The command is definitely
pico_set_float_implementation(PicoMite pico_dcp) and it makes the difference, if I use pico_float_dcp the linker fails with library not found
Setting that bit makes no difference
The HSTX handling is all done on core 1, all in RAM, with nothing else on that core and a dedicated stack.
I've confirmed no FP activity in any of my interrupts - 1mSec timer, UART RX/TX and 44100Hz PWM wrap
The main code doesn't use single precision FP except in the MP3 codec (github dr_mp3)
pico_set_float_implementation(PicoMite pico_dcp) and it makes the difference, if I use pico_float_dcp the linker fails with library not found
Setting that bit makes no difference
The HSTX handling is all done on core 1, all in RAM, with nothing else on that core and a dedicated stack.
I've confirmed no FP activity in any of my interrupts - 1mSec timer, UART RX/TX and 44100Hz PWM wrap
The main code doesn't use single precision FP except in the MP3 codec (github dr_mp3)
Interesting, but I don't think it is that because the impact at 372MHz (720P) is less than 324MHz (XGA)For what it's worth, in my overclocking tests when pushing too hard it's the ARM FPU (single precision) failing first by loosing precision, maybe delaying too.
Statistics: Posted by matherp — Mon Apr 21, 2025 6:37 pm