This updates the configuration script to set the default page size to
64KiB on Aarch64 Linux. This is motivated by compatibility as a build
configured for a 64KiB page will work on kernels that use the smaller
4KiB or 16KiB pages, whereas the reverse is not true.
To make the configured page size setting more visible, the script now
displays the page size when printing the configuration results.
Users that want to override the page size in to choose a smaller value
can still do so with the --with-lg-pagesize configuration option.
The static inline definition made more sense when these functions just
dispatched to a syscall wrapper. Since they acquired a retry loop, a
non-inline definition makes more sense.
An include of unistd.h is needed to make the declaration of the
syscall function visible to the compiler. The include of sys/mman.h
is not used at all.
When the dehugify functionality was retired in an previous commit, a
dehugify-related function and global variable in a test was
accidentally left in-place causing builds that add -Werror to CFLAGS
to fail.
Giving the advice MADV_DONTNEED to a range of virtual memory backed by
a transparent huge page already causes that range of virtual memory to
become backed by regular pages.
When hugepage is enabled and PAGESIZE is large, the test could
ask for a stack size larger than user limit. Allocating the
memory instead can avoid the failure.
any future changes to the underlying data type for bin sizes
(such as upgrading from `uint16_t` to `uint32_t`) can be achieved
by modifying only the `cache_bin_sz_t` definition.
Signed-off-by: Xin Yang <yangxin.dev@bytedance.com>
The maximum allowed value for `nflush_batch` is
`CACHE_BIN_NFLUSH_BATCH_MAX`. However, `tcache_bin_flush_impl_small`
could potentially declare an array of `emap_batch_lookup_result_t`
of size `CACHE_BIN_NFLUSH_BATCH_MAX + 1`. leads to a `VARIABLE_ARRAY`
assertion failure, observed when `tcache_nslots_small_max` is
configured to 2048. This patch ensures the array size does not exceed
the allowed maximum.
Signed-off-by: Xin Yang <yangxin.dev@bytedance.com>
in the dirty ecache has been limited. This patch was tested with real
workloads using ClickHouse (Clickbench Q35) on a system with 2x240 vCPUs.
The results showed a 2X in query per second (QPS) performance and
a reduction in page faults to 29% of the previous rate. Additionally,
microbenchmark testing involved 256 memory reallocations resizing
from 4KB to 16KB in one arena, which demonstrated a 5X performance
improvement.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
