ARM newbie question here (and this isn't necessarily directly
BeagleBoard related, but this is the most helpful group of ARM experts I
know):
If an ARM chip has an FPU, does the main part of the chip have a defined
endianness? I know some ARM chips can switch endianness, but the ones
I'm familiar with that can do that don't have an FPU. Also, I'm
assumming the FPU has a fixed endianness.
For example, the OMAP3530 specs say that the ARM data endianness is
configurable. Does that include both integers and floats?
Also, I'm interested in the answer if in the context of SIMD as well as FPU.
Thank you,
Andrew
Andrew Straw wrote:
ARM newbie question here (and this isn't necessarily directly
BeagleBoard related, but this is the most helpful group of ARM experts I
know):
If an ARM chip has an FPU, does the main part of the chip have a defined
endianness? I know some ARM chips can switch endianness, but the ones
I'm familiar with that can do that don't have an FPU. Also, I'm
assumming the FPU has a fixed endianness.
For example, the OMAP3530 specs say that the ARM data endianness is
configurable. Does that include both integers and floats?
Also, I'm interested in the answer if in the context of SIMD as well as FPU.
Seeing as my question generated exactly zero replies, let me ask the
original question I was trying to gather data to answer myself. Please
replace "the architectures we want to support" with "OMAP3/ beagleboard"
in this text:
While working on portable macros for NAN, INF and co, I was
wondering why the current version of my code was working
(http://projects.scipy.org/numpy/browser/trunk/numpy/core/include/numpy/npy_math.h,
first lines). I then realized that IEEE 754 did not impose an
endianness, contrary to my belief. The macros would fail if the FPU and
the ALU were using a different endianness. Is this still a possibility
on the architectures we want to support ?
I do see from the data sheet that the DSP is little endian.
Thanks,
Andrew