On Thu, 11 Mar 2021 at 10:40, Jisheng Zhang <Jisheng.Zhang@xxxxxxxxxxxxx> wrote: > > Clean up the code to use the "mmc" directly instead of "host->mmc". > If the code sits in hot code path, this clean up also brings trvial > performance improvement. Take the sdhci_post_req() for example: > > before the patch: > ... > 8d0: a9be7bfd stp x29, x30, [sp, #-32]! > 8d4: 910003fd mov x29, sp > 8d8: f9000bf3 str x19, [sp, #16] > 8dc: f9400833 ldr x19, [x1, #16] > 8e0: b9404261 ldr w1, [x19, #64] > 8e4: 34000161 cbz w1, 910 <sdhci_post_req+0x50> > 8e8: f9424400 ldr x0, [x0, #1160] > 8ec: d2800004 mov x4, #0x0 // #0 > 8f0: b9401a61 ldr w1, [x19, #24] > 8f4: b9403262 ldr w2, [x19, #48] > 8f8: f9400000 ldr x0, [x0] > 8fc: f278003f tst x1, #0x100 > 900: f9401e61 ldr x1, [x19, #56] > 904: 1a9f17e3 cset w3, eq // eq = none > 908: 11000463 add w3, w3, #0x1 > 90c: 94000000 bl 0 <dma_unmap_sg_attrs> > ... > > After the patch: > ... > 8d0: a9be7bfd stp x29, x30, [sp, #-32]! > 8d4: 910003fd mov x29, sp > 8d8: f9000bf3 str x19, [sp, #16] > 8dc: f9400833 ldr x19, [x1, #16] > 8e0: b9404261 ldr w1, [x19, #64] > 8e4: 34000141 cbz w1, 90c <sdhci_post_req+0x4c> > 8e8: b9401a61 ldr w1, [x19, #24] > 8ec: d2800004 mov x4, #0x0 // #0 > 8f0: b9403262 ldr w2, [x19, #48] > 8f4: f9400000 ldr x0, [x0] > 8f8: f278003f tst x1, #0x100 > 8fc: f9401e61 ldr x1, [x19, #56] > 900: 1a9f17e3 cset w3, eq // eq = none > 904: 11000463 add w3, w3, #0x1 > 908: 94000000 bl 0 <dma_unmap_sg_attrs> > ... > > We saved one ldr instruction: "ldr x0, [x0, #1160]" Nice! Even if I think the cleanup of code makes sense alone. > > Signed-off-by: Jisheng Zhang <Jisheng.Zhang@xxxxxxxxxxxxx> > --- > drivers/mmc/host/sdhci.c | 33 ++++++++++++++++----------------- > 1 file changed, 16 insertions(+), 17 deletions(-) > [...] > @@ -2489,14 +2489,14 @@ void sdhci_enable_sdio_irq(struct mmc_host *mmc, int enable) > unsigned long flags; > > if (enable) > - pm_runtime_get_noresume(host->mmc->parent); > + pm_runtime_get_noresume(mmc->parent); Maybe use mmc_dev(mmc) instead? At least I think I would appreciate consistency in the entire c-file, today it seems like both "mmc->parent" and mmc_dev(mmc) are being used. [...] Kind regards Uffe
