ARM THUMB: cpy instruction mustn't be emitted by default

[llvmbot]

Bugzilla Link	1388
Resolution	FIXED
Resolved on	May 18, 2015 07:11
Version	trunk
OS	All
Depends On	llvm/llvm-bugzilla-archive#1401
Reporter	LLVM Bugzilla Contributor
CC	@john-brawn-arm

Extended Description

The cpy instruction is only defined in ARMv6.

[llvmbot]

A possible solution for this bug is to emit 'mov' instead of 'cpy'. The problem
is that the mov sets the flags and I'm afraid that it can break something in
rematerialization.

Cheng?

[llvmbot]

If all mov are changed to cpy, the backend emits wrong code:

    cmp r5, r2
    mov r0, r5
    bgt .LBB1_6     @​cond_true99

[llvmbot]

The proper fix is to model CC as explicit register(s). Then you can fix the scheduler and register allocator,
i.e. any machine instruction level module that insert move, load, store ops, to insert instructions at the
right spot to avoid clobbering the CC registers. Not only is it useful for this bug, it also makes it easier to
implement a pass that optimize away cmp / test, etc.

Perhaps it's ok to just add instruction properties that tells us whether a instruction sets / reads CC. It's
hard to say.

[lattner]

removing me from cc list. I am aka unassignedbugs@nondot, so I'm getting two copies :)

[lattner]

FWIW, I think that it is about time we start modelling CC's correctly. This would require, at the minimum,
adding a new physreg to represent the CC, then marking each instruction that sets it as clobbering it.

[llvmbot]

It's better to model each flag individually e.g. N, Z, C, V, Q, the GE bits (collectively) etc. That way you can correctly model instructions that set N and Z but leave C and V unaffected. This allows you to move a MOV in between a CMP and a BCS, or between an ADDS and an ADC.

[john-brawn-arm]

This was fixed by r216138:

Author: Jonathan Roelofs jonathan@codesourcery.com
Date: Wed Aug 20 23:38:50 2014 +0000

Lower thumbv4t & thumbv5 lo->lo copies through a push-pop sequence

On pre-v6 hardware, 'MOV lo, lo' gives undefined results, so such copies need to
be avoided. This patch trades simplicity for implementation time at the expense
of performance... As they say: correctness first, then performance.

See http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-August/075998.html for a few
ideas on how to make this better.

AsmParser also doesn't permit MOV lo, lo (or CPY, as it used to be known), so you can't use it in embedded asm, but MC does permit it, so using clang to assemble a .s file with --target=thumbv4t-none-eabi gives no error but should. That's less of a problem though.

[lattner]

mentioned in issue llvm/llvm-bugzilla-archive#1401