Does it really work with any GCC version earlier than 7?

@ikawrakow I can't answer this, do you want me to set the condition to only meet version 7?

do you want me to set the condition to only meet version 7?

I have no strong feelings about this, was just surprised that a) someone is still using GCC 7, and b) the condition was less < 8, so included even earlier versions.

What is the use case for GCC 7 in the year 2024 anyway?

I will mention again that it seems like this PR is adding yet another implementation of _mm256_set_m128i
(MM256_SET_M128I is defined above). I don't know which is preferred or why one is conditionally emulated and the other is unconditionally emulated.

do you want me to set the condition to only meet version 7?

I have no strong feelings about this, was just surprised that a) someone is still using GCC 7, and b) the condition was less < 8, so included even earlier versions.

What is the use case for GCC 7 in the year 2024 anyway?

Hi super @ikawrakow
Because I intended to include llama.cpp natively in openSUSE and SUSE Enterprise, and the distribution natively comes with version 7

@ikawrakow Do I need to do anything else?

I think we should use the MM256_SET_M128I macro everywhere instead of adding another workaround, it was added precisely for this purpose. #2405

Fixed in #5889.