Thing Done

← The fleet

Feather M4

Cortex-M4F (SAMD51) · Arm Cortex-M · 120 MHz

Feather M4 Express
Board
Feather M4 Express
Core
Cortex-M4F (SAMD51)
Clock
120 MHz
float / double
32-bit / 64-bit
uint8_t add
8.98 M/s
rank 10 of 16 · 111 ns
uint32_t add
9.71 M/s
rank 10 of 16 · 103 ns
uint32_t divide
6.53 M/s
rank 8 of 16 · 153 ns
float multiply
8.98 M/s
rank 6 of 16 · 111 ns
Cycles per uint32 add
12.4 cycles
rank 5 of 16 — the clock-independent number

What every operation costs

Every numeric type, all five operations, as clock cycles per operation: the shape of this part's cost independent of its clock. The bars are the story of where it falls off a cliff — divide, 64-bit, floating point.

The same cells as nanoseconds per operation, median of three samples. Darker is slower; the small number is this board's rank in the fleet for that cell (1 = fastest of 16).

AddSubtractMultiplyDivideModulo
uint8_t 111 ns #10 of 16 111 ns #10 of 16 111 ns #10 of 16 143 ns #8 of 16 157 ns #8 of 16
uint16_t 111 ns #10 of 16 111 ns #10 of 16 111 ns #10 of 16 164 ns #8 of 16 186 ns #8 of 16
uint32_t 103 ns #10 of 16 103 ns #10 of 16 103 ns #10 of 16 153 ns #8 of 16 161 ns #8 of 16
uint64_t 161 ns #10 of 14 161 ns #9 of 14 178 ns #7 of 14 1.13 µs #7 of 14 1.13 µs #7 of 14
float 111 ns #6 of 16 111 ns #6 of 16 111 ns #6 of 16 211 ns #5 of 16 1.15 µs #10 of 16
double 835 ns #10 of 16 839 ns #10 of 16 587 ns #7 of 16 794 ns #5 of 16 2.35 µs #10 of 16

In the fleet

Head to head