[PATCH] x86: Add {load}/{store} tests for apx instructions.

[Cui, Lili]

> > --- /dev/null
> > +++ b/gas/testsuite/gas/i386/x86-64-pseudos-apx.s
> > @@ -0,0 +1,43 @@
> > +# Check 64bit APX instructions with pseudo prefixes for encoding
> > +
> > +.text
> > +_start:
> > +        #APX REX2
> > +        .irp m, mov, adc, add, and, cmp, or, sbb, sub, test, xor, xchg
> > +        \m %ecx, %r31d
> > +        {load}  \m %ecx, %r31d
> > +        {store} \m %ecx, %r31d
> > +        {load}  \m (%ecx), %r31d
> > +        {store} \m %r31d, (%ecx)
> > +        .endr
> > +
> > +        movaps (%r31),%xmm2
> > +        {load} movaps (%r31),%xmm2
> > +        {store} movaps %xmm2, (%r31)
> > +
> > +        #APX EVEX promoted from legacy
> > +        .irp m, adc, add, and, or, sbb, sub, xor
> > +        {evex}         \m %ecx, %r31d
> > +        {evex} {load}  \m %ecx, %r31d
> > +        {evex} {store} \m %ecx, %r31d
> > +        {evex} {load}  \m (%ecx), %r31d
> > +        {evex} {store} \m %r31d, (%ecx)
> > +        .endr
> 
> Like you have it here, imo the other sets would also benefit from having an
> instance with neither {load} nor {store}.
> 

Done.

> > +        #APX NDD
> > +        .irp m, add, or, adc, sbb, and, sub, xor
> > +        {load}  \m %ecx, %r31d, %eax
> > +        {store} \m %ecx, %r31d, %eax
> > +        {load}  \m %r31,(%r8),%r16
> > +        {load}  \m (%r8),%r31,%r16
> > +        {store} \m %r31,(%r8),%r16
> > +        {store} \m (%r8),%r31,%r16
> > +        .endr
> > +
> > +        #APX NF
> > +        .irp m, add, or, and, sub, xor
> > +        {nf} {load}  \m %ecx, %r31d
> > +        {nf} {store} \m %ecx, %r31d
> > +        {nf} {load}  \m (%r8d), %r31d
> > +        {nf} {store} \m %r31d, (%r8d)
> > +        .endr
> 
> I'm not convinced this last set is particularly useful. If it's being kept, question
> would be why NF and NDD together aren't tested.
> 
> Having reached the end of the file: What about CTESTcc and CCMPcc?
> 

I met some issues when adding CTESTcc and CCMPcc. 

For example, the disassembler for "{load} ccmpbl %edx, %eax" now looks a bit weird. 

"ccmpbl {dfv=}.s %edx,%eax"

Thanks,
Lili.