Build issue with building for Android - Cygwin about glm HOT 9 CLOSED

Hi,

There is no specific support for NEON instructions yet.

Which version of the NDK are you using?

Thanks,
Christophe

from glm.

Hi,

I am using r8c NDK.

Regards,
Subhransu

from glm.

Tried with r7b also. I get the same error, so guess it's independent of NDK version.

from glm.

I wonder whether it's not a compiler issue:
http://stackoverflow.com/questions/3775991/gdb-error-invalid-offset-value-too-big-0x00000400-unable-to-build-app-in-debu

Do you use GLM inside a big switch here?

from glm.

Yes, I saw this link before logging a bug report.
There is no switch case in my code, the error is reproducible with only the following 4 lines of code,

glm::mat4 m1(1.0f);
glm::mat4 m2(1.0f);
glm::mat4 m3(1.0f);

m3 = m1*m2;

There is no big switch case in glm either (my prelim investigation) ..
Wondering what's going wrong!

from glm.

Erm... me too :/

from glm.

Just added those lines to my Android code and everything seems to work fine, I don't even get a warning (NDK r8b).

BTW adding explicit NEON support in GLM might not give as huge performance boost as one could expect. GLM is template-based and included directly into user code so there is a pretty good chance that GCC will automatically perform NEON-wise optimizations when NEON support is enabled in Android.mk. But, of course, I may be wrong on that one, it might be worth looking into.

from glm.

Ok. Then I am going to close this issue. Please reopen it if you find more clues on what generates this compatibility issue.

Thanks,
Christophe

from glm.

@jimvonmoon, I am sharing more or less the same thought than you for say 80% of the cases. However, some operations can be significantly improve using SIMD instructions directly. The SSE implementation of matrix inverse and multiplication show a significant gain.

from glm.