Benchmarks to compare LGT8F328p and ATmega328p

[SuperUserNameMan]

Hello,

Here is a list of benchmarks you can use to compare performance between LGT8f328p and ATmega328p :

• https://github.com/SuperUserNameMan/arduino_fbench : quick and dirty port of the "John Walker's Floating Point Benchmark" (aka Fbench) to Arduino. Can be used to compare float performance between LGT and ATmega, as well as GCC compiler optimization levels.

• https://github.com/SuperUserNameMan/Arduino_prime_number_benchmark : forked from Ralph Bacon's repo. This version was compatible with the original Larduino_HSP SDK and could also compile directly using the standard Arduino Nano SDK.

• https://github.com/SuperUserNameMan/arduino_flops_benchmark : this is a port of the "Linpack Floating Point benchmark" to evaluates the performance in FLOPS. (I wanted to compare how far away we are from the TOP500 supercomputers)

If you have other benchmarks, please share them !

[dbuezas]

I really like these benchmarks! do you mind if I paste your tables here?

[SuperUserNameMan]

I don't mind. :-)

[dbuezas]

Work by @SuperUserNameMan

arduino_fbench

source: https://github.com/SuperUserNameMan/arduino_fbench

optimization	LGT8F328p @ 32MHz	LGT8F328p @ 16MHz	ATmega328p @ 16MHz
Os	4.670s
O0	5.105s
O1	4.657s
O2	4.651s
O3	4.651s	9.301s	10.894s
Ofast	4.605s	9.209s	10.774s

Arduino_prime_number_benchmark

source: https://github.com/SuperUserNameMan/arduino_flops_benchmark
forked from : https://github.com/RalphBacon/LGT8F328P-Arduino-Clone-Chip-ATMega328P

Calculate 10,000 prime numbers :

MCU	Clock	Benchmark	SDK
ATmega328p	16MHz	121 s.
ATmega32u4	16MHz	120 s.
ATmega328p	8MHz	242 s.
LGT8F328p	16MHz	118 s.	default Arduino Nano SDK (old bootloader)
LGT8F328p	16MHz	118 s.	Larduino_HSP SDK
LGT8F328p	32MHz	59 s.	Larduino_HSP SDK
STM32F103	72MHz	3 s.	Arduino Core STM32 SDK

arduino_flops_benchmark

source: https://github.com/SuperUserNameMan/arduino_flops_benchmark
original port from : https://gist.github.com/projectgus/8279947

Arduino FLOPS benchmark based on "Linpack Floating Point benchmark"

optimization level	LGT8F328p @ 32MHz	LGT8F328p @ 16MHz	ATmega328p @ 16 MHz
Os	205 kFLOPS
O0	160 kFLOPS
O1	204 kFLOPS
O2	205 kFLOPS
O3	220 kFLOPS	109kFLOPS	90 kFLOPS
Ofast	226 kFLOPS	112kFLOPS	92 kFLOPS

[youxiaojie]

like it!try 328d!why there is not a operation system for avr?32Mhz is similar to i486!
my 328d

21:22:10.343 -> F_CPU value: 16000000
21:22:10.343 -> Clock cycles per µsecond: 16
21:22:10.376 -> LGT8Fx detectedCLKPR : 1 -> 1
21:22:10.410 -> GUID of this device: F1437BFE
21:22:10.443 -> Prime calculation starting.
21:24:09.991 ->
21:24:09.991 -> Found 10000 primes in 118 seconds
21:24:10.024 -> Highest prime found was: 104729
21:25:10.708 -> Prime calculation starting (again)
21:27:10.518 ->
21:27:10.518 -> Found 10000 primes in 118 seconds
21:27:10.551 -> Highest prime found was: 104729
21:28:11.301 -> Prime calculation starting (again)

by the way, how to select sdk?
the clkpr correct?

21:34:46.484 -> Measured run time in seconds should be divided by 1.00
21:34:46.550 -> to normalise for reporting results. For archival results,
21:34:46.616 -> adjust iteration count so the benchmark runs about five minutes.
21:34:46.649 ->
21:34:55.919 -> Timing: 9209 ms
21:34:55.919 -> Marginal ray 47.09478759765 0.04178471088
21:34:55.985 -> Paraxial ray 47.08371734619 0.04177865028
21:34:56.018 -> Longitudinal spherical aberration: -0.01107025146
21:34:56.084 -> (Maximum permissible): 0.05306754112
21:34:56.150 -> Offense against sine condition (coma): 0.00008904934
21:34:56.184 -> (Maximum permissible): 0.00249999976
21:34:56.250 -> Axial chromatic aberration: 0.00451278686
21:34:56.316 -> (Maximum permissible): 0.05306754112

21:41:44.243 ->
21:41:44.243 -> The LINPACK benchmark.
21:41:44.276 -> Language: C
21:41:44.276 -> Datatype: Double precision real
21:41:44.309 -> Matrix order N = 8
21:41:44.375 -> Leading matrix dimension LDA = 9
21:41:44.408 ->
21:41:44.408 -> Norm. Resid Resid MACHEP X[1] X[N]
21:41:44.474 ->
21:41:44.474 -> inf 0.00000053644180 0.00000000000000 1.00000047683715 1.00000035762786
21:41:44.541 -> Factor Solve Total MFLOPS Unit Cray-Ratio
21:41:44.640 ->
21:41:44.640 -> 0.003012657 0.001155853 0.004168510 0.112590181 17.763536453 0.074437675
21:41:44.706 -> LINPACK_BENCH
21:41:44.739 -> Normal end of execution.
21:41:44.772 ->
21:41:44.772 -> Starting benchmark...
21:41:44.772 ->
21:41:44.772 -> LINPACK_BENCH
21:41:44.805 -> C version
21:41:44.805 ->
21:41:44.805 -> The LINPACK benchmark.
21:41:44.839 -> Language: C
21:41:44.839 -> Datatype: Double precision real
21:41:44.872 -> Matrix order N = 8
21:41:44.938 -> Leading matrix dimension LDA = 9
21:41:44.971 ->
21:41:44.971 -> Norm. Resid Resid MACHEP X[1] X[N]
21:41:45.037 ->
21:41:45.037 -> inf 0.00000053644180 0.00000000000000 1.00000047683715 1.00000035762786
21:41:45.103 -> Factor Solve Total MFLOPS Unit Cray-Ratio
21:41:45.203 ->
21:41:45.203 -> 0.003012657 0.001155853 0.004168510 0.112590181 17.763536453 0.074437675
21:41:45.269 -> LINPACK_BENCH
21:41:45.302 -> Normal end of execution.
21:41:45.302 ->