gvsafronov/futriix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
futriix/deps/memkind/src/man/hbwmalloc.3

418 lines
13 KiB
Groff
Raw Normal View History

Squashed 'deps/memkind/src/' content from commit bb9f19dd1 git-subtree-dir: deps/memkind/src git-subtree-split: bb9f19dd1b3ed6cc5e1b35919564ccf6f4b32f69
2019-02-01 14:45:39 -05:00
.\"
.\" Copyright (C) 2014 - 2018 Intel Corporation.
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions are met:
.\" 1. Redistributions of source code must retain the above copyright notice(s),
.\" this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright notice(s),
.\" this list of conditions and the following disclaimer in the documentation
.\" and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
.\" OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
.\" MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
.\" EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
.\" LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
.\" PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
.\" LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
.\" OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
.\" ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.\"
.TH "HBWMALLOC" 3 "2015-03-31" "Intel Corporation" "HBWMALLOC" \" -*- nroff -*-
.SH "NAME"
hbwmalloc \- The high bandwidth memory interface
.br
Note: hbwmalloc.h functionality is considered as stable API (STANDARD API).
.SH "SYNOPSIS"
.nf
.B #include <hbwmalloc.h>
.sp
.B Link with -lmemkind
.sp
.B int hbw_check_available(void);
.br
.BI "void* hbw_malloc(size_t " "size" );
.br
.BI "void* hbw_calloc(size_t " "nmemb" ", size_t " "size" );
.br
.BI "void* hbw_realloc (void " "*ptr" ", size_t " "size" );
.br
.BI "void hbw_free(void " "*ptr" );
.br
.BI "int hbw_posix_memalign(void " "**memptr" ", size_t " "alignment" ", size_t " "size" );
.br
.BI "int hbw_posix_memalign_psize(void " "**memptr" ", size_t " "alignment" ", size_t " "size" ", hbw_pagesize_t " "pagesize" );
.br
.B hbw_policy_t hbw_get_policy(void);
.br
.BI "int hbw_set_policy(hbw_policy_t " "mode" );
.br
.BI "int hbw_verify_memory_region(void " "*addr" ", size_t " "size" ", int " "flags" );
.fi
.SH "DESCRIPTION"
.BR hbw_check_available ()
returns zero if high bandwidth memory is available or an error code
described in the
.B ERRORS
section if not.
.PP
.BR hbw_malloc ()
allocates
.I size
bytes of uninitialized high bandwidth memory. The allocated space is
suitably aligned (after possible pointer coercion) for storage of any
type of object. If
.I size
is zero then
.BR hbw_malloc ()
returns NULL.
.PP
.BR hbw_calloc ()
allocates space for
.I nmemb
objects in high bandwidth memory, each
.I size
bytes in length. The result is identical to calling
.BR hbw_malloc ()
with an argument of
.IR nmemb * size
, with the exception that the allocated memory is explicitly
initialized to zero bytes. If
.I nmemb
or
.I size
is 0, then
.BR hbw_calloc ()
returns NULL.
.PP
.BR hbw_realloc ()
changes the size of the previously allocated high bandwidth memory
referenced by
.I ptr
to
.I size
bytes. The contents of the memory are unchanged up to the lesser of
the new and old sizes. If the new size is larger, the contents of the
newly allocated portion of the memory are undefined. Upon success, the
memory referenced by
.I ptr
is freed and a pointer to the newly allocated high bandwidth memory is
returned.
.B Note:
.BR hbw_realloc ()
may move the memory allocation, resulting in a different return value
than
.IR "ptr" .
If
.I ptr
is NULL, the
.BR hbw_realloc ()
function behaves identically to
.BR hbw_malloc ()
for the specified size.
The address
.IR "ptr" ,
if not NULL, was returned by a previous call to
.BR hbw_malloc (),
.BR hbw_calloc (),
.BR hbw_realloc (),
or
.BR hbw_posix_memalign ().
Otherwise, or if
.I hbw_free(ptr)
was called before, undefined behavior occurs.
.B Note:
.BR hbw_realloc ()
cannot be used with a pointer returned by
.BR hbw_posix_memalign_psize ().
.PP
.BR hbw_free ()
causes the allocated memory referenced by
.I ptr
to be made available for future allocations. If
.I ptr
is NULL, no action occurs.
The address
.IR "ptr" ,
if not NULL, must have been returned by a previous call to
.BR hbw_malloc (),
.BR hbw_calloc (),
.BR hbw_realloc (),
.BR hbw_posix_memalign (),
or
.BR hbw_posix_memalign_psize ().
Otherwise, if
.I hbw_free(ptr)
was called before, undefined behavior occurs.
.PP
.BR hbw_posix_memalign ()
allocates
.I size
bytes of high bandwidth memory such that the allocation's base address
is an even multiple of
.IR "alignment" ,
and returns the allocation in the value pointed to by
.IR "memptr" .
The requested
.I alignment
must be a power of 2 at least as large as
.IR "sizeof(void *)" .
.PP
.BR hbw_posix_memalign_psize ()
allocates
.I size
bytes of high bandwidth memory such that the allocation's base address
is an even multiple of
.IR "alignment" ,
and returns the allocation in the value pointed to by
.IR "memptr" .
The requested
.I alignment
must be a power of 2 at least as large as
.IR "sizeof(void *)" .
The memory will be allocated using pages determined by the
.IR "pagesize"
variable which may be one of the following enumerated values:
.TP
.B HBW_PAGESIZE_4KB
The four kilobyte page size option. Note that with transparent huge
pages enabled these allocations may be promoted by the operating
system to two megabyte pages.
.TP
.B HBW_PAGESIZE_2MB
The two megabyte page size option. Note: This page size requires
huge pages configuration described in SYSTEM CONFIGURATION section.
.TP
.B HBW_PAGESIZE_1GB (DEPRECATED)
This option allows the user to specify arbitrary sizes backed by
1GB chunks of huge pages. Huge pages are allocated even if the
size is not a modulo of 1GB. Note: This page size requires
huge pages configuration described in SYSTEM CONFIGURATION section.
.TP
.B HBW_PAGESIZE_1GB_STRICT (DEPRECATED)
The total size of the allocation must be a multiple of 1GB with
this option, otherwise the allocation will fail. Note: This page
size requires huge pages configuration described in SYSTEM
CONFIGURATION section.
.TP
HBW_PAGESIZE_2MB, HBW_PAGESIZE_1GB and HBW_PAGESIZE_1GB_STRICT options are not supported with HBW_POLICY_INTERLEAVE policy.
.PP
.BR hbw_get_policy ()
returns the current fallback policy when insufficient high bandwidth
memory is available.
.PP
.BR hbw_set_policy ()
sets the current fallback policy. The policy can be modified only once in the lifetime of an application and before calling hbw_*alloc() or hbw_posix_memalign*() function.
.br
Note: If the policy is not set, than HBW_POLICY_PREFERRED will be used by default.
.TP
.B HBW_POLICY_BIND
If insufficient high bandwidth memory from the nearest NUMA node is
available to satisfy a request, the allocated pointer is set to NULL
and
.I errno
is set to ENOMEM. If insufficient high bandwidth memory pages are
available at fault time the Out Of Memory (OOM) killer is triggered.
Note that pages are faulted exclusively from the high bandwidth NUMA
node nearest at time of allocation, not at time of fault.
.TP
.B HBW_POLICY_BIND_ALL
If insufficient high bandwidth memory is available to satisfy a request,
the allocated pointer is set to NULL and
.I errno
is set to ENOMEM. If insufficient high bandwidth memory pages are
available at fault time the Out Of Memory (OOM) killer is triggered.
Note that pages are faulted from the high bandwidth NUMA nodes.
Nearest NUMA node is selected at time of page fault.
.TP
.B HBW_POLICY_PREFERRED
If insufficient memory is available from the high bandwidth NUMA node
closest at allocation time, fall back to standard memory (default)
with the smallest NUMA distance.
.TP
.B HBW_POLICY_INTERLEAVE
Interleave faulted pages from across all high bandwidth NUMA nodes
using standard size pages (the Transparent Huge Page feature is
disabled).
.PP
.BR hbw_verify_memory_region ()
verifies if memory region fully fall into high bandwidth memory. Returns
0 if memory address range from
.IR "addr"
to
.IR "addr"
+
.IR "size"
is allocated in high bandwidth memory,
-1 if any fragment of memory was not backed by high bandwidth memory [e.g. when memory is not initialized]
or one of error codes described in ERRORS section.
Using this function in production code may result in serious performance penalty.
.IR Flags
argument may include optional flags that modifies function behaviour:
.TP
.B HBW_TOUCH_PAGES
Before checking pages, function will touch first byte of all pages in address range starting from
.IR "addr"
to
.IR "addr"
+
.IR "size"
by read and write (so the content will be overwritten by the same data as it was read).
Using this option may trigger Out Of Memory killer.
.SH "RETURN VALUE"
.BR hbw_get_policy ()
returns
.B HBW_POLICY_BIND,
.B HBW_POLICY_BIND_ALL,
.B HBW_POLICY_PREFERRED
or
.B HBW_POLICY_INTERLEAVE
which represents the current high bandwidth policy.
.BR hbw_free ()
do not have return value.
.BR hbw_malloc ()
.BR hbw_calloc (),
and
.BR hbw_realloc ()
return the pointer to the allocated memory, or NULL if the request
fails.
.BR hbw_posix_memalign (),
.BR hbw_posix_memalign_psize ()
and
.BR hbw_set_policy ()
return zero on success and return an error code
as described in the
.B ERRORS
section below on failure.
.SH ERRORS
.TP
Error codes described here are the POSIX standard error codes as defined in <errno.h>
.TP
.BR hbw_check_available ()
returns
.BR ENODEV
if high-bandwidth memory is unavailable.
.TP
.BR "hbw_posix_memalign" "() and " "hbw_posix_memalign_psize" "()"
If the
.I alignment
parameter is not a power of two, or was not a multiple of
.IR "sizoeof(void *)" ,
then
.B EINVAL
is returned.
If the policy and
.I pagesize
combination is unsupported then
.B EINVAL
is returned.
If there was insufficient memory to satisfy the request then
.B ENOMEM
is returned.
.TP
.BR hbw_set_policy ()
returns
.B EPERM
if hbw_set_policy () was called more than once, or
.B EINVAL if
.I mode
argument was neither
.B HBW_POLICY_PREFERRED,
.B HBW_POLICY_BIND,
.B HBW_POLICY_BIND_ALL
nor
.B HBW_POLICY_INTERLEAVE.
.TP
.BR hbw_verify_memory_region ()
returns
.B EINVAL
if
.IR "addr"
is NULL,
.IR "size"
equals 0 or
.IR "flags"
contained unsupported bit set. If memory pointed by
.IR "addr"
could not be verified then
.B EFAULT
is returned.
.SH "NOTES"
The
.I hbwmalloc.h
file defines the external functions and enumerations for the hbwmalloc
library. These interfaces define a heap manager that targets high
bandwidth memory numa nodes.
.SH "FILES"
.TP
.I /usr/bin/memkind-hbw-nodes
Prints a comma separated list of high bandwidth nodes.
.SH "ENVIRONMENT"
.TP
.B MEMKIND_HBW_NODES
This environment variable is a comma separated list of NUMA nodes that
are treated as high bandwidth. Uses the
.I libnuma
routine
.BR numa_parse_nodestring ()
for parsing, so the syntax described in the
.BR numa (3)
man page for this routine applies for example: 1-3,5 is a valid setting.
.TP
.B MEMKIND_ARENA_NUM_PER_KIND
This environment variable allows leveraging internal mechanism of
the library for setting number of arenas per kind. Value should be
a positive integer (not greater than INT_MAX defined in limits.h).
The user should set the value based on the characteristics
of application that is using the library. Higher value can
provide better performance in extremely multithreaded applications at
the cost of memory overhead. See section "IMPLEMENTATION NOTES" of
.BR jemalloc (3)
for more details about arenas.
.TP
.B MEMKIND_HEAP_MANAGER
Controls heap management behavior in memkind library by switching to one of the available heap managers.
.br
Values:
.br
JEMALLOC – sets the jemalloc heap manager
.br
TBB – sets the Intel Threading Building Blocks heap manager. This option requires installed
Intel Threading Building Blocks library.
If the MEMKIND_HEAP_MANAGER is not set then the jemalloc heap manager will be used by default.
.SH "SYSTEM CONFIGURATION"
Interfaces for obtaining 2MB (HUGETLB) need allocated
huge pages in the kernel's huge page pool.
.TP
.B HUGETLB (huge pages)
Current number of "persistent" huge pages can be read from /proc/sys/vm/nr_hugepages file.
Proposed way of setting hugepages is: "sudo sysctl vm.nr_hugepages=<number_of_hugepages>".
More information can be found here:
https://www.kernel.org/doc/Documentation/vm/hugetlbpage.txt
.SH "KNOWN ISSUES"
.TP
.B HUGETLB (huge pages)
There might be some overhead in huge pages consumption caused by heap management.
If your allocation fails because of OOM, please try to allocate extra huge pages (e.g. 8 huge pages).
.SH "COPYRIGHT"
Copyright (C) 2014 - 2018 Intel Corporation. All rights reserved.
.SH "SEE ALSO"
.BR malloc (3),
.BR numa (3),
.BR numactl (8),
.BR mbind (2),
.BR mmap (2),
.BR move_pages (2)
.BR jemalloc (3)
.BR memkind (3)
Reference in New Issue Copy Permalink