[llvm-commits] [llvm] r61030 - in /llvm/trunk: include/llvm/Attributes.h lib/AsmParser/llvmAsmParser.y lib/Bitcode/Writer/BitcodeWriter.cpp lib/VMCore/Attributes.cpp

[Nick Lewycky]

Duncan Sands wrote:
> Hi Nick,
> 
>> It turns out that "align 1" and unaligned are different. Add a bias to the
>> alignment attribute such that 0 means unaligned.
> 
> what happens if a function is aligned 1024 or something huge like that?
> Also, I don't think 0 means unaligned, I think it means default alignment.

You're right, by unaligned I meant default alignment. I was trying to 
distinguish it from byte aligned (align 1).

We support "huge" alignments like 1024 just fine. It's a 5-bit integer 
[0..31]  whose mapping looks like this:

0 - default alignment
1 - align 1
2 - align 2
3 - align 4
4 - align 8
...
30 - align 2^29
31 - align 2^30

I should mention that previously, we were silently dropping any 
alignment larger than 32768.

Nick

> Ciao,
> 
> Duncan.
> 
>> +                                     // 0 means unaligned different from align 1
> 
> -> 0 means default alignment
> 
> Ciao,
> 
> Duncan.
>