Turn on the memprof-limits engine for the duration of a scope, e.g. at the start of your program. Simply doing so should have no impact on performance. This is necessary before using any of the functions below.

A result is either the return value of a successful computation, or, in case the computation has been interrupted, a (private) exception raised by memprof-limits or any other exception (bug, uncaught exception, or other interrupt) that arose during clean-up after the first exception was raised. (This exception can be discarded.)

Once interrupted, the exception will keep being raised by memprof-limits regularly until a result is produced.

Any exception raised in the absence of an interruption by the corresponding limit is unaffected and arises normally.

All functions producing a result raise Failure if the engine is not started.

Executes the argument function under the threat of being interrupted if the global memory (major heap word count) exceeds the limit.

If interrupted, then the program is likely in a state where subsequent checks of memory limits will result in further interrupts being raised in other monitored parts of the program. You can either let other monitored threads be interrupted too until the situation is resolved (for instance by the program shutting down), or you can call Gc.compact () to attempt to free up space. The chance for a monitored thread to be interrupted relative to others is proportional to the allocation rate.

Set the limit in KiB for the major heap word count over which one or more threads (selected with with_global_memory_limit) will be interrupted. A special value of 0 disables the limit.

Default: 0

Executes the argument function under the threat of being interrupted if it allocates more than the limit (in KiB). It counts allocations (approximately) but not deallocations; as such it is a measure of work done which is more portable than wall clock, and can be more suitable (for instance it does not count system calls).

Refer to the documentation for the accuracy of the counting method and for advice about good values for limit.

If two calls to with_allocation_limit are nested, then allocations are counted for both. A call to with_allocation_limit cannot be used to go beyond a previous limit.

The maximal value for ~limit, which is equal to 72'057'594'037'927'935 KiB.

Use the following reimplementation of the Memprof interface for profiling needs.