The majority of float operations are implemented using SSE but some are using the old X87 FP co-processor of ancient 486. While this guarantee and old retrocompatibility (all X86 has X87 instruction set implemented) forces Cosmos to that lowest common denominator.
This is the list of opcodes that are implemented using X87 instead of SSE:
The majority of float operations are implemented using SSE but some are using the old X87 FP co-processor of ancient 486. While this guarantee and old retrocompatibility (all X86 has X87 instruction set implemented) forces Cosmos to that lowest common denominator.
This is the list of opcodes that are implemented using X87 instead of SSE: