3 ms·
Historically, float to integer conversions in C on x86 were very slow due to the compiler reconfiguring the FPU rounding mode for each conversion.
by RossBencina 3y ago
Historically, float to integer conversions in C on x86 were very slow due to the compiler reconfiguring the FPU rounding mode for each conversion.
- raphlinus 3y agoHistorically yes. This is most definitely not true since processors had SSE3 (~2004), using the FISTTP instruction, or I think you can also use the packed float to integer instructions like CVTTPS2PI as far back as SSE (1999).