Conversation

Ryan Castellucci (they/them) nonbinary_flag

Hmm. Is there any point in supporting big endian CPUs in new software?

4
1
0

@ryanc s390x is basically the last one standing

2
0
0

@ryanc Do you want to start the Third Endian War? Cause questions like that are how you start the Third Endian War.

0
2
0

@aschmitz πŸ–₯οΈπŸ€ΎπŸŒ‹

0
2
0

@ariadne That sounds like "not worth my time".

1
1
0

@ryanc yeah. they're just doing it in a very IBM way.

0
0
0

@noisytoot @ryanc yes, and nobody cares about it.

1
0
0
@ariadne @ryanc except me, and people with older powerpc hardware
1
0
0

@noisytoot @ryanc yes, people who are not IBM's target customer

1
0
0
@ariadne @ryanc when I get my power9 server working I intend to use it to try to find and fix some big endian portability bugs (I will still use little endian on the host though for practical reasons, because I want a working web browser and GPU)
1
0
0

@ariadne @noisytoot Power9 seems genuinely cool, but endianness bugs are some bullshit.

1
2
0

@ryanc @ariadne @noisytoot I use power9 and... Well... ppc64le. I believe it was added to support all the little endian linux stuff there is in power8. The cpu can run in either mode, i successfully ran some big endian vms just for fun.

I guess the answer isn't so much the software but what is the spec of what you're reading in.

0
0
0

@ryanc ... gonna have to remind myself if left and right shift operators on unsigned integers in c++ switch their numeric effect on values when running on big endian architectures.

if so, ohhhh I bet there are a ton of bugs out there.

1
0
0

@gsuberland @ryanc most of the PC code i write these days implicitly assumes little endian, 8 bit bytes, twos complement integers, and ieee754 floats.

Any platform that breaks with that baseline probably doesn't have working Vulkan and this isn't interesting to me :p

1
1
0

@gsuberland @ryanc realistically i have little time or motivation to support anything other than x86-64, aarch64, and maybe rv64 in the future

0
0
0