MATLAB Coder - ld: 32-bit RIP relative reference out of range

Question

Vadim Bertrand 2021-6-30

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/868798-matlab-coder-ld-32-bit-rip-relative-reference-out-of-range

评论： Vadim Bertrand 2021-7-22

在 MATLAB Online 中打开

Hi,

I'm trying to compile MATLAB functions to C++ using MATLAB Coder and I'm getting the following error message:

ld: 32-bit RIP relative reference out of range

I should mention that when running the entrypoint function whitin MATLAB it consumes ~15GB of RAM because it works on large matrices.

I'm under OSX and I use clang++ shipped with the lastest version of XCode. MATLAB version is 2020a.

I don't have experience with C++ compiling and linking proces so I'm a bit out of clue on how to solve this error. If anyone can provide explainations and ideas to fix it will be really helpfull.

Thanks.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Adit Calambur 2021-7-21

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/868798-matlab-coder-ld-32-bit-rip-relative-reference-out-of-range#answer_751414

Hi Vadim,

To summarise your issue, you are unable to generate code for an entry-point function of considerable size. On further reading about this error message, it seems as if your compiler is having trouble allocating memory above a certain threshold. Possible solutions might be:

Try increasing the dynamic memory allocation threshold from its current value. See Control Memory Allocation for Variable-Size Arrays
Ensure that your compiler can handle memory allocation above 32-bit addresses.
If your compiler supports data above 4GiB, you might have to change certain compiler options. See StackOverflow ld: 32-bit RIP.

.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Vadim Bertrand 2021-7-22

Hi,

Thanks for your answer.

Indeed the problem was related to memory allocation, and to my understanding to static or fix memory allocation. I forced MATLAB Coder to consider the big matrices causing issues to be dynamically allocated by passing their dimensions as parameters to my function instead of having them constant. Maybe it adds some computation time overhead but it was the only workaround I found to my original issue. (Before I tried to use -mcmodel=large without success)

Regards,

Vadim

请先登录，再进行评论。