/* * Dummy soundcard for virtual rawmidi devices * * Copyright (c) 2000 by Takashi Iwai * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* * VIRTUAL RAW MIDI DEVICE CARDS * * This dummy card contains up to 4 virtual rawmidi devices. * They are not real rawmidi devices but just associated with sequencer * clients, so that any input/output sources can be connected as a raw * MIDI device arbitrary. * Also, multiple access is allowed to a single rawmidi device. * * Typical usage is like following: * - Load snd-virmidi module. * # modprobe snd-virmidi index=2 * Then, sequencer clients 72:0 to 75:0 will be created, which are * mapped from /dev/snd/midiC1D0 to /dev/snd/midiC1D3, respectively. * * - Connect input/output via aconnect. * % aconnect 64:0 72:0 # keyboard input redirection 64:0 -> 72:0 * % aconnect 72:0 65:0 # output device redirection 72:0 -> 65:0 * * - Run application using a midi device (eg. /dev/snd/midiC1D0) */ #include #include #include #include #include #include #include #include #include /* hack: OSS defines midi_devs, so undefine it (versioned symbols) */ #undef midi_devs MODULE_AUTHOR("Takashi Iwai "); MODULE_DESCRIPTION("Dummy soundcard for virtual rawmidi devices"); MODULE_LICENSE("GPL"); MODULE_SUPPORTED_DEVICE("{{ALSA,Virtual rawmidi device}}"); #define MAX_MIDI_DEVICES 4 static int index[SNDRV_CARDS] = SNDRV_DEFAULT_IDX; /* Index 0-MAX */ static char *id[SNDRV_CARDS] = SNDRV_DEFAULT_STR; /* ID for this card */ static bool enable[SNDRV_CARDS] = {1, [1 ... (SNDRV_CARDS - 1)] = 0}; static int midi_devs[SNDRV_CARDS] = {[0 ... (SNDRV_CARDS - 1)] = 4}; module_param_array(index, int, NULL, 0444); MODULE_PARM_DESC(index, "Index value for virmidi soundcard."); module_param_array(id, charp, NULL, 0444); MODULE_PARM_DESC(id, "ID string for virmidi soundcard."); module_param_array(enable, bool, NULL, 0444); MODULE_PARM_DESC(enable, "Enable this soundcard."); module_param_array(midi_devs, int, NULL, 0444); MODULE_PARM_DESC(midi_devs, "MIDI devices # (1-4)"); struct snd_card_virmidi { struct snd_card *card; struct snd_rawmidi *midi[MAX_MIDI_DEVICES]; }; static struct platform_device *devices[SNDRV_CARDS]; static int snd_virmidi_probe(struct platform_device *devptr) { struct snd_card *card; struct snd_card_virmidi *vmidi; int idx, err; int dev = devptr->id; err = snd_card_new(&devptr->dev, index[dev], id[dev], THIS_MODULE, sizeof(struct snd_card_virmidi), &card); if (err < 0) return err; vmidi = card->private_data; vmidi->card = card; if (midi_devs[dev] > MAX_MIDI_DEVICES) { snd_printk(KERN_WARNING "too much midi devices for virmidi %d: force to use %d\n", dev, MAX_MIDI_DEVICES); midi_devs[dev] = MAX_MIDI_DEVICES; } for (idx = 0; idx < midi_devs[dev]; idx++) { struct snd_rawmidi *rmidi; struct snd_virmidi_dev *rdev; err = snd_virmidi_new(card, idx, &rmidi); if (err < 0) goto __nodev; rdev = rmidi->private_data; vmidi->midi[idx] = rmidi; strcpy(rmidi->name, "Virtual Raw MIDI"); rdev->seq_mode = SNDRV_VIRMIDI_SEQ_DISPATCH; } strcpy(card->driver, "VirMIDI"); strcpy(card->shortname, "VirMIDI"); sprintf(card->longname, "Virtual MIDI Card %i", dev + 1); err = snd_card_register(card); if (!err) { platform_set_drvdata(devptr, card); return 0; } __nodev: snd_card_free(card); return err; } static int snd_virmidi_remove(struct platform_device *devptr) { snd_card_free(platform_get_drvdata(devptr)); return 0; } #define SND_VIRMIDI_DRIVER "snd_virmidi" static struct platform_driver snd_virmidi_driver = { .probe = snd_virmidi_probe, .remove = snd_virmidi_remove, .driver = { .name = SND_VIRMIDI_DRIVER, }, }; static void snd_virmidi_unregister_all(void) { int i; for (i = 0; i < ARRAY_SIZE(devices); ++i) platform_device_unregister(devices[i]); platform_driver_unregister(&snd_virmidi_driver); } static int __init alsa_card_virmidi_init(void) { int i, cards, err; err = platform_driver_register(&snd_virmidi_driver); if (err < 0) return err; cards = 0; for (i = 0; i < SNDRV_CARDS; i++) { struct platform_device *device; if (!enable[i]) continue; device = platform_device_register_simple(SND_VIRMIDI_DRIVER, i, NULL, 0); if (IS_ERR(device)) continue; if (!platform_get_drvdata(device)) { platform_device_unregister(device); continue; } devices[i] = device; cards++; } if (!cards) { #ifdef MODULE printk(KERN_ERR "Card-VirMIDI soundcard not found or device busy\n"); #endif snd_virmidi_unregister_all(); return -ENODEV; } return 0; } static void __exit alsa_card_virmidi_exit(void) { snd_virmidi_unregister_all(); } module_init(alsa_card_virmidi_init) module_exit(alsa_card_virmidi_exit) cf84fd1ec749b64dd8245195e277d5f804ca12 /tools/perf/pmu-events/arch/x86 parent2d706e790f0508dff4fb72eca9b4892b79757feb (diff)
mm: optimize PageWaiters bit use for unlock_page()
In commit 62906027091f ("mm: add PageWaiters indicating tasks are waiting for a page bit") Nick Piggin made our page locking no longer unconditionally touch the hashed page waitqueue, which not only helps performance in general, but is particularly helpful on NUMA machines where the hashed wait queues can bounce around a lot. However, the "clear lock bit atomically and then test the waiters bit" sequence turns out to be much more expensive than it needs to be, because you get a nasty stall when trying to access the same word that just got updated atomically. On architectures where locking is done with LL/SC, this would be trivial to fix with a new primitive that clears one bit and tests another atomically, but that ends up not working on x86, where the only atomic operations that return the result end up being cmpxchg and xadd. The atomic bit operations return the old value of the same bit we changed, not the value of an unrelated bit. On x86, we could put the lock bit in the high bit of the byte, and use "xadd" with that bit (where the overflow ends up not touching other bits), and look at the other bits of the result. However, an even simpler model is to just use a regular atomic "and" to clear the lock bit, and then the sign bit in eflags will indicate the resulting state of the unrelated bit #7. So by moving the PageWaiters bit up to bit #7, we can atomically clear the lock bit and test the waiters bit on x86 too. And architectures with LL/SC (which is all the usual RISC suspects), the particular bit doesn't matter, so they are fine with this approach too. This avoids the extra access to the same atomic word, and thus avoids the costly stall at page unlock time. The only downside is that the interface ends up being a bit odd and specialized: clear a bit in a byte, and test the sign bit. Nick doesn't love the resulting name of the new primitive, but I'd rather make the name be descriptive and very clear about the limitation imposed by trying to work across all relevant architectures than make it be some generic thing that doesn't make the odd semantics explicit. So this introduces the new architecture primitive clear_bit_unlock_is_negative_byte(); and adds the trivial implementation for x86. We have a generic non-optimized fallback (that just does a "clear_bit()"+"test_bit(7)" combination) which can be overridden by any architecture that can do better. According to Nick, Power has the same hickup x86 has, for example, but some other architectures may not even care. All these optimizations mean that my page locking stress-test (which is just executing a lot of small short-lived shell scripts: "make test" in the git source tree) no longer makes our page locking look horribly bad. Before all these optimizations, just the unlock_page() costs were just over 3% of all CPU overhead on "make test". After this, it's down to 0.66%, so just a quarter of the cost it used to be. (The difference on NUMA is bigger, but there this micro-optimization is likely less noticeable, since the big issue on NUMA was not the accesses to 'struct page', but the waitqueue accesses that were already removed by Nick's earlier commit). Acked-by: Nick Piggin <npiggin@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Bob Peterson <rpeterso@redhat.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Andrew Lutomirski <luto@kernel.org> Cc: Andreas Gruenbacher <agruenba@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mel Gorman <mgorman@techsingularity.net> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'tools/perf/pmu-events/arch/x86')