aboutsummaryrefslogtreecommitdiffstats
path: root/arch
diff options
context:
space:
mode:
authorHiro Yoshioka <hyoshiok@miraclelinux.com>2006-06-23 05:04:16 -0400
committerLinus Torvalds <torvalds@g5.osdl.org>2006-06-23 10:42:56 -0400
commitc22ce143d15eb288543fe9873e1c5ac1c01b69a1 (patch)
treedc7d457b8952fc50dfc90df659b35de4117c61fc /arch
parent7dbdf43cfa635ddc3701cc8d1eab07597cd731c0 (diff)
[PATCH] x86: cache pollution aware __copy_from_user_ll()
Use the x86 cache-bypassing copy instructions for copy_from_user(). Some performance data are Total of GLOBAL_POWER_EVENTS (CPU cycle samples) 2.6.12.4.orig 1921587 2.6.12.4.nt 1599424 1599424/1921587=83.23% (16.77% reduction) BSQ_CACHE_REFERENCE (L3 cache miss) 2.6.12.4.orig 57427 2.6.12.4.nt 20858 20858/57427=36.32% (63.7% reduction) L3 cache miss reduction of __copy_from_user_ll samples % 37408 65.1412 vmlinux __copy_from_user_ll 23 0.1103 vmlinux __copy_user_zeroing_intel_nocache 23/37408=0.061% (99.94% reduction) Top 5 of 2.6.12.4.nt Counted GLOBAL_POWER_EVENTS events (time during which processor is not stopped) with a unit mask of 0x01 (mandatory) count 100000 samples % app name symbol name 128392 8.0274 vmlinux __copy_user_zeroing_intel_nocache 64206 4.0143 vmlinux journal_add_journal_head 59746 3.7355 vmlinux do_get_write_access 47674 2.9807 vmlinux journal_put_journal_head 46021 2.8774 vmlinux journal_dirty_metadata pattern9-0-cpu4-0-09011728/summary.out Counted BSQ_CACHE_REFERENCE events (cache references seen by the bus unit) with a unit mask of 0x3f (multiple flags) count 3000 samples % app name symbol name 69755 4.2861 vmlinux __copy_user_zeroing_intel_nocache 55685 3.4215 vmlinux journal_add_journal_head 52371 3.2179 vmlinux __find_get_block 45504 2.7960 vmlinux journal_put_journal_head 36005 2.2123 vmlinux journal_stop pattern9-0-cpu4-0-09011744/summary.out Counted BSQ_CACHE_REFERENCE events (cache references seen by the bus unit) with a unit mask of 0x200 (read 3rd level cache miss) count 3000 samples % app name symbol name 1147 5.4994 vmlinux journal_add_journal_head 881 4.2240 vmlinux journal_dirty_data 872 4.1809 vmlinux blk_rq_map_sg 734 3.5192 vmlinux journal_commit_transaction 617 2.9582 vmlinux radix_tree_delete pattern9-0-cpu4-0-09011731/summary.out iozone results are original 2.6.12.4 CPU time = 207.768 sec cache aware CPU time = 184.783 sec (three times run) 184.783/207.768=88.94% (11.06% reduction) original: pattern9-0-cpu4-0-08191720/iozone.out: CPU Utilization: Wall time 45.997 CPU time 64.527 CPU utilization 140.28 % pattern9-0-cpu4-0-08191741/iozone.out: CPU Utilization: Wall time 46.878 CPU time 71.933 CPU utilization 153.45 % pattern9-0-cpu4-0-08191743/iozone.out: CPU Utilization: Wall time 45.152 CPU time 71.308 CPU utilization 157.93 % cache awre: pattern9-0-cpu4-0-09011728/iozone.out: CPU Utilization: Wall time 44.842 CPU time 62.465 CPU utilization 139.30 % pattern9-0-cpu4-0-09011731/iozone.out: CPU Utilization: Wall time 44.718 CPU time 59.273 CPU utilization 132.55 % pattern9-0-cpu4-0-09011744/iozone.out: CPU Utilization: Wall time 44.367 CPU time 63.045 CPU utilization 142.10 % Signed-off-by: Hiro Yoshioka <hyoshiok@miraclelinux.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'arch')
-rw-r--r--arch/i386/lib/usercopy.c137
1 files changed, 129 insertions, 8 deletions
diff --git a/arch/i386/lib/usercopy.c b/arch/i386/lib/usercopy.c
index 4cf981d70f45..6979297ce278 100644
--- a/arch/i386/lib/usercopy.c
+++ b/arch/i386/lib/usercopy.c
@@ -425,15 +425,121 @@ __copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size)
425 : "eax", "edx", "memory"); 425 : "eax", "edx", "memory");
426 return size; 426 return size;
427} 427}
428
429/*
430 * Non Temporal Hint version of __copy_user_zeroing_intel. It is cache aware.
431 * hyoshiok@miraclelinux.com
432 */
433
434static unsigned long __copy_user_zeroing_intel_nocache(void *to,
435 const void __user *from, unsigned long size)
436{
437 int d0, d1;
438
439 __asm__ __volatile__(
440 " .align 2,0x90\n"
441 "0: movl 32(%4), %%eax\n"
442 " cmpl $67, %0\n"
443 " jbe 2f\n"
444 "1: movl 64(%4), %%eax\n"
445 " .align 2,0x90\n"
446 "2: movl 0(%4), %%eax\n"
447 "21: movl 4(%4), %%edx\n"
448 " movnti %%eax, 0(%3)\n"
449 " movnti %%edx, 4(%3)\n"
450 "3: movl 8(%4), %%eax\n"
451 "31: movl 12(%4),%%edx\n"
452 " movnti %%eax, 8(%3)\n"
453 " movnti %%edx, 12(%3)\n"
454 "4: movl 16(%4), %%eax\n"
455 "41: movl 20(%4), %%edx\n"
456 " movnti %%eax, 16(%3)\n"
457 " movnti %%edx, 20(%3)\n"
458 "10: movl 24(%4), %%eax\n"
459 "51: movl 28(%4), %%edx\n"
460 " movnti %%eax, 24(%3)\n"
461 " movnti %%edx, 28(%3)\n"
462 "11: movl 32(%4), %%eax\n"
463 "61: movl 36(%4), %%edx\n"
464 " movnti %%eax, 32(%3)\n"
465 " movnti %%edx, 36(%3)\n"
466 "12: movl 40(%4), %%eax\n"
467 "71: movl 44(%4), %%edx\n"
468 " movnti %%eax, 40(%3)\n"
469 " movnti %%edx, 44(%3)\n"
470 "13: movl 48(%4), %%eax\n"
471 "81: movl 52(%4), %%edx\n"
472 " movnti %%eax, 48(%3)\n"
473 " movnti %%edx, 52(%3)\n"
474 "14: movl 56(%4), %%eax\n"
475 "91: movl 60(%4), %%edx\n"
476 " movnti %%eax, 56(%3)\n"
477 " movnti %%edx, 60(%3)\n"
478 " addl $-64, %0\n"
479 " addl $64, %4\n"
480 " addl $64, %3\n"
481 " cmpl $63, %0\n"
482 " ja 0b\n"
483 " sfence \n"
484 "5: movl %0, %%eax\n"
485 " shrl $2, %0\n"
486 " andl $3, %%eax\n"
487 " cld\n"
488 "6: rep; movsl\n"
489 " movl %%eax,%0\n"
490 "7: rep; movsb\n"
491 "8:\n"
492 ".section .fixup,\"ax\"\n"
493 "9: lea 0(%%eax,%0,4),%0\n"
494 "16: pushl %0\n"
495 " pushl %%eax\n"
496 " xorl %%eax,%%eax\n"
497 " rep; stosb\n"
498 " popl %%eax\n"
499 " popl %0\n"
500 " jmp 8b\n"
501 ".previous\n"
502 ".section __ex_table,\"a\"\n"
503 " .align 4\n"
504 " .long 0b,16b\n"
505 " .long 1b,16b\n"
506 " .long 2b,16b\n"
507 " .long 21b,16b\n"
508 " .long 3b,16b\n"
509 " .long 31b,16b\n"
510 " .long 4b,16b\n"
511 " .long 41b,16b\n"
512 " .long 10b,16b\n"
513 " .long 51b,16b\n"
514 " .long 11b,16b\n"
515 " .long 61b,16b\n"
516 " .long 12b,16b\n"
517 " .long 71b,16b\n"
518 " .long 13b,16b\n"
519 " .long 81b,16b\n"
520 " .long 14b,16b\n"
521 " .long 91b,16b\n"
522 " .long 6b,9b\n"
523 " .long 7b,16b\n"
524 ".previous"
525 : "=&c"(size), "=&D" (d0), "=&S" (d1)
526 : "1"(to), "2"(from), "0"(size)
527 : "eax", "edx", "memory");
528 return size;
529}
530
428#else 531#else
532
429/* 533/*
430 * Leave these declared but undefined. They should not be any references to 534 * Leave these declared but undefined. They should not be any references to
431 * them 535 * them
432 */ 536 */
433unsigned long 537unsigned long __copy_user_zeroing_intel(void *to, const void __user *from,
434__copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size); 538 unsigned long size);
435unsigned long 539unsigned long __copy_user_intel(void __user *to, const void *from,
436__copy_user_intel(void __user *to, const void *from, unsigned long size); 540 unsigned long size);
541unsigned long __copy_user_zeroing_intel_nocache(void *to,
542 const void __user *from, unsigned long size);
437#endif /* CONFIG_X86_INTEL_USERCOPY */ 543#endif /* CONFIG_X86_INTEL_USERCOPY */
438 544
439/* Generic arbitrary sized copy. */ 545/* Generic arbitrary sized copy. */
@@ -515,8 +621,8 @@ do { \
515 : "memory"); \ 621 : "memory"); \
516} while (0) 622} while (0)
517 623
518 624unsigned long __copy_to_user_ll(void __user *to, const void *from,
519unsigned long __copy_to_user_ll(void __user *to, const void *from, unsigned long n) 625 unsigned long n)
520{ 626{
521 BUG_ON((long) n < 0); 627 BUG_ON((long) n < 0);
522#ifndef CONFIG_X86_WP_WORKS_OK 628#ifndef CONFIG_X86_WP_WORKS_OK
@@ -576,8 +682,8 @@ survive:
576} 682}
577EXPORT_SYMBOL(__copy_to_user_ll); 683EXPORT_SYMBOL(__copy_to_user_ll);
578 684
579unsigned long 685unsigned long __copy_from_user_ll(void *to, const void __user *from,
580__copy_from_user_ll(void *to, const void __user *from, unsigned long n) 686 unsigned long n)
581{ 687{
582 BUG_ON((long)n < 0); 688 BUG_ON((long)n < 0);
583 if (movsl_is_ok(to, from, n)) 689 if (movsl_is_ok(to, from, n))
@@ -588,6 +694,21 @@ __copy_from_user_ll(void *to, const void __user *from, unsigned long n)
588} 694}
589EXPORT_SYMBOL(__copy_from_user_ll); 695EXPORT_SYMBOL(__copy_from_user_ll);
590 696
697unsigned long __copy_from_user_ll_nocache(void *to, const void __user *from,
698 unsigned long n)
699{
700 BUG_ON((long)n < 0);
701#ifdef CONFIG_X86_INTEL_USERCOPY
702 if ( n > 64 && cpu_has_xmm2)
703 n = __copy_user_zeroing_intel_nocache(to, from, n);
704 else
705 __copy_user_zeroing(to, from, n);
706#else
707 __copy_user_zeroing(to, from, n);
708#endif
709 return n;
710}
711
591/** 712/**
592 * copy_to_user: - Copy a block of data into user space. 713 * copy_to_user: - Copy a block of data into user space.
593 * @to: Destination address, in user space. 714 * @to: Destination address, in user space.
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-19 12:20:38 -0500