Welcome to My Whetstone Performance Benchmarks Across Languages!

September 5, 2025 · View on GitHub

In this repository, you'll find simple implementations of the popular Whetstone benchmark. Why Whetstone? It's easy to understand, widely adopted, and relatively accurate. This makes it ideal for comparing performance across various technologies—different languages, compilers, and hardware. With Whetstone, you can feel confident you're comparing "oranges to oranges."

I personally use this benchmark every time I buy a new computer, install a new compiler, or update my environment.

Java

You may want to run the benchmark in your favorite IDE (e.g., Eclipse). In that case, configure your run settings to include two arguments—inner loop count and outer loop count—separated by a space in the Arguments tab.

Try starting with: 1000 300 (adjust depending on your CPU speed).

  • Source file: Dwhet.java
  • Easy to adapt for server-side use, such as wrapping into a servlet.

C

  • Source file: dwhet.c
  • Compiles cleanly under GCC 15.2 Some systems may require minor adjustments to the timing functions.

Compile using:

gcc -Ofast dwhet.c -o dwhet_run_fast -lm

Or with Intel Compiler:

icc dwhet.c -o Dwhet_optimized_with_intel_c.exe -O3 -ipo -no-prec-div

Suggested arguments: 3000 1000

C++ (Microsoft Visual Studio 2017–2022)

  • Source file: Dwhet.cpp

Instructions:

  1. Create a new project: Visual C++ Console Application (e.g., name it Dwhet)
  2. Replace the default Dwhet.cpp with the version from this repo.
  3. Go to Project → Properties, and add _CRT_SECURE_NO_WARNINGS; to Preprocessor Definitions

Repeat step 3 for both Debug and Release configurations.

Suggested arguments: 3000 1000 (e.g., for Intel Core i7 @ 2.8 GHz)

Fortran

  • Most compilers use dwhet.f
  • For NVIDIA nvfortran, use: whet_cuda_parallel.cuf

Suggested arguments for dwhet.f: 6000 2000, then adjust.

Example compilation:

Intel Fortran:

ifort dwhet.f -o Dwhet_optimized_with_intel -O3 -ipo -no-prec-div

GFortran:

gfortran dwhet.f -o Dwhet_gcc -Ofast

NVIDIA CUDA Fortran:

If your system supports CUDA 12.7 and you're using nvhpc-25-5-cuda-multi (e.g., in Ubuntu 24.04 on WSL2 under Windows 11 pro), compile with:

nvfortran -cuda whet_cuda_parallel.cuf -o whet_cuda_parallel

Suggested inputs (for GPUs like RTX 4070 with 8GB VRAM):

Enter inner loop count: 3600000 Enter outer loop count: 1200000 Enter number of GPU threads: 8192

Adjust these for your GPU capabilities.

Free Pascal

Tested on Windows 11 pro and Ubuntu 20.04 (via WSL2).

  • Source: whet.pas

Windows (64-bit with Core i7-1360P):

ppcrossx64 -O3 -OoFASTMATH -CpCOREAVX2 -Xs -XX -Sh whet.pas

Ubuntu:

fpc whet.pas

Julia

  • Source: whestone.jl

Compile with:

julia --optimize=3 --check-bounds=no whestone.jl

Try: 1000 3000, then adjust as needed.

Rust

  • File: main.rs under src/
  • Project structure supports Cargo

Compile:

cargo build --release

Run:

cargo run --release

Try: 1000 3000 as initial loop parameters.

My Articles in Java Developer's Journal

http://web.archive.org/web/20190703063724/http://anatolykrivitsky.sys-con.com/

Legal Text Related to the Code

(c) 2002–2025 Anatoly S. Krivitsky, Ph.D. All rights reserved.

You are granted permission to use, distribute, and publish this code freely, provided the above copyright is retained.

Contact: akrivitsky@yahoo.com akrivitsky@gmail.com

Disclaimer of Liability

This code is provided "as is" without warranty of any kind. Use at your own risk. The author is not liable for any damages, direct or indirect, including (but not limited to) lost profits, data corruption, or hardware failure resulting from use or inability to use of the code.