From 35d75aeda29ede36e187d8b6f6fca0e1ed7569a8 Mon Sep 17 00:00:00 2001 From: Jonathan Peyton Date: Mon, 20 Mar 2017 22:11:31 +0000 Subject: Minor improvement of KMP_YIELD_NOW() macro. This change slightly improves performance of KMP_YIELD_NOW() macro, by using _rdtsc() intrinsic function if possible. Patch by Hansang Bae Differential Revision: https://reviews.llvm.org/D31008 llvm-svn: 298314 --- openmp/runtime/src/kmp.h | 12 ++++++++---- openmp/runtime/src/z_Linux_util.cpp | 6 +++--- 2 files changed, 11 insertions(+), 7 deletions(-) (limited to 'openmp/runtime/src') diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h index 885495218ea..c89506e9254 100644 --- a/openmp/runtime/src/kmp.h +++ b/openmp/runtime/src/kmp.h @@ -894,10 +894,14 @@ extern int __kmp_place_num_threads_per_core; #else # if KMP_OS_UNIX && (KMP_ARCH_X86 || KMP_ARCH_X86_64) // HW TSC is used to reduce overhead (clock tick instead of nanosecond). - extern double __kmp_ticks_per_nsec; -# define KMP_NOW() __kmp_hardware_timestamp() -# define KMP_NOW_MSEC() ((kmp_uint64)(KMP_NOW()/__kmp_ticks_per_nsec)/KMP_USEC_PER_SEC) -# define KMP_BLOCKTIME_INTERVAL() (__kmp_dflt_blocktime * KMP_USEC_PER_SEC * __kmp_ticks_per_nsec) + extern kmp_uint64 __kmp_ticks_per_msec; +# if KMP_COMPILER_ICC +# define KMP_NOW() _rdtsc() +# else +# define KMP_NOW() __kmp_hardware_timestamp() +# endif +# define KMP_NOW_MSEC() (KMP_NOW()/__kmp_ticks_per_msec) +# define KMP_BLOCKTIME_INTERVAL() (__kmp_dflt_blocktime * __kmp_ticks_per_msec) # define KMP_BLOCKING(goal, count) ((goal) > KMP_NOW()) # else // System time is retrieved sporadically while blocking. diff --git a/openmp/runtime/src/z_Linux_util.cpp b/openmp/runtime/src/z_Linux_util.cpp index 34114bcbb02..2c034f9c560 100644 --- a/openmp/runtime/src/z_Linux_util.cpp +++ b/openmp/runtime/src/z_Linux_util.cpp @@ -90,7 +90,7 @@ static pthread_mutexattr_t __kmp_suspend_mutex_attr; static kmp_cond_align_t __kmp_wait_cv; static kmp_mutex_align_t __kmp_wait_mx; -double __kmp_ticks_per_nsec; +kmp_uint64 __kmp_ticks_per_msec = 1000000; /* ------------------------------------------------------------------------ */ /* ------------------------------------------------------------------------ */ @@ -2157,7 +2157,7 @@ __kmp_now_nsec() } #if KMP_ARCH_X86 || KMP_ARCH_X86_64 -/* Measure clock tick per nanosecond */ +/* Measure clock ticks per millisecond */ void __kmp_initialize_system_tick() { @@ -2166,7 +2166,7 @@ __kmp_initialize_system_tick() kmp_uint64 goal = __kmp_hardware_timestamp() + delay; kmp_uint64 now; while ((now = __kmp_hardware_timestamp()) < goal); - __kmp_ticks_per_nsec = 1.0 * (delay + (now - goal)) / (__kmp_now_nsec() - nsec); + __kmp_ticks_per_msec = (kmp_uint64)(1e6 * (delay + (now - goal)) / (__kmp_now_nsec() - nsec)); } #endif -- cgit v1.2.3