oneAPI Deep Neural Network Library (oneDNN)
Performance library for Deep Learning
1.96.0
Service

A set of functions that aid in oneDNN debugging and profiling. More...

Classes

struct  dnnl_version_t
 Structure containing version information as per Semantic Versioning More...
 

Macros

#define DNNL_RUNTIME_NONE   0u
 No runtime (disabled)
 
#define DNNL_RUNTIME_SEQ   1u
 Sequential runtime (CPU only)
 
#define DNNL_RUNTIME_OMP   2u
 OpenMP runtime (CPU only)
 
#define DNNL_RUNTIME_TBB   4u
 TBB runtime (CPU only)
 
#define DNNL_RUNTIME_THREADPOOL   8u
 Threadpool runtime (CPU only)
 
#define DNNL_RUNTIME_OCL   256u
 OpenCL runtime.
 
#define DNNL_RUNTIME_SYCL   512u
 SYCL runtime.
 
#define DNNL_RUNTIME_DPCPP   DNNL_RUNTIME_SYCL
 DPC++ runtime.
 
#define DNNL_JIT_PROFILE_NONE   0u
 Disable profiling completely.
 
#define DNNL_JIT_PROFILE_VTUNE   1u
 Enable VTune Amplifier integration.
 
#define DNNL_JIT_PROFILE_LINUX_PERFMAP   2u
 Enable Linux perf integration via perfmap files.
 
#define DNNL_JIT_PROFILE_LINUX_JITDUMP   4u
 Enable Linux perf integration via jitdump files.
 
#define DNNL_JIT_PROFILE_LINUX_JITDUMP_USE_TSC   8u
 Instruct Linux perf integration via jitdump files to use TSC. More...
 
#define DNNL_JIT_PROFILE_LINUX_PERF   (DNNL_JIT_PROFILE_LINUX_JITDUMP | DNNL_JIT_PROFILE_LINUX_PERFMAP)
 Enable Linux perf integration (both jitdump and perfmap)
 

Typedefs

using dnnl::version_t = dnnl_version_t
 Structure containing version information as per Semantic Versioning More...
 

Enumerations

enum  dnnl::status
 Status values returned by the library functions. More...
 
enum  dnnl::cpu_isa
 CPU instruction set flags. More...
 
enum  dnnl_cpu_isa_t
 CPU instruction set flags. More...
 

Functions

dnnl_status_t DNNL_API dnnl_set_verbose (int level)
 Configures verbose output to stdout. More...
 
dnnl_status_t DNNL_API dnnl_set_jit_dump (int enable)
 Configures dumping of JIT-generated code. More...
 
const dnnl_version_t DNNL_API * dnnl_version (void)
 Returns library version information. More...
 
dnnl_status_t DNNL_API dnnl_set_jit_profiling_flags (unsigned flags)
 Sets library profiling flags. More...
 
dnnl_status_t DNNL_API dnnl_set_jit_profiling_jitdumpdir (const char *dir)
 Sets JIT dump output path. More...
 
dnnl_status_t DNNL_API dnnl_set_max_cpu_isa (dnnl_cpu_isa_t isa)
 Sets the maximal ISA the library can dispatch to on the CPU. More...
 
dnnl_cpu_isa_t DNNL_API dnnl_get_effective_cpu_isa (void)
 Gets the maximal ISA the library can dispatch to on the CPU. More...
 
status dnnl::set_verbose (int level)
 Configures verbose output to stdout. More...
 
const version_tdnnl::version ()
 Returns library version information. More...
 
status dnnl::set_jit_dump (int enable)
 Configures dumping of JIT-generated code. More...
 
status dnnl::set_jit_profiling_flags (unsigned flags)
 Sets library profiling flags. More...
 
status dnnl::set_jit_profiling_jitdumpdir (const std::string &dir)
 Sets JIT dump output path. More...
 
status dnnl::set_max_cpu_isa (cpu_isa isa)
 Sets the maximal ISA the library can dispatch to on the CPU. More...
 
cpu_isa dnnl::get_effective_cpu_isa ()
 Gets the maximal ISA the library can dispatch to on the CPU. More...
 

Detailed Description

A set of functions that aid in oneDNN debugging and profiling.

Macro Definition Documentation

◆ DNNL_JIT_PROFILE_LINUX_JITDUMP_USE_TSC

#define DNNL_JIT_PROFILE_LINUX_JITDUMP_USE_TSC   8u

Instruct Linux perf integration via jitdump files to use TSC.

DNNL_JIT_PROFILE_LINUX_JITDUMP must be set too for this to take effect.

Typedef Documentation

◆ version_t

Structure containing version information as per Semantic Versioning

Enumeration Type Documentation

◆ status

enum dnnl::status
strong

Status values returned by the library functions.

Enumerator
success 

The operation was successful.

out_of_memory 

The operation failed due to an out-of-memory condition.

invalid_arguments 

The operation failed because of incorrect function arguments.

unimplemented 

The operation failed because requested functionality is not implemented.

iterator_ends 

Primitive iterator passed over last primitive descriptor.

runtime_error 

Primitive or engine failed on execution.

not_required 

Queried element is not required for given primitive.

◆ cpu_isa

enum dnnl::cpu_isa
strong

CPU instruction set flags.

Enumerator
all 

Any ISA (excepting those listed as initial support)

sse41 

Intel Streaming SIMD Extensions 4.1 (Intel SSE4.1)

avx 

Intel Advanced Vector Extensions (Intel AVX)

avx2 

Intel Advanced Vector Extensions 2 (Intel AVX2)

avx512_mic 

Intel Advanced Vector Extensions 512 (Intel AVX-512) subset for Intel Xeon Phi processors x200 Series.

avx512_mic_4ops 

Intel AVX-512 subset for Intel Xeon Phi processors 7235, 7285, 7295 Series.

avx512_core 

Intel AVX-512 subset for Intel Xeon Scalable processor family and Intel Core processor family.

avx512_core_vnni 

Intel AVX-512 and Intel Deep Learning Boost (Intel DL Boost) support for Intel Xeon Scalable processor family and Intel Core processor family.

avx512_core_bf16 

Intel AVX-512, Intel DL Boost and bfloat16 support for Intel Xeon Scalable processor family and Intel Core processor family.

avx512_core_amx 

Intel AVX-512, Intel DL Boost and bfloat16 support and Intel AMX with 8-bit integer and bfloat16 support (initial support)

avx2_vnni 

Intel AVX2 and Intel Deep Learning Boost (Intel DL Boost) support.

◆ dnnl_cpu_isa_t

CPU instruction set flags.

Enumerator
dnnl_cpu_isa_all 

Any ISA (excepting those listed as initial support)

dnnl_cpu_isa_sse41 

Intel Streaming SIMD Extensions 4.1 (Intel SSE4.1)

dnnl_cpu_isa_avx 

Intel Advanced Vector Extensions (Intel AVX)

dnnl_cpu_isa_avx2 

Intel Advanced Vector Extensions 2 (Intel AVX2)

dnnl_cpu_isa_avx512_mic 

Intel Advanced Vector Extensions 512 (Intel AVX-512) subset for Intel Xeon Phi processors x200 Series.

dnnl_cpu_isa_avx512_mic_4ops 

Intel AVX-512 subset for Intel Xeon Phi processors 7235, 7285, 7295 Series.

dnnl_cpu_isa_avx512_core 

Intel AVX-512 subset for Intel Xeon Scalable processor family and Intel Core processor family.

dnnl_cpu_isa_avx512_core_vnni 

Intel AVX-512 and Intel Deep Learning Boost (Intel DL Boost) support for Intel Xeon Scalable processor family and Intel Core processor family.

dnnl_cpu_isa_avx512_core_bf16 

Intel AVX-512, Intel DL Boost and bfloat16 support for Intel Xeon Scalable processor family and Intel Core processor family.

dnnl_cpu_isa_avx512_core_amx 

Intel AVX-512, Intel DL Boost and bfloat16 support and Intel AMX with 8-bit integer and bfloat16 support (initial support)

dnnl_cpu_isa_avx2_vnni 

Intel AVX2 and Intel Deep Learning Boost (Intel DL Boost) support.

Function Documentation

◆ dnnl_set_verbose()

dnnl_status_t DNNL_API dnnl_set_verbose ( int  level)

Configures verbose output to stdout.

Note
Enabling verbose output affects performance. This setting overrides the DNNL_VERBOSE environment variable.
Parameters
levelVerbosity level:
  • 0: no verbose output (default),
  • 1: primitive information at execution,
  • 2: primitive information at creation and execution.
Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the level value is invalid, and dnnl_success/dnnl::status::success on success.

◆ dnnl_set_jit_dump()

dnnl_status_t DNNL_API dnnl_set_jit_dump ( int  enable)

Configures dumping of JIT-generated code.

Note
This setting overrides the DNNL_JIT_DUMP environment variable.
Parameters
enableFlag value. Set to 0 to disable and set to 1 to enable.
Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the flag value is invalid, and dnnl_success/dnnl::status::success on success.

◆ dnnl_version()

const dnnl_version_t DNNL_API* dnnl_version ( void  )

Returns library version information.

Returns
Pointer to a constant structure containing
  • major: major version number,
  • minor: minor version number,
  • patch: patch release number,
  • hash: git commit hash.

◆ dnnl_set_jit_profiling_flags()

dnnl_status_t DNNL_API dnnl_set_jit_profiling_flags ( unsigned  flags)

Sets library profiling flags.

The flags define which profilers are supported.

Note
This setting overrides DNNL_JIT_PROFILE environment variable.
See also
Profiling oneDNN Performance
Parameters
flagsProfiling flags that can contain the following bits:

Passing DNNL_JIT_PROFILE_NONE disables profiling completely.

Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the flags value is invalid, and dnnl_success/dnnl::status::success on success.

◆ dnnl_set_jit_profiling_jitdumpdir()

dnnl_status_t DNNL_API dnnl_set_jit_profiling_jitdumpdir ( const char *  dir)

Sets JIT dump output path.

Only applicable to Linux and is only used when profiling flags have DNNL_JIT_PROFILE_LINUX_PERF bit set.

After the first JIT kernel is generated, the jitdump output will be placed into temporary directory created using the mkdtemp template 'dir/.debug/jit/dnnl.XXXXXX'.

See also
Profiling oneDNN Performance
Note
This setting overrides JITDUMPDIR environment variable. If JITDUMPDIR is not set, and this function is never called, the path defaults to HOME. Passing NULL reverts the value to default.
The directory is accessed only when the first JIT kernel is being created. JIT profiling will be disabled in case of any errors accessing or creating this directory.
Parameters
dirJIT dump output path.
Returns
dnnl_success/dnnl::status::success if the output directory was set correctly and an error status otherwise.
dnnl_unimplemented/dnnl::status::unimplemented on Windows.

◆ dnnl_set_max_cpu_isa()

dnnl_status_t DNNL_API dnnl_set_max_cpu_isa ( dnnl_cpu_isa_t  isa)

Sets the maximal ISA the library can dispatch to on the CPU.

See dnnl_cpu_isa_t and dnnl::cpu_isa for the list of the values accepted by the C and C++ API functions respectively.

This function has effect only before the first JIT kernel is generated and will return an error afterwards.

This function overrides the DNNL_MAX_CPU_ISA environment variable. The environment variable can be set to the desired maximal ISA name in upper case and with dnnl_cpu_isa prefix removed. For example: DNNL_MAX_CPU_ISA=AVX2.

Note
The ISAs are only partially ordered:
  • SSE41 < AVX < AVX2,
  • AVX2 < AVX512_MIC < AVX512_MIC_4OPS,
  • AVX2 < AVX512_CORE < AVX512_CORE_VNNI < AVX512_CORE_BF16 < AVX512_CORE_AMX,
  • AVX2 < AVX2_VNNI.
See also
CPU Dispatcher Control for more details
Parameters
isaMaximal ISA the library should dispatch to. Pass dnnl_cpu_isa_all/dnnl::cpu_isa::all to remove ISA restrictions (except for ISAs with initial support in the library).
Returns
dnnl_success/dnnl::status::success on success and a dnnl_invalid_arguments/dnnl::status::invalid_arguments if the isa parameter is invalid or the ISA cannot be changed at this time.
dnnl_unimplemented/dnnl::status::unimplemented if the feature was disabled at build time (see Build Options for more details).

◆ dnnl_get_effective_cpu_isa()

dnnl_cpu_isa_t DNNL_API dnnl_get_effective_cpu_isa ( void  )

Gets the maximal ISA the library can dispatch to on the CPU.

See dnnl_cpu_isa_t and dnnl::cpu_isa for the list of the values returned by the C and C++ API functions respectively.

See also
CPU Dispatcher Control for more details
Returns
dnnl_cpu_isa_t value reflecting the maximal ISA the library may dispatch to.

◆ set_verbose()

status dnnl::set_verbose ( int  level)
inline

Configures verbose output to stdout.

Note
Enabling verbose output affects performance. This setting overrides the DNNL_VERBOSE environment variable.
Parameters
levelVerbosity level:
  • 0: no verbose output (default),
  • 1: primitive information at execution,
  • 2: primitive information at creation and execution.
Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the level value is invalid, and dnnl_success/dnnl::status::success on success.

◆ version()

const version_t* dnnl::version ( )
inline

Returns library version information.

Returns
Pointer to a constant structure containing
  • major: major version number,
  • minor: minor version number,
  • patch: patch release number,
  • hash: git commit hash.

◆ set_jit_dump()

status dnnl::set_jit_dump ( int  enable)
inline

Configures dumping of JIT-generated code.

Note
This setting overrides the DNNL_JIT_DUMP environment variable.
Parameters
enableFlag value. Set to 0 to disable and set to 1 to enable.
Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the flag value is invalid, and dnnl_success/dnnl::status::success on success.

◆ set_jit_profiling_flags()

status dnnl::set_jit_profiling_flags ( unsigned  flags)
inline

Sets library profiling flags.

The flags define which profilers are supported.

Note
This setting overrides DNNL_JIT_PROFILE environment variable.
See also
Profiling oneDNN Performance
Parameters
flagsProfiling flags that can contain the following bits:

Passing DNNL_JIT_PROFILE_NONE disables profiling completely.

Returns
dnnl_invalid_arguments/dnnl::status::invalid_arguments if the flags value is invalid, and dnnl_success/dnnl::status::success on success.

◆ set_jit_profiling_jitdumpdir()

status dnnl::set_jit_profiling_jitdumpdir ( const std::string &  dir)
inline

Sets JIT dump output path.

Only applicable to Linux and is only used when profiling flags have DNNL_JIT_PROFILE_LINUX_PERF bit set.

After the first JIT kernel is generated, the jitdump output will be placed into temporary directory created using the mkdtemp template 'dir/.debug/jit/dnnl.XXXXXX'.

See also
Profiling oneDNN Performance
Note
This setting overrides JITDUMPDIR environment variable. If JITDUMPDIR is not set, and this function is never called, the path defaults to HOME. Passing NULL reverts the value to default.
The directory is accessed only when the first JIT kernel is being created. JIT profiling will be disabled in case of any errors accessing or creating this directory.
Parameters
dirJIT dump output path.
Returns
dnnl_success/dnnl::status::success if the output directory was set correctly and an error status otherwise.
dnnl_unimplemented/dnnl::status::unimplemented on Windows.

◆ set_max_cpu_isa()

status dnnl::set_max_cpu_isa ( cpu_isa  isa)
inline

Sets the maximal ISA the library can dispatch to on the CPU.

See dnnl_cpu_isa_t and dnnl::cpu_isa for the list of the values accepted by the C and C++ API functions respectively.

This function has effect only before the first JIT kernel is generated and will return an error afterwards.

This function overrides the DNNL_MAX_CPU_ISA environment variable. The environment variable can be set to the desired maximal ISA name in upper case and with dnnl_cpu_isa prefix removed. For example: DNNL_MAX_CPU_ISA=AVX2.

Note
The ISAs are only partially ordered:
  • SSE41 < AVX < AVX2,
  • AVX2 < AVX512_MIC < AVX512_MIC_4OPS,
  • AVX2 < AVX512_CORE < AVX512_CORE_VNNI < AVX512_CORE_BF16 < AVX512_CORE_AMX,
  • AVX2 < AVX2_VNNI.
See also
CPU Dispatcher Control for more details
Parameters
isaMaximal ISA the library should dispatch to. Pass dnnl_cpu_isa_all/dnnl::cpu_isa::all to remove ISA restrictions (except for ISAs with initial support in the library).
Returns
dnnl_success/dnnl::status::success on success and a dnnl_invalid_arguments/dnnl::status::invalid_arguments if the isa parameter is invalid or the ISA cannot be changed at this time.
dnnl_unimplemented/dnnl::status::unimplemented if the feature was disabled at build time (see Build Options for more details).

◆ get_effective_cpu_isa()

cpu_isa dnnl::get_effective_cpu_isa ( )
inline

Gets the maximal ISA the library can dispatch to on the CPU.

See dnnl_cpu_isa_t and dnnl::cpu_isa for the list of the values returned by the C and C++ API functions respectively.

See also
CPU Dispatcher Control for more details
Returns
dnnl_cpu_isa_t value reflecting the maximal ISA the library may dispatch to.