|
From: | Avi Kivity |
Subject: | Re: [Qemu-devel] [RFC 1/2] pci-dma-api-v1 |
Date: | Sat, 29 Nov 2008 21:48:22 +0200 |
User-agent: | Thunderbird 2.0.0.18 (X11/20081119) |
Blue Swirl wrote:
The previous similar attempt by Anthony for generic DMA using vectored IO was abandoned because the malloc/free overhead was more than the performance gain. Have you made any performance measurements? How does this version compare to the previous ones?
The pointers you gave don't blame malloc/free, instead the lack of aio readv/writev.
Since Andrea's patches contain emulation for aio readv/writev, the performance degradation will not occur (though we will not see the benefit either).
I doubt you can get measure malloc overhead with anything less a thousand disks (even there, other overheads are likely to drown that malloc).
-- I have a truly marvellous patch that fixes the bug which this signature is too narrow to contain.
[Prev in Thread] | Current Thread | [Next in Thread] |