However, using the large SIMD registers in OS kernels leads to performance problems. The reason is that the kernel needs to restore all registers to their original state before continuing an interrupted program. This means that the kernel has to save the complete SIMD state to main memory on each system call or hardware interrupt. Since the SIMD state is very large (512–1600 bytes) and interrupts can occur very often, these additional save/restore operations considerably harm performance. To avoid this, we disable SIMD for our kernel (not for applications running on top!).

A problem with disabling SIMD is that floating point operations on x86_64 require SIMD registers by default. To solve this problem, we add the soft-float feature, which emulates all floating point operations through software functions based on normal integers.

MMX: The Multi Media Extension instruction set was introduced in 1997 and defines eight 64 bit registers called mm0 through mm7. These registers are just aliases for the registers of the x87 floating point unit.

SSE is used by default by clang (and apparently rustc) when targeting x86_64 because being able to execute some level of SSE instruction (i think SSE2) is a requirement for x86_64 compatibility. Floating point math without SSE has severe performance issues, mostly because of a bad design of the original x87 coprocessor instruction set (which was designed when optimizations of modern cpus did not exist yet). For example, converting a floating point number to an integer without SSE (discarding the decimal part) is a slow operation that flushes pipelines and causes a lot of other bad things.