nfinit / ansibench Goto Github PK

BYTE magazine's BYTEmark/Native Mode Benchmarks are an extremely portable and reasonably comprehensive collection of benchmarks that focus on CPU, FPU and memory performance in single-processor systems.

Uwe F. Mayer at the University of Utah hosts a Linux-friendly port of version 2 of these benchmarks on his website that would make a great addition to this package. Because it is a little more complex than the simple benchmarks packaged thus far, re-packaging it in a consistent and modular way will probably take some time.

i want to run this for wasm. Thank you

specifically this line :

CARGS += $(OPTLEVEL) $(CFLAGS) $(XCFLAGS) $(INPUT) -o$(OUT)

adding space between -o and $(OUT) fixed this for me ....

CoreMark was designed by Shay Gal-On of the Embedded Microprocessor Benchmark Consortium as a modern alternative to the classic integer-focused Dhrystone benchmark still frequently used in the industry to measure and compare the performance of microcontrollers and other small embedded devices. It aims to improve on Dhrystone by eliminating library calls in timed portions of the program, providing a standardized set of run and reporting rules and testing proficiency in applications more relevant to embedded systems.

The CoreMark sources are released by the EEMBC under an Apache license and hosted and maintained at this repository. The microcontroller-oriented nature of this benchmark makes it potentially useful for evaluating older and less powerful systems, as well as an alternative to Dhrystone for measuring general integer performance that is less dependent on standard library quality and vulnerable to aggressive compiler optimization.

First of all, thank you for your benchmarks selection!
I encountered a bug when building the nbench and whetstone benchmarks using make, which was that it failed to link libm. I was using gcc 13.2.0 and ubuntu 24.04.
This happened because the -lm flag (in the CFLAGS variable) is before the input files (in the INPUT variable).

I have left a pull request with a solution for this problem.

The Hierarchical INTegration benchmark is built around a novel metric of "quality improvement" over time rather than operations/loop iterations over time as in many other standard benchmarks, and aims to evaluate a complete system, including memory and disk subsystems, rather than the performance of the processor alone.

HINT was previously hosted on an official site that has since been taken down that included a number of source packages, and sources still float elsewhere around the web.

Could you consider adding LICENSE files?

Especially I'd like to know the license of https://github.com/nfinit/ansibench/blob/master/linpack/src/linpack.c