[LLVMdev] [PATCH] x86/asm: avoid mnemonics without type suffix

[Jeremy Fitzhardinge]

(Resent without HTML)

On 07/14/2013 10:19 AM, Linus Torvalds wrote:
> Now, there are possible cases where you want to make the size explicit
> because you are mixing memory operand sizes and there can be nasty
> performance implications of doing a 32-bit write and then doing a
> 64-bit read of the result. I'm not actually aware of us having ever
> worried/cared about it, but it's a possible source of trouble: mixing
> bitop instructions with non-bitop instructions can have some subtle
> interactions, and you need to be careful, since the size of the
> operand affects both the offset *and* the memory access size.
The SDM entry for BT mentions that the instruction may touch 2 or 4
bytes depending on the operand size, but doesn't specifically mention
that a 64 bit operation size touches 8 bytes - and it doesn't mention
anything at all about operand size and access size in BTR/BTS/BTC
(unless it's implied as part of the discussion about encoding the MSBs
of a constant bit offset in the offset of the addressing mode). Is that
an oversight?

>  The
> access size generally is meaningless from a semantic standpoint
> (little-endian being the only sane model), but the access size *can*
> have performance implications for the write queue forwarding.

It looks like that if the base address isn't aligned then neither is the
generated access, so you could get a protection fault if it overlaps a
page boundary, which is a semantic rather than purely operational
difference.

    J