2018-07-30 7:14 GMT+03:00 Tomasz Figa <tfiga@xxxxxxxxxxxx>: > On Wed, Jul 25, 2018 at 10:47 PM Matwey V. Kornilov <matwey@xxxxxxxxxx> wrote: >> >> 2018-07-24 23:55 GMT+03:00 Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>: >> > On Tue, 24 Jul 2018, Matwey V. Kornilov wrote: >> > >> >> 2018-07-23 21:57 GMT+03:00 Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>: >> >> > On Mon, 23 Jul 2018, Matwey V. Kornilov wrote: >> >> > >> >> >> I've tried to strategies: >> >> >> >> >> >> 1) Use dma_unmap and dma_map inside the handler (I suppose this is >> >> >> similar to how USB core does when there is no URB_NO_TRANSFER_DMA_MAP) >> >> > >> >> > Yes. >> >> > >> >> >> 2) Use sync_cpu and sync_device inside the handler (and dma_map only >> >> >> once at memory allocation) >> >> >> >> >> >> It is interesting that dma_unmap/dma_map pair leads to the lower >> >> >> overhead (+1us) than sync_cpu/sync_device (+2us) at x86_64 platform. >> >> >> At armv7l platform using dma_unmap/dma_map leads to ~50 usec in the >> >> >> handler, and sync_cpu/sync_device - ~65 usec. >> >> >> >> >> >> However, I am not sure is it mandatory to call >> >> >> dma_sync_single_for_device for FROM_DEVICE direction? >> >> > >> >> > According to Documentation/DMA-API-HOWTO.txt, the CPU should not write >> >> > to a DMA_FROM_DEVICE-mapped area, so dma_sync_single_for_device() is >> >> > not needed. >> >> >> >> Well, I measured the following at armv7l. The handler execution time >> >> (URB_NO_TRANSFER_DMA_MAP is used for all cases): >> >> >> >> 1) coherent DMA: ~3000 usec (pwc is not functional) >> >> 2) explicit dma_unmap and dma_map in the handler: ~52 usec >> >> 3) explicit dma_sync_single_for_cpu (no dma_sync_single_for_device): ~56 usec > > That's very strange because on ARM dma_unmap_single() does exactly the > same thing as dma_sync_single_for_cpu(): > > arm_dma_map_page() > https://elixir.bootlin.com/linux/v4.18-rc7/source/arch/arm/mm/dma-mapping.c#L129 > arm_dma_unmap_page() > https://elixir.bootlin.com/linux/v4.18-rc7/source/arch/arm/mm/dma-mapping.c#L159 > > arm_dma_sync_single_for_cpu() > https://elixir.bootlin.com/linux/v4.18-rc7/source/arch/arm/mm/dma-mapping.c#L167 > > Could you post the code you used for testing of cases 2) and 3)? I remeasured the timings (see my other mail). The code you asked is here: https://github.com/matwey/linux/tree/pwc_dma_v3 but I am going to rebase this branch. > > Best regards, > Tomasz > -- With best regards, Matwey V. Kornilov. Sternberg Astronomical Institute, Lomonosov Moscow State University, Russia 119234, Moscow, Universitetsky pr-k 13, +7 (495) 9392382
