This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / arch / m32r / lib / delay.c
diff --git a/arch/m32r/lib/delay.c b/arch/m32r/lib/delay.c
new file mode 100644 (file)
index 0000000..7afe66e
--- /dev/null
@@ -0,0 +1,126 @@
+/*
+ * linux/arch/m32r/lib/delay.c
+ *
+ * Copyright (c) 2002  Hitoshi Yamamoto, Hirokazu Takata
+ * Copyright (c) 2004  Hirokazu Takata
+ */
+
+/* $Id$ */
+
+#include <linux/config.h>
+#include <linux/param.h>
+#ifdef CONFIG_SMP
+#include <linux/sched.h>
+#include <asm/current.h>
+#include <asm/smp.h>
+#endif  /* CONFIG_SMP */
+#include <asm/processor.h>
+
+void __delay(unsigned long loops)
+{
+#ifdef CONFIG_ISA_DUAL_ISSUE
+       __asm__ __volatile__ (
+               "beqz   %0, 2f                  \n\t"
+               "addi   %0, #-1                 \n\t"
+
+               " .fillinsn                     \n\t"
+               "1:                             \n\t"
+               "cmpz   %0  ||  addi  %0, #-1   \n\t"
+               "bc     2f  ||  cmpz  %0        \n\t"
+               "bc     2f  ||  addi  %0, #-1   \n\t"
+               "cmpz   %0  ||  addi  %0, #-1   \n\t"
+               "bc     2f  ||  cmpz  %0        \n\t"
+               "bnc    1b  ||  addi  %0, #-1   \n\t"
+               " .fillinsn                     \n\t"
+               "2:                             \n\t"
+               : "+r" (loops)
+               : "r" (0)
+               : "cbit"
+       );
+#else
+       __asm__ __volatile__ (
+               "beqz   %0, 2f                  \n\t"
+               " .fillinsn                     \n\t"
+               "1:                             \n\t"
+               "addi   %0, #-1                 \n\t"
+               "blez   %0, 2f                  \n\t"
+               "addi   %0, #-1                 \n\t"
+               "blez   %0, 2f                  \n\t"
+               "addi   %0, #-1                 \n\t"
+               "blez   %0, 2f                  \n\t"
+               "addi   %0, #-1                 \n\t"
+               "bgtz   %0, 1b                  \n\t"
+               " .fillinsn                     \n\t"
+               "2:i                            \n\t"
+               : "+r" (loops)
+               : "r" (0)
+       );
+#endif
+}
+
+void __const_udelay(unsigned long xloops)
+{
+#if defined(CONFIG_ISA_M32R2) && defined(CONFIG_ISA_DSP_LEVEL2)
+       /*
+        * loops [1] = (xloops >> 32) [sec] * loops_per_jiffy [1/jiffy]
+        *            * HZ [jiffy/sec]
+        *          = (xloops >> 32) [sec] * (loops_per_jiffy * HZ) [1/sec]
+        *          = (((xloops * loops_per_jiffy) >> 32) * HZ) [1]
+        *
+        * NOTE:
+        *   - '[]' depicts variable's dimension in the above equation.
+        *   - "rac" instruction rounds the accumulator in word size.
+        */
+       __asm__ __volatile__ (
+               "srli   %0, #1                          \n\t"
+               "mulwhi %0, %1  ; a0                    \n\t"
+               "mulwu1 %0, %1  ; a1                    \n\t"
+               "sadd           ; a0 += (a1 >> 16)      \n\t"
+               "rac    a0, a0, #1                      \n\t"
+               "mvfacmi %0, a0                         \n\t"
+               : "+r" (xloops)
+               : "r" (current_cpu_data.loops_per_jiffy)
+               : "a0", "a1"
+       );
+#elif defined(CONFIG_ISA_M32R2) || defined(CONFIG_ISA_M32R)
+       /*
+        * u64 ull;
+        * ull = (u64)xloops * (u64)current_cpu_data.loops_per_jiffy;
+        * xloops = (ull >> 32);
+        */
+       __asm__ __volatile__ (
+               "and3   r4, %0, #0xffff         \n\t"
+               "and3   r5, %1, #0xffff         \n\t"
+               "mul    r4, r5                  \n\t"
+               "srl3   r6, %0, #16             \n\t"
+               "srli   r4, #16                 \n\t"
+               "mul    r5, r6                  \n\t"
+               "add    r4, r5                  \n\t"
+               "and3   r5, %0, #0xffff         \n\t"
+               "srl3   r6, %1, #16             \n\t"
+               "mul    r5, r6                  \n\t"
+               "add    r4, r5                  \n\t"
+               "srl3   r5, %0, #16             \n\t"
+               "srli   r4, #16                 \n\t"
+               "mul    r5, r6                  \n\t"
+               "add    r4, r5                  \n\t"
+               "mv     %0, r4                  \n\t"
+               : "+r" (xloops)
+               : "r" (current_cpu_data.loops_per_jiffy)
+               : "r4", "r5", "r6"
+       );
+#else
+#error unknown isa configuration
+#endif
+       __delay(xloops * HZ);
+}
+
+void __udelay(unsigned long usecs)
+{
+       __const_udelay(usecs * 0x000010c7);  /* 2**32 / 1000000 (rounded up) */
+}
+
+void __ndelay(unsigned long nsecs)
+{
+       __const_udelay(nsecs * 0x00005);  /* 2**32 / 1000000000 (rounded up) */
+}