On 2/3/26 2:38 PM, BALATON Zoltan wrote:
On Tue, 3 Feb 2026, Philippe Mathieu-Daudé wrote:
On 3/2/26 21:56, Pierrick Bouvier wrote:
From: Philippe Mathieu-Daudé <[email protected]>
Endianness access is constant between device resets.
Use the field instead of calling the same function.
Maybe we should justify here, simply adding from your v3 comment:
When doing automated benchmark with fio (see instructions
in [*]) this resulted in a +1.4% speedup vs existing code.
[*] https://lore.kernel.org/qemu-devel/20260202185233.GC405548@fedora/
What does existing code refer to here? According to cover letter it's
within 0.003% of existing speed so not actually faster but only a tiny bit
slower. The speed up above maybe from previous version of the series
without this patch so maybe clarify that more if you include this in the
commit message.
Existing is upstream/master.
+1.4% is with this patch.
End result is -0.3% with the full series applied.
Regards,
BALATON Zoltan
Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
Signed-off-by: Pierrick Bouvier <[email protected]>
---
include/hw/virtio/virtio-access.h | 24 ++++++++++++------------
hw/virtio/virtio.c | 4 ++--
2 files changed, 14 insertions(+), 14 deletions(-)
