Andreas,
Thank you for your response. I would like to look into this, but if you are
already working on a patch, I will avoid duplicating effort. On the other
hand, if you think my going down the route with Python and gdb would
possibly provide additional insights for this problem, I can try to make
time to do this.
Thanks again,
Patrick
On Tue, Dec 9, 2014 at 1:44 AM, Andreas Hansson <andreas.hans...@arm.com>
wrote:
> Hi Patrick,
>
> I don’t dare say exactly. I suspect it is the interplay of the
> multi-channel DRAM and the oddity of having one set of controllers per
> memory range (and the chopped up ranges you get with >2 GB worth of
> memory). If you relly need this working now then a few Python prints along
> with gdb is probably the best way to go.
>
> The more long term solution to this is probably proper support for
> memory aliasing. In short we allocate a big chunk of contiguous address
> space on some higher addresses for the entire DRAM (4 GB in your case), and
> them map part of that (2 GB most likely) to a lower range. I’m working on a
> patch for this, and it should not be too long before it is on RB.
>
> Andreas
>
> From: Patrick via gem5-users <gem5-users@gem5.org>
> Reply-To: Patrick <plafr...@gmail.com>, gem5 users mailing list <
> gem5-users@gem5.org>
> Date: Monday, 8 December 2014 20:36
>
> To: gem5 users mailing list <gem5-users@gem5.org>
> Subject: Re: [gem5-users] Why does gem5 create twice the number of memory
> controllers and simulate 1GB memory more while sepecified memory size is
> greater than 3GB
>
> Andreas,
>
> Below is the command line I am using.
>
> ./build/X86/gem5.opt --debug-flags=PatrickTLB,PatrickPageTableWalker
> --debug-file=out ./configs/example/fs.py --cpu-type=timing --caches
> --l2cache --mem-channels=4 -b boot
>
> "boot" is a benchmark I added that initializes SysConfig to use 4GB of
> memory (i.e. SysConfig('boot.rcS','4GB') ).
>
> The "PatrickTLB" and "PatrickPageTableWalker" are debug flags I added,
> but I assume they aren't relevant to my question here.
>
> Any help is appreciated.
>
> Regards,
> Patrick
>
> On Mon, Dec 8, 2014 at 11:09 AM, Andreas Hansson <andreas.hans...@arm.com>
> wrote:
>
>> Patrick,
>>
>> Could you elaborate on the command line you are using when experiencing
>> this error?
>>
>> Thanks,
>>
>> Andreas
>>
>> From: Patrick via gem5-users <gem5-users@gem5.org>
>> Reply-To: Patrick <plafr...@gmail.com>, gem5 users mailing list <
>> gem5-users@gem5.org>
>> Date: Monday, 8 December 2014 17:51
>> To: gem5 users mailing list <gem5-users@gem5.org>
>> Subject: Re: [gem5-users] Why does gem5 create twice the number of
>> memory controllers and simulate 1GB memory more while sepecified memory
>> size is greater than 3GB
>>
>> Sandnon,
>>
>> When I try to run with > 3GB, I get an error message saying:
>>
>> "panic: Cannot test intersection of interleaved range [0x100000000 :
>> 0x13fffffff], [14 : 13] = 1"
>>
>> Were you able to get around this somehow, or did you not see this?
>>
>> Any help is appreciated.
>>
>> Regards,
>> Patrick
>>
>> On Tue, Oct 14, 2014 at 8:51 PM, sandnon via gem5-users <
>> gem5-users@gem5.org> wrote:
>>
>>> Hi all,
>>>
>>> When I ran full system mode of gem5 with --mem-size=4GB and
>>> --disk-image=/dist/m5/system/disks/linux-x86.img, it gave me a warn
>>> message that
>>>
>>> > warn: Physical memory size specified is 4GB which is greater than
>>> 3GB. Twice the number of memory controllers would be created.
>>>
>>> And the slave terminal said
>>>
>>> > Memory: 4043292k/5242880k available (3742k kernel code, 150376k
>>> reserved, 1874k data, 232k init)
>>>
>>> It seems gem5 simulated 5GB for me while I requested for 4GB.
>>>
>>> I searched it and found a [changeset](http://repo.gem5.
>>> org/gem5?cmd=changeset;node=fae4550d2103) which said
>>>
>>> > On the PC platform, the memory region 0xC0000000-0xFFFFFFFF is
>>> reserved for various devices. Hence, if the physical memory size is
>>> greater than 3GB, we need to split it into two parts.
>>>
>>> I still have no idea why gem5 do that even I saw the comments
>>> aforementioned. I know linux kernel will reserve 0xC0000000-0XFFFFFFFF
>>> virtual address for kernel. But I'm not sure if it is related.
>>>
>>> I conclude my questions again
>>>
>>> - Why does gem5 create twice the number of memory controllers while
>>> physical memory size is greater than 3GB
>>> - Why does gem5 simulate 1GB memory more while specified memory size is
>>> greater than 3GB
>>>
>>> ---
>>> Sincerely,
>>> Sandnon
>>> _______________________________________________
>>> gem5-users mailing list
>>> gem5-users@gem5.org
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>>
>>
>>
>> -- IMPORTANT NOTICE: The contents of this email and any attachments are
>> confidential and may also be privileged. If you are not the intended
>> recipient, please notify the sender immediately and do not disclose the
>> contents to any other person, use it for any purpose, or store or copy the
>> information in any medium. Thank you.
>>
>> ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ,
>> Registered in England & Wales, Company No: 2557590
>> ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ,
>> Registered in England & Wales, Company No: 2548782
>>
>
>
> -- IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
>
> ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ,
> Registered in England & Wales, Company No: 2557590
> ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ,
> Registered in England & Wales, Company No: 2548782
>
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
