Thread safety of ALP: Should static arrays in encoder be thread_local for multithreaded compression?

[deepakg-2203]

Hi,
I am using ALP for multithreaded compression and noticed that in "encoder.hpp",the function "encode_simdized()" uses static arrays like this:

`alignas(64) static PT ENCODED_VALUE_ARR[1024];

alignas(64) static PT VALUE_ARR_WITHOUT_SPECIALS[1024];

alignas(64) static UT TMP_INDEX_ARR[1024];`

Since these are static, they are shared between threads and could cause data races during parallel compression.
Would it be correct to change these to thread_local static to make the function thread-safe? Or is there a recommended way to use ALP safely in a multithreaded context?

Thanks!

[azimafroozeh]

Hi,

You are completely right. The static arrays should be changed to thread_local to make the encoder safe in multithreaded contexts. This change will not cause any issues for ALP.

The recommended long-term approach is to allocate these buffers through a buffer manager to monitor and limit total memory usage.

I’d be happy to review a PR that makes the initial change (static → thread_local).

Thanks for catching this!