Segmentation fault in dfloor benchmark at large input size (length =1,000,000)

[Deepak91168]

Description

When running the benchmark ./benchmark.length.out for @stdlib/math/strided/special/dfloor on Fedora 42, the benchmark crashes with a segmentation fault at the largest tested length (1,000,000). Smaller lengths (up to 100,000) complete successfully without error.

To Reproduce this error:

cd /lib/node_modules/@stdlib/math/strided/special/dfloor/benchmark/c

Run:

./benchmark.length.out

Last few line of output:

  ---
  iterations: 1000
  elapsed: 0.221682072
  rate: 4510.964700011
  ...
ok 14 benchmark finished
# c::dfloor:len=100000
  ---
  iterations: 1000
  elapsed: 0.184779882
  rate: 5411.844551710
  ...
ok 15 benchmark finished
# c::dfloor:len=1000000
[1]    98416 segmentation fault (core dumped)  ./benchmark.length.out

Reason for this issue non-dynamic implementation of:
/lib/node_modules/@stdlib/math/strided/special/dfloor/src/dfloor.c

Temporally can be fixed by increasing stack memory limit to unlimited using:

ulimit -s unlimited

Potential Permanent fix:
Dynamic implementation of dfloor.c using malloc

Related issue

Any similar C file with non dynamic implementation can cause similar issue on other systems

Questions

Can we refactor all C files to use dynamic memory allocation wherever necessary to improve scalability and prevent issues similar segmentation faults?

[stdlib-bot]

👋 Hi there! 👋

And thank you for opening your first issue! We will get back to you shortly. 🏃 💨

If you have any further questions while waiting for a response, please join our Gitter channel to chat with project maintainers and other community members.

[Deepak91168]

@Neerajpathak07

[kgryte]

This is a duplicate of #369.

[navyansh007]

I have raised a PR #7317 that resolves this issue.

[Neerajpathak07]

@Deepak91168 A word of advice while creating a issue check if there is any issue or PR that aims to resolves it!!
It's a big help to the maintainers as they are loaded with work recently XD
Going ahead and closing this one!!

[Deepak91168]

@Deepak91168 A word of advice while creating a issue check if there is any issue or PR that aims to resolves it!! It's a big help to the maintainers as they are loaded with work recently XD Going ahead and closing this one!!

Ok! I will make sure for any future work!