#include <compile_time_options.h>

Include dependency graph for instrumentation.h:

This graph shows which files directly or indirectly include this file:

Macros
Likwid support
#define	LIKWID_MARKER_INIT

#define	LIKWID_MARKER_THREADINIT

#define	LIKWID_MARKER_CLOSE

#define	LIKWID_MARKER_START(opt)

#define	LIKWID_MARKER_STOP(opt)

NVTX support
#define	NVTX_MARKER_INIT

#define	NVTX_MARKER_CLOSE

#define	NVTX_MARKER_START(opt)

#define	NVTX_MARKER_STOP(opt)

Clang LSAN support
#define	LSAN_DISABLE

#define	LSAN_ENABLE

Macro Definition Documentation

◆ LIKWID_MARKER_INIT

#define LIKWID_MARKER_INIT

Macro initializing likwid instrumentation. Used in main().

Definition at line 22 of file instrumentation.h.

◆ LIKWID_MARKER_THREADINIT

#define LIKWID_MARKER_THREADINIT

Macro initializing likwid instrumentation on a worker thread. Used in main().

Definition at line 28 of file instrumentation.h.

◆ LIKWID_MARKER_CLOSE

#define LIKWID_MARKER_CLOSE

Macro finalizing likwid instrumentation. Used in main().

Definition at line 33 of file instrumentation.h.

◆ LIKWID_MARKER_START

#define LIKWID_MARKER_START ( opt )

A set of macros that start and stop likwid instrumentation (if support for likwid is enabled). We currently annotate all named host compute loops, i.e., cpu_simd_loop() and cpu_reduction_loop() in loop.h. Usage:

LIKWID_MARKER_START("string identifier")
// critical compute kernel section
LIKWID_MARKER_STOP("string identifier")

Definition at line 46 of file instrumentation.h.

◆ LIKWID_MARKER_STOP

#define LIKWID_MARKER_STOP ( opt )

A set of macros that start and stop likwid instrumentation (if support for likwid is enabled). We currently annotate all named host compute loops, i.e., cpu_simd_loop() and cpu_reduction_loop() in loop.h. Usage:

LIKWID_MARKER_START("string identifier")
// critical compute kernel section
LIKWID_MARKER_STOP("string identifier")

Definition at line 51 of file instrumentation.h.

◆ NVTX_MARKER_INIT

#define NVTX_MARKER_INIT

Macro starting the CUDA profiler. Used in main().

Definition at line 75 of file instrumentation.h.

◆ NVTX_MARKER_CLOSE

#define NVTX_MARKER_CLOSE

Macro stopping the CUDA profiler. Used in main().

Definition at line 80 of file instrumentation.h.

◆ NVTX_MARKER_START

#define NVTX_MARKER_START ( opt )

A set of macros that push and pop a named NVTX range (if support for NVTX is enabled). Such a range shows up as a labeled interval in the timeline of the CUDA profiler and allows to attribute individual kernel launches to the compute loop they originate from. We currently annotate all named device compute loops, i.e., gpu_loop() and gpu_reduction_loop() in loop.h. Usage:

NVTX_MARKER_START("string identifier")
// critical compute kernel section
NVTX_MARKER_STOP("string identifier")

Definition at line 96 of file instrumentation.h.

◆ NVTX_MARKER_STOP

#define NVTX_MARKER_STOP ( opt )

A set of macros that push and pop a named NVTX range (if support for NVTX is enabled). Such a range shows up as a labeled interval in the timeline of the CUDA profiler and allows to attribute individual kernel launches to the compute loop they originate from. We currently annotate all named device compute loops, i.e., gpu_loop() and gpu_reduction_loop() in loop.h. Usage:

NVTX_MARKER_START("string identifier")
// critical compute kernel section
NVTX_MARKER_STOP("string identifier")

Definition at line 101 of file instrumentation.h.

◆ LSAN_DISABLE

#define LSAN_DISABLE

Explicitly disable/enable the LLVM/Clang LeakSanitiver

LSAN_DISABLE
// Calling some external code path that is leaky and that we cannot
// control...
LSAN_ENABLE

Definition at line 121 of file instrumentation.h.

◆ LSAN_ENABLE

#define LSAN_ENABLE