[1] https://www.intel.com/content/www/us/en/docs/fortran-compiler/developer-guide-reference/2023-0/code-coverage-tool.html
链接已复制
Is there any indication when the compiler team will implement this feture in IFX?
Since IFORT is no longer available, it is a problem to properly implement detailed unit tests.
Hi, Fran:
We try to survey our customer community about different feature requests. If possible, would you please let us know your company/institute name, and the name of the app you plan to apply the code coverage functionality to? Is there a public website for your target app? Thanks,
Shiquan
Hi, Shiquan,
I am more of a private user. I used to work in a company where Fortran was a key lnguage for space system applicaiton development and I am certain that they suffer the same issue with the code coverage using IFX.
There are some libraries in Github (https://github.com/ffadrique) that include all unit tests that I have always used to analyse the code coverage. I have another couple of libraries in the pipeline that I will only publish when I can complete the unit testing.
Cheers,
Fran
Interesting - LLVM compilers provide source code coverage capability from the command line. Are you saying that ifx is not able to use this yet, or is the aim to provide reporting similar to ifort in OneAPI 2026?
ifx is not able to use it, and the aim is to provide a code coverage feature similar to ifort.
Thanks for the clarification @Devorah_H_Intel
For the OP, @Paul_Levold , a couple of suggestions:
If you still have OneAPI 2025.0 (or earlier) installed, you could still use IFORT and its code coverage tool.
Alternatively, you could try using GNU Fortran and the gcov tool to provide code coverage.
It's a good idea to get your code building with alternative compilers anyway, because they can pick up bugs in your code that the other compiler doesn't.
Hi,
indeed, I use gfortran with gcov for the unit test coverage, but this forces me to maintain two environments, which is a bit of an unnecessary burden. Still useful for portability and debugging.
When code coverage or speed is "Can we do an entire cycle in 8.192 seconds, 16834 divided by 2000 steps per second?" if yes, then it does not matter, if not change the FFT engine first.
Some people have simple lives.
And then the data tells you what happens when a large truck hits a larger bridge. The bridge moves. No idea about the truck.
Doesn't look like it, according to the release notes: Intel® Fortran Compiler for oneAPI Release Notes 2026. Though I would treat them with a pinch of salt; the MKL release notes originally omitted the removal of some deprecated procedures.