You can customize your system environment by specifying paths where the compiler searches for certain files such as libraries, include files, configuration files, and certain settings.
The following table shows the compile-time environment variables that affect the compiler:
Compile-Time Environment Variable |
Description |
---|---|
CL (Windows) _CL_ (Windows) |
Define the files and options you use most often with the CL variable. Note: You cannot set the CL environment variable to a string that contains an equal sign. You can use the pound sign instead. In the following example, the pound sign (#) is used as a substitute for an equal sign in the assigned string: SET CL=/Dtest#100 |
IA32ROOT (IA-32 architecture and Intel® 64 architecture) |
Points to the directories containing the include and library files for a non-standard installation structure. NoteIA-32 is only available for C++; it is not available for DPC++. |
ICXCFG |
Specifies the configuration file for customizing compilations when invoking the compiler using icx. NoteThis environment variable is only available for C++; it is not available for DPC++. |
ICPXCFG |
Specifies the configuration file for customizing compilations when invoking the compiler using icpx. NoteThis environment variable is only available for C++; it is not available for DPC++. |
ICXCFG |
Specifies a configuration file, which the compiler should use instead of the default configuration file. |
__INTEL_PRE_CFLAGS __INTEL_POST_CFLAGS |
Specifies a set of compiler options to add to the compile line. This is an extension to the facility already provided in the compiler configuration file icx.cfg. NoteBy default, a configuration file named icx.cfg (Windows, Linux), or icpx.cfg (Linux) is used. This file is in the same directory as the compiler executable. To use another configuration file in another location, you can use the ICXCFG (Windows, Linux), ICPXCFG (Linux) environment variable to assign the directory and file name for the configuration file.You can insert command line options in the prefix position using __INTEL_PRE_CFLAGS , or in the suffix position using __INTEL_POST_CFLAGS. The command line is built as follows: Syntax:icx <PRE flags> <flags from configuration file> <flags from the compiler invocation> <POST flags> NoteThe driver issues a warning that the compiler is overriding an option because of an environment variable, but only when you include the option /W5 (Windows) or -w3 (Linux). |
PATH |
Specifies the directories the system searches for binary executable files. NoteOn Windows, this also affects the search for Dynamic Link Libraries (DLLs). |
TMP TMPDIR TEMP |
Specifies the location for temporary files. If none of these are specified, or writeable, or found, the compiler stores temporary files in /tmp (Linux) or the current directory (Windows). The compiler searches for these variables in the following order: TMP, TMPDIR, and TEMP. NoteOn Windows, these environment variables cannot be set from Visual Studio. |
LD_LIBRARY_PATH (Linux) |
Specifies the location for shared objects (.so files). |
INCLUDE (Windows) |
Specifies the directories for the source header files (include files). |
LIB (Windows) |
Specifies the directories for all libraries used by the compiler and linker. |
GNU Environment Variables and Extensions | |
CPATH (Linux) |
Specifies the path to include directory for C/C++ compilations. |
C_INCLUDE_PATH (Linux) |
Specifies path to include directory for C compilations. |
CPLUS_INCLUDE_PATH (Linux) |
Specifies path to include directory for C++ compilations. |
DEPENDENCIES_OUTPUT (Linux) |
Specifies how to output dependencies for make based on the non-system header files processed by the compiler. System header files are ignored in the dependency output. |
GCC_EXEC_PREFIX (Linux) |
Specifies alternative names for the linker (ld) and assembler (as). |
GCCROOT (Linux) |
Specifies the location of the GCC* binaries. Set this variable only when the compiler cannot locate the GCC binaries when using the -gcc-name option. NoteThis environment variable is only available for C++; it is not available for DPC++. |
GXX_INCLUDE (Linux) |
Specifies the location of the GCC headers. Set this variable to specify the locations of the GCC installed files when the compiler does not find the needed values as specified by the use of -gcc-name=directory-name/gcc or -gxx-name=directory-name/g++. NoteThis environment variable is only available for C++; it is not available for DPC++. |
GXX_ROOT (Linux) |
Specifies the location of the GCC binaries. Set this variable to specify the locations of the GCC installed files when the compiler does not find the needed values as specified by the use of -gcc-name=directory-name/gcc or -gxx-name=directory-name/g++. NoteThis environment variable is only available for C++; it is not available for DPC++. |
LIBRARY_PATH (Linux) |
Specifies the path for libraries to be used during the link phase. |
SUNPRO_DEPENDENCIES (Linux) |
This variable is the same as DEPENDENCIES_OUTPUT, except that system header files are not ignored. |
The following table summarizes compiler environment variables that are recognized at run time.
Run-Time Environment Variable |
Description |
---|---|
GNU extensions (recognized by the Intel OpenMP* compatibility library) | |
GOMP_CPU_AFFINITY (Linux) |
GNU extension recognized by the Intel OpenMP* compatibility library. Specifies a list of OS processor IDs. You must set this environment variable before the first parallel region or before certain API calls including omp_get_max_threads(), omp_get_num_procs() and any affinity API calls. For detailed information on this environment variable, see Thread Affinity Interface. Default: Affinity is disabled |
GOMP_STACKSIZE (Linux) |
GNU extension recognized by the Intel OpenMP compatibility library. Same as OMP_STACKSIZE.KMP_STACKSIZE overrides GOMP_STACKSIZE, which overrides OMP_STACKSIZE. Default: See the description for OMP_STACKSIZE. |
OpenMP* Environment Variables (OMP_) and Extensions (KMP_) | |
OMP_CANCELLATION |
Activates cancellation of the innermost enclosing region of the type specified. If set to TRUE, the effects of the cancel construct and of cancellation points are enabled and cancellation is activated. If set to FALSE, cancellation is disabled and the cancel construct and cancellation points are effectively ignored. NoteInternal barrier code will work differently depending on whether the cancellation is enabled. Barrier code should repeatedly check the global flag to figure out if the cancellation had been triggered. If a thread observes the cancellation it should leave the barrier prematurely with the return value 1 (may wake up other threads). Otherwise, it should leave the barrier with the return value 0. Enables (TRUE) or disables (FALSE) cancellation of the innermost enclosing region of the type specified. Default: FALSE Example: OMP_CANCELLATION=TRUE |
OMP_DISPLAY_ENV |
Enables (TRUE) or disables (FALSE) the printing to stderr of the OpenMP version number and the values associated with the OpenMP environment variable. Possible values are: TRUE, FALSE, or VERBOSE. Default: FALSE Example: OMP_DISPLAY_ENV=TRUE |
OMP_DEFAULT_DEVICE |
Sets the device that will be used in a target region. The OpenMP routine omp_set_default_device or a device clause in a parallelpragma can override this variable. If no device with the specified device number exists, the code is executed on the host. If this environment variable is not set, device number 0 is used. |
OMP_DYNAMIC |
Enables (TRUE) or disables (FALSE) the dynamic adjustment of the number of threads. Default: FALSE Example: OMP_DYNAMIC=TRUE |
OMP_MAX_ACTIVE_LEVELS |
The maximum number of levels of parallel nesting for the program. Default: 1 Syntax: OMP_MAX_ACTIVE_LEVELS=TRUE |
OMP_NESTED |
Enables (TRUE) or disables (FALSE) nested parallelism. Default: FALSE Example: OMP_NESTED=TRUE |
OMP_NUM_THREADS |
Sets the maximum number of threads to use for OpenMP* parallel regions if no other value is specified in the application. The value can be a single integer, in which case it specifies the number of threads for all parallel regions. The value can also be a comma-separated list of integers, in which case each integer specifies the number of threads for a parallel region at a nesting level. The first position in the list represents the outer-most parallel nesting level, the second position represents the next-inner parallel nesting level, and so on. At any level, the integer can be left out of the list. If the first integer in a list is left out, it implies the normal default value for threads is used at the outer-most level. If the integer is left out of any other level, the number of threads for that level is inherited from the previous level. This environment variable applies to the options Qopenmp (Windows) or qopenmp (Linux)). Default: The number of processors visible to the operating system on which the program is executed. Syntax: OMP_NUM_THREADS=value[,value]* |
OMP_PLACES |
Specifies an explicit ordered list of places, either as an abstract name describing a set of places or as an explicit list of places described by nonnegative numbers. An exclusion operator “!” can also be used to exclude the number or place immediately following the operator. For explicit lists, the meaning of the numbers and how the numbering is done for a list of nonnegative numbers are implementation defined. Generally, the numbers represent the smallest unit of execution exposed by the execution environment, typically a hardware thread. Intervals can be specified using the <lower-bound> : <length> : <stride> notation to represent the following list of numbers: "<lower-bound>, <lower-bound> + <stride>, ..., <lower-bound> +(<length>-1)*<stride>."When <stride> is omitted, a unit stride is assumed. Intervals can specify numbers within a place as well as sequences of places. # EXPLICIT LIST EXAMPLE setenv OMP_PLACES "{0,1,2,3},{4,5,6,7},{8,9,10,11},{12,13,14,15}" setenv OMP_PLACES "{0:4},{4:4},{8:4},{12:4}" setenv OMP_PLACES "{0:4}:4:4" The abstract names listed below should be understood by the execution and run-time environment:
When requesting fewer places or more resources than available on the system, the determination of which resources of type abstract_name are to be included in the place list is implementation-defined. The precise definitions of the abstract names are implementation defined. An implementation may also add abstract names as appropriate for the target platform. The abstract name may be appended by a positive number in parentheses to denote the length of the place list to be created, that is abstract_name(num-places). # ABSTRACT NAMES EXAMPLE setenv OMP_PLACES threads setenv OMP_PLACES threads(4) NoteIf any numerical values cannot be mapped to a processor on the target platform the behavior is implementation-defined. The behavior is also implementation-defined when the OMP_PLACES environment variable is defined using an abstract name. |
OMP_PROC_BIND (Windows, Linux) |
Sets the thread affinity policy to be used for parallel regions at the corresponding nested level. Enables (TRUE) or disables (FALSE) the binding of threads to processor contexts. If enabled, this is the same as specifying KMP_AFFINITY=scatter. If disabled, this is the same as specifying KMP_AFFINITY=none. Acceptable values: TRUE, FALSE, or a comma separated list, each element of which is one of the following values: MASTER, CLOSE, SPREAD. Default: FALSE If set to FALSE, the execution environment may move OpenMP* threads between OpenMP* places, thread affinity is disabled, and proc_bind clauses on parallel constructs are ignored. Otherwise, the execution environment should not move OpenMP* threads between OpenMP* places, thread affinity is enabled, and the initial thread is bound to the first place in the OpenMP* place list. If set to MASTER, all threads are bound to the same place as the master thread. If set to CLOSE, threads are bound to successive places, close to where the master thread is bound. If set to SPREAD, the master thread's partition is subdivided and threads are bound to single place successive sub-partitions. NoteKMP_AFFINITY takes precedence over GOMP_CPU_AFFINITY and OMP_PROC_BIND. GOMP_CPU_AFFINITY takes precedence over OMP_PROC_BIND. |
OMP_SCHEDULE |
Sets the run-time schedule type and an optional chunk size. Default: STATIC, no chunk size specified Example syntax: OMP_SCHEDULE="kind[,chunk_size]" NoteSome environment variables are available for both Intel® microprocessors and non-Intel microprocessors, but may perform additional optimizations for Intel® microprocessors than for non-Intel microprocessors. |
OMP_STACKSIZE |
Sets the number of bytes to allocate for each OpenMP* thread to use as the private stack for the thread. Recommended size is 16M. Use the optional suffixes to specify byte units: B (bytes), K (Kilobytes), M (Megabytes), G (Gigabytes), or T (Terabytes) to specify the units. If you specify a value without a suffix, the byte unit is assumed to be K (Kilobytes). This variable does not affect the native operating system threads created by the user program, or the thread executing the sequential part of an OpenMP* program. The kmp_{set,get}_stacksize_s() routines set/retrieve the value. The kmp_set_stacksize_s() routine must be called from sequential part, before first parallel region is created. Otherwise, calling kmp_set_stacksize_s() has no effect. Default (IA-32 architecture): 2M Default (Intel® 64 architecture): 4M Related environment variables: KMP_STACKSIZE (overrides OMP_STACKSIZE). Syntax: OMP_STACKSIZE=value |
OMP_THREAD_LIMIT |
Limits the number of simultaneously-executing threads in an OpenMP* program. If this limit is reached and another native operating system thread encounters OpenMP* API calls or constructs, the program can abort with an error message. If this limit is reached when an OpenMP* parallel region begins, a one-time warning message might be generated indicating that the number of threads in the team was reduced, but the program will continue. This environment variable is only used for programs compiled with the following option: Qopenmp (Windows) or qopenmp (Linux). The omp_get_thread_limit() routine returns the value of the limit. Default: No enforced limit Related environment variable: KMP_ALL_THREADS (overrides OMP_THREAD_LIMIT). Example syntax: OMP_THREAD_LIMIT=value |
OMP_WAIT_POLICY |
Decides whether threads spin (active) or yield (passive) while they are waiting. OMP_WAIT_POLICY=ACTIVE is an alias for KMP_LIBRARY=turnaround, and OMP_WAIT_POLICY=PASSIVE is an alias for KMP_LIBRARY=throughput. Default: Passive Syntax: OMP_WAIT_POLICY=value |
KMP_AFFINITY (Windows, Linux) |
Enables run-time library to bind threads to physical processing units. You must set this environment variable before the first parallel region, or certain API calls including omp_get_max_threads(), omp_get_num_procs() and any affinity API calls. For detailed information on this environment variable, see Thread Affinity Interface. Default: noverbose,warnings,respect,granularity=core,none Default (Windows with multiple processor groups): noverbose,warnings,norespect,granularity=group,compact,0,0 NoteOn Windows with multiple processor groups, the norespect affinity modifier is assumed when the process affinity mask equals a single processor group (which is default on Windows). Otherwise, the respect affinity modifier is used. |
KMP_ALL_THREADS |
Limits the number of simultaneously-executing threads in an OpenMP* program. If this limit is reached and another native operating system thread encounters OpenMP* API calls or constructs, then the program may abort with an error message. If this limit is reached at the time an OpenMP* parallel region begins, a one-time warning message may be generated indicating that the number of threads in the team was reduced, but the program will continue execution. This environment variable is only used for programs compiled with the Qopenmp(Windows) or qopenmp (Linux) option. Default: No enforced limit. |
KMP_BLOCKTIME |
Sets the time, in milliseconds, that a thread should wait, after completing the execution of a parallel region, before sleeping. Use the optional character suffixes: s (seconds), m (minutes), h (hours), or d (days) to specify the units. Specify infinite for an unlimited wait time. Default: 200 milliseconds Related Environment Variable: KMP_LIBRARY environment variable. |
KMP_CPUINFO_FILE |
Specifies an alternate file name for a file containing the machine topology description. The file must be in the same format as /proc/cpuinfo. Default: None |
KMP_DETERMINISTIC_REDUCTION |
Enables (TRUE) or disables (FALSE) the use of a specific ordering of the reduction operations for implementing the reduction clause for an OpenMP* parallel region. This has the effect that, for a given number of threads, in a given parallel region, for a given data set and reduction operation, a floating point reduction done for an OpenMP* reduction clause has a consistent floating point result from run to run, since round-off errors are identical. Default: FALSE |
KMP_DYNAMIC_MODE |
Selects the method used to determine the number of threads to use for a parallel region when OMP_DYNAMIC=TRUE. Possible values: (asat | load_balance | thread_limit), where,
Default (Intel® 64 architecture): load_balance (on all supported OSes) |
KMP_HOT_TEAMS_MAX_LEVEL |
Sets the maximum nested level to which teams of threads will be hot. NoteA hot team is a team of threads optimized for faster reuse by subsequent parallel regions. In a hot team, threads are kept ready for execution of the next parallel region, in contrast to the cold team, which is freed after each parallel region, with its threads going into a common pool of threads. For values of 2 and above, nested parallelism should be enabled. Default: 1 |
KMP_HOT_TEAMS_MODE |
Specifies the run-time behavior when the number of threads in a hot team is reduced. Possible values:
Default: 0 |
KMP_HW_SUBSET |
Specifies the subset of available hardware resources for uniform hardware topology hierarchy. It is ignored for non-uniform topology. The subset is specified in terms of number of units per upper layer unit starting from top layer downwards. E.g. the number of sockets (top layer units), cores per socket, and the threads per core, to use with an OpenMP* application, as an alternative to writing complicated explicit affinity settings or a limiting process affinity mask. You can also specify an offset value to set which resources to use. An extended syntax is available when KMP_TOPOLOGY_METHOD=hwloc. Depending on what resources are detected, you may be able to specify additional resources, such as NUMA nodes and groups of hardware resources that share certain cache levels. Basic syntax: num_unitsID[@offset] [,num_unitsID[@offset]...] Supported unit IDs are not are not case-sensitive.
NoteThe hardware cache can be specified as a unit, e.g. L2 for L2 cache, or LL for last level cache.Extended syntax when KMP_TOPOLOGY_METHOD=hwloc: Additional IDs can be specified if detected. For example:
When any numa or tile units are specified in KMP_HW_SUBSET, the KMP_TOPOLOGY_METHOD will be automatically set to hwloc,so there is no need to set it explicitly. If you don't specify one or more types of resource, such as socket or thread, all available resources of that type are used. The run-time library prints a warning, and the setting of KMP_HW_SUBSET is ignored if:
This variable does not work if the OpenMP* affinity is set to disabled. Default: If omitted, the default value is to use all the available hardware resources. Examples:
To see the result of the setting, you can specify verbosemodifier in KMP_AFFINITY environment variable. The OpenMP run-time library will output to stderr stream the information about discovered HW topology before and after the KMP_HW_SUBSET setting was setting applied. For example, on Intel® Xeon Phi™ 7210 cpu in SNC-4 Clustering Mode, the setting KMP_AFFINITY=verbose KMP_HW_SUBSET=1N,1L2,1L1,1T outputs various verbose information to stderr, including the following lines about discovered HW topology before and after KMP_HW_SUBSET was applied:
|
KMP_INHERIT_FP_CONTROL |
Enables (TRUE) or disables (FALSE) the copying of the floating-point control settings of the master thread to the floating-point control settings of the OpenMP* worker threads at the start of each parallel region. Default: TRUE |
KMP_LIBRARY |
Selects the OpenMP* run-time library execution mode. The values for this variable are serial, turnaround, or throughput. Default: throughput |
KMP_PLACE_THREADS |
Deprecated; use KMP_HW_SUBSET instead. |
KMP_SETTINGS |
Enables (TRUE) or disables (FALSE) the printing of OpenMP* run-time library environment variables during program execution. Two lists of variables are printed: user-defined environment variables settings and effective values of variables used by OpenMP* run-time library. Default: FALSE |
KMP_STACKSIZE |
Sets the number of bytes to allocate for each OpenMP* thread to use as its private stack. Recommended size is 16m. Use the optional suffixes to specify byte units: B (bytes), K (Kilobytes), M (Megabytes), G (Gigabytes), or T (Terabytes) to specify the units. If you specify a value without a suffix, the byte unit is assumed to be K (Kilobytes). KMP_STACKSIZE overrides GOMP_STACKSIZE, which overrides OMP_STACKSIZE.Default (IA-32 architecture): 2m Default (Intel® 64 architecture): 4m |
KMP_TOPOLOGY_METHOD |
Forces OpenMP* to use a particular machine topology modeling method. Possible values are:
|
KMP_VERSION |
Enables (TRUE) or disables (FALSE) the printing of OpenMP run-time library version information during program execution. Default: FALSE |
KMP_WARNINGS |
Enables (TRUE) or disables (FALSE) displaying warnings from the OpenMP run-time library during program execution. Default: TRUE |
DPC++ Environment Variables |
|
DPCPP_CPU_CU_AFFINITY |
Set thread affinity to CPU. The value and meaning is the following:
This environment variable is similar to the OMP_PROC_BIND variable used by OpenMP*. Default: Not set |
DPCPP_CPU_SCHEDULE |
Specify the algorithm for scheduling work-groups by the scheduler. Currently, DPC++ uses TBB for scheduling. The value selects the petitioner used by the TBB scheduler. The value and meaning is the following:
Default: dynamic |
DPCPP_CPU_NUM_CUS |
Set the numbers threads used for kernel execution. To avoid over subscription, maximum value of DPCPP_CPU_NUM_CUS should be the number of hardware threads. If DPCPP_CPU_NUM_CUS is 1, all the workgroups are executed sequentially by a single thread and this is useful for debugging. This environment variable is similar to OMP_NUM_THREADS variable used by OpenMP*. Default: Not set. Determined by TBB. |
DPCPP_CPU_PLACES |
Specify the places that affinities are set. The value is { sockets | numa_domains | cores | threads }. This environment variable is similar to the OMP_PLACES variable used by OpenMP*. If value is numa_domains, TBB NUMA API will be used. This is analogous to OMP_PLACES=numa_domains in the OpenMP 5.1 Specification. TBB task arena is bound to numa node and SYCL nd range is uniformly distributed to task arenas. DPCPP_CPU_PLACES is suggested to be used together with DPCPP_CPU_CU_AFFINITY. Default: cores |
The following table summarizes CPU environment variables that are recognized at run time.
Runtime configuration |
Default value |
Description |
---|---|---|
CL_CONFIG_CPU_FORCE_PRIVATE_MEM_SIZE |
32KB |
Forces CL_DEVICE_PRIVATE_MEM_SIZE for the CPU device to be the given value. The value must include the unit; for example: 8MB, 8192KB, 8388608B. NoteYou must compile your host application with sufficient stack size. |
CL_CONFIG_CPU_FORCE_LOCAL_MEM_SIZE |
32KB |
Forces CL_DEVICE_LOCAL_MEM_SIZE for CPU device to be the given value. The value needs to be set with size including units, examples: 8MB, 8192KB, 8388608B. NoteYou must compile your host application with sufficient stack size. Our recommendation is to set the stack size equal to twice the local memory size to cover possible application and OpenCL Runtime overheads. |
CL_CONFIG_CPU_EXPENSIVE_MEM_OPT |
0 |
A bitmap indicating enabled expensive memory optimizations. These optimizations may lead to more JIT compilation time, but give some performance benefit. NoteCurrently, only the least significant bit is available.Available bits:
|
CL_CONFIG_CPU_STREAMING_ALWAYS |
False |
Controls whether non-temporal instructions are used. |