From d1a6cb195f610978ba5d2351e60f938f7f261d59 Mon Sep 17 00:00:00 2001 From: Henrik Gramner Date: Wed, 8 Jul 2015 22:33:53 +0200 Subject: x86: Serialize rdtsc in read_time() Improves the accuracy of measurements, especially in short sections. To quote the Intel 64 and IA-32 Architectures Software Developer's Manual: "The RDTSC instruction is not a serializing instruction. It does not necessarily wait until all previous instructions have been executed before reading the counter. Similarly, subsequent instructions may begin execution before the read operation is performed. If software requires RDTSC to be executed only after all previous instructions have completed locally, it can either use RDTSCP (if the processor supports that instruction) or execute the sequence LFENCE;RDTSC." SSE2 is a requirement for lfence so only use it on SSE2-capable systems. Prefer lfence;rdtsc over rdtscp since rdtscp is supported on fewer systems. Signed-off-by: Luca Barbato --- libavutil/x86/timer.h | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'libavutil/x86') diff --git a/libavutil/x86/timer.h b/libavutil/x86/timer.h index cdd67ddbb5..bb7c341341 100644 --- a/libavutil/x86/timer.h +++ b/libavutil/x86/timer.h @@ -30,7 +30,12 @@ static inline uint64_t read_time(void) { uint32_t a, d; - __asm__ volatile("rdtsc" : "=a" (a), "=d" (d)); + __asm__ volatile( +#if ARCH_X86_64 || defined(__SSE2__) + "lfence \n\t" +#endif + "rdtsc \n\t" + : "=a" (a), "=d" (d)); return ((uint64_t)d << 32) + a; } -- cgit v1.2.3