Stefan Schmidt escreveu: > Hello. > > On Sat, 2008-07-12 at 13:01, Antonio Ospite wrote: >> On Wed, 09 Jul 2008 06:17:16 +0000 >> [EMAIL PROTECTED] wrote: >> >>> ++ int val = (bits == MMC_VDD_165_195) ? 6 : fls(bits) / 2 + 3; >> Ok, the formula has been tested, and it works. > > Great, thanks for testing it out. > >> Just as a side-note, we have this call: >> ezx_pcap_mmcsd_voltage((1 << vdd) & pdata->ocr_mask); >> we shift here and then "unshift" the value with fls(). >> I don't know how much fls() costs, is it worth to avoid it? >> If yes, I could make the needed changes. > > It could be a left over as we used fls() already before Daniel made the > changes > based on Eric comments. Daniel, your opinion on this?
Shift left and just after shift right again seems stupid. Please fix this. > >> Ah, and is there a reason why we are not using msecs_to_jiffies() when >> setting the detect_delay field in ezx_mci_platform_data? We have: >> .detect_delay = 150 / (1000/HZ), > > Hmm, perhaps Daniel did just not know about it? Of course only a guess. Eric Miao wrote: > Why not just initialize this detect_delay within the ezx_mci_platform_data > definition below? And thus avoid that static structure declaration above? Using msecs_to_jiffies() on the declaration of ezx_mci_platform_data leads to: arch/arm/mach-pxa/ezx.c:199: error: initializer element is not constant arch/arm/mach-pxa/ezx.c:199: error: (near initialization for ‘ezx_mci_platform_data.detect_delay’) To make this simple, i think we should simply use: .detect_delay = 15, -- Daniel Ribeiro