site_archiver@lists.apple.com Delivered-To: darwin-dev@lists.apple.com User-agent: Microsoft-Entourage/10.1.1.2418 Folk: To make a *very* long story as short as I can, I am running into a performance issue that has me scratching my head a little bit (and might save somebody else some time in the future when list searching reappears). The issue: when journaling is on, I see poor fread() performance. Shark: all time is spent in semaphores in the kernel. poor == 10x slower. The code in question (see below) creates the following output: *** Moe:~/proj/journal pdc$ time ./tj /{pathcut}/test.data file size: 1984 kb (1.9375 meg) total reads: 2031616 real 0m4.880s user 0m0.120s sys 0m1.130s *** Do nothing else but this, using Disk Utility: Turn Journaling Off on data file volume. Unmount Volume. Mount Volume. *** Moe:~/proj/journal pdc$ time ./tj /{pathcut}test.data file size: 1984 kb (1.9375 meg) total reads: 2031616 real 0m0.413s user 0m0.070s sys 0m0.240s *** 4.9s vs 0.4s And this extra 4.4s is not a constant time -- it's at least linear. So, what gives? Is this normal behaviour? Any Ideas? Yes, I have re-implemented my app to use mmap(), but issues I've run into with that are forcing me to re-evaluate the original issue: slow fread() when Journaling is on. Thanks for any ideas and sorry so verbose, -pdc --- snip --- /* compiled with "cc -o jt jt.c" */ #include <stdio.h> static void scan(const char *fname); int main(int argc, char *argv[]) { if (argc != 2) { printf("Usage: %s file\n", argv[0]); exit(1); } scan(argv[1]); return 0; } #define READ_SIZE 4096 void scan(const char *fname) { FILE *fptr = fopen(fname, "rb"); long size = 0; char buff[READ_SIZE]; long idx; if (!fptr) { printf("can't open '%s'\n", fname); exit(1); } fseek(fptr, 0, SEEK_END); size = ftell(fptr); fseek(fptr, 0, SEEK_SET); printf("file size:\t%ld kb (%g meg)\n", size/1024, (float)size/(1024.0*1024.0)); for (idx = 0; idx < size/128 ; idx++) { long pos = rand() % (size - READ_SIZE); pos &= 0xFFFF1000; // benefit of the doubt... fseek(fptr, pos, SEEK_SET); if (fread(buff, READ_SIZE, 1, fptr) != 1) { printf("cant read at #%ld (pos: %ld)\n", idx, pos); exit(1); } } printf("total reads:\t%ld\n", size); fclose(fptr); } _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.appl... This email sent to site_archiver@lists.apple.com