+/* Coalesced MMIO regions are areas where write operations can be reordered.
+ * This usually implies that write operations are side-effect free. This allows
+ * batching which can make a major impact on performance when using
+ * virtualization.
+ */
+void qemu_register_coalesced_mmio(target_phys_addr_t addr, ram_addr_t size);
+
+void qemu_unregister_coalesced_mmio(target_phys_addr_t addr, ram_addr_t size);
+
+/*******************************************/
+/* host CPU ticks (if available) */
+
+#if defined(_ARCH_PPC)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ int64_t retval;
+#ifdef _ARCH_PPC64
+ /* This reads timebase in one 64bit go and includes Cell workaround from:
+ http://ozlabs.org/pipermail/linuxppc-dev/2006-October/027052.html
+ */
+ __asm__ __volatile__ (
+ "mftb %0\n\t"
+ "cmpwi %0,0\n\t"
+ "beq- $-8"
+ : "=r" (retval));
+#else
+ /* http://ozlabs.org/pipermail/linuxppc-dev/1999-October/003889.html */
+ unsigned long junk;
+ __asm__ __volatile__ (
+ "mftbu %1\n\t"
+ "mftb %L0\n\t"
+ "mftbu %0\n\t"
+ "cmpw %0,%1\n\t"
+ "bne $-16"
+ : "=r" (retval), "=r" (junk));
+#endif
+ return retval;
+}
+
+#elif defined(__i386__)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ int64_t val;
+ asm volatile ("rdtsc" : "=A" (val));
+ return val;
+}
+
+#elif defined(__x86_64__)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ uint32_t low,high;
+ int64_t val;
+ asm volatile("rdtsc" : "=a" (low), "=d" (high));
+ val = high;
+ val <<= 32;
+ val |= low;
+ return val;
+}
+
+#elif defined(__hppa__)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ int val;
+ asm volatile ("mfctl %%cr16, %0" : "=r"(val));
+ return val;
+}
+
+#elif defined(__ia64)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ int64_t val;
+ asm volatile ("mov %0 = ar.itc" : "=r"(val) :: "memory");
+ return val;
+}
+
+#elif defined(__s390__)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+ int64_t val;
+ asm volatile("stck 0(%1)" : "=m" (val) : "a" (&val) : "cc");
+ return val;
+}
+
+#elif defined(__sparc_v8plus__) || defined(__sparc_v8plusa__) || defined(__sparc_v9__)
+
+static inline int64_t cpu_get_real_ticks (void)
+{
+#if defined(_LP64)
+ uint64_t rval;
+ asm volatile("rd %%tick,%0" : "=r"(rval));
+ return rval;
+#else
+ union {
+ uint64_t i64;
+ struct {
+ uint32_t high;
+ uint32_t low;
+ } i32;
+ } rval;
+ asm volatile("rd %%tick,%1; srlx %1,32,%0"
+ : "=r"(rval.i32.high), "=r"(rval.i32.low));
+ return rval.i64;
+#endif
+}
+
+#elif defined(__mips__)
+
+static inline int64_t cpu_get_real_ticks(void)
+{
+#if __mips_isa_rev >= 2
+ uint32_t count;
+ static uint32_t cyc_per_count = 0;
+
+ if (!cyc_per_count)
+ __asm__ __volatile__("rdhwr %0, $3" : "=r" (cyc_per_count));
+
+ __asm__ __volatile__("rdhwr %1, $2" : "=r" (count));
+ return (int64_t)(count * cyc_per_count);
+#else
+ /* FIXME */
+ static int64_t ticks = 0;
+ return ticks++;
+#endif
+}
+
+#else
+/* The host CPU doesn't have an easily accessible cycle counter.
+ Just return a monotonically increasing value. This will be
+ totally wrong, but hopefully better than nothing. */
+static inline int64_t cpu_get_real_ticks (void)
+{
+ static int64_t ticks = 0;
+ return ticks++;
+}
+#endif
+
+/* profiling */
+#ifdef CONFIG_PROFILER
+static inline int64_t profile_getclock(void)
+{
+ return cpu_get_real_ticks();
+}
+
+extern int64_t kqemu_time, kqemu_time_start;
+extern int64_t qemu_time, qemu_time_start;
+extern int64_t tlb_flush_time;
+extern int64_t kqemu_exec_count;
+extern int64_t dev_time;
+extern int64_t kqemu_ret_int_count;
+extern int64_t kqemu_ret_excp_count;
+extern int64_t kqemu_ret_intr_count;
+#endif
+