapping? --STeve Andre' I don't know enough to answer this. There might be other limitations kicking in. But currently you at least have a chance. -Otto
Re: new upper limit with BIGMEM
On Mon, Apr 04, 2011 at 11:59:29PM +, Miod Vallat wrote: driver level? The limit of profanity comments in the kernel source code? There's a limit for that ? possibly in bisounours-land, not in our country !
Re: new upper limit with BIGMEM
On 04/05/11 03:49, Otto Moerbeek wrote: On Tue, Apr 05, 2011 at 02:11:02AM -0400, STeve Andre' wrote: All right, let me ask again, or a little differently. I understand what you are saying, but the 32 bit limit on amd64 has surprised a lot of people lately (I know, they didn't read up). How much physical memory can I stuff into an amd64 box now? Could I have a 64G machine and run ten 6G processes without swapping? --STeve Andre' I don't know enough to answer this. There might be other limitations kicking in. But currently you at least have a chance. -Otto Realistically, the only question that anyone cares about is this: OpenBSD 4.9-beta (GENERIC.MP) #793: Tue Feb 1 18:45:20 MST 2011 t...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP real mem = 3486384128 (3324MB) avail mem = 3379560448 (3223MB) Just like SMP, what people DO with the RAM (or processors) is irrelevant, it's what shows up in the dmesg that matters! (oops. Need to upgrade that machine! Hm. I thought this machine only had 2G in it...) (the memory lesson is appreciated, at least here, though) Nick.
Re: new upper limit with BIGMEM
OpenBSD 4.9-current (GENERIC.MP) #36: Mon Apr 4 09:39:35 EST 2011 d...@hotspare.eait.uq.edu.au:/home/dlg/src/sys/arch/amd64/compile/GENERIC. MP real mem = 137428045824 (131061MB) avail mem = 133755703296 (127559MB) seems to work ok...
Re: new upper limit with BIGMEM
On Tue, Apr 5, 2011 at 7:33 AM, David Gwynne dlgwy...@gmail.com wrote: OpenBSD 4.9-current (GENERIC.MP) #36: Mon Apr 4 09:39:35 EST 2011 d...@hotspare.eait.uq.edu.au:/home/dlg/src/sys/arch/amd64/compile/GENERIC. MP real mem = 137428045824 (131061MB) avail mem = 133755703296 (127559MB) seems to work ok... But have you hit the limit?
Re: new upper limit with BIGMEM
real mem = 137428045824 (131061MB) avail mem = 133755703296 (127559MB) seems to work ok... But have you hit the limit? The sky is the limit, but his is not a flying machine. Miod
Re: new upper limit with BIGMEM
- Original Message - | real mem = 137428045824 (131061MB) | avail mem = 133755703296 (127559MB) | | seems to work ok... | | But have you hit the limit? | | The sky is the limit, but his is not a flying machine. | | Miod Umm, we conquered the skies a while ago. Really the solar system is the limit currently. -- James A. Peltier IT Services - Research Computing Group Simon Fraser University - Burnaby Campus Phone : 778-782-6573 Fax : 778-782-3045 E-Mail : jpelt...@sfu.ca Website : http://www.sfu.ca/itservices http://blogs.sfu.ca/people/jpeltier
Re: new upper limit with BIGMEM
On Tue, Apr 05, 2011 at 02:02:10PM -0700, James A. Peltier wrote: - Original Message - | real mem = 137428045824 (131061MB) | avail mem = 133755703296 (127559MB) | | seems to work ok... | | But have you hit the limit? | | The sky is the limit, but his is not a flying machine. | | Miod Umm, we conquered the skies a while ago. Really the solar system is the limit currently. What we, biped? -- His Highness, The Holy Space Kraken -- James A. Peltier IT Services - Research Computing Group Simon Fraser University - Burnaby Campus Phone : 778-782-6573 Fax : 778-782-3045 E-Mail : jpelt...@sfu.ca Website : http://www.sfu.ca/itservices http://blogs.sfu.ca/people/jpeltier
new upper limit with BIGMEM
So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit?
Re: new upper limit with BIGMEM
So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod
Re: new upper limit with BIGMEM
On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre'
Re: new upper limit with BIGMEM
On Mon, Apr 4, 2011 at 6:59 PM, Miod Vallat m...@online.fr wrote: The limit of profanity comments in the kernel source code? I think this is the truly important metric here. When will this limit have to be bumped?
Re: new upper limit with BIGMEM
So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? Seriously, http://marc.info/?l=openbsd-cvsm=129114552517946w=2 and then http://marc.info/?l=openbsd-cvsm=129098908815042w=2 48-64 cores usually don't come with 8GB. So 8GB is not the max. The max is practically what somebody can actually get their hands on, and post dmesg pron. TB? Maybe if somebody can donate a capable machine, then they can get it done. A TB class machine will come with at least 128-512 cores :-) This machine will also come equipped with 40/100 GBE ethernet adapter. And cost a bomb right now. It will therefore be practically supported in a few years for all of the BSDs. Read this for the physical limits, scroll to Larger physical address space http://en.wikipedia.org/wiki/X86-64#Architectural_features driver level? The limit of profanity comments in the kernel source code? Not as much as some of the other codebases. Open source is constrained, I bet closed code is more likely to be littered with it.
Re: new upper limit with BIGMEM
On Mon, Apr 04, 2011 at 09:23:48PM -0400, STeve Andre' wrote: On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre' As vague as the first question. Addressing memory is done from a context. -Otto
Re: new upper limit with BIGMEM
On 04/05/11 00:52, Otto Moerbeek wrote: On Mon, Apr 04, 2011 at 09:23:48PM -0400, STeve Andre' wrote: On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre' As vague as the first question. Addressing memory is done from a context. -Otto Hmm. OK -- can amd64 then handle 256G of memory?
Re: new upper limit with BIGMEM
On Tue, Apr 05, 2011 at 12:57:28AM -0400, STeve Andre' wrote: On 04/05/11 00:52, Otto Moerbeek wrote: On Mon, Apr 04, 2011 at 09:23:48PM -0400, STeve Andre' wrote: On 04/04/11 19:59, Miod Vallat wrote: So, now that BIGMEM is up, what is the new max? are we talking TB? or is 8GB the new upper limit? The limit of the number of vague questions for which there is only vague answers is infinite. Which limit are you talking about? The limit of the number of open files in the kernel? The limit of the number of distinct PCI busses in the kernel? The limit of the number of distinct keys handled at the keyboard driver level? The limit of profanity comments in the kernel source code? Miod I think the (first) question is how much memory can be addressed? --STeve Andre' As vague as the first question. Addressing memory is done from a context. -Otto Hmm. OK -- can amd64 then handle 256G of memory? Still missing context. There's the kernel and userland, physical and virtual memory to name a few things that tell a bit about context. MAXDSIZE is the maximum of virtual memory a process can allocate using malloc and similar mechanisms. MAXDSIZE has already been at 8GB for a while. That didn't change with bigmem, which talks about addressable physical memory from the kernel. The two limits have no immediate connection. So the answer is: no, enabling bigmem did not change MAXDSIZE. Processes cannot allocate more virtual memory from the heap with bigmem enabled. Of course now that the kernel can address more physical memory, it might consider backing more of the allocated virtual memory with physical pages at any point in time. -Otto
