Age | Commit message (Collapse) | Author |
|
Remote revalidating requires to update file size on every write
on a file opened with O_APPEND. If host FD exists, it can be
used to update the size and skip round trip to the gofer. With
this change, O_APPEND writes with remote revalidating is almost
as fast as exclusive mode:
BM_Append
VFS1 60.7us
VFS2 56.8us
VFS2 exclusive 14.2us
This change 15.8us
Updates #1792
PiperOrigin-RevId: 369486801
|
|
This is the most often pattern of calling system calls in real applications.
PiperOrigin-RevId: 367320048
|
|
This benchmark currently takes > 15 minutes to run in that case.
PiperOrigin-RevId: 366891726
|
|
This benchmark currently takes > 15 minutes to run in that case.
PiperOrigin-RevId: 366817185
|
|
PiperOrigin-RevId: 366573366
|
|
Run all tests (or a given test partition) in a single sandbox.
Previously, each individual unit test executed in a new
sandbox, which takes much longer to execute.
Before After
Syscall tests: 37m22.768s 14m5.272s
PiperOrigin-RevId: 361661726
|
|
PiperOrigin-RevId: 345399936
|
|
This test regularly times out when "shared" filesystem is enabled.
PiperOrigin-RevId: 329950622
|
|
This is to cover the common pattern: open->read/write->close,
where SetAttr needs to be called to update atime/mtime before
the file is closed.
Benchmark results:
BM_OpenReadClose/10240 CPU
setattr+clunk: 63783 ns
VFS2: 68109 ns
VFS1: 72507 ns
Updates #1198
PiperOrigin-RevId: 329628461
|
|
This is done to ease troubleshooting when tests fail. runsc
logs are not stored when tests passe, so this will only
affect failing tests and should not increase log storage
too badly.
PiperOrigin-RevId: 327717551
|
|
PiperOrigin-RevId: 317941748
|
|
* Increase a buffer size for getdents64
* Increase a number of shards
PiperOrigin-RevId: 304670004
|
|
- Fix definitions of Futex* wrappers.
- Correctly handle glibc syscall() (which returns -1 and sets errno instead of
returning the raw syscall return value).
- De-parameterize FutexWaitBitset, which was apparently intended to test with
deadlines of between 0 and 100000 nanoseconds after the Unix epoch, but was
broken due to the preceding two issues.
- Use wall time to measure the durations of tests that are expected to block
(and thus stop accumulating CPU time).
- Require 5s for all tests to improve robustness in the presence of sentry GC.
- Remove FutexContend and FutexContendDeadline; it's unclear what these are
supposed to measure, given that (1) FutexLock is unrealistically inefficient
and (2) the benchmark rewards slow scheduling (since this reduces
contention).
PiperOrigin-RevId: 302925246
|
|
Without this change, the assembly code of this test compiled without
optimizations:
mov -0x150(%rbp),%rax
movl $0x77777777,(%rax)
lea -0x128(%rbp),%rax
with optimizations:
movl $0x77777777,0x0
This code doesn't work properly, because the test changes rax in the segv
handler.
PiperOrigin-RevId: 299896117
|
|
PiperOrigin-RevId: 297915815
|
|
PiperOrigin-RevId: 297009116
|
|
PiperOrigin-RevId: 296104390
|