Messages in this thread | | | From | Rusty Russell <> | Subject | Re: [PATCH 08/16] virtio_ring: virtqueue_add_outbuf / virtqueue_add_inbuf. | Date | Wed, 06 Mar 2013 17:03:42 +1100 |
| |
"Michael S. Tsirkin" <mst@redhat.com> writes: >> Without the inline keywords, it doesn't inline virtqueue_add, and thus >> sg_next_chained and sg_next_add aren't inlined: >> >> $ for i in `seq 50`; do /usr/bin/time --format=%U ./vringh_test --indirect --eventidx --parallel; done 2>&1 | stats --trim-outliers >> Using CPUS 0 and 3 >> Guest: notified 39102-39145(39105), pinged 39060-39063(39063) >> Host: notified 39060-39063(39063), pinged 19551-19581(19553) >> 3.050000-3.220000(3.136875) >> >> With inline: >> >> $ for i in `seq 50`; do /usr/bin/time --format=%U ./vringh_test --indirect --eventidx --parallel; done 2>&1 | stats --trim-outliers >> Using CPUS 0 and 3 >> Guest: notified 39084-39148(39099), pinged 39062-39063(39062) >> Host: notified 39062-39063(39062), pinged 19542-19574(19550) >> 2.940000-3.140000(3.014583) >> >> Cheers, >> Rusty. > > Cool and did it actually unroll all loops?
Sorry for the delay in answering... I spent a day chasing red-herrings, as my tests became limited by the vringh side, so optimizations on the virtio side were having no effect :(
I'll answer a related question based on the current tree, where virtio_add_outbuf (and virtio_add_inbuf) are the stars:
return virtqueue_add(vq, &sg, sg_next_arr, num, 0, 1, 0, data, gfp);
Ideally, gcc would eliminated the input-descriptor loop altogether (that 0 for in_sgs), and unroll the 1-iteration output-descriptor loop into straightline code.
Which it seems to have done, by my reading of the asm (at least, for gcc 4.7 on 32-bit x86).
Cheers, Rusty.
| |