Using Itanium ?
Or do you guys have PCI domains on Altix UV ?
Brice
Le 08/10/2014 07:37, James Custer a écrit :
> Well, I just started looking at hwloc today as a potential replacement for
> SGI's custom topology command and ran into a problem where when I run lstopo
> I get the following:
> lstopo: pci-common.c:121: hwloc_pci_compare_busids: Assertion `0' failed.
>
> So I looked into it and tracked it down to this change. It was pretty easy to
> fix so I thought I'd pass it along. :)
>
> Regards,
> James
> ________________________________________
> From: hwloc-devel [hwloc-devel-boun...@open-mpi.org] on behalf of Brice
> Goglin [brice.gog...@inria.fr]
> Sent: Wednesday, October 08, 2014 12:32 AM
> To: hwloc-de...@open-mpi.org
> Subject: Re: [hwloc-devel] [PATCH] Fix code relying on nonexistent macros
>
> Applied thanks.
> Interesting to see somebody who cares about PCi domains :)
> Brice
>
>
> Le 08/10/2014 07:24, James Custer a écrit :
>> Commit 1e57af5 removed macros HWLOC_HAVE_PCI_FIND_CAP,
>> HWLOC_HAVE_PCIDEV_DOMAIN, and HWLOC_HAVE_LIBPCIACCESS.
>> However, these are still referenced in src/topology-pci.c.
>> Specifically, the check to see if HWLOC_HAVE_LIBPCIACCESS or
>> HWLOC_HAVE_PCIDEV_DOMAIN will always be false, resulting in
>> domain being set to zero, rather than always being pcidev->domain.
>>
>> The other change is not a bug, just removal of dead code.
>> ---
>> src/topology-pci.c | 12 ------------
>> 1 file changed, 12 deletions(-)
>>
>> diff --git a/src/topology-pci.c b/src/topology-pci.c
>> index f8a3cec..1382025 100644
>> --- a/src/topology-pci.c
>> +++ b/src/topology-pci.c
>> @@ -117,9 +117,6 @@ hwloc_look_pci(struct hwloc_backend *backend)
>> unsigned short tmp16;
>> char name[128];
>> unsigned offset;
>> -#ifdef HWLOC_HAVE_PCI_FIND_CAP
>> - struct pci_cap *cap;
>> -#endif
>>
>> /* initialize the config space in case we fail to read it (missing
>> permissions, etc). */
>> memset(config_space_cache, 0xff, CONFIG_SPACE_CACHESIZE);
>> @@ -127,11 +124,7 @@ hwloc_look_pci(struct hwloc_backend *backend)
>> pci_device_cfg_read(pcidev, config_space_cache, 0,
>> CONFIG_SPACE_CACHESIZE, NULL);
>>
>> /* try to read the domain */
>> -#if (defined HWLOC_HAVE_LIBPCIACCESS) || (defined HWLOC_HAVE_PCIDEV_DOMAIN)
>> domain = pcidev->domain;
>> -#else
>> - domain = 0; /* default domain number */
>> -#endif
>>
>> /* try to read the device_class */
>> device_class = pcidev->device_class >> 8;
>> @@ -150,12 +143,7 @@ hwloc_look_pci(struct hwloc_backend *backend)
>> obj->attr->pcidev.revision = config_space_cache[PCI_REVISION_ID];
>>
>> obj->attr->pcidev.linkspeed = 0; /* unknown */
>> -#ifdef HWLOC_HAVE_PCI_FIND_CAP
>> - cap = pci_find_cap(pcidev, PCI_CAP_ID_EXP, PCI_CAP_NORMAL);
>> - offset = cap ? cap->addr : 0;
>> -#else
>> offset = hwloc_pci_find_cap(config_space_cache, PCI_CAP_ID_EXP);
>> -#endif /* HWLOC_HAVE_PCI_FIND_CAP */
>>
>> if (0xffff == pcidev->vendor_id && 0xffff == pcidev->device_id) {
>> /* SR-IOV puts ffff:ffff in Virtual Function config space.
> _______________________________________________
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
> Link to this post:
> http://www.open-mpi.org/community/lists/hwloc-devel/2014/10/4263.php
> _______________________________________________
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
> Link to this post:
> http://www.open-mpi.org/community/lists/hwloc-devel/2014/10/4263.php
