chrono | High-performance C++ library for multiphysics and multibody

The non-OpenMP vectorizer is defeating your benchmark with loop inversion.
Make your function __attribute__((noinline, noclone)) to stop GCC from inlining it into the repeat loop. For cases like this with large enough functions that call/ret overhead is minor, and constant propagation isn't important, this is a pretty good way to make sure that the compiler doesn't hoist work out of the loop.

And in future, check the asm, and/or make sure the benchmark time scales linearly with the iteration count. e.g. increasing 500 up to 1000 should give the same average time in a benchmark that's working properly, but it won't with -O3. (Although it's surprisingly close here, so that smell test doesn't definitively detect the problem!)

After adding the missing #pragma omp simd to the code, yeah I can reproduce this. On i7-6700k Skylake (3.9GHz with DDR4-2666) with GCC 10.2 -O3 (without -march=native or -fopenmp), I get 18266, but with -O3 -fopenmp I get avg time 39772.

With the OpenMP vectorized version, if I look at top while it runs, memory usage (RSS) is steady at 771 MiB. (As expected: init code faults in the two inputs, and the first iteration of the timed region writes to result, triggering page-faults for it, too.)

But with the "normal" vectorizer (not OpenMP), I see the memory usage climb from ~500 MiB until it exits just as it reaches the max 770MiB.

So it looks like gcc -O3 performed some kind of loop inversion after inlining and defeated the memory-bandwidth-intensive aspect of your benchmark loop, only touching each array element once.

The asm shows the evidence: GCC 9.3 -O3 on Godbolt doesn't vectorize, and it leaves an empty inner loop instead of repeating the work.

I talked the problem in more steps, first thing I made the run reproducible:

In order to fix this error, you should precise the path to g++. In my case, here is the updated content of my compilers.yaml file:

As you can see in the error, compiler 'gcc@10.2.0' does not support compiling C++ programs.

In order to display the compilers, use the command:

Firstly, your matrix multiply algorithm is over complex than a normal one(Or it's just wrong), you may reference the wiki for a typical algorithm:

Input: matrices A and B

Let C be a new matrix of the appropriate size

For i from 1 to n:

For j from 1 to p:

Let sum = 0

For k from 1 to m:

Set sum ← sum + Aik × Bkj

Set Cij ← sum

Return C

There is a critical bug in your code, you haven't initialized the result matrix.

So the fixed code may like this:

Those are not a common field, they are completely unrelated fields. The fact that they share a name is, as far as the Rust compiler is concerned, an insignificant coincidence. You need to use pattern matching to get the field from either case

There's a bug in the spec that is in the process of being fixed. And VS2019 faithfully reproduced the spec. Wrap your format string in string{}, or give it a trailing s literal to turn it into a string, and this will work around the bug.

In essence you don't need to include either header into the other. This fixes the circular includes. In ROS_Topic_Thread.h, friend class Master_Thread; already forward declares Mater_Thread, so you don't need the Master_Thread.h header. In Master_Thread.h, you can also forward declare class ROS_Topic_Thread; before the declaration of the Master_Thread class, since you only use references to ROS_Topic_Thread in the transfer_to_master_buffer method.

It is likely that the Python module is bound to functions compiled in a language such as c++ or using cython. The implementation of the python module may thus have better performance, depending on how it reads from the file or processes data.