Note: This bug is displayed in read-only format because
the product is no longer active in Red Hat Bugzilla.
RHEL Engineering is moving the tracking of its product development work on RHEL 6 through RHEL 9 to Red Hat Jira (issues.redhat.com). If you're a Red Hat customer, please continue to file support cases via the Red Hat customer portal. If you're not, please head to the "RHEL project" in Red Hat Jira and file new tickets here. Individual Bugzilla bugs in the statuses "NEW", "ASSIGNED", and "POST" are being migrated throughout September 2023. Bugs of Red Hat partners with an assigned Engineering Partner Manager (EPM) are migrated in late September as per pre-agreed dates. Bugs against components "kernel", "kernel-rt", and "kpatch" are only migrated if still in "NEW" or "ASSIGNED". If you cannot log in to RH Jira, please consult article #7032570. That failing, please send an e-mail to the RH Jira admins at rh-issues@redhat.com to troubleshoot your issue as a user management inquiry. The email creates a ServiceNow ticket with Red Hat. Individual Bugzilla bugs that are migrated will be moved to status "CLOSED", resolution "MIGRATED", and set with "MigratedToJIRA" in "Keywords". The link to the successor Jira issue will be found under "Links", have a little "two-footprint" icon next to it, and direct you to the "RHEL project" in Red Hat Jira (issue links are of type "https://issues.redhat.com/browse/RHEL-XXXX", where "X" is a digit). This same link will be available in a blue banner at the top of the page informing you that that bug has been migrated.
Description of problem:
Currently the code needlessly tries to check if SSE2 is available and then use it. However, in all supported QEMU CPUs, we expose SSE2.
In addition, it looks like this hack has #ifndef _WIN64 , which means in WIN64 we don't use the fast (manually implemented) memcpy.
The compiler knows how to do it memcpy fast using SSE2 if we have the flag in the compiler to target SSE2.
(Note: need to see if it can do it even if the memory is not aligned, but in any case this can't hurt).
Version-Release number of selected component (if applicable):
How reproducible:
Steps to Reproduce:
1.
2.
3.
Actual results:
Expected results:
Additional info:
The driver doesn't use memcopy but rather RtlCopyMemory, which I think is not affected by this flag.
Implementing SSE for win64 will probably help video performance there.
(In reply to comment #11)
> ack the bug. we can re-compile, the new binary has to be certified
Please see comment #8. This flag is not relevant for kernel drivers. We implement the SSE2 memory copy by ourselves for the 32-bit driver. But we are missing this implementation for the 64-bit driver. I would have tried to compare hd video on windows 7 32-guest to windoes 7 64-guest, and if there is a major difference, consider implementing the SSE2 code for 64-guest. Or wait till we have a user space driver...
I highly doubt win7 does not use SSE for memcpy, in kernel. In any case, I don't see the harm in adding this compilation switch.
Regardless, the fact we don't have it for Win7/64 is documented at https://bugs.freedesktop.org/show_bug.cgi?id=37457
(In reply to comment #13)
> (In reply to comment #11)
> > ack the bug. we can re-compile, the new binary has to be certified
>
> Please see comment #8. This flag is not relevant for kernel drivers.
1. In this case we need to close this bug, or change the subject to "Implement the SSE2 code for the 64-bit kernel space driver"
>We
> implement the SSE2 memory copy by ourselves for the 32-bit driver.
2. Was it worse doing it? Any performance improvements were observed on Win32?
>But we are
> missing this implementation for the 64-bit driver. I would have tried to
> compare hd video on windows 7 32-guest to windoes 7 64-guest,
3. it is reasonable and easy step.
> and if there is a
> major difference, consider implementing the SSE2 code for 64-guest.
4. if difference, then the implementation is justified!
> Or wait
> till we have a user space driver...
posted patches on spice-devel removing all SSE2 using fast_memcpy_* code after benchmarks showed it was slower then the internal RtlCopyMemory (which is ifdeffed to memcpy in current Windows development kits).
http://patchwork.freedesktop.org/patch/25265/
Comment 21Christophe Fergeau
2015-06-04 14:43:59 UTC