Hi,
> Subject: [Qemu-devel] [PATCH v2 00/10] virtio/vring: optimization patches
>
> This includes two optimization of virtio:
>
> - "slimming down" VirtQueueElements by not including room for
> 1024 buffers. This makes malloc much faster.
>
> - optimizations to limit the number of address_space_translate
> calls in virtio.c, from Vincenzo and myself.
>
Very nice! After apply those optimizations (patch 8, 9, 10), I got 4MB/sec
bonus of speed.
Based on my virtio-crypto device benchmark, ase-128-cbc algorithm. I haven't
rebase
other patches of this patch set yet, maybe I can get other bonus, can I?
Before applying those three patches:
Testing AES-128-CBC cipher:
Encrypting in chunks of 256 bytes: done. 246.84 MiB in 5.02 secs: 49.13
MiB/sec (1011061 packets)
Encrypting in chunks of 256 bytes: done. 247.03 MiB in 5.02 secs: 49.16
MiB/sec (1011840 packets)
Encrypting in chunks of 256 bytes: done. 246.98 MiB in 5.02 secs: 49.17
MiB/sec (1011636 packets)
Encrypting in chunks of 256 bytes: done. 247.14 MiB in 5.02 secs: 49.19
MiB/sec (1012270 packets)
Encrypting in chunks of 256 bytes: done. 246.96 MiB in 5.02 secs: 49.16
MiB/sec (1011565 packets)
Encrypting in chunks of 256 bytes: done. 246.97 MiB in 5.02 secs: 49.18
MiB/sec (1011594 packets)
Encrypting in chunks of 256 bytes: done. 246.89 MiB in 5.02 secs: 49.15
MiB/sec (1011259 packets)
Encrypting in chunks of 256 bytes: done. 246.96 MiB in 5.02 secs: 49.15
MiB/sec (1011561 packets)
'Perf top' shows:
23.61% qemu-kvm [.] address_space_translate
14.49% qemu-kvm [.] qemu_get_ram_ptr
4.65% qemu-kvm [.] phys_page_find
4.31% qemu-kvm [.] address_space_translate_internal
3.18% libpthread-2.19.so [.] __pthread_mutex_unlock_usercnt
2.83% qemu-kvm [.] qemu_ram_addr_from_host
2.40% qemu-kvm [.] address_space_map
2.34% libc-2.19.so [.] _int_malloc
2.22% libc-2.19.so [.] _int_free
1.96% libc-2.19.so [.] malloc
1.71% libpthread-2.19.so [.] pthread_mutex_lock
1.40% qemu-kvm [.] find_next_zero_bit
1.38% libc-2.19.so [.] malloc_consolidate
1.31% qemu-kvm [.] lduw_le_phys
1.27% libc-2.19.so [.] __memcpy_sse2_unaligned
1.05% qemu-kvm [.] qemu_get_ram_block
1.05% qemu-kvm [.] object_unref
1.04% qemu-kvm [.] memory_region_get_ram_addr
After applying those optimizations:
Encrypting in chunks of 256 bytes: done. 267.92 MiB in 5.03 secs:
53.31 MiB/sec (1097399 packets)
Encrypting in chunks of 256 bytes: done. 268.05 MiB in 5.02 secs: 53.35
MiB/sec (1097935 packets)
Encrypting in chunks of 256 bytes: done. 265.40 MiB in 5.02 secs: 52.82
MiB/sec (1087091 packets)
Encrypting in chunks of 256 bytes: done. 263.18 MiB in 5.01 secs: 52.50
MiB/sec (1077999 packets)
Encrypting in chunks of 256 bytes: done. 266.85 MiB in 5.01 secs: 53.29
MiB/sec (1093010 packets)
Encrypting in chunks of 256 bytes: done. 267.64 MiB in 5.02 secs: 53.28
MiB/sec (1096251 packets)
Encrypting in chunks of 256 bytes: done. 267.30 MiB in 5.02 secs: 53.24
MiB/sec (1094861 packets)
Encrypting in chunks of 256 bytes: done. 267.29 MiB in 5.02 secs: 53.25
MiB/sec (1094833 packets)
'Perf top' shows:
22.56% qemu-kvm [.] address_space_translate
13.29% qemu-kvm [.] qemu_get_ram_ptr
4.71% qemu-kvm [.] phys_page_find
4.43% qemu-kvm [.] address_space_translate_internal
3.47% libpthread-2.19.so [.] __pthread_mutex_unlock_usercnt
3.08% qemu-kvm [.] qemu_ram_addr_from_host
2.62% qemu-kvm [.] address_space_map
2.61% libc-2.19.so [.] _int_malloc
2.58% libc-2.19.so [.] _int_free
2.38% libc-2.19.so [.] malloc
2.06% libpthread-2.19.so [.] pthread_mutex_lock
1.68% libc-2.19.so [.] malloc_consolidate
1.35% libc-2.19.so [.] __memcpy_sse2_unaligned
1.23% qemu-kvm [.] lduw_le_phys
1.18% qemu-kvm [.] find_next_zero_bit
1.02% qemu-kvm [.] object_unref
Regards,
-Gonglei
