[Documentation] memory arena, reuse, and pattern

[mohsenMahmoodzadeh]

As described in Python API Doc, there are some params in session options coressponding to memory configuration such as:

• enable_cpu_mem_arena
• enable_mem_usage
• enable_mem_pattern

There are some descriptions for them tough, but I can not understaned their usage and the concepts behind them precisely.

Can anyone give me some clarity about these parameters?

[faxu]

Mem arena:

onnxruntime/onnxruntime/core/framework/session_options.h

Line 75 in 7d58332

// enable the memory arena on CPU

Memory reuse:

onnxruntime/onnxruntime/core/framework/session_options.h

Line 70 in 7d58332

// Enable memory resue in memory planning. Allows to reuse tensor buffer between tensors if they are of

Memory pattern:

onnxruntime/onnxruntime/core/framework/session_options.h

Line 63 in 7d58332

// enable the memory pattern optimization.

More reading on memory management:

• https://onnxruntime.ai/docs/get-started/with-c.html#features
• https://onnxruntime.ai/docs/performance/tune-performance.html#tips-for-tuning-performance

[mohsenMahmoodzadeh]

Thank you for your responses. But I have some problems with Memory Reuse and Memory Pattern Optimization.

Can you give some conceptual background about what they are and what they do? I read the references you mentioned but couldn't understand precisely their usages.