Greetings,

I'm trying to switch my home-grown iphone math library to eigen, but ran into compilation errors I can't decipher. I am able to compile it to target the intel-based simulator, but when I target the arm chip for the actual iphone, i see these errors. I apologize in advance for any obvious problems - I'm a C++ template neophyte!

errors:

Code: Select all

include/arm_neon.h:7860: error: cannot convert 'const float32_t*' to 'const float*' for argument '1' to '__builtin_neon_sf __vector__ __builtin_neon_vld1v2sf(const float*)'
include/arm_neon.h:7926: error: cannot convert 'const float32_t*' to 'const float*' for argument '1' to '__builtin_neon_sf __vector__ __builtin_neon_vld1v4sf(const float*)'
include/arm_neon.h:7986: error: cannot convert 'const float32_t*' to 'const float*' for argument '1' to '__builtin_neon_sf __vector__ __builtin_neon_vld1_lanev2sf(const float*, __builtin_neon_sf __vector__, int)'
.... (1142 more!)

I suspect it has to do with the arm_neon.h header & the float32_t type - but that's where I'm stuck.

You need Konstantinos here --- our ARM specialist. Find his email address in the Eigen/src/Core/arch/NEON/PacketMath.h file. You could CC him in a bug report, or in an email to the list.

Thanks for the quick response. I mailed Konstantinos and will reply with a followup when/if I make progress.

Jimjones,
I'm faced with similar issue, and would appreciate if you could share the fix for the compilation issues. Thanks.

Hi,

seems like there has been a change of the float32_t type from custom data type to a typedef for float:
http://www.pubbs.net/200911/gcc/3874-ne ... -type.html

This was one year ago, are your compilers/libs up to date?