| Commit message (Collapse) | Author | Age |
... | |
| | | |
|
| | | |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| | |
and get rid of the dependency loop on the grpc shutdown path. Make sure
all background closures are complete before shutting down the other grpc
modules.
Avoid using the backup poller in TCP endpoints if using the background
poller.
|
| | |
|
|\| |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
Callers are updated to properly initialize the memory.
This behavior can be overridden using GRPC_ARENA_INIT_STRATEGY
environment variable.
|
| |\
| | |
| | | |
Do not make TCP_USER_TIMEOUT failure an error
|
| | | |
|
| |\ \
| | | |
| | | | |
Account the memory usage of channel and call by resource quota
|
| | | | |
|
| |/ / |
|
| | |
| | |
| | |
| | | |
Suggested-by: vjpai@google.com
|
| | |
| | |
| | |
| | |
| | |
| | | |
Make sure paddings are GPR_CACHELINE_SIZE minus the
size of fields in that cache line. Otherwise, we will waste an
additional cache line for no good reason.
|
| | |
| | |
| | |
| | | |
Explain that the input parameter cannot be nullptr.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This commit contains a few improvements:
1. Using a consequetive range of [0..4], will allow us to merge all
branches of error_is_special into one comparison.
2. With (1), we can remove the for loops to find entries in
error_status_map with a single O(1) lookup.
3. grpc_error_is_special() code paths should be inlined for ref
and unref to avoid callq for the majority of cases where speical
error is used.
4. grpc_error_get_int() should never accept a nullptr argument to
avoid an expensive branch in the hot path. Callers should all
allocate a dummy int on the stack when calling.
|
| |\ \
| | | |
| | | | |
Protect Spammy Log with Tracer
|
| |\ \ \
| | | | |
| | | | | |
fix: while using eventfd, a random fd will be closed when error occurs during creating eventfd
|
| | |/ /
| |/| | |
|
| |\ \ \
| | | | |
| | | | | |
Simple optimizations in TCP read
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
We overallocate by 2x for target. Unless buffer is more than half
full, we should not delay read for more allocation.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
If we have a continous stream of bytes on the socket, we will
never grow the buffer, because we will never get EAGAIN, and
call finish. This is a serious performance issue, which can
be misued.
As soon as we have a full buffer, update the estimate.
|
| |\ \ \ \
| | |/ / /
| |/| | | |
Unify fallthrough comment in switch statements
|
| | | | | |
|
| |/ / / |
|
| |\ \ \
| | | | |
| | | | | |
Clean some timer code
|
| |\ \ \ \
| | | | | |
| | | | | | |
Disable error tracking until Fathom changes are commited
|
| |\ \ \ \ \
| | | | | | |
| | | | | | | |
Core infrastructure for timer manager debug
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | | |
|
| | |/ / / /
| |/| | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | |/ /
| | | | |
| | | | |
| | | | | |
while using eventfd, when error occurs during creating eventfd, a random fd will be closed]
|
| |\ \ \ \
| | | |/ /
| | |/| | |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| |/ / / |
|
| | | | |
|
| |\ \ \
| | | | |
| | | | | |
Increase the maximum number of timer shards to 32.
|
| |\ \ \ \
| | | | | |
| | | | | | |
Avoid extra branches in grpc_error_get_(str|int).
|
| | |_|/ /
| |/| | | |
|
| | | | | |
|
| | |/ /
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | | |
Commit 82f9886e accidentally sets the maximum number of timer shards 1, from
previously 32.
We probably want to increase the max shards further.
|