On Thursday 16 August 2012, Chao Xie wrote:

> +enum {
> +     clk32, vctcxo, pll1, pll1_2, pll1_4, pll1_8, pll1_16, pll1_6, pll1_12,
> +     pll1_24, pll1_48, pll1_96, pll1_13, pll1_13_1_5, pll1_2_1_5,
> +     pll1_3_16, uart_pll, twsi0, twsi1, gpio, kpc, rtc, pwm0, pwm1, pwm2,
> +     pwm3, uart0_mux, uart0, uart1_mux, uart1, uart2_mux, uart2,
> +     ssp0_mux, ssp0, ssp1_mux, ssp1, dfc, sdh0_mux, sdh0, sdh1_mux, sdh1,
> +     usb, sph, disp0_mux, disp0, ccic0_mux, ccic0, ccic0_phy_mux,
> +     ccic0_phy, ccic0_sphy_div, ccic0_sphy, clk_max
> +};

I wonder whether you should just get rid of this enum

> +void __init pxa910_clk_init(void)
> +{
> +     struct clk *clocks[clk_max];

and this array,

> +     clocks[clk32] =
> +         clk_register_fixed_rate(NULL, "clk32", NULL, CLK_IS_ROOT, 3200);
> +     clk_register_clkdev(clocks[clk32], "clk32", NULL);
> +
> +     clocks[vctcxo] =
> +         clk_register_fixed_rate(NULL, "vctcxo", NULL, CLK_IS_ROOT,
> +                                 26000000);
> +     clk_register_clkdev(clocks[vctcxo], "vctcxo", NULL);

because now that the "obfuscation" macro is gone, it's clear that each index
is used only once here and then passed right into the next function. If you
write it like:

        struct clk *clk;

        clk = clk_register_fixed_rate(NULL, "clk32", NULL, CLK_IS_ROOT, 3200);
        clk_register_clkdev(clk, "clk32", NULL);

it becomes much shorter, partly because things start fitting into one
line again. The only exception in this file is 

> +     clocks[uart_pll] =
> +         mmp_clk_register_factor("uart_pll", "pll1_4", 0,
> +                                 mpmu_base + MPMU_UART_PLL,
> +                                 &uart_factor_masks, uart_factor_tbl,
> +                                 ARRAY_SIZE(uart_factor_tbl));
> +     clk_set_rate(clocks[uart_pll], 14745600);
> +     clk_register_clkdev(clocks[uart_pll], "uart_pll", NULL);

with

> +     clocks[uart0_mux] =
> +         clk_register_mux(NULL, "uart0_mux", uart_parent,
> +                          ARRAY_SIZE(uart_parent), CLK_SET_RATE_PARENT,
> +                          apbc_base + APBC_UART0, 4, 3, 0, &clk_lock);
> +     clk_set_parent(clocks[uart0_mux], clocks[uart_pll]);
> +     clk_register_clkdev(clocks[uart0_mux], "uart_mux.0", NULL);
> +
> +     clocks[uart0] =
> +         mmp_clk_register_apbc("uart0", "uart0_mux", apbc_base + APBC_UART0,
> +                               10, 0, &clk_lock);
> +     clk_register_clkdev(clocks[uart0], NULL, "pxa2xx-uart.0");

so just add another

        struct clk *clk_uart;

        clk_uart = mmp_clk_register_factor("uart_pll", "pll1_4", 0,
                        mpmu_base + MPMU_UART_PLL, &uart_factor_masks, 
uart_factor_tbl,
                        ARRAY_SIZE(uart_factor_tbl));



        Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Reply via email to