Reputation: 1891
I would like to write an ARM SIMD computer vision code on an intel based windows machine. I would like to know what are the different options for doing that. I know for example that I can run a rasperrpi-emulator on windows, but that's really slow, and not productive. I hope if for example there is an IDE like Visuatl C++ that compiles code for ARM and test it like I'm exactly on an ARM platform without buying a real ARM board CPU.
Is it the only way to achieve that to buy an ARM based board ?
If I'm going to use an Emulator like QEMU, is there a faster or elegant way to develop computer vision software on it ?
I'm targeting ARM11 CPUs, that are quad-cores,..etc. Not the old ARM Architecture like M Series or ARM7.
Upvotes: 4
Views: 2371
Reputation: 1684
Answer depends on what you really have as "ARM SIMD" code baseline. Normally ARM SIMD = NEON SIMD extensions.
If this is the case and your NEON codebase is written using NEON intrinsics, then you can try recently introduced "automated porting NEON -> SSE solution", posted by Intel here.
Looks like easy to use: just include NEONtoSSE.h to your code and re-compile (I expect various C/C++ compilers compatibility).
Upvotes: 1
Reputation: 5211
By far, your best bet in terms of shortening your test-debug-compile cycle will be buying an ARM board and just programming natively on it or will a cross-compiler. I have had good luck with Gumstix, which generally are able to take a Linux distro pretty easily. From there, you have all of your GNU tools and you can just program directly on it.
Upvotes: 0