Update asm to use the new format for inline assembly.

Also adds Arm support

https://github.com/rust-lang/rfcs/blob/master/text/2843-llvm-asm.md
rust-lang/rfcs#2873

src/cpucounter.c

@@ -4,14 +4,18 @@
 uint64_t cpucounter(void)
 {
     uint64_t low, high;
-    __asm__ __volatile__ ("rdtscp" : "=a" (low), "=d" (high) : : "%ecx");
+    __asm__ __volatile__("rdtscp"
+                         : "=a"(low), "=d"(high)
+                         :
+                         : "%ecx");
     return (high << 32) | low;
 }
 #elif defined(__aarch64__)
 uint64_t cpucounter(void)
 {
     uint64_t virtual_timer_value;
-    __asm__ __volatile__ ("mrs %0, cntvct_el0" : "=r"(virtual_timer_value));
+    __asm__ __volatile__("mrs %0, cntvct_el0"
+                         : "=r"(virtual_timer_value));
     return virtual_timer_value;
 }
 #endif

src/cpucounter.rs

@@ -6,18 +6,17 @@ pub(crate) struct CPUCounter;
 #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
 unsafe fn cpucounter() -> u64 {
     let (low, high): (u64, u64);
-    asm!("rdtscp" :  "={eax}" (low), "={edx}" (high) : : "ecx");
+    //asm!("rdtscp" :  "={eax}" (low), "={edx}" (high) : : "ecx");
+    asm!("rdtscp", out("rax") low, out("rdx") high, out("rcx") _);
     (high << 32) | low
 }
 
-
-// https://github.com/google/benchmark/blob/v1.1.0/src/cycleclock.h#L116
 #[cfg(asm)]
 #[inline]
 #[cfg(any(target_arch = "aarch64"))]
 unsafe fn cpucounter() -> u64 {
     let (vtm): (u64);
-    asm!("mrs %0, cntvct_el0" : "=r"(vtm));
+    asm!("mrs {}, cntvct_el0", out(reg) vtm);
     vtm
 }