At 2022-09-21 18:15:02, "王志强" <00107082@163.com> wrote:

But glibc profiling (total 29072 samples) is way much lighter, pagefault is the most cost while glibc spend significat time on "free"

pthread_attr_setschedparam?(82.021% 23845/29072)
asm_exc_page_fault(1.657% 395/23845)
_dl_catch_error?(16.714% 4859/29072) __libc_start_main(100.000% 4859/4859)
cfree(58.839% 2859/4859)
main(31.138% 1513/4859)
asm_exc_page_fault(2.115% 32/1513)

pthread_attr_setschedparam?(3.725% 181/4859)
random(2.099% 102/4859)
random_r(1.832% 89/4859)
__libc_malloc(1.420% 69/4859)

update: glibc spend most of its time within malloc, 82% samples, not the pagefault routine (only 1%), I misread the report in my last mail.

__libc_malloc?(82.195% 20700/25184)
asm_exc_page_fault(1.589% 329/20700)
_dl_catch_error??(16.717% 4210/25184) __libc_start_main(100.000% 4210/4210)
cfree(56.508% 2379/4210)
main(32.660% 1375/4210)
asm_exc_page_fault(2.327% 32/1375)

__libc_malloc?(4.418% 186/4210)
random(2.375% 100/4210)
random_r(1.805% 76/4210)
__libc_malloc(1.116% 47/4210)

David