From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: MIME-Version: 1.0 In-Reply-To: <146914709427.8780.16475894348537760808@resonance> References: <1467302331-9663-1-git-send-email-wens@csie.org> <1467302331-9663-5-git-send-email-wens@csie.org> <146800297598.73491.13472152266331488346@resonance> <146914709427.8780.16475894348537760808@resonance> From: Chen-Yu Tsai Date: Fri, 22 Jul 2016 09:22:25 +0800 Message-ID: Subject: Re: [PATCH v4 4/7] rtc: ac100: Add clk output support To: Michael Turquette Cc: Chen-Yu Tsai , Lee Jones , Alessandro Zummo , Alexandre Belloni , Rob Herring , Mark Rutland , Maxime Ripard , Stephen Boyd , linux-arm-kernel , rtc-linux@googlegroups.com, linux-clk , devicetree , linux-kernel Content-Type: text/plain; charset=UTF-8 List-ID: On Fri, Jul 22, 2016 at 8:24 AM, Michael Turquette wrote: > Quoting Chen-Yu Tsai (2016-07-13 20:34:34) >> On Sat, Jul 9, 2016 at 2:36 AM, Michael Turquette >> wrote: >> > Quoting Chen-Yu Tsai (2016-06-30 08:58:48) >> >> +static long ac100_clkout_round_rate(struct clk_hw *hw, unsigned long rate, >> >> + unsigned long prate) >> >> +{ >> >> + unsigned long best_rate = 0, tmp_rate, tmp_prate; >> >> + int i; >> >> + >> >> + if (prate == AC100_RTC_32K_RATE) >> >> + return divider_round_rate(hw, rate, &prate, NULL, >> >> + AC100_CLKOUT_DIV_WIDTH, >> >> + CLK_DIVIDER_POWER_OF_TWO); >> >> + >> >> + for (i = 0; ac100_clkout_prediv[i].div; i++) { >> >> + tmp_prate = DIV_ROUND_UP(prate, ac100_clkout_prediv[i].val); >> >> + tmp_rate = divider_round_rate(hw, rate, &tmp_prate, NULL, >> >> + AC100_CLKOUT_DIV_WIDTH, >> >> + CLK_DIVIDER_POWER_OF_TWO); >> >> + >> >> + if (tmp_rate > rate) >> >> + continue; >> >> + if (rate - tmp_rate < best_rate - tmp_rate) >> >> + best_rate = tmp_rate; >> >> + } >> >> + >> >> + return best_rate; >> >> +} >> >> + >> >> +static int ac100_clkout_determine_rate(struct clk_hw *hw, >> >> + struct clk_rate_request *req) >> >> +{ >> >> + struct clk_hw *best_parent; >> >> + unsigned long best = 0; >> >> + int i, num_parents = clk_hw_get_num_parents(hw); >> >> + >> >> + for (i = 0; i < num_parents; i++) { >> >> + struct clk_hw *parent = clk_hw_get_parent_by_index(hw, i); >> >> + unsigned long tmp, prate = clk_hw_get_rate(parent); >> >> + >> >> + tmp = ac100_clkout_round_rate(hw, req->rate, prate); >> >> + >> >> + if (tmp > req->rate) >> >> + continue; >> >> + if (req->rate - tmp < req->rate - best) { >> >> + best = tmp; >> >> + best_parent = parent; >> >> + } >> >> + } >> >> + >> >> + if (!best) >> >> + return -EINVAL; >> >> + >> >> + req->best_parent_hw = best_parent; >> >> + req->best_parent_rate = best; >> >> + req->rate = best; >> >> + >> >> + return 0; >> >> +} >> > >> > You only need one of the two functions above. Keep the .determine_rate >> > callback and drop .round_rate. >> >> Only .round_rate is set in the clk_ops. ac100_clkout_round_rate just >> serves to split out some of the calculations done per parent. > > If you only use .round_rate, shouldn't you drop .determine_rate then? Oops! I meant we only set .determine_rate. ac100_clkout_round_rate is called in the per parent loop in .determine_rate. ChenYu > > Regards, > Mike > >> >> > Otherwise the rest of the clk support looks great to me. Feel free to >> > add: >> > >> > Reviewed-by: Michael Turquette >> >> Thanks! >> >> ChenYu