allmemory(1)						    BSD General Commands Manual 					      allmemory(1)

NAME

     allmemory -- User-program and framework memory analysis tool.

SYNOPSIS

     allmemory [-f] [-sections] [-v] [-vf v_filter] [-p] [-n] [-F] [-FD] [-purge] [-noframework] [-noprocess] [-proc pid | process_name]
	       [-noproc pid | process_name] [-32bit] [-64bit] [-i path] [-o path] [-d]

     allmemory diff path1 path2 [-p] [-f] [-F] [-noframework] [-noprocess] [-ff framework_filter] [-pf process_filter] [-proc pid | process_name]
	       [-noproc pid | process_name] [-32bit] [-64bit] [-no32v64]

DESCRIPTION (Normal mode)
     In its normal operating mode, the allmemory tool collects and analyses data on the resident memory usage of user processes.  Data from each
     run of allmemory is saved and can be viewed and analyzed at a later time.	The save location of the collected data can be specified by the
     user. (See the -i and -o flags below).  By default allmemory presents top level summary information on all user processes and frameworks
     found on the system.  More or less data can be displayed using the Data Reporting options specified below.

     IMPORTANT NOTES:

     This is a completely new version of the allmemory tool.  To use the old version of allmemory, run /usr/local/bin/allmemory_old.

     If fresh data is being collected, new version of allmemory takes significantly longer to run than the old version (~30 seconds).  If the
     state of the system has not significantly changed since the last time allmemory data was collected, make sure to use the -i or -d flags to
     save time when viewing data.

     Data Reporting options (Specify zero or more):

     -f       Display a per-segment residency information breakdown for each framework.

     -sections
	      Display a per-section residency information breakdown for each framework.

     -v       Display information on address space layouts of each process (similar to vmmap).

     -vf v_filter
	      Filter the address space output from the -v flag to only show VM Regions with more resident pages than v_filter.

     -p       Display detailed total residency information for each process on a per VM region category basis.

     -F       Display summary information on per-process private pages that contribute to the total resident pages for each framework.

     -FD      Display per-segment information on per-process private pages that contribute to the total resident pages for each framework.

     -purge   Displays the NoVolatile resident column for all processes.  Purgeable volatile pages are not counted towards this total.

     -n       Do NOT display collected data (i.e. when an automated test is running allmemory).  This lowers allmemory's highwater memory usage
	      mark.

     -noframework
	      Do NOT display data collected on frameworks.

     -noprocess
	      Do NOT display data collected on processes.

     -proc pid | process_name
	      Display data about the process with the specified pid/name.  This option can be specified more than once to display data on more
	      than one process.

     -32bit   Display data only for 32-bit processes.  Cannot be used in conjunction with the -64bit flag.

     -64bit   Display data only for 64-bit processes.  Cannot be used in conjunction with the -32bit flag.

     Data Saving/Restoring options: (Specify one at most):

     -d       Load and display the data from the default save directory ( /tmp/allmemoryDataFiles/ )

     -i path  Load and display the data from a previous run of allmemory stored in the <path> directory

     -o path  Specify a directory in which to store the data collected in this run of allmemory.  Defaults to: /tmp/allmemoryDataFiles/

DESCRIPTION (diff Mode)
     In diff mode, allmemory is used to compare data taken at two different times.  This is useful when trying to track down the specifics of a
     memory-related regression and saves time that would be otherwise spent on tedious and error-prone arithmetic.  Since allmemory has knowledge
     of its own data structures, this mode is robust against output format changes.

     Required (specify all):

     <path1> <path2>
     Paths to the two data sets to compare.  Options (Specify zero or more):

     -proc pid | process_name
	      Display data about the process with the specified pid/name.  This option can be specified more than once to display data on more
	      than one process.

     -f       Display a per-segment residency information breakdown for each framework.

     -F       Display summary information on per-process private pages that contribute to the total resident pages for each framework.

     -noframework
	      Do NOT display data collected on frameworks.

     -noprocess
	      Do NOT display data collected on processes.

     -ff framework_filter
	      Only displays frameworks for which the difference number of resident pages is equal to or above the framework_filter value.

     -pf process_filter
	      Only displays processes for which the difference number of resident pages is equal to or above the process_filter value.

     -p       Provides a detailed breakdown of the differences between the specified data sets on a per VM-region category basis.

     -32bit   Display data only for 32-bit processes.  Cannot be used in conjunction with the -64bit flag.

     -64bit   Display data only for 64-bit processes.  Cannot be used in conjunction with the -32bit flag.

     -no32v64
	      Do not compare 32-bit processes and frameworks versus 64-bit processes and frameworks. By default, the comparison will happen.

EXPLANATION OF OUTPUT

     Individual Process Data Columns:

     PrivateRes (In pages):

     The number of pages resident in physical memory in a given process's address space that are not shared with other processes.

     This total includes private pages as well as copied pages in COW regions and resident pages in shared regions that are not used by other pro-
     cesses.

     NoSpec PrivateRes is the number of pages resident in physical memory in a process's address space not counting pages brought in specula-
     tively.

     If one were to simply count the total number of pages resident in a process's address space, that total would be the sum of Shared and Resi-
     dent pages.

     Copied (In pages):

     The number of copied pages resident in physical memory in a given process's address space.

     Copied pages are assumed to be private to the process and are counted towards the Resident total.	The Copied and Dirty columns are non-
     exclusive, as pages can be both dirty and copied.

     Dirty (In pages):

     The number of dirty pages resident in physical memory in a given process's address space.

     Dirty pages are not assumed to be private to the process and so do not automatically count towards the Resident total.  The Copied and Dirty
     columns are non-exclusive, as pages can be both dirty and copied.

     Swapped (In pages):

     Pages swapped out of a given process's address space.  These pages were previously resident in physical memory and are now swapped to disk.

     Swapped pages do not count towards any other totals.

     Shared (In pages):

     The number of shared pages resident in physical memory in a given process's address space.

     This total does not count towards the Resident pages category or any other category.

     As was the case with NoSpec Resident pages, the NoSpec Shared pages total ignores shared pages brought in speculatively.

     If one were to simply count the total number of pages resident in a process's address space, that total would be the sum of Shared and Resi-
     dent pages.

     Total Process Data Columns:

     TotalRes (In pages):

     The number of pages resident in physical memory across the entire system.

     This total includes private pages as well as copied pages in COW regions.	Shared pages are counted towards this total once (i.e. double
     counting is avoided).

     NoSpec TotalRes is the number of pages resident in physical memory across the system not counting pages brought in speculatively.

     Copied (In pages):

     The number of copied pages resident in physical memory in a given process's address space.

     Copied pages are assumed to be private to the process and are counted towards the Resident total.	The Copied and Dirty columns are non-
     exclusive, as pages can be both dirty and copied.

     Dirty (In pages):

     The number of dirty pages resident in physical memory in a given process's address space.

     Dirty pages are not assumed to be private to the process and so do not automatically count towards the Resident total.  The Copied and Dirty
     columns are non-exclusive, as pages can be both dirty and copied.

     Swapped (In pages):

     Pages swapped out of a given process's address space.  These pages were previously resident in physical memory and are now swapped to disk.

     Swapped pages do not count towards any other totals.

     Shared (In pages):

     The number of shared pages resident in physical memory across the entire system.

     This total counts towards the TotalRes pages category and its subcategories (i.e. Copied, Dirty, Swapped).

     As was the case with NoSpec Resident pages, the NoSpec Shared pages total ignores shared pages brought in speculatively.

     Framework Data Columns:

     Resident (In pages):

     The number of pages resident in physical memory attributed to a given framework across the address spaces of all processes.  A heuristic is
     used to avoid double counting pages shared across several processes.

     Any copied pages encountered are assumed to be private to the process being examined and are counted towards the Resident total.  In other
     words, Resident is the total number of private pages attributed to a given framework, plus the number of shared resident pages (avoiding dou-
     ble counting).

     NoSpec Resident total is calculated in the same way, ignoring speculative pages.

     Copied (In pages):

     The number of copied pages resident in physical memory attributed to the given framework across the address spaces of all processes.

     As described above, this total is included in the total Resident and NoSpec Resident framework totals.

     The Copied and Dirty columns are non-exclusive, as pages can be both dirty and copied.

     Dirty (In pages):

     The number of dirty pages resident in physical memory attributed to the given framework across the address spaces of all processes.

     The Copied and Dirty columns are non-exclusive, as pages can be both dirty and copied.

     Swapped (In pages):

     The number of pages swapped out of physical memory attributed to the given framework.  These pages were previously resident in physical mem-
     ory and are now swapped to disk.

     These pages do not count towrds any other total.

     Filesize (In pages):

     The size in pages of the framework on disk.

SEE ALSO

     heap(1), leaks(1), top(1), vmmap(1)

Apple Inc.							   June 2, 2019 							Apple Inc.