Changes in / [2802824:33e62f1b]
- Files:
-
- 5 edited
-
benchmark/io/readv.cfa (modified) (3 diffs)
-
libcfa/src/concurrency/io.cfa (modified) (21 diffs)
-
libcfa/src/concurrency/kernel.cfa (modified) (1 diff)
-
libcfa/src/concurrency/kernel.hfa (modified) (2 diffs)
-
libcfa/src/concurrency/kernel_private.hfa (modified) (1 diff)
Legend:
- Unmodified
- Added
- Removed
-
benchmark/io/readv.cfa
r2802824 r33e62f1b 59 59 unsigned long int nthreads = 2; 60 60 unsigned long int nprocs = 1; 61 unsigned flags = 0; 62 unsigned sublen = 16; 61 int flags = 0; 63 62 64 63 arg_loop: 65 64 for(;;) { 66 65 static struct option options[] = { 67 {"duration", required_argument, 0, 'd'}, 68 {"nthreads", required_argument, 0, 't'}, 69 {"nprocs", required_argument, 0, 'p'}, 70 {"bufsize", required_argument, 0, 'b'}, 71 {"userthread", no_argument , 0, 'u'}, 72 {"submitthread", no_argument , 0, 's'}, 73 {"submitlength", required_argument, 0, 'l'}, 66 {"duration", required_argument, 0, 'd'}, 67 {"nthreads", required_argument, 0, 't'}, 68 {"nprocs", required_argument, 0, 'p'}, 69 {"bufsize", required_argument, 0, 'b'}, 70 {"userthread", no_argument , 0, 'u'}, 74 71 {0, 0, 0, 0} 75 72 }; 76 73 77 74 int idx = 0; 78 int opt = getopt_long(argc, argv, "d:t:p:b:u sl:", options, &idx);75 int opt = getopt_long(argc, argv, "d:t:p:b:u", options, &idx); 79 76 80 77 const char * arg = optarg ? optarg : ""; … … 116 113 flags |= CFA_CLUSTER_IO_POLLER_USER_THREAD; 117 114 break; 118 case 's':119 flags |= CFA_CLUSTER_IO_POLLER_THREAD_SUBMITS;120 break;121 case 'l':122 sublen = strtoul(arg, &end, 10);123 if(*end != '\0' && sublen < 16) {124 fprintf(stderr, "Submit length must be at least 16, was %s\n", arg);125 goto usage;126 }127 flags |= (sublen << CFA_CLUSTER_IO_BUFFLEN_OFFSET);128 break;129 115 // Other cases 130 116 default: /* ? */ … … 137 123 fprintf(stderr, " -p, --nprocs=NPROCS Number of kernel threads\n"); 138 124 fprintf(stderr, " -b, --buflen=SIZE Number of bytes to read per request\n"); 139 fprintf(stderr, " -u, --userthread If set, cluster uses user-thread to poll I/O\n");140 fprintf(stderr, " -s, --submitthread If set, cluster uses polling thread to submit I/O\n");141 125 exit(EXIT_FAILURE); 142 126 } -
libcfa/src/concurrency/io.cfa
r2802824 r33e62f1b 18 18 19 19 #include "kernel.hfa" 20 #include "bitmanip.hfa"21 20 22 21 #if !defined(HAVE_LINUX_IO_URING_H) 23 void __kernel_io_startup( cluster &, unsigned, bool ) {22 void __kernel_io_startup( cluster &, int, bool ) { 24 23 // Nothing to do without io_uring 25 24 } … … 92 91 struct __io_poller_fast { 93 92 struct __io_data * ring; 93 bool waiting; 94 94 $thread thrd; 95 95 }; … … 97 97 void ?{}( __io_poller_fast & this, struct cluster & cltr ) { 98 98 this.ring = cltr.io; 99 this.waiting = true; 99 100 (this.thrd){ "Fast I/O Poller", cltr }; 100 101 } … … 125 126 // Like head/tail but not seen by the kernel 126 127 volatile uint32_t alloc; 127 volatile uint32_t * ready; 128 uint32_t ready_cnt; 128 volatile uint32_t ready; 129 129 130 130 __spinlock_t lock; … … 145 145 volatile unsigned long long int block; 146 146 } submit_avg; 147 struct {148 volatile unsigned long long int val;149 volatile unsigned long long int cnt;150 volatile unsigned long long int block;151 } look_avg;152 147 } stats; 153 148 #endif … … 197 192 void * stack; 198 193 pthread_t kthrd; 199 volatile bool blocked;200 194 } slow; 201 195 __io_poller_fast fast; … … 207 201 // I/O Startup / Shutdown logic 208 202 //============================================================================================= 209 void __kernel_io_startup( cluster & this, unsignedio_flags, bool main_cluster ) {203 void __kernel_io_startup( cluster & this, int io_flags, bool main_cluster ) { 210 204 this.io = malloc(); 211 205 … … 280 274 sq.array = ( uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.array); 281 275 sq.alloc = *sq.tail; 282 283 if( io_flags & CFA_CLUSTER_IO_POLLER_THREAD_SUBMITS ) { 284 /* paranoid */ verify( is_pow2( io_flags >> CFA_CLUSTER_IO_BUFFLEN_OFFSET ) || ((io_flags >> CFA_CLUSTER_IO_BUFFLEN_OFFSET) < 8) ); 285 sq.ready_cnt = max(io_flags >> CFA_CLUSTER_IO_BUFFLEN_OFFSET, 8); 286 sq.ready = alloc_align( 64, sq.ready_cnt ); 287 for(i; sq.ready_cnt) { 288 sq.ready[i] = -1ul32; 289 } 290 } 291 else { 292 sq.ready_cnt = 0; 293 sq.ready = 0p; 294 } 276 sq.ready = *sq.tail; 295 277 296 278 // completion queue … … 325 307 this.io->submit_q.stats.submit_avg.cnt = 0; 326 308 this.io->submit_q.stats.submit_avg.block = 0; 327 this.io->submit_q.stats.look_avg.val = 0;328 this.io->submit_q.stats.look_avg.cnt = 0;329 this.io->submit_q.stats.look_avg.block = 0;330 309 this.io->completion_q.stats.completed_avg.val = 0; 331 310 this.io->completion_q.stats.completed_avg.slow_cnt = 0; … … 347 326 // Create the poller thread 348 327 __cfadbg_print_safe(io_core, "Kernel I/O : Creating slow poller for cluter %p\n", &this); 349 this.io->poller.slow.blocked = false;350 328 this.io->poller.slow.stack = __create_pthread( &this.io->poller.slow.kthrd, __io_poller_slow, &this ); 351 329 } … … 369 347 if( this.io->cltr_flags & CFA_CLUSTER_IO_POLLER_USER_THREAD ) { 370 348 with( this.io->poller.fast ) { 349 /* paranoid */ verify( waiting ); // The thread shouldn't be in a system call 371 350 /* paranoid */ verify( this.procs.head == 0p || &this == mainCluster ); 372 351 /* paranoid */ verify( this.idles.head == 0p || &this == mainCluster ); 373 352 374 353 // We need to adjust the clean-up based on where the thread is 375 if( thrd. state == Ready || thrd.preempted != __NO_PREEMPTION ) {354 if( thrd.preempted != __NO_PREEMPTION ) { 376 355 377 356 // This is the tricky case 378 357 // The thread was preempted and now it is on the ready queue 379 358 /* paranoid */ verify( thrd.state == Active ); // The thread better be in this state 380 359 /* paranoid */ verify( thrd.next != 0p ); // The thread should be the last on the list 381 360 /* paranoid */ verify( this.ready_queue.head == &thrd ); // The thread should be the only thing on the list … … 426 405 if(this.print_stats) { 427 406 with(this.io->submit_q.stats, this.io->completion_q.stats) { 428 double lavgv = 0; 429 double lavgb = 0; 430 if(look_avg.cnt != 0) { 431 lavgv = ((double)look_avg.val ) / look_avg.cnt; 432 lavgb = ((double)look_avg.block) / look_avg.cnt; 433 } 434 435 __cfaabi_bits_print_safe( STDOUT_FILENO, 407 __cfaabi_bits_print_safe( STDERR_FILENO, 436 408 "----- I/O uRing Stats -----\n" 437 "- total submit calls : %'15llu\n" 438 "- avg submit : %'18.2lf\n" 439 "- pre-submit block %% : %'18.2lf\n" 440 "- total ready search : %'15llu\n" 441 "- avg ready search len : %'18.2lf\n" 442 "- avg ready search block : %'18.2lf\n" 443 "- total wait calls : %'15llu (%'llu slow, %'llu fast)\n" 444 "- avg completion/wait : %'18.2lf\n", 409 "- total submit calls : %'15llu\n" 410 "- avg submit : %'18.2lf\n" 411 "- pre-submit block %% : %'18.2lf\n" 412 "- total wait calls : %'15llu (%'llu slow, %'llu fast)\n" 413 "- avg completion/wait : %'18.2lf\n", 445 414 submit_avg.cnt, 446 415 ((double)submit_avg.val) / submit_avg.cnt, 447 416 (100.0 * submit_avg.block) / submit_avg.cnt, 448 look_avg.cnt,449 lavgv,450 lavgb,451 417 completed_avg.slow_cnt + completed_avg.fast_cnt, 452 418 completed_avg.slow_cnt, completed_avg.fast_cnt, … … 475 441 close(this.io->fd); 476 442 477 free( this.io->submit_q.ready ); // Maybe null, doesn't matter478 443 free( this.io ); 479 444 } … … 489 454 // Process a single completion message from the io_uring 490 455 // This is NOT thread-safe 491 static [int, bool] __drain_io( & struct __io_data ring, * sigset_t mask, int waitcnt, bool in_kernel ) { 492 unsigned to_submit = 0; 493 if( ring.cltr_flags & CFA_CLUSTER_IO_POLLER_THREAD_SUBMITS ) { 494 495 // If the poller thread also submits, then we need to aggregate the submissions which are ready 496 uint32_t * tail = ring.submit_q.tail; 497 const uint32_t mask = *ring.submit_q.mask; 498 499 // Go through the list of ready submissions 500 for( i; ring.submit_q.ready_cnt ) { 501 // replace any submission with the sentinel, to consume it. 502 uint32_t idx = __atomic_exchange_n( &ring.submit_q.ready[i], -1ul32, __ATOMIC_RELAXED); 503 504 // If it was already the sentinel, then we are done 505 if( idx == -1ul32 ) continue; 506 507 // If we got a real submission, append it to the list 508 ring.submit_q.array[ ((*tail) + to_submit) & mask ] = idx & mask; 509 to_submit++; 510 } 511 512 // Increment the tail based on how many we are ready to submit 513 __atomic_fetch_add(tail, to_submit, __ATOMIC_SEQ_CST); 514 515 // update statistics 516 #if !defined(__CFA_NO_STATISTICS__) 517 ring.submit_q.stats.submit_avg.val += to_submit; 518 ring.submit_q.stats.submit_avg.cnt += 1; 519 #endif 520 } 521 522 int ret = syscall( __NR_io_uring_enter, ring.fd, to_submit, waitcnt, IORING_ENTER_GETEVENTS, mask, _NSIG / 8); 456 static int __drain_io( struct __io_data & ring, sigset_t * mask, int waitcnt, bool in_kernel ) { 457 int ret = syscall( __NR_io_uring_enter, ring.fd, 0, waitcnt, IORING_ENTER_GETEVENTS, mask, _NSIG / 8); 523 458 if( ret < 0 ) { 524 459 switch((int)errno) { … … 562 497 __atomic_fetch_add( ring.completion_q.head, count, __ATOMIC_RELAXED ); 563 498 564 return [count, count > 0 || to_submit > 0];499 return count; 565 500 } 566 501 … … 584 519 if( ring.cltr_flags & CFA_CLUSTER_IO_POLLER_USER_THREAD ) { 585 520 while(!__atomic_load_n(&ring.done, __ATOMIC_SEQ_CST)) { 586 587 __atomic_store_n( &ring.poller.slow.blocked, true, __ATOMIC_SEQ_CST );588 589 521 // In the user-thread approach drain and if anything was drained, 590 522 // batton pass to the user-thread 591 int count; 592 bool again; 593 [count, again] = __drain_io( ring, &mask, 1, true ); 594 595 __atomic_store_n( &ring.poller.slow.blocked, false, __ATOMIC_SEQ_CST ); 523 int count = __drain_io( ring, &mask, 1, true ); 596 524 597 525 // Update statistics … … 601 529 #endif 602 530 603 if( again) {531 if(count > 0) { 604 532 __cfadbg_print_safe(io_core, "Kernel I/O : Moving to ring %p to fast poller\n", &ring); 605 533 __unpark( &ring.poller.fast.thrd __cfaabi_dbg_ctx2 ); … … 611 539 while(!__atomic_load_n(&ring.done, __ATOMIC_SEQ_CST)) { 612 540 //In the naive approach, just poll the io completion queue directly 613 int count; 614 bool again; 615 [count, again] = __drain_io( ring, &mask, 1, true ); 541 int count = __drain_io( ring, &mask, 1, true ); 616 542 617 543 // Update statistics … … 640 566 // Then loop until we need to start 641 567 while(!__atomic_load_n(&this.ring->done, __ATOMIC_SEQ_CST)) { 642 643 568 // Drain the io 644 int count; 645 bool again; 646 [count, again] = __drain_io( *this.ring, 0p, 0, false ); 647 648 if(!again) reset++; 569 this.waiting = false; 570 int count = __drain_io( *this.ring, 0p, 0, false ); 571 reset += count > 0 ? 1 : 0; 649 572 650 573 // Update statistics … … 654 577 #endif 655 578 656 // If we got something, just yield and check again579 this.waiting = true; 657 580 if(reset < 5) { 581 // If we got something, just yield and check again 658 582 yield(); 659 583 } 660 // We didn't get anything baton pass to the slow poller661 584 else { 585 // We didn't get anything baton pass to the slow poller 662 586 __cfadbg_print_safe(io_core, "Kernel I/O : Moving to ring %p to slow poller\n", &this.ring); 587 post( this.ring->poller.sem ); 588 park( __cfaabi_dbg_ctx ); 663 589 reset = 0; 664 665 // wake up the slow poller666 post( this.ring->poller.sem );667 668 // park this thread669 park( __cfaabi_dbg_ctx );670 590 } 671 591 } 672 592 673 593 __cfadbg_print_safe(io_core, "Kernel I/O : Fast poller for ring %p stopping\n", &this.ring); 674 }675 676 static inline void __wake_poller( struct __io_data & ring ) __attribute__((artificial));677 static inline void __wake_poller( struct __io_data & ring ) {678 if(!__atomic_load_n( &ring.poller.slow.blocked, __ATOMIC_SEQ_CST)) return;679 680 sigval val = { 1 };681 pthread_sigqueue( ring.poller.slow.kthrd, SIGUSR1, val );682 594 } 683 595 … … 720 632 uint32_t idx = __atomic_fetch_add(&ring.submit_q.alloc, 1ul32, __ATOMIC_SEQ_CST); 721 633 722 // Mask the idx now to allow make everything easier to check 723 idx &= *ring.submit_q.mask; 634 // Validate that we didn't overflow anything 635 // Check that nothing overflowed 636 /* paranoid */ verify( true ); 637 638 // Check that it goes head -> tail -> alloc and never head -> alloc -> tail 639 /* paranoid */ verify( true ); 724 640 725 641 // Return the sqe 726 return [&ring.submit_q.sqes[ idx ], idx];642 return [&ring.submit_q.sqes[ idx & (*ring.submit_q.mask)], idx]; 727 643 } 728 644 729 645 static inline void __submit( struct __io_data & ring, uint32_t idx ) { 730 // Get now the data we definetely need 731 uint32_t * const tail = ring.submit_q.tail; 646 // get mutual exclusion 647 lock(ring.submit_q.lock __cfaabi_dbg_ctx2); 648 649 // Append to the list of ready entries 650 uint32_t * tail = ring.submit_q.tail; 732 651 const uint32_t mask = *ring.submit_q.mask; 733 652 734 // There are 2 submission schemes, check which one we are using 735 if( ring.cltr_flags & CFA_CLUSTER_IO_POLLER_THREAD_SUBMITS ) { 736 // If the poller thread submits, then we just need to add this to the ready array 737 738 /* paranoid */ verify( idx <= mask ); 739 /* paranoid */ verify( idx != -1ul32 ); 740 741 // We need to find a spot in the ready array 742 __attribute((unused)) int len = 0; 743 __attribute((unused)) int block = 0; 744 uint32_t expected = -1ul32; 745 uint32_t ready_mask = ring.submit_q.ready_cnt - 1; 746 uint32_t off = __tls_rand(); 747 LOOKING: for() { 748 for(i; ring.submit_q.ready_cnt) { 749 uint32_t ii = (i + off) & ready_mask; 750 if( __atomic_compare_exchange_n( &ring.submit_q.ready[ii], &expected, idx, true, __ATOMIC_SEQ_CST, __ATOMIC_RELAXED ) ) { 751 break LOOKING; 752 } 753 754 len ++; 755 } 756 757 block++; 758 yield(); 653 ring.submit_q.array[ (*tail) & mask ] = idx & mask; 654 __atomic_fetch_add(tail, 1ul32, __ATOMIC_SEQ_CST); 655 656 // Submit however, many entries need to be submitted 657 int ret = syscall( __NR_io_uring_enter, ring.fd, 1, 0, 0, 0p, 0); 658 if( ret < 0 ) { 659 switch((int)errno) { 660 default: 661 abort( "KERNEL ERROR: IO_URING SUBMIT - %s\n", strerror(errno) ); 759 662 } 760 761 __wake_poller( ring ); 762 763 // update statistics 764 #if !defined(__CFA_NO_STATISTICS__) 765 __atomic_fetch_add( &ring.submit_q.stats.look_avg.val, len, __ATOMIC_RELAXED ); 766 __atomic_fetch_add( &ring.submit_q.stats.look_avg.block, block, __ATOMIC_RELAXED ); 767 __atomic_fetch_add( &ring.submit_q.stats.look_avg.cnt, 1, __ATOMIC_RELAXED ); 768 #endif 769 770 __cfadbg_print_safe( io, "Kernel I/O : Added %u to ready for %p\n", idx, active_thread() ); 771 } 772 else { 773 // get mutual exclusion 774 lock(ring.submit_q.lock __cfaabi_dbg_ctx2); 775 776 // Append to the list of ready entries 777 778 /* paranoid */ verify( idx <= mask ); 779 780 ring.submit_q.array[ (*tail) & mask ] = idx & mask; 781 __atomic_fetch_add(tail, 1ul32, __ATOMIC_SEQ_CST); 782 783 // Submit however, many entries need to be submitted 784 int ret = syscall( __NR_io_uring_enter, ring.fd, 1, 0, 0, 0p, 0); 785 if( ret < 0 ) { 786 switch((int)errno) { 787 default: 788 abort( "KERNEL ERROR: IO_URING SUBMIT - %s\n", strerror(errno) ); 789 } 790 } 791 792 // update statistics 793 #if !defined(__CFA_NO_STATISTICS__) 794 ring.submit_q.stats.submit_avg.val += 1; 795 ring.submit_q.stats.submit_avg.cnt += 1; 796 #endif 797 798 unlock(ring.submit_q.lock); 799 800 __cfadbg_print_safe( io, "Kernel I/O : Performed io_submit for %p, returned %d\n", active_thread(), ret ); 801 } 663 } 664 665 // update statistics 666 #if !defined(__CFA_NO_STATISTICS__) 667 ring.submit_q.stats.submit_avg.val += 1; 668 ring.submit_q.stats.submit_avg.cnt += 1; 669 #endif 670 671 unlock(ring.submit_q.lock); 672 // Make sure that idx was submitted 673 // Be careful to not get false positive if we cycled the entire list or that someone else submitted for us 674 __cfadbg_print_safe( io, "Kernel I/O : Performed io_submit for %p, returned %d\n", active_thread(), ret ); 802 675 } 803 676 -
libcfa/src/concurrency/kernel.cfa
r2802824 r33e62f1b 258 258 } 259 259 260 void ?{}(cluster & this, const char name[], Duration preemption_rate, unsignedio_flags) with( this ) {260 void ?{}(cluster & this, const char name[], Duration preemption_rate, int io_flags) with( this ) { 261 261 this.name = name; 262 262 this.preemption_rate = preemption_rate; -
libcfa/src/concurrency/kernel.hfa
r2802824 r33e62f1b 117 117 struct __io_data; 118 118 119 #define CFA_CLUSTER_IO_POLLER_USER_THREAD 1 << 0 // 0x1 120 #define CFA_CLUSTER_IO_POLLER_THREAD_SUBMITS 1 << 1 // 0x2 121 // #define CFA_CLUSTER_IO_POLLER_KERNEL_SIDE 1 << 2 // 0x4 122 #define CFA_CLUSTER_IO_BUFFLEN_OFFSET 16 119 #define CFA_CLUSTER_IO_POLLER_USER_THREAD 1 << 0 120 // #define CFA_CLUSTER_IO_POLLER_KERNEL_SIDE 1 << 1 123 121 124 122 … … 312 310 extern Duration default_preemption(); 313 311 314 void ?{} (cluster & this, const char name[], Duration preemption_rate, unsignedflags);312 void ?{} (cluster & this, const char name[], Duration preemption_rate, int flags); 315 313 void ^?{}(cluster & this); 316 314 317 static inline void ?{} (cluster & this) 318 static inline void ?{} (cluster & this, Duration preemption_rate) 319 static inline void ?{} (cluster & this, const char name[]) 320 static inline void ?{} (cluster & this, unsignedflags) { this{"Anonymous Cluster", default_preemption(), flags}; }321 static inline void ?{} (cluster & this, Duration preemption_rate, unsignedflags) { this{"Anonymous Cluster", preemption_rate, flags}; }322 static inline void ?{} (cluster & this, const char name[], unsignedflags) { this{name, default_preemption(), flags}; }315 static inline void ?{} (cluster & this) { this{"Anonymous Cluster", default_preemption(), 0}; } 316 static inline void ?{} (cluster & this, Duration preemption_rate) { this{"Anonymous Cluster", preemption_rate, 0}; } 317 static inline void ?{} (cluster & this, const char name[]) { this{name, default_preemption(), 0}; } 318 static inline void ?{} (cluster & this, int flags) { this{"Anonymous Cluster", default_preemption(), flags}; } 319 static inline void ?{} (cluster & this, Duration preemption_rate, int flags) { this{"Anonymous Cluster", preemption_rate, flags}; } 320 static inline void ?{} (cluster & this, const char name[], int flags) { this{name, default_preemption(), flags}; } 323 321 324 322 static inline [cluster *&, cluster *& ] __get( cluster & this ) __attribute__((const)) { return this.node.[next, prev]; } -
libcfa/src/concurrency/kernel_private.hfa
r2802824 r33e62f1b 77 77 //----------------------------------------------------------------------------- 78 78 // I/O 79 void __kernel_io_startup ( cluster &, unsigned, bool );79 void __kernel_io_startup ( cluster &, int, bool ); 80 80 void __kernel_io_finish_start( cluster & ); 81 81 void __kernel_io_prepare_stop( cluster & );
Note:
See TracChangeset
for help on using the changeset viewer.
