Common Arm processor Peak performance

Introduce

The following table shows the GFLOPS of single-core of some common ARM CPUs using MegPeak tests.

Build(see here) MegPeak and copy MegPeak to the device you want to test, then just run(see here) it!

Platform	CPU	Architecture	Frequence(GHz)	GFLOPS	FLOPS/Cycle
/	Cortex-A7	ARMV7	1.0	1.902002	1.9020
/	Cortex-A53	ARM64	1.3	9.844987	7.5731
Xiaomi Mi 9 Core 0	Cortex-A55	ARM64	1.8	13.566583	7.537
raspberry-pi 4b	Cortex-A72	ARM64	1.5	11.664233	7.776
Xiaomi Mi 9 Core 6	Cortex-A76	ARM64	2.42	38.691399	15.988
Realme X7 Pro Core 4	Cortex-A77	ARM64	2.6	41.318005	15.892
Xiaomi Mi 11 Core 6	Cortex-A78	ARM64	2.4	33.781448	14.076
Apple Mac	Apple M1	ARM64	3.2	102.424	32.008

The GFLOPS can tell you the peak computing performance of your device. More important, the FLOPS/Cycle can help you deduce hardware features.

Taking the A55 as an example, considering that the instruction fmla performs two floating-point operations(one multiplication and one addtion) and that the FLOPS/Cycle metric is close to 8, we can deduce that EUs（Execution Engine） have one 128-bit vector multiplication and addtion unit or two 64-bit vector multiplication and addtion units.
Taking A77 as an example, as the FLOPS/Cycle metric is about 16, this means the processor can execute 2 SIMD fmla instruction at one cycle, so the backend of A77 processor must have two SIMD EUs and it must be double issue at least!
Taking Apple M1 as example, it FLOPS/Cycle metric reach 32, so it must have 4 SIMD EUS, the more SIMD EU, the more power it will consume when it works, but Apple M1 is used in laptop, it can bear the power consume.

If you have any question about the above testing results or want to contribute your testing results, please contact us by issue.