There are two bytecode entry points in src/hotspot/cpu/x86/templateTable_x86.cpp where new AVX10.2 floating point instructions can be used:

1. TemplateTable::convert()
2. TemplateTable::float_cmp()

#1 involves float to integer, float to long, double to integer, and double to long conversions. #2 uses floating point comparisons to determine which of three (-1, 0, 1) output values is provided.

The new AVX10.2 floating point conversion and comparison instructions will require fewer instructions to properly post-process the initial outputs. So, it's worth using them if a platform supports it.