Platforms with no analigned write

[LemonParty]

Are there a platforms that supports by FPC where you can't do analigned write?

[LeP]

Are there a platforms that supports by FPC where you can't do analigned write?

This may be a start of knowledge: https://blog.vitlabuda.cz/2025/01/22/unaligned-memory-access-on-various-cpu-architectures.html

[LemonParty]

Good, LeP.
Maybe FPC has compiler defines to mark platforms with no unaligned access?

[MathMan]

Good, LeP.
Maybe FPC has compiler defines to mark platforms with no unaligned access?

No need for that, there is already UnAligned - use it like i.e.

Code: Pascal  [Select][+][-]

1.   Value: DWord;
2.   P: pByte;
3.  
4.   Val := $ffffffff;
5.   UnAligned( pDWord( P )^ ) := Val;
6.  

[LeP]

No need for that, there is already UnAligned - use it like i.e.
...............

@LemonParty asked about knowing platforms where you CAN'T do unaligned operations, not how to do an UNALIGNED operations.

[MarkMLl]

Are there a platforms that supports by FPC where you can't do analigned write?

SPARC. I can't speak for recent work, but traditionally it would either fail fatally or rely on the OS to simulate the unaligned access with a considerable performance penalty.

MarkMLl

[creaothceann]

https://wiki.freepascal.org/m68k

[MathMan]

No need for that, there is already UnAligned - use it like i.e.
...............

@LemonParty asked about knowing platforms where you CAN'T do unaligned operations, not how to do an UNALIGNED operations.

Yes, I understood as much. And what is the consequence if you know these platforms? Or asked different - what is the background to the question? I could only think of one - and towards that I answered. So maybe not an answer to the question asked but maybe an answer to the motivation behind?

[LemonParty]

Yes, I understood as much. And what is the consequence if you know these platforms? Or asked different - what is the background to the question? I could only think of one - and towards that I answered. So maybe not an answer to the question asked but maybe an answer to the motivation behind?

The purpouse is to generate different code for platforms where unaligned is not supported.

[PascalDragon]

Good, LeP.
Maybe FPC has compiler defines to mark platforms with no unaligned access?

Yes, it has: FPC_REQUIRES_PROPER_ALIGNMENT. (though this also includes platforms that can allow unaligned accesses but where they might be slower, like Aarch64)

[LemonParty]

Thank you, PascalDragon.

[MathMan]

Yes, I understood as much. And what is the consequence if you know these platforms? Or asked different - what is the background to the question? I could only think of one - and towards that I answered. So maybe not an answer to the question asked but maybe an answer to the motivation behind?

The purpouse is to generate different code for platforms where unaligned is not supported.

Which is what I assumed was the background. Now, with the answer from PascalDragon, you have two choices to handle it. I use UnAligned where I have too - and checked that, i.e. on x86-64, that it is simply ignored by FPC if the target platform supports unaligned access.

After reading PascalDragon's answer carefully I have to admit that I don't know how the compiler handles it on Aarch64 - I simply do not have a system to check.