mirror of
https://github.com/gperftools/gperftools
synced 2025-01-11 09:09:49 +00:00
lower default transfer batch size down to 512
Some workloads get much slower with too large batch size. This closes bug #678. binary_trees benchmark benefits from larger batch size. And I found that 512 is not much slower than huge value that we had.
This commit is contained in:
parent
6fdfc5a7f4
commit
64892ae730
@ -44,7 +44,7 @@ namespace tcmalloc {
|
||||
// thread and central caches.
|
||||
static int32 FLAGS_tcmalloc_transfer_num_objects;
|
||||
|
||||
static const int32 kDefaultTransferNumObjecs = 32768;
|
||||
static const int32 kDefaultTransferNumObjecs = 512;
|
||||
|
||||
// The init function is provided to explicit initialize the variable value
|
||||
// from the env. var to avoid C++ global construction that might defer its
|
||||
|
Loading…
Reference in New Issue
Block a user