summaryrefslogtreecommitdiff
path: root/arch/arm/plat-samsung/include/plat/pll.h
diff options
context:
space:
mode:
authorTony Lindgren <tony@atomide.com>2011-11-08 00:27:23 +0400
committerTony Lindgren <tony@atomide.com>2011-11-08 00:27:23 +0400
commitd30cc16c8e48368e0518f4975a78711e53e14a0f (patch)
tree26b57f7ab5a963cc3d6c57dff6951bd930875583 /arch/arm/plat-samsung/include/plat/pll.h
parent41eb2d813f558900884e240c2f723e36c7bd151f (diff)
parenta1bcc1dcef8451b4291ea2a1b2677cb194102952 (diff)
downloadlinux-d30cc16c8e48368e0518f4975a78711e53e14a0f.tar.xz
Merge branch 'fixes-modulesplit' into fixes
Diffstat (limited to 'arch/arm/plat-samsung/include/plat/pll.h')
-rw-r--r--arch/arm/plat-samsung/include/plat/pll.h323
1 files changed, 323 insertions, 0 deletions
diff --git a/arch/arm/plat-samsung/include/plat/pll.h b/arch/arm/plat-samsung/include/plat/pll.h
new file mode 100644
index 000000000000..357af7c1c664
--- /dev/null
+++ b/arch/arm/plat-samsung/include/plat/pll.h
@@ -0,0 +1,323 @@
+/* linux/arch/arm/plat-samsung/include/plat/pll.h
+ *
+ * Copyright (c) 2009-2011 Samsung Electronics Co., Ltd.
+ * http://www.samsung.com/
+ *
+ * Copyright 2008 Openmoko, Inc.
+ * Copyright 2008 Simtec Electronics
+ * Ben Dooks <ben@simtec.co.uk>
+ * http://armlinux.simtec.co.uk/
+ *
+ * Samsung PLL codes
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+*/
+
+#include <asm/div64.h>
+
+#define S3C24XX_PLL_MDIV_MASK (0xFF)
+#define S3C24XX_PLL_PDIV_MASK (0x1F)
+#define S3C24XX_PLL_SDIV_MASK (0x3)
+#define S3C24XX_PLL_MDIV_SHIFT (12)
+#define S3C24XX_PLL_PDIV_SHIFT (4)
+#define S3C24XX_PLL_SDIV_SHIFT (0)
+
+static inline unsigned int s3c24xx_get_pll(unsigned int pllval,
+ unsigned int baseclk)
+{
+ unsigned int mdiv, pdiv, sdiv;
+ uint64_t fvco;
+
+ mdiv = (pllval >> S3C24XX_PLL_MDIV_SHIFT) & S3C24XX_PLL_MDIV_MASK;
+ pdiv = (pllval >> S3C24XX_PLL_PDIV_SHIFT) & S3C24XX_PLL_PDIV_MASK;
+ sdiv = (pllval >> S3C24XX_PLL_SDIV_SHIFT) & S3C24XX_PLL_SDIV_MASK;
+
+ fvco = (uint64_t)baseclk * (mdiv + 8);
+ do_div(fvco, (pdiv + 2) << sdiv);
+
+ return (unsigned int)fvco;
+}
+
+#define S3C2416_PLL_MDIV_MASK (0x3FF)
+#define S3C2416_PLL_PDIV_MASK (0x3F)
+#define S3C2416_PLL_SDIV_MASK (0x7)
+#define S3C2416_PLL_MDIV_SHIFT (14)
+#define S3C2416_PLL_PDIV_SHIFT (5)
+#define S3C2416_PLL_SDIV_SHIFT (0)
+
+static inline unsigned int s3c2416_get_pll(unsigned int pllval,
+ unsigned int baseclk)
+{
+ unsigned int mdiv, pdiv, sdiv;
+ uint64_t fvco;
+
+ mdiv = (pllval >> S3C2416_PLL_MDIV_SHIFT) & S3C2416_PLL_MDIV_MASK;
+ pdiv = (pllval >> S3C2416_PLL_PDIV_SHIFT) & S3C2416_PLL_PDIV_MASK;
+ sdiv = (pllval >> S3C2416_PLL_SDIV_SHIFT) & S3C2416_PLL_SDIV_MASK;
+
+ fvco = (uint64_t)baseclk * mdiv;
+ do_div(fvco, (pdiv << sdiv));
+
+ return (unsigned int)fvco;
+}
+
+#define S3C6400_PLL_MDIV_MASK (0x3FF)
+#define S3C6400_PLL_PDIV_MASK (0x3F)
+#define S3C6400_PLL_SDIV_MASK (0x7)
+#define S3C6400_PLL_MDIV_SHIFT (16)
+#define S3C6400_PLL_PDIV_SHIFT (8)
+#define S3C6400_PLL_SDIV_SHIFT (0)
+
+static inline unsigned long s3c6400_get_pll(unsigned long baseclk,
+ u32 pllcon)
+{
+ u32 mdiv, pdiv, sdiv;
+ u64 fvco = baseclk;
+
+ mdiv = (pllcon >> S3C6400_PLL_MDIV_SHIFT) & S3C6400_PLL_MDIV_MASK;
+ pdiv = (pllcon >> S3C6400_PLL_PDIV_SHIFT) & S3C6400_PLL_PDIV_MASK;
+ sdiv = (pllcon >> S3C6400_PLL_SDIV_SHIFT) & S3C6400_PLL_SDIV_MASK;
+
+ fvco *= mdiv;
+ do_div(fvco, (pdiv << sdiv));
+
+ return (unsigned long)fvco;
+}
+
+#define PLL6553X_MDIV_MASK (0x7F)
+#define PLL6553X_PDIV_MASK (0x1F)
+#define PLL6553X_SDIV_MASK (0x3)
+#define PLL6553X_KDIV_MASK (0xFFFF)
+#define PLL6553X_MDIV_SHIFT (16)
+#define PLL6553X_PDIV_SHIFT (8)
+#define PLL6553X_SDIV_SHIFT (0)
+
+static inline unsigned long s3c_get_pll6553x(unsigned long baseclk,
+ u32 pll_con0, u32 pll_con1)
+{
+ unsigned long result;
+ u32 mdiv, pdiv, sdiv, kdiv;
+ u64 tmp;
+
+ mdiv = (pll_con0 >> PLL6553X_MDIV_SHIFT) & PLL6553X_MDIV_MASK;
+ pdiv = (pll_con0 >> PLL6553X_PDIV_SHIFT) & PLL6553X_PDIV_MASK;
+ sdiv = (pll_con0 >> PLL6553X_SDIV_SHIFT) & PLL6553X_SDIV_MASK;
+ kdiv = pll_con1 & PLL6553X_KDIV_MASK;
+
+ /*
+ * We need to multiple baseclk by mdiv (the integer part) and kdiv
+ * which is in 2^16ths, so shift mdiv up (does not overflow) and
+ * add kdiv before multiplying. The use of tmp is to avoid any
+ * overflows before shifting bac down into result when multipling
+ * by the mdiv and kdiv pair.
+ */
+
+ tmp = baseclk;
+ tmp *= (mdiv << 16) + kdiv;
+ do_div(tmp, (pdiv << sdiv));
+ result = tmp >> 16;
+
+ return result;
+}
+
+#define PLL35XX_MDIV_MASK (0x3FF)
+#define PLL35XX_PDIV_MASK (0x3F)
+#define PLL35XX_SDIV_MASK (0x7)
+#define PLL35XX_MDIV_SHIFT (16)
+#define PLL35XX_PDIV_SHIFT (8)
+#define PLL35XX_SDIV_SHIFT (0)
+
+static inline unsigned long s5p_get_pll35xx(unsigned long baseclk, u32 pll_con)
+{
+ u32 mdiv, pdiv, sdiv;
+ u64 fvco = baseclk;
+
+ mdiv = (pll_con >> PLL35XX_MDIV_SHIFT) & PLL35XX_MDIV_MASK;
+ pdiv = (pll_con >> PLL35XX_PDIV_SHIFT) & PLL35XX_PDIV_MASK;
+ sdiv = (pll_con >> PLL35XX_SDIV_SHIFT) & PLL35XX_SDIV_MASK;
+
+ fvco *= mdiv;
+ do_div(fvco, (pdiv << sdiv));
+
+ return (unsigned long)fvco;
+}
+
+#define PLL36XX_KDIV_MASK (0xFFFF)
+#define PLL36XX_MDIV_MASK (0x1FF)
+#define PLL36XX_PDIV_MASK (0x3F)
+#define PLL36XX_SDIV_MASK (0x7)
+#define PLL36XX_MDIV_SHIFT (16)
+#define PLL36XX_PDIV_SHIFT (8)
+#define PLL36XX_SDIV_SHIFT (0)
+
+static inline unsigned long s5p_get_pll36xx(unsigned long baseclk,
+ u32 pll_con0, u32 pll_con1)
+{
+ unsigned long result;
+ u32 mdiv, pdiv, sdiv, kdiv;
+ u64 tmp;
+
+ mdiv = (pll_con0 >> PLL36XX_MDIV_SHIFT) & PLL36XX_MDIV_MASK;
+ pdiv = (pll_con0 >> PLL36XX_PDIV_SHIFT) & PLL36XX_PDIV_MASK;
+ sdiv = (pll_con0 >> PLL36XX_SDIV_SHIFT) & PLL36XX_SDIV_MASK;
+ kdiv = pll_con1 & PLL36XX_KDIV_MASK;
+
+ tmp = baseclk;
+
+ tmp *= (mdiv << 16) + kdiv;
+ do_div(tmp, (pdiv << sdiv));
+ result = tmp >> 16;
+
+ return result;
+}
+
+#define PLL45XX_MDIV_MASK (0x3FF)
+#define PLL45XX_PDIV_MASK (0x3F)
+#define PLL45XX_SDIV_MASK (0x7)
+#define PLL45XX_MDIV_SHIFT (16)
+#define PLL45XX_PDIV_SHIFT (8)
+#define PLL45XX_SDIV_SHIFT (0)
+
+enum pll45xx_type_t {
+ pll_4500,
+ pll_4502,
+ pll_4508
+};
+
+static inline unsigned long s5p_get_pll45xx(unsigned long baseclk, u32 pll_con,
+ enum pll45xx_type_t pll_type)
+{
+ u32 mdiv, pdiv, sdiv;
+ u64 fvco = baseclk;
+
+ mdiv = (pll_con >> PLL45XX_MDIV_SHIFT) & PLL45XX_MDIV_MASK;
+ pdiv = (pll_con >> PLL45XX_PDIV_SHIFT) & PLL45XX_PDIV_MASK;
+ sdiv = (pll_con >> PLL45XX_SDIV_SHIFT) & PLL45XX_SDIV_MASK;
+
+ if (pll_type == pll_4508)
+ sdiv = sdiv - 1;
+
+ fvco *= mdiv;
+ do_div(fvco, (pdiv << sdiv));
+
+ return (unsigned long)fvco;
+}
+
+/* CON0 bit-fields */
+#define PLL46XX_MDIV_MASK (0x1FF)
+#define PLL46XX_PDIV_MASK (0x3F)
+#define PLL46XX_SDIV_MASK (0x7)
+#define PLL46XX_LOCKED_SHIFT (29)
+#define PLL46XX_MDIV_SHIFT (16)
+#define PLL46XX_PDIV_SHIFT (8)
+#define PLL46XX_SDIV_SHIFT (0)
+
+/* CON1 bit-fields */
+#define PLL46XX_MRR_MASK (0x1F)
+#define PLL46XX_MFR_MASK (0x3F)
+#define PLL46XX_KDIV_MASK (0xFFFF)
+#define PLL4650C_KDIV_MASK (0xFFF)
+#define PLL46XX_MRR_SHIFT (24)
+#define PLL46XX_MFR_SHIFT (16)
+#define PLL46XX_KDIV_SHIFT (0)
+
+enum pll46xx_type_t {
+ pll_4600,
+ pll_4650,
+ pll_4650c,
+};
+
+static inline unsigned long s5p_get_pll46xx(unsigned long baseclk,
+ u32 pll_con0, u32 pll_con1,
+ enum pll46xx_type_t pll_type)
+{
+ unsigned long result;
+ u32 mdiv, pdiv, sdiv, kdiv;
+ u64 tmp;
+
+ mdiv = (pll_con0 >> PLL46XX_MDIV_SHIFT) & PLL46XX_MDIV_MASK;
+ pdiv = (pll_con0 >> PLL46XX_PDIV_SHIFT) & PLL46XX_PDIV_MASK;
+ sdiv = (pll_con0 >> PLL46XX_SDIV_SHIFT) & PLL46XX_SDIV_MASK;
+ kdiv = pll_con1 & PLL46XX_KDIV_MASK;
+
+ if (pll_type == pll_4650c)
+ kdiv = pll_con1 & PLL4650C_KDIV_MASK;
+ else
+ kdiv = pll_con1 & PLL46XX_KDIV_MASK;
+
+ tmp = baseclk;
+
+ if (pll_type == pll_4600) {
+ tmp *= (mdiv << 16) + kdiv;
+ do_div(tmp, (pdiv << sdiv));
+ result = tmp >> 16;
+ } else {
+ tmp *= (mdiv << 10) + kdiv;
+ do_div(tmp, (pdiv << sdiv));
+ result = tmp >> 10;
+ }
+
+ return result;
+}
+
+#define PLL90XX_MDIV_MASK (0xFF)
+#define PLL90XX_PDIV_MASK (0x3F)
+#define PLL90XX_SDIV_MASK (0x7)
+#define PLL90XX_KDIV_MASK (0xffff)
+#define PLL90XX_LOCKED_SHIFT (29)
+#define PLL90XX_MDIV_SHIFT (16)
+#define PLL90XX_PDIV_SHIFT (8)
+#define PLL90XX_SDIV_SHIFT (0)
+#define PLL90XX_KDIV_SHIFT (0)
+
+static inline unsigned long s5p_get_pll90xx(unsigned long baseclk,
+ u32 pll_con, u32 pll_conk)
+{
+ unsigned long result;
+ u32 mdiv, pdiv, sdiv, kdiv;
+ u64 tmp;
+
+ mdiv = (pll_con >> PLL90XX_MDIV_SHIFT) & PLL90XX_MDIV_MASK;
+ pdiv = (pll_con >> PLL90XX_PDIV_SHIFT) & PLL90XX_PDIV_MASK;
+ sdiv = (pll_con >> PLL90XX_SDIV_SHIFT) & PLL90XX_SDIV_MASK;
+ kdiv = pll_conk & PLL90XX_KDIV_MASK;
+
+ /*
+ * We need to multiple baseclk by mdiv (the integer part) and kdiv
+ * which is in 2^16ths, so shift mdiv up (does not overflow) and
+ * add kdiv before multiplying. The use of tmp is to avoid any
+ * overflows before shifting bac down into result when multipling
+ * by the mdiv and kdiv pair.
+ */
+
+ tmp = baseclk;
+ tmp *= (mdiv << 16) + kdiv;
+ do_div(tmp, (pdiv << sdiv));
+ result = tmp >> 16;
+
+ return result;
+}
+
+#define PLL65XX_MDIV_MASK (0x3FF)
+#define PLL65XX_PDIV_MASK (0x3F)
+#define PLL65XX_SDIV_MASK (0x7)
+#define PLL65XX_MDIV_SHIFT (16)
+#define PLL65XX_PDIV_SHIFT (8)
+#define PLL65XX_SDIV_SHIFT (0)
+
+static inline unsigned long s5p_get_pll65xx(unsigned long baseclk, u32 pll_con)
+{
+ u32 mdiv, pdiv, sdiv;
+ u64 fvco = baseclk;
+
+ mdiv = (pll_con >> PLL65XX_MDIV_SHIFT) & PLL65XX_MDIV_MASK;
+ pdiv = (pll_con >> PLL65XX_PDIV_SHIFT) & PLL65XX_PDIV_MASK;
+ sdiv = (pll_con >> PLL65XX_SDIV_SHIFT) & PLL65XX_SDIV_MASK;
+
+ fvco *= mdiv;
+ do_div(fvco, (pdiv << sdiv));
+
+ return (unsigned long)fvco;
+}