diff mbox series

[4/5] x86/APX: extend SSE2AVX coverage

Message ID 155929a3-eb8b-4b82-a4ca-84ab6de34b97@suse.com
State New
Headers
Series x86/APX: respect -msse2avx |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_binutils_build--master-arm success Testing passed
linaro-tcwg-bot/tcwg_binutils_build--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_binutils_check--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_binutils_check--master-arm success Testing passed

Commit Message

Jan Beulich March 22, 2024, 9:28 a.m. UTC 
  Legacy encoded SIMD insns are converted to AVX ones in that mode. When
eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
available; there are quite a few which can't be converted).

For GFNI alter the gfni template such that VexW would be emitted even
for the SSE templates: There the attribute is simply meaningless, but
it simplifies the template quite a bit.

Note that LDDQU is represented as VMOVDQU32 (and the prior use of the
sse3 template needs dropping, to get the order right).

Note further that in a few cases, due to the use of templates, AVX512VL
is used when AVX512F would suffice. Since AVX10 is the main reference,
this shouldn't be too much of a problem.
---
To preempt the question: If we weren't to do this (i.e. leave legacy-
encoded SIMD insns using eGPR-s alone), I'd raise the counter question
of why these insns are supported by APX then in the first place.

I'm tempted to "convert" legacy encoded insns in maps 2 and 3 even
without -msse2avx. Thoughts?

Should we also cover RCP{P,S}S and RSQRT{P,S}S (via VRCP14{P,S}S /
VSQRT14{P,S}S)? The latter provide slightly better precision results,
but is that a potential issue?

By using a mask register (which supposedly shouldn't be used by legacy
SIMD code) we could likely convert further insns (by emitting a pair of
replacement ones).

What about SHA and KeyLocker insns not using eGPR-s? Their legacy
encodings could be replaced by EVEX ones, too, provided that's a gain:
Version 003 of the doc doesn't clarify whether, like other VEX/EVEX
insns and unlike legacy ones, register bits beyond bit 127 would be
cleared. That's the whole purpose of the SSE2AVX insns, after all. Yet
of course there's the problem here that then such insns (not using any
eGPR in their operands) would suddenly gain a dependency of the
resulting code on APX_F (and not AVX512* / AVX10). Perhaps for these
we'd really need -msse2apx then.

Should we also convert %xmm<N>-only templates (to consistently permit
use of {evex})? Or should we reject use of {evex}, but then also that of
{vex}/{vex3}?

Comments

Cui, Lili March 29, 2024, 9:10 a.m. UTC | #1 
> Legacy encoded SIMD insns are converted to AVX ones in that mode. When
> eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
> available; there are quite a few which can't be converted).
> 
> For GFNI alter the gfni template such that VexW would be emitted even
> for the SSE templates: There the attribute is simply meaningless, but
> it simplifies the template quite a bit.
> 

For this part, although adding VexW to the SSE template is more concise, it also breaks the rules and creates hidden dangers, it feels a bit unworthy.  GFNI SSE does not support eGPR-s, I'm not sure if we should give it an error instead of converting it.

> Note that LDDQU is represented as VMOVDQU32 (and the prior use of the
> sse3 template needs dropping, to get the order right).
> 

This conversion is clever, although the mnemonic has changed, but considering it is controlled by -msse2avx, maybe we can mention in the option that it might change the mnemonic. Judging from the option name alone, it is difficult for users to predict that the mnemonic will change (traditionally, it seems to just add V).

> Note further that in a few cases, due to the use of templates, AVX512VL
> is used when AVX512F would suffice. Since AVX10 is the main reference,
> this shouldn't be too much of a problem.
> ---

Yes, agreed.

> To preempt the question: If we weren't to do this (i.e. leave legacy-
> encoded SIMD insns using eGPR-s alone), I'd raise the counter question
> of why these insns are supported by APX then in the first place.
> 
> I'm tempted to "convert" legacy encoded insns in maps 2 and 3 even
> without -msse2avx. Thoughts?

I was a little worried about this conversion, so I asked a few people for their opinions, they found this approach a bit unacceptable, here are some ideas I collected.

1. The compiler will do this during the backend instruction selection phase. Binutils should only do instruction translation, not instruction selection.
2. We can only convert some instructions, not all instructions. When users use eGPR-s illegally, some will report an error, while others will not, which is very confusing.
3. Binutils needs to report errors for illegal instructions to ensure the correctness of the compiler.
4. I don't know if there are any machines in the future that don't expect to generate EVEX instructions.

> 
> Should we also cover RCP{P,S}S and RSQRT{P,S}S (via VRCP14{P,S}S /
> VSQRT14{P,S}S)? The latter provide slightly better precision results,
> but is that a potential issue?
> 

As far as I know, there should be no problem.

> By using a mask register (which supposedly shouldn't be used by legacy
> SIMD code) we could likely convert further insns (by emitting a pair of
> replacement ones).
> 
> What about SHA and KeyLocker insns not using eGPR-s? Their legacy
> encodings could be replaced by EVEX ones, too, provided that's a gain:
> Version 003 of the doc doesn't clarify whether, like other VEX/EVEX
> insns and unlike legacy ones, register bits beyond bit 127 would be
> cleared. That's the whole purpose of the SSE2AVX insns, after all. Yet
> of course there's the problem here that then such insns (not using any
> eGPR in their operands) would suddenly gain a dependency of the
> resulting code on APX_F (and not AVX512* / AVX10). Perhaps for these
> we'd really need -msse2apx then.
> 

If the CPU does not support the avx512 instruction, Binutils directly convert it, which will cause segment fault. It does seem like a new option is needed. These two instructions may change in the future, we can wait and see if this is necessary.

> Should we also convert %xmm<N>-only templates (to consistently permit
> use of {evex})? Or should we reject use of {evex}, but then also that of
> {vex}/{vex3}?

Do you mean SHA and KeyLocker?

Regards,
Lili.
Jan Beulich April 2, 2024, 8:48 a.m. UTC | #2 
On 29.03.2024 10:10, Cui, Lili wrote:
>> Legacy encoded SIMD insns are converted to AVX ones in that mode. When
>> eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
>> available; there are quite a few which can't be converted).
>>
>> For GFNI alter the gfni template such that VexW would be emitted even
>> for the SSE templates: There the attribute is simply meaningless, but
>> it simplifies the template quite a bit.
>>
> 
> For this part, although adding VexW to the SSE template is more concise, it also breaks the rules and creates hidden dangers, it feels a bit unworthy.  GFNI SSE does not support eGPR-s, I'm not sure if we should give it an error instead of converting it.

We should convert whatever's possible to convert. I'll re-consider the VexW
part following your comment (without promising that I'll undo it; in
particular I don't see any hidden dangers).

>> Note that LDDQU is represented as VMOVDQU32 (and the prior use of the
>> sse3 template needs dropping, to get the order right).
> 
> This conversion is clever, although the mnemonic has changed, but considering it is controlled by -msse2avx, maybe we can mention in the option that it might change the mnemonic. Judging from the option name alone, it is difficult for users to predict that the mnemonic will change (traditionally, it seems to just add V).

I don't think doc adjustment is needed here. We already have at least one
example where the mnemonic also changes: CVTPI2PD -> VCVTDQ2PD.

>> I'm tempted to "convert" legacy encoded insns in maps 2 and 3 even
>> without -msse2avx. Thoughts?
> 
> I was a little worried about this conversion, so I asked a few people for their opinions, they found this approach a bit unacceptable, here are some ideas I collected.
> 
> 1. The compiler will do this during the backend instruction selection phase. Binutils should only do instruction translation, not instruction selection.
> 2. We can only convert some instructions, not all instructions. When users use eGPR-s illegally, some will report an error, while others will not, which is very confusing.
> 3. Binutils needs to report errors for illegal instructions to ensure the correctness of the compiler.
> 4. I don't know if there are any machines in the future that don't expect to generate EVEX instructions.

Okay, I'll bin this (vague) plan then.

>> What about SHA and KeyLocker insns not using eGPR-s? Their legacy
>> encodings could be replaced by EVEX ones, too, provided that's a gain:
>> Version 003 of the doc doesn't clarify whether, like other VEX/EVEX
>> insns and unlike legacy ones, register bits beyond bit 127 would be
>> cleared. That's the whole purpose of the SSE2AVX insns, after all. Yet
>> of course there's the problem here that then such insns (not using any
>> eGPR in their operands) would suddenly gain a dependency of the
>> resulting code on APX_F (and not AVX512* / AVX10). Perhaps for these
>> we'd really need -msse2apx then.
> 
> If the CPU does not support the avx512 instruction, Binutils directly convert it, which will cause segment fault. It does seem like a new option is needed. These two instructions may change in the future, we can wait and see if this is necessary.

Well, this has resolved itself by the insn groups having been removed in
version 4 of the spec.

>> Should we also convert %xmm<N>-only templates (to consistently permit
>> use of {evex})? Or should we reject use of {evex}, but then also that of
>> {vex}/{vex3}?
> 
> Do you mean SHA and KeyLocker?

No, I mean templates with all XMM operands and no memory ones. Such don't
use eGPR-s, yet could be converted to their EVEX counterparts, too (by way
of the programmer adding {evex} to the _legacy_ insn). Hence the question
on how to treat {evex} there, and then also {vex} / {vex3}. Take, for
example, MOVHLPS or MOVLHPS.

Jan
Cui, Lili April 3, 2024, 7:59 a.m. UTC | #3 
> > This conversion is clever, although the mnemonic has changed, but
> considering it is controlled by -msse2avx, maybe we can mention in the
> option that it might change the mnemonic. Judging from the option name
> alone, it is difficult for users to predict that the mnemonic will change
> (traditionally, it seems to just add V).
> 
> I don't think doc adjustment is needed here. We already have at least one
> example where the mnemonic also changes: CVTPI2PD -> VCVTDQ2PD.
> 

Oh, there has been such a conversion before. Another thing that comes to mind is that sse2avx was previously used to support sse to vex conversion. This option works on machines that don't support evex. We now extend sse to evex, which makes this option unavailable on machines that do not support the evex instruction (e.g. hybrid machines like Alderlake). Do you think we should add a new option?

> >> Should we also convert %xmm<N>-only templates (to consistently permit
> >> use of {evex})? Or should we reject use of {evex}, but then also that
> >> of {vex}/{vex3}?
> >
> > Do you mean SHA and KeyLocker?
> 
> No, I mean templates with all XMM operands and no memory ones. Such
> don't use eGPR-s, yet could be converted to their EVEX counterparts, too (by
> way of the programmer adding {evex} to the _legacy_ insn). Hence the
> question on how to treat {evex} there, and then also {vex} / {vex3}. Take, for
> example, MOVHLPS or MOVLHPS.

I'm not sure if you want to support this conversion under -sse2avx. I think this conversion is only used by people writing assembler by hand. As for adding a prefix to convert sse to vex or evex, I think this requirement doesn't make much sense at the moment, maybe in the future if evex is faster than the vex instruction we can provide an option like sse2avx512 to achieve this conversion.

Thanks,
Lili.
Jan Beulich April 3, 2024, 8:19 a.m. UTC | #4 
On 03.04.2024 09:59, Cui, Lili wrote:
>>> This conversion is clever, although the mnemonic has changed, but
>> considering it is controlled by -msse2avx, maybe we can mention in the
>> option that it might change the mnemonic. Judging from the option name
>> alone, it is difficult for users to predict that the mnemonic will change
>> (traditionally, it seems to just add V).
>>
>> I don't think doc adjustment is needed here. We already have at least one
>> example where the mnemonic also changes: CVTPI2PD -> VCVTDQ2PD.
>>
> 
> Oh, there has been such a conversion before. Another thing that comes to mind is that sse2avx was previously used to support sse to vex conversion. This option works on machines that don't support evex. We now extend sse to evex, which makes this option unavailable on machines that do not support the evex instruction (e.g. hybrid machines like Alderlake). Do you think we should add a new option?

That's a question I've tentatively answered with "No". SSE => VEX requires
systems supporting AVX. SSE-with-eGPR requires systems with APX.
SSE-with-eGPR => EVEX similarly can rely on APX being there, and I expect
all such systems will support at least AVX10/128. If that is deemed a wrong
assumption, then indeed we may need to consider adding a new option (but
not -msse2avx512 as you suggest further down, as SSE only ever covers 128-
bit operations; -msse2avx10 maybe).

>>>> Should we also convert %xmm<N>-only templates (to consistently permit
>>>> use of {evex})? Or should we reject use of {evex}, but then also that
>>>> of {vex}/{vex3}?
>>>
>>> Do you mean SHA and KeyLocker?
>>
>> No, I mean templates with all XMM operands and no memory ones. Such
>> don't use eGPR-s, yet could be converted to their EVEX counterparts, too (by
>> way of the programmer adding {evex} to the _legacy_ insn). Hence the
>> question on how to treat {evex} there, and then also {vex} / {vex3}. Take, for
>> example, MOVHLPS or MOVLHPS.
> 
> I'm not sure if you want to support this conversion under -sse2avx. I think this conversion is only used by people writing assembler by hand.

Aiui -msse2avx is there mainly for hand-written assembly. Compilers will
do better insn selection on their own anyway.

> As for adding a prefix to convert sse to vex or evex, I think this requirement doesn't make much sense at the moment, maybe in the future if evex is faster than the vex instruction we can provide an option like sse2avx512 to achieve this conversion.

That's not my point. Consider this example:

	.text
sse2avx:
		movlhps	%xmm0, %xmm1
	{vex}	movlhps	%xmm0, %xmm1
	{vex3}	movlhps	%xmm0, %xmm1
	{evex}	movlhps	%xmm0, %xmm1

		movlps	(%rax), %xmm1
	{vex}	movlps	(%rax), %xmm1
	{vex3}	movlps	(%rax), %xmm1
	{evex}	movlps	(%rax), %xmm1

Other than the {evex}-prefixed lines, everything assembles smoothly prior
to the patch here. IOW even {vex3} has an effect on the non-VEX mnemonic.
With my patch as it is now, the 2nd {evex}-prefixed line assembles fine,
while the 1st doesn't. This is simply inconsistent. Hence why I see two
options: Disallow all three pseudo-prefixes on legacy mnemonics, or permit
{evex} consistently, too.

Jan
Jiang, Haochen April 3, 2024, 9:17 a.m. UTC | #5 
> On 03.04.2024 09:59, Cui, Lili wrote:
> >>> This conversion is clever, although the mnemonic has changed, but
> >> considering it is controlled by -msse2avx, maybe we can mention in the
> >> option that it might change the mnemonic. Judging from the option name
> >> alone, it is difficult for users to predict that the mnemonic will change
> >> (traditionally, it seems to just add V).
> >>
> >> I don't think doc adjustment is needed here. We already have at least one
> >> example where the mnemonic also changes: CVTPI2PD -> VCVTDQ2PD.
> >>
> >
> > Oh, there has been such a conversion before. Another thing that comes to
> mind is that sse2avx was previously used to support sse to vex conversion.
> This option works on machines that don't support evex. We now extend sse to
> evex, which makes this option unavailable on machines that do not support
> the evex instruction (e.g. hybrid machines like Alderlake). Do you think we
> should add a new option?
> 
> That's a question I've tentatively answered with "No". SSE => VEX requires
> systems supporting AVX. SSE-with-eGPR requires systems with APX.
> SSE-with-eGPR => EVEX similarly can rely on APX being there, and I expect
> all such systems will support at least AVX10/128. If that is deemed a wrong
> assumption, then indeed we may need to consider adding a new option (but
> not -msse2avx512 as you suggest further down, as SSE only ever covers 128-
> bit operations; -msse2avx10 maybe).

Just to clarify one thing, is the behavior change for -msse2avx happen only with
EGPRs?

If answer is no, no matter the assumption is right or not, I suppose the problem is
if we change the behavior for legacy -msse2avx to also promote to EVEX silently,
from my understanding, it will cause on current hybrid clients and upcoming Sierra
Forest and Clearwater Forest, EVEX instructions will appear under these platforms,
which is not aligned with previous behavior for -msse2avx. Eventually it will crash
since on those machines, there are no EVEX instructions. It might be a surprise to
users.

Thx,
Haochen
Cui, Lili April 3, 2024, 9:22 a.m. UTC | #6 
> On 03.04.2024 09:59, Cui, Lili wrote:
> >>> This conversion is clever, although the mnemonic has changed, but
> >> considering it is controlled by -msse2avx, maybe we can mention in
> >> the option that it might change the mnemonic. Judging from the option
> >> name alone, it is difficult for users to predict that the mnemonic
> >> will change (traditionally, it seems to just add V).
> >>
> >> I don't think doc adjustment is needed here. We already have at least
> >> one example where the mnemonic also changes: CVTPI2PD ->
> VCVTDQ2PD.
> >>
> >
> > Oh, there has been such a conversion before. Another thing that comes to
> mind is that sse2avx was previously used to support sse to vex conversion.
> This option works on machines that don't support evex. We now extend sse
> to evex, which makes this option unavailable on machines that do not
> support the evex instruction (e.g. hybrid machines like Alderlake). Do you
> think we should add a new option?
> 
> That's a question I've tentatively answered with "No". SSE => VEX requires
> systems supporting AVX. SSE-with-eGPR requires systems with APX.
> SSE-with-eGPR => EVEX similarly can rely on APX being there, and I expect all
> such systems will support at least AVX10/128. If that is deemed a wrong
> assumption, then indeed we may need to consider adding a new option (but
> not -msse2avx512 as you suggest further down, as SSE only ever covers 128-
> bit operations; -msse2avx10 maybe).
> 

Yes, I was wrong, only Egprs trigger sse to evex conversion. Your assumption is correct.

> >>>> Should we also convert %xmm<N>-only templates (to consistently
> >>>> permit use of {evex})? Or should we reject use of {evex}, but then
> >>>> also that of {vex}/{vex3}?
> >>>
> >>> Do you mean SHA and KeyLocker?
> >>
> >> No, I mean templates with all XMM operands and no memory ones. Such
> >> don't use eGPR-s, yet could be converted to their EVEX counterparts,
> >> too (by way of the programmer adding {evex} to the _legacy_ insn).
> >> Hence the question on how to treat {evex} there, and then also {vex}
> >> / {vex3}. Take, for example, MOVHLPS or MOVLHPS.
> >
> > I'm not sure if you want to support this conversion under -sse2avx. I think
> this conversion is only used by people writing assembler by hand.
> 
> Aiui -msse2avx is there mainly for hand-written assembly. Compilers will do
> better insn selection on their own anyway.
> 
> > As for adding a prefix to convert sse to vex or evex, I think this requirement
> doesn't make much sense at the moment, maybe in the future if evex is faster
> than the vex instruction we can provide an option like sse2avx512 to achieve
> this conversion.
> 
> That's not my point. Consider this example:
> 
> 	.text
> sse2avx:
> 		movlhps	%xmm0, %xmm1
> 	{vex}	movlhps	%xmm0, %xmm1
> 	{vex3}	movlhps	%xmm0, %xmm1
> 	{evex}	movlhps	%xmm0, %xmm1
> 
> 		movlps	(%rax), %xmm1
> 	{vex}	movlps	(%rax), %xmm1
> 	{vex3}	movlps	(%rax), %xmm1
> 	{evex}	movlps	(%rax), %xmm1
> 
> Other than the {evex}-prefixed lines, everything assembles smoothly prior to
> the patch here. IOW even {vex3} has an effect on the non-VEX mnemonic.
> With my patch as it is now, the 2nd {evex}-prefixed line assembles fine, while
> the 1st doesn't. This is simply inconsistent. Hence why I see two
> options: Disallow all three pseudo-prefixes on legacy mnemonics, or permit
> {evex} consistently, too.
> 

Oh, I got you, thank you for your detailed explanation. This increases the robustness of binutils, and since we already support some of them, I thought it would be nice to support it if it didn't require too much effort.

Thanks,
Lili.
Cui, Lili April 3, 2024, 9:29 a.m. UTC | #7 
> -----Original Message-----
> From: Jiang, Haochen <haochen.jiang@intel.com>
> Sent: Wednesday, April 3, 2024 5:17 PM
> To: Beulich, Jan <JBeulich@suse.com>; Cui, Lili <lili.cui@intel.com>
> Cc: H.J. Lu <hjl.tools@gmail.com>; Binutils <binutils@sourceware.org>
> Subject: RE: [PATCH 4/5] x86/APX: extend SSE2AVX coverage
> 
> 
> 
> > On 03.04.2024 09:59, Cui, Lili wrote:
> > >>> This conversion is clever, although the mnemonic has changed, but
> > >> considering it is controlled by -msse2avx, maybe we can mention in
> > >> the option that it might change the mnemonic. Judging from the
> > >> option name alone, it is difficult for users to predict that the
> > >> mnemonic will change (traditionally, it seems to just add V).
> > >>
> > >> I don't think doc adjustment is needed here. We already have at
> > >> least one example where the mnemonic also changes: CVTPI2PD ->
> VCVTDQ2PD.
> > >>
> > >
> > > Oh, there has been such a conversion before. Another thing that
> > > comes to
> > mind is that sse2avx was previously used to support sse to vex conversion.
> > This option works on machines that don't support evex. We now extend
> > sse to evex, which makes this option unavailable on machines that do
> > not support the evex instruction (e.g. hybrid machines like
> > Alderlake). Do you think we should add a new option?
> >
> > That's a question I've tentatively answered with "No". SSE => VEX
> > requires systems supporting AVX. SSE-with-eGPR requires systems with
> APX.
> > SSE-with-eGPR => EVEX similarly can rely on APX being there, and I
> > expect all such systems will support at least AVX10/128. If that is
> > deemed a wrong assumption, then indeed we may need to consider adding
> > a new option (but not -msse2avx512 as you suggest further down, as SSE
> > only ever covers 128- bit operations; -msse2avx10 maybe).
> 
> Just to clarify one thing, is the behavior change for -msse2avx happen only
> with EGPRs?
> 
> If answer is no, no matter the assumption is right or not, I suppose the
> problem is if we change the behavior for legacy -msse2avx to also promote to
> EVEX silently, from my understanding, it will cause on current hybrid clients
> and upcoming Sierra Forest and Clearwater Forest, EVEX instructions will
> appear under these platforms, which is not aligned with previous behavior
> for -msse2avx. Eventually it will crash since on those machines, there are no
> EVEX instructions. It might be a surprise to users.
> 

For all sse2avx instructions, we put AVX before the AVX512 template, so only Egpr will trigger sse to evex. My previous understanding was wrong.

Lili.
Jan Beulich April 3, 2024, 10:22 a.m. UTC | #8 
On 03.04.2024 11:17, Jiang, Haochen wrote:
> 
> 
>> On 03.04.2024 09:59, Cui, Lili wrote:
>>>>> This conversion is clever, although the mnemonic has changed, but
>>>> considering it is controlled by -msse2avx, maybe we can mention in the
>>>> option that it might change the mnemonic. Judging from the option name
>>>> alone, it is difficult for users to predict that the mnemonic will change
>>>> (traditionally, it seems to just add V).
>>>>
>>>> I don't think doc adjustment is needed here. We already have at least one
>>>> example where the mnemonic also changes: CVTPI2PD -> VCVTDQ2PD.
>>>>
>>>
>>> Oh, there has been such a conversion before. Another thing that comes to
>> mind is that sse2avx was previously used to support sse to vex conversion.
>> This option works on machines that don't support evex. We now extend sse to
>> evex, which makes this option unavailable on machines that do not support
>> the evex instruction (e.g. hybrid machines like Alderlake). Do you think we
>> should add a new option?
>>
>> That's a question I've tentatively answered with "No". SSE => VEX requires
>> systems supporting AVX. SSE-with-eGPR requires systems with APX.
>> SSE-with-eGPR => EVEX similarly can rely on APX being there, and I expect
>> all such systems will support at least AVX10/128. If that is deemed a wrong
>> assumption, then indeed we may need to consider adding a new option (but
>> not -msse2avx512 as you suggest further down, as SSE only ever covers 128-
>> bit operations; -msse2avx10 maybe).
> 
> Just to clarify one thing, is the behavior change for -msse2avx happen only with
> EGPRs?

Yes, or if the programmer uses {evex}.

Jan

> If answer is no, no matter the assumption is right or not, I suppose the problem is
> if we change the behavior for legacy -msse2avx to also promote to EVEX silently,
> from my understanding, it will cause on current hybrid clients and upcoming Sierra
> Forest and Clearwater Forest, EVEX instructions will appear under these platforms,
> which is not aligned with previous behavior for -msse2avx. Eventually it will crash
> since on those machines, there are no EVEX instructions. It might be a surprise to
> users.
> 
> Thx,
> Haochen
Jan Beulich April 5, 2024, 7:09 a.m. UTC | #9 
On 29.03.2024 10:10, Cui, Lili wrote:
>> Legacy encoded SIMD insns are converted to AVX ones in that mode. When
>> eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
>> available; there are quite a few which can't be converted).
>>
>> For GFNI alter the gfni template such that VexW would be emitted even
>> for the SSE templates: There the attribute is simply meaningless, but
>> it simplifies the template quite a bit.
> 
> For this part, although adding VexW to the SSE template is more concise, it also breaks the rules and creates hidden dangers, it feels a bit unworthy.

Coming back to this: What rules is this breaking?

I can see your point of "hidden dangers", yet I'm unsure about drawing
the conclusion that it's "unworthy": The <gfni> template would need
dropping altogether in the alternative approach, doubling the number of
insn template lines. That's not _that_ much, but still worthwhile to
weigh against those latent risks.

>  GFNI SSE does not support eGPR-s, I'm not sure if we should give it an error instead of converting it.

Well, this would mean not converting various others either. One of the
goals, however, was to allow conversion of as many of the insns as
possible, even if their legacy forms has no REX2 representation.

Jan
Cui, Lili April 7, 2024, 1:48 a.m. UTC | #10 
> On 29.03.2024 10:10, Cui, Lili wrote:
> >> Legacy encoded SIMD insns are converted to AVX ones in that mode.
> >> When eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
> >> available; there are quite a few which can't be converted).
> >>
> >> For GFNI alter the gfni template such that VexW would be emitted even
> >> for the SSE templates: There the attribute is simply meaningless, but
> >> it simplifies the template quite a bit.
> >
> > For this part, although adding VexW to the SSE template is more concise, it
> also breaks the rules and creates hidden dangers, it feels a bit unworthy.
> 
> Coming back to this: What rules is this breaking?
> 

Normally we only add VexW to VEX and EVEX formats, not to legacy formats. Once someone wants to use VexW to distinguish them, he will find that there is a special one there. 
But this is just an assumption, we have other flags to distinguish them. I won't insist on this issue, it won't have much impact.

> I can see your point of "hidden dangers", yet I'm unsure about drawing the
> conclusion that it's "unworthy": The <gfni> template would need dropping
> altogether in the alternative approach, doubling the number of insn
> template lines. That's not _that_ much, but still worthwhile to weigh against
> those latent risks.
> 
> >  GFNI SSE does not support eGPR-s, I'm not sure if we should give it an error
> instead of converting it.
> 
> Well, this would mean not converting various others either. One of the goals,
> however, was to allow conversion of as many of the insns as possible, even if
> their legacy forms has no REX2 representation.
> 

Since they are under the sse2mmx option, I think this is OK.

Thanks,
Lili.
Jan Beulich April 8, 2024, 6:25 a.m. UTC | #11 
On 07.04.2024 03:48, Cui, Lili wrote:
>> On 29.03.2024 10:10, Cui, Lili wrote:
>>>> Legacy encoded SIMD insns are converted to AVX ones in that mode.
>>>> When eGPR-s are in use, i.e. with APX, convert to AVX10 insns (where
>>>> available; there are quite a few which can't be converted).
>>>>
>>>> For GFNI alter the gfni template such that VexW would be emitted even
>>>> for the SSE templates: There the attribute is simply meaningless, but
>>>> it simplifies the template quite a bit.
>>>
>>> For this part, although adding VexW to the SSE template is more concise, it
>> also breaks the rules and creates hidden dangers, it feels a bit unworthy.
>>
>> Coming back to this: What rules is this breaking?
>>
> 
> Normally we only add VexW to VEX and EVEX formats, not to legacy formats. Once someone wants to use VexW to distinguish them, he will find that there is a special one there. 
> But this is just an assumption, we have other flags to distinguish them. I won't insist on this issue, it won't have much impact.

Just to mention it - I've already changed this, in less ugly a way than I
first feared it would be.

And just to mention this as well: From midday today I'll be gone for a
little over a week; please expect delayed responses (should any be needed
anywhere).

Jan
Cui, Lili April 8, 2024, 7:38 a.m. UTC | #12 
> -----Original Message-----
> From: Jan Beulich <jbeulich@suse.com>
> Sent: Monday, April 8, 2024 2:25 PM
> To: Cui, Lili <lili.cui@intel.com>
> Cc: H.J. Lu <hjl.tools@gmail.com>; Binutils <binutils@sourceware.org>
> Subject: Re: [PATCH 4/5] x86/APX: extend SSE2AVX coverage
> 
> On 07.04.2024 03:48, Cui, Lili wrote:
> >> On 29.03.2024 10:10, Cui, Lili wrote:
> >>>> Legacy encoded SIMD insns are converted to AVX ones in that mode.
> >>>> When eGPR-s are in use, i.e. with APX, convert to AVX10 insns
> >>>> (where available; there are quite a few which can't be converted).
> >>>>
> >>>> For GFNI alter the gfni template such that VexW would be emitted
> >>>> even for the SSE templates: There the attribute is simply
> >>>> meaningless, but it simplifies the template quite a bit.
> >>>
> >>> For this part, although adding VexW to the SSE template is more
> >>> concise, it
> >> also breaks the rules and creates hidden dangers, it feels a bit unworthy.
> >>
> >> Coming back to this: What rules is this breaking?
> >>
> >
> > Normally we only add VexW to VEX and EVEX formats, not to legacy
> formats. Once someone wants to use VexW to distinguish them, he will find
> that there is a special one there.
> > But this is just an assumption, we have other flags to distinguish them. I
> won't insist on this issue, it won't have much impact.
> 
> Just to mention it - I've already changed this, in less ugly a way than I first
> feared it would be.
> 

Ok, thanks.

> And just to mention this as well: From midday today I'll be gone for a little
> over a week; please expect delayed responses (should any be needed
> anywhere).
> 

Have a nice holiday and I'll send out the APX ZU patch when you get back.

Lili.
diff mbox series

Patch

--- a/gas/config/tc-i386.c
+++ b/gas/config/tc-i386.c
@@ -4099,7 +4099,7 @@  build_evex_prefix (void)
   /* Check the REX.W bit and VEXW.  */
   if (i.tm.opcode_modifier.vexw == VEXWIG)
     w = (evexwig == evexw1 || (i.rex & REX_W)) ? 1 : 0;
-  else if (i.tm.opcode_modifier.vexw)
+  else if (i.tm.opcode_modifier.vexw && !(i.rex & REX_W))
     w = i.tm.opcode_modifier.vexw == VEXW1 ? 1 : 0;
   else
     w = (flag_code == CODE_64BIT ? i.rex & REX_W : evexwig == evexw1) ? 1 : 0;
@@ -8223,7 +8223,12 @@  check_VecOperands (const insn_template *
   if ((is_cpu (t, CpuXOP) && t->operands == 5)
       || (t->opcode_space == SPACE_0F3A
 	  && (t->base_opcode | 3) == 0x0b
-	  && is_cpu (t, CpuAPX_F)))
+	  && (is_cpu (t, CpuAPX_F)
+	   || (t->opcode_modifier.sse2avx && t->opcode_modifier.evex
+	       && (!t->opcode_modifier.vex
+		   || (i.encoding != encoding_default
+		       && i.encoding != encoding_vex
+		       && i.encoding != encoding_vex3))))))
     {
       if (i.op[0].imms->X_op != O_constant
 	  || !fits_in_imm4 (i.op[0].imms->X_add_number))
--- a/gas/doc/c-i386.texi
+++ b/gas/doc/c-i386.texi
@@ -312,7 +312,9 @@  when respective checks fail.
 @cindex @samp{-msse2avx} option, x86-64
 @item -msse2avx
 This option specifies that the assembler should encode SSE instructions
-with VEX prefix.
+with VEX prefix, requiring AVX to be available.  SSE instructions using
+extended GPRs will be encoded with EVEX prefix, requiring AVX512 or AVX10 to
+be available.
 
 @cindex @samp{-muse-unaligned-vector-move} option, i386
 @cindex @samp{-muse-unaligned-vector-move} option, x86-64
--- /dev/null
+++ b/gas/testsuite/gas/i386/sse2avx-apx.d
@@ -0,0 +1,257 @@ 
+#as: -msse2avx
+#objdump: -dw
+#name: x86-64 SSE+ with APX encoding
+
+.*:     file format .*
+
+Disassembly of section .text:
+
+0+ <sse2avx>:
+[ 	]*[a-f0-9]+:	62 f9 7c 08 5b 60 01 	vcvtdq2ps 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 ff 08 e6 60 01 	vcvtpd2dqx 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 5a 60 01 	vcvtpd2psx 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7d 08 5b 60 01 	vcvtps2dq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 e6 60 01 	vcvttpd2dqx 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 5b 60 01 	vcvttps2dq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 6f 60 01 	vmovdqu32 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 28 60 01 	vmovapd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7c 08 28 60 01 	vmovaps 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7d 08 6f 60 01 	vmovdqa32 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 6f 60 01 	vmovdqu32 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 2a 60 01 	vmovntdqa 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 16 60 01 	vmovshdup 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 12 60 01 	vmovsldup 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 10 60 01 	vmovupd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7c 08 10 60 01 	vmovups 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 1c 60 01 	vpabsb 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 1d 60 01 	vpabsw 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 1e 60 01 	vpabsd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 51 60 01 	vsqrtpd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7c 08 51 60 01 	vsqrtps 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 29 60 01 	vmovapd %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7c 08 29 60 01 	vmovaps %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7f 60 01 	vmovdqa32 %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7e 08 7f 60 01 	vmovdqu32 %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7d 08 e7 60 01 	vmovntdq %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 fd 08 2b 60 01 	vmovntpd %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7c 08 2b 60 01 	vmovntps %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 fd 08 11 60 01 	vmovupd %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7c 08 11 60 01 	vmovups %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 cd 08 58 70 01 	vaddpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 58 70 01 	vaddps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 dc 70 01 	vaesenc 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 dd 70 01 	vaesenclast 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 de 70 01 	vaesdec 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 df 70 01 	vaesdeclast 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 55 70 01 	vandnpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 55 70 01 	vandnps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 54 70 01 	vandpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 54 70 01 	vandps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 5e 70 01 	vdivpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 5e 70 01 	vdivps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 cf 70 01 	vgf2p8mulb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 5f 70 01 	vmaxpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 5f 70 01 	vmaxps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 5d 70 01 	vminpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 5d 70 01 	vminps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 59 70 01 	vmulpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 59 70 01 	vmulps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 56 70 01 	vorpd  0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 56 70 01 	vorps  0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 63 70 01 	vpacksswb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 6b 70 01 	vpackssdw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 67 70 01 	vpackuswb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 2b 70 01 	vpackusdw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 fc 70 01 	vpaddb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 fd 70 01 	vpaddw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 fe 70 01 	vpaddd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 d4 70 01 	vpaddq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 ec 70 01 	vpaddsb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 ed 70 01 	vpaddsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 dc 70 01 	vpaddusb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 dd 70 01 	vpaddusw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 db 70 01 	vpandd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 df 70 01 	vpandnd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e0 70 01 	vpavgb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e3 70 01 	vpavgw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 44 70 01 00 	vpclmullqlqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 44 70 01 01 	vpclmulhqlqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 44 70 01 10 	vpclmullqhqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 44 70 01 11 	vpclmulhqhqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f5 70 01 	vpmaddwd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 04 70 01 	vpmaddubsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3c 70 01 	vpmaxsb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 ee 70 01 	vpmaxsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3d 70 01 	vpmaxsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 de 70 01 	vpmaxub 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3e 70 01 	vpmaxuw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3f 70 01 	vpmaxud 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 38 70 01 	vpminsb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 ea 70 01 	vpminsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 39 70 01 	vpminsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 da 70 01 	vpminub 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3a 70 01 	vpminuw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 3b 70 01 	vpminud 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa cd 08 28 70 01 	vpmuldq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e4 70 01 	vpmulhuw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 0b 70 01 	vpmulhrsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e5 70 01 	vpmulhw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 d5 70 01 	vpmullw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 40 70 01 	vpmulld 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 f4 70 01 	vpmuludq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 eb 70 01 	vpord  0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f6 70 01 	vpsadbw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 4d 08 00 70 01 	vpshufb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f1 70 01 	vpsllw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f2 70 01 	vpslld 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 f3 70 01 	vpsllq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e1 70 01 	vpsraw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e2 70 01 	vpsrad 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 d1 70 01 	vpsrlw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 d2 70 01 	vpsrld 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 d3 70 01 	vpsrlq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f8 70 01 	vpsubb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 f9 70 01 	vpsubw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 fa 70 01 	vpsubd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 fb 70 01 	vpsubq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e8 70 01 	vpsubsb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 e9 70 01 	vpsubsw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 d8 70 01 	vpsubusb 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 d9 70 01 	vpsubusw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 68 70 01 	vpunpckhbw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 69 70 01 	vpunpckhwd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 6a 70 01 	vpunpckhdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 6d 70 01 	vpunpckhqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 60 70 01 	vpunpcklbw 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 61 70 01 	vpunpcklwd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 62 70 01 	vpunpckldq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 6c 70 01 	vpunpcklqdq 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4d 08 ef 70 01 	vpxord 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 5c 70 01 	vsubpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 5c 70 01 	vsubps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 15 70 01 	vunpckhpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 15 70 01 	vunpckhps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 14 70 01 	vunpcklpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 14 70 01 	vunpcklps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 57 70 01 	vxorpd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 57 70 01 	vxorps 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 7d 08 70 70 01 64 	vpshufd \$0x64,0x10\(%r16\),%xmm6
+[ 	]*[a-f0-9]+:	62 f9 7e 08 70 70 01 64 	vpshufhw \$0x64,0x10\(%r16\),%xmm6
+[ 	]*[a-f0-9]+:	62 f9 7f 08 70 70 01 64 	vpshuflw \$0x64,0x10\(%r16\),%xmm6
+[ 	]*[a-f0-9]+:	62 fb fd 08 09 70 01 04 	vrndscalepd \$(0x)?4,0x10\(%r16\),%xmm6
+[ 	]*[a-f0-9]+:	62 fb 7d 08 08 70 01 04 	vrndscaleps \$(0x)?4,0x10\(%r16\),%xmm6
+[ 	]*[a-f0-9]+:	62 fb cd 08 ce 70 01 64 	vgf2p8affineqb \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb cd 08 cf 70 01 64 	vgf2p8affineinvqb \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 0f 70 01 64 	vpalignr \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 44 70 01 64 	vpclmulqdq \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cd 08 c6 70 01 64 	vshufpd \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4c 08 c6 70 01 64 	vshufps \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 fd 08 2f 60 02 	vcomisd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 e6 60 02 	vcvtdq2pd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 e6 60 02 	vcvtdq2pd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7c 08 5a 60 02 	vcvtps2pd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 ff 08 12 60 02 	vmovddup 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7f 08 10 60 02 	vmovs{bad} 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 20 60 02 	vpmovsxbw 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 23 60 02 	vpmovsxwd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 25 60 02 	vpmovsxdq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 30 60 02 	vpmovzxbw 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 33 60 02 	vpmovzxwd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 35 60 02 	vpmovzxdq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 2e 60 02 	vucomisd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 13 60 02 	vmovlpd %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7c 08 13 60 02 	vmovlps %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 fd 08 17 60 02 	vmovhpd %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7c 08 17 60 02 	vmovhps %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7f 08 11 60 02 	vmovs{bad} %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 dd 08 12 60 02 	vmovlpd 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5c 08 12 60 02 	vmovlps 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 dd 08 16 60 02 	vmovhpd 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5c 08 16 60 02 	vmovhps 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 7e e0    	vmovq  %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 fd 08 6e e0    	vmovq  %r16,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 7e e0    	vmovq  %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 fd 08 6e e0    	vmovq  %r16,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 fd 08 d6 60 02 	vmovq  %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 fe 08 7e 60 02 	vmovq  0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 e1 7f 08 2d c4    	vcvtsd2si %xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7f 08 2d 48 02 	vcvtsd2si 0x10\(%r16\),%ecx
+[ 	]*[a-f0-9]+:	62 e1 7f 08 2c c4    	vcvttsd2si %xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7f 08 2c 48 02 	vcvttsd2si 0x10\(%r16\),%ecx
+[ 	]*[a-f0-9]+:	62 e1 ff 08 2d c4    	vcvtsd2si %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 ff 08 2d 48 02 	vcvtsd2si 0x10\(%r16\),%rcx
+[ 	]*[a-f0-9]+:	62 e1 ff 08 2c c4    	vcvttsd2si %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 ff 08 2c 48 02 	vcvttsd2si 0x10\(%r16\),%rcx
+[ 	]*[a-f0-9]+:	62 f9 df 08 2a e0    	vcvtsi2sd %r16,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 df 08 2a 60 02 	vcvtsi2sdq 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 de 08 2a e0    	vcvtsi2ss %r16,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 de 08 2a 60 02 	vcvtsi2ssq 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb dd 08 22 e0 64 	vpinsrq \$0x64,%r16,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb dd 08 22 60 02 64 	vpinsrq \$0x64,0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb fd 08 16 e0 64 	vpextrq \$0x64,%xmm4,%r16
+[ 	]*[a-f0-9]+:	62 fb fd 08 16 60 02 64 	vpextrq \$0x64,%xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 fb cd 08 0b 70 02 04 	vrndscalesd \$(0x)?4,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 58 70 02 	vaddsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 5a 70 02 	vcvtsd2ss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 5e 70 02 	vdivsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 5f 70 02 	vmaxsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 5d 70 02 	vminsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 59 70 02 	vmulsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 51 70 02 	vsqrtsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 cf 08 5c 70 02 	vsubsd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 58 70 04 	vaddss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 5a 70 04 	vcvtss2sd 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 5e 70 04 	vdivss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 5f 70 04 	vmaxss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 5d 70 04 	vminss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 59 70 04 	vmulss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 51 70 04 	vsqrtss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 4e 08 5c 70 04 	vsubss 0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 f9 7c 08 2f 60 04 	vcomiss 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 10 60 04 	vmovss 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 21 60 04 	vpmovsxbd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 24 60 04 	vpmovsxwq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 31 60 04 	vpmovzxbd 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 34 60 04 	vpmovzxwq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7c 08 2e 60 04 	vucomiss 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7e 08 11 60 04 	vmovss %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e e0    	vmovd  %xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e 60 04 	vmovd  %xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 7d 08 6e e0    	vmovd  %r16d,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 7d 08 6e 60 04 	vmovd  0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 e1 7e 08 2d c4    	vcvtss2si %xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7e 08 2d 48 04 	vcvtss2si 0x10\(%r16\),%ecx
+[ 	]*[a-f0-9]+:	62 e1 7e 08 2c c4    	vcvttss2si %xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7e 08 2c 48 04 	vcvttss2si 0x10\(%r16\),%ecx
+[ 	]*[a-f0-9]+:	62 e1 fe 08 2d c4    	vcvtss2si %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 fe 08 2d 48 04 	vcvtss2si 0x10\(%r16\),%rcx
+[ 	]*[a-f0-9]+:	62 e1 fe 08 2c c4    	vcvttss2si %xmm4,%r16
+[ 	]*[a-f0-9]+:	62 f9 fe 08 2c 48 04 	vcvttss2si 0x10\(%r16\),%rcx
+[ 	]*[a-f0-9]+:	62 fb fd 08 17 e0 64 	vextractps \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 16 e0 64 	vpextrd \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 16 60 04 64 	vpextrd \$0x64,%xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 fb 7d 08 17 e0 64 	vextractps \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 17 60 04 64 	vextractps \$0x64,%xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 fb 5d 08 22 e0 64 	vpinsrd \$0x64,%r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb 5d 08 22 60 04 64 	vpinsrd \$0x64,0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5f 08 2a e0    	vcvtsi2sd %r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5f 08 2a 60 04 	vcvtsi2sdl 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5e 08 2a e0    	vcvtsi2ss %r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5e 08 2a 60 04 	vcvtsi2ssl 0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb 4d 08 21 70 04 64 	vinsertps \$0x64,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fb 4d 08 0a 70 04 04 	vrndscaless \$(0x)?4,0x10\(%r16\),%xmm6,%xmm6
+[ 	]*[a-f0-9]+:	62 fa 7d 08 22 60 08 	vpmovsxbq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 fa 7d 08 32 60 08 	vpmovzxbq 0x10\(%r16\),%xmm4
+[ 	]*[a-f0-9]+:	62 e1 7d 08 c5 c4 64 	vpextrw \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 e1 7d 08 c5 c4 64 	vpextrw \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 15 60 08 64 	vpextrw \$0x64,%xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 f9 5d 08 c4 e0 64 	vpinsrw \$0x64,%r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5d 08 c4 e0 64 	vpinsrw \$0x64,%r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 f9 5d 08 c4 60 08 64 	vpinsrw \$0x64,0x10\(%r16\),%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb 7d 08 14 e0 64 	vpextrb \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 14 e0 64 	vpextrb \$0x64,%xmm4,%r16d
+[ 	]*[a-f0-9]+:	62 fb 7d 08 14 60 10 64 	vpextrb \$0x64,%xmm4,0x10\(%r16\)
+[ 	]*[a-f0-9]+:	62 fb 5d 08 20 e0 64 	vpinsrb \$0x64,%r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb 5d 08 20 e0 64 	vpinsrb \$0x64,%r16d,%xmm4,%xmm4
+[ 	]*[a-f0-9]+:	62 fb 5d 08 20 60 10 64 	vpinsrb \$0x64,0x10\(%r16\),%xmm4,%xmm4
+#pass
--- /dev/null
+++ b/gas/testsuite/gas/i386/sse2avx-apx.s
@@ -0,0 +1,316 @@ 
+	.text
+	.sse_check warning
+sse2avx:
+
+# Tests for op mem128, xmm
+	cvtdq2ps 16(%r16),%xmm4
+	cvtpd2dq 16(%r16),%xmm4
+	cvtpd2ps 16(%r16),%xmm4
+	cvtps2dq 16(%r16),%xmm4
+	cvttpd2dq 16(%r16),%xmm4
+	cvttps2dq 16(%r16),%xmm4
+	lddqu 16(%r16),%xmm4
+	movapd 16(%r16),%xmm4
+	movaps 16(%r16),%xmm4
+	movdqa 16(%r16),%xmm4
+	movdqu 16(%r16),%xmm4
+	movntdqa 16(%r16),%xmm4
+	movshdup 16(%r16),%xmm4
+	movsldup 16(%r16),%xmm4
+	movupd 16(%r16),%xmm4
+	movups 16(%r16),%xmm4
+	pabsb 16(%r16),%xmm4
+	pabsw 16(%r16),%xmm4
+	pabsd 16(%r16),%xmm4
+	sqrtpd 16(%r16),%xmm4
+	sqrtps 16(%r16),%xmm4
+
+# Tests for op xmm, mem128
+	movapd %xmm4,16(%r16)
+	movaps %xmm4,16(%r16)
+	movdqa %xmm4,16(%r16)
+	movdqu %xmm4,16(%r16)
+	movntdq %xmm4,16(%r16)
+	movntpd %xmm4,16(%r16)
+	movntps %xmm4,16(%r16)
+	movupd %xmm4,16(%r16)
+	movups %xmm4,16(%r16)
+
+# Tests for op mem128, xmm[, xmm]
+	addpd 16(%r16),%xmm6
+	addps 16(%r16),%xmm6
+	aesenc 16(%r16),%xmm6
+	aesenclast 16(%r16),%xmm6
+	aesdec 16(%r16),%xmm6
+	aesdeclast 16(%r16),%xmm6
+	andnpd 16(%r16),%xmm6
+	andnps 16(%r16),%xmm6
+	andpd 16(%r16),%xmm6
+	andps 16(%r16),%xmm6
+	divpd 16(%r16),%xmm6
+	divps 16(%r16),%xmm6
+	gf2p8mulb 16(%r16),%xmm6
+	maxpd 16(%r16),%xmm6
+	maxps 16(%r16),%xmm6
+	minpd 16(%r16),%xmm6
+	minps 16(%r16),%xmm6
+	mulpd 16(%r16),%xmm6
+	mulps 16(%r16),%xmm6
+	orpd 16(%r16),%xmm6
+	orps 16(%r16),%xmm6
+	packsswb 16(%r16),%xmm6
+	packssdw 16(%r16),%xmm6
+	packuswb 16(%r16),%xmm6
+	packusdw 16(%r16),%xmm6
+	paddb 16(%r16),%xmm6
+	paddw 16(%r16),%xmm6
+	paddd 16(%r16),%xmm6
+	paddq 16(%r16),%xmm6
+	paddsb 16(%r16),%xmm6
+	paddsw 16(%r16),%xmm6
+	paddusb 16(%r16),%xmm6
+	paddusw 16(%r16),%xmm6
+	pand 16(%r16),%xmm6
+	pandn 16(%r16),%xmm6
+	pavgb 16(%r16),%xmm6
+	pavgw 16(%r16),%xmm6
+	pclmullqlqdq 16(%r16),%xmm6
+	pclmulhqlqdq 16(%r16),%xmm6
+	pclmullqhqdq 16(%r16),%xmm6
+	pclmulhqhqdq 16(%r16),%xmm6
+	pmaddwd 16(%r16),%xmm6
+	pmaddubsw 16(%r16),%xmm6
+	pmaxsb 16(%r16),%xmm6
+	pmaxsw 16(%r16),%xmm6
+	pmaxsd 16(%r16),%xmm6
+	pmaxub 16(%r16),%xmm6
+	pmaxuw 16(%r16),%xmm6
+	pmaxud 16(%r16),%xmm6
+	pminsb 16(%r16),%xmm6
+	pminsw 16(%r16),%xmm6
+	pminsd 16(%r16),%xmm6
+	pminub 16(%r16),%xmm6
+	pminuw 16(%r16),%xmm6
+	pminud 16(%r16),%xmm6
+	pmuldq 16(%r16),%xmm6
+	pmulhuw 16(%r16),%xmm6
+	pmulhrsw 16(%r16),%xmm6
+	pmulhw 16(%r16),%xmm6
+	pmullw 16(%r16),%xmm6
+	pmulld 16(%r16),%xmm6
+	pmuludq 16(%r16),%xmm6
+	por 16(%r16),%xmm6
+	psadbw 16(%r16),%xmm6
+	pshufb 16(%r16),%xmm6
+	psllw 16(%r16),%xmm6
+	pslld 16(%r16),%xmm6
+	psllq 16(%r16),%xmm6
+	psraw 16(%r16),%xmm6
+	psrad 16(%r16),%xmm6
+	psrlw 16(%r16),%xmm6
+	psrld 16(%r16),%xmm6
+	psrlq 16(%r16),%xmm6
+	psubb 16(%r16),%xmm6
+	psubw 16(%r16),%xmm6
+	psubd 16(%r16),%xmm6
+	psubq 16(%r16),%xmm6
+	psubsb 16(%r16),%xmm6
+	psubsw 16(%r16),%xmm6
+	psubusb 16(%r16),%xmm6
+	psubusw 16(%r16),%xmm6
+	punpckhbw 16(%r16),%xmm6
+	punpckhwd 16(%r16),%xmm6
+	punpckhdq 16(%r16),%xmm6
+	punpckhqdq 16(%r16),%xmm6
+	punpcklbw 16(%r16),%xmm6
+	punpcklwd 16(%r16),%xmm6
+	punpckldq 16(%r16),%xmm6
+	punpcklqdq 16(%r16),%xmm6
+	pxor 16(%r16),%xmm6
+	subpd 16(%r16),%xmm6
+	subps 16(%r16),%xmm6
+	unpckhpd 16(%r16),%xmm6
+	unpckhps 16(%r16),%xmm6
+	unpcklpd 16(%r16),%xmm6
+	unpcklps 16(%r16),%xmm6
+	xorpd 16(%r16),%xmm6
+	xorps 16(%r16),%xmm6
+
+# Tests for op imm8, mem128, xmm
+	pshufd $100,16(%r16),%xmm6
+	pshufhw $100,16(%r16),%xmm6
+	pshuflw $100,16(%r16),%xmm6
+	roundpd $4,16(%r16),%xmm6
+	roundps $4,16(%r16),%xmm6
+
+# Tests for op imm8, mem128, xmm[, xmm]
+	gf2p8affineqb $100,16(%r16),%xmm6
+	gf2p8affineinvqb $100,16(%r16),%xmm6
+	palignr $100,16(%r16),%xmm6
+	pclmulqdq $100,16(%r16),%xmm6
+	shufpd $100,16(%r16),%xmm6
+	shufps $100,16(%r16),%xmm6
+
+# Tests for op mem64, xmm
+	comisd 16(%r16),%xmm4
+	cvtdq2pd 16(%r16),%xmm4
+	cvtpi2pd 16(%r16),%xmm4
+	cvtps2pd 16(%r16),%xmm4
+	movddup 16(%r16),%xmm4
+	movsd 16(%r16),%xmm4
+	pmovsxbw 16(%r16),%xmm4
+	pmovsxwd 16(%r16),%xmm4
+	pmovsxdq 16(%r16),%xmm4
+	pmovzxbw 16(%r16),%xmm4
+	pmovzxwd 16(%r16),%xmm4
+	pmovzxdq 16(%r16),%xmm4
+	ucomisd 16(%r16),%xmm4
+
+# Tests for op xmm, mem64
+	movlpd %xmm4,16(%r16)
+	movlps %xmm4,16(%r16)
+	movhpd %xmm4,16(%r16)
+	movhps %xmm4,16(%r16)
+	movsd %xmm4,16(%r16)
+
+# Tests for op mem64, xmm[, xmm]
+	movlpd 16(%r16),%xmm4
+	movlps 16(%r16),%xmm4
+	movhpd 16(%r16),%xmm4
+	movhps 16(%r16),%xmm4
+
+# Tests for op xmm, regq/mem64
+# Tests for op regq/mem64, xmm
+	movd %xmm4,%r16
+	movd %r16,%xmm4
+	movq %xmm4,%r16
+	movq %r16,%xmm4
+	movq %xmm4,16(%r16)
+	movq 16(%r16),%xmm4
+
+# Tests for op xmm/mem64, regl
+	cvtsd2si %xmm4,%r16d
+	cvtsd2si 16(%r16),%ecx
+	cvttsd2si %xmm4,%r16d
+	cvttsd2si 16(%r16),%ecx
+
+# Tests for op xmm/mem64, regq
+	cvtsd2si %xmm4,%r16
+	cvtsd2si 16(%r16),%rcx
+	cvttsd2si %xmm4,%r16
+	cvttsd2si 16(%r16),%rcx
+
+# Tests for op regq/mem64, xmm[, xmm]
+	cvtsi2sdq %r16,%xmm4
+	cvtsi2sdq 16(%r16),%xmm4
+	cvtsi2ssq %r16,%xmm4
+	cvtsi2ssq 16(%r16),%xmm4
+
+# Tests for op imm8, regq/mem64, xmm[, xmm]
+	pinsrq $100,%r16,%xmm4
+	pinsrq $100,16(%r16),%xmm4
+
+# Tests for op imm8, xmm, regq/mem64
+	pextrq $100,%xmm4,%r16
+	pextrq $100,%xmm4,16(%r16)
+
+# Tests for op imm8, mem64, xmm[, xmm]
+	roundsd $4,16(%r16),%xmm6
+
+# Tests for op mem64, xmm[, xmm]
+	addsd 16(%r16),%xmm6
+	cvtsd2ss 16(%r16),%xmm6
+	divsd 16(%r16),%xmm6
+	maxsd 16(%r16),%xmm6
+	minsd 16(%r16),%xmm6
+	mulsd 16(%r16),%xmm6
+	sqrtsd 16(%r16),%xmm6
+	subsd 16(%r16),%xmm6
+
+# Tests for op mem32, xmm[, xmm]
+	addss 16(%r16),%xmm6
+	cvtss2sd 16(%r16),%xmm6
+	divss 16(%r16),%xmm6
+	maxss 16(%r16),%xmm6
+	minss 16(%r16),%xmm6
+	mulss 16(%r16),%xmm6
+	sqrtss 16(%r16),%xmm6
+	subss 16(%r16),%xmm6
+
+# Tests for op mem32, xmm
+	comiss 16(%r16),%xmm4
+	movss 16(%r16),%xmm4
+	pmovsxbd 16(%r16),%xmm4
+	pmovsxwq 16(%r16),%xmm4
+	pmovzxbd 16(%r16),%xmm4
+	pmovzxwq 16(%r16),%xmm4
+	ucomiss 16(%r16),%xmm4
+
+# Tests for op xmm, mem32
+	movss %xmm4,16(%r16)
+
+# Tests for op xmm, regl/mem32
+# Tests for op regl/mem32, xmm
+	movd %xmm4,%r16d
+	movd %xmm4,16(%r16)
+	movd %r16d,%xmm4
+	movd 16(%r16),%xmm4
+
+# Tests for op xmm/mem32, regl
+	cvtss2si %xmm4,%r16d
+	cvtss2si 16(%r16),%ecx
+	cvttss2si %xmm4,%r16d
+	cvttss2si 16(%r16),%ecx
+
+# Tests for op xmm/mem32, regq
+	cvtss2si %xmm4,%r16
+	cvtss2si 16(%r16),%rcx
+	cvttss2si %xmm4,%r16
+	cvttss2si 16(%r16),%rcx
+
+# Tests for op imm8, xmm, regq/mem32
+	extractps $100,%xmm4,%r16
+
+# Tests for op imm8, xmm, regl/mem32
+	pextrd $100,%xmm4,%r16d
+	pextrd $100,%xmm4,16(%r16)
+	extractps $100,%xmm4,%r16d
+	extractps $100,%xmm4,16(%r16)
+
+# Tests for op imm8, regl/mem32, xmm[, xmm]
+	pinsrd $100,%r16d,%xmm4
+	pinsrd $100,16(%r16),%xmm4
+
+# Tests for op regl/mem32, xmm[, xmm]
+	cvtsi2sd %r16d,%xmm4
+	cvtsi2sd 16(%r16),%xmm4
+	cvtsi2ss %r16d,%xmm4
+	cvtsi2ss 16(%r16),%xmm4
+
+# Tests for op imm8, mem32, xmm[, xmm]
+	insertps $100,16(%r16),%xmm6
+	roundss $4,16(%r16),%xmm6
+
+# Tests for op mem16, xmm
+	pmovsxbq 16(%r16),%xmm4
+	pmovzxbq 16(%r16),%xmm4
+
+# Tests for op imm8, xmm, regl/mem16
+	pextrw $100,%xmm4,%r16d
+	pextrw $100,%xmm4,%r16
+	pextrw $100,%xmm4,16(%r16)
+
+# Tests for op imm8, regl/mem16, xmm[, xmm]
+	pinsrw $100,%r16d,%xmm4
+	pinsrw $100,%r16,%xmm4
+	pinsrw $100,16(%r16),%xmm4
+
+# Tests for op imm8, xmm, regl/mem8
+	pextrb $100,%xmm4,%r16d
+	pextrb $100,%xmm4,%r16
+	pextrb $100,%xmm4,16(%r16)
+
+# Tests for op imm8, regl/mem8, xmm[, xmm]
+	pinsrb $100,%r16d,%xmm4
+	pinsrb $100,%r16,%xmm4
+	pinsrb $100,16(%r16),%xmm4
--- a/gas/testsuite/gas/i386/x86-64.exp
+++ b/gas/testsuite/gas/i386/x86-64.exp
@@ -386,6 +386,7 @@  run_dump_test "x86-64-apx-ndd-wig"
 run_dump_test "x86-64-apx-jmpabs"
 run_dump_test "x86-64-apx-jmpabs-intel"
 run_dump_test "x86-64-apx-jmpabs-inval"
+run_dump_test "sse2avx-apx"
 run_dump_test "x86-64-avx512f-rcigrz-intel"
 run_dump_test "x86-64-avx512f-rcigrz"
 run_dump_test "x86-64-clwb"
--- a/gas/testsuite/gas/i386/x86-64-sse2avx.d
+++ b/gas/testsuite/gas/i386/x86-64-sse2avx.d
@@ -746,9 +746,19 @@  Disassembly of section .text:
 [ 	]*[a-f0-9]+:	c5 79 7e c8          	vmovd  %xmm9,%eax
 [ 	]*[a-f0-9]+:	c4 a1 79 7e c8       	vmovd  %xmm1,%eax
 [ 	]*[a-f0-9]+:	c4 e1 f9 7e c8       	vmovq  %xmm1,%rax
+[ 	]*[a-f0-9]+:	62 f1 7d 08 7e c8    	\{evex\} vmovd %xmm1,%eax
 [ 	]*[a-f0-9]+:	c5 f9 7e c8          	vmovd  %xmm1,%eax
 [ 	]*[a-f0-9]+:	c5 f9 7e c8          	vmovd  %xmm1,%eax
 [ 	]*[a-f0-9]+:	c4 e1 79 7e c8       	vmovd  %xmm1,%eax
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e c8    	vmovd  %xmm1,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e c8    	vmovd  %xmm1,%r16d
+[ 	]*[a-f0-9]+:	62 d9 7d 08 7e c8    	vmovd  %xmm1,%r24d
+[ 	]*[a-f0-9]+:	62 79 7d 08 7e c8    	vmovd  %xmm9,%r16d
+[ 	]*[a-f0-9]+:	62 b9 7d 08 7e c8    	vmovd  %xmm1,%r16d
+[ 	]*[a-f0-9]+:	62 f9 fd 08 7e c8    	vmovq  %xmm1,%r16
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e c8    	vmovd  %xmm1,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e c8    	vmovd  %xmm1,%r16d
+[ 	]*[a-f0-9]+:	62 f9 7d 08 7e c8    	vmovd  %xmm1,%r16d
 [ 	]*[a-f0-9]+:	c5 f8 ae 11          	vldmxcsr \(%rcx\)
 [ 	]*[a-f0-9]+:	c5 f8 ae 19          	vstmxcsr \(%rcx\)
 [ 	]*[a-f0-9]+:	c5 f8 5b f4          	vcvtdq2ps %xmm4,%xmm6
--- a/gas/testsuite/gas/i386/x86-64-sse2avx.s
+++ b/gas/testsuite/gas/i386/x86-64-sse2avx.s
@@ -847,10 +847,21 @@  _start:
 	rex.r	movd	%xmm1, %eax
 	rex.x	movd	%xmm1, %eax
 	rex.w	movd	%xmm1, %eax
+	{evex}	movd	%xmm1, %eax
 	{rex}	movd	%xmm1, %eax
 	{rex2}	movd	%xmm1, %eax
 	{vex3}	movd	%xmm1, %eax
 
+		movd	%xmm1, %r16d
+	rex	movd	%xmm1, %r16d
+	rex.b	movd	%xmm1, %r16d
+	rex.r	movd	%xmm1, %r16d
+	rex.x	movd	%xmm1, %r16d
+	rex.w	movd	%xmm1, %r16d
+	{evex}	movd	%xmm1, %r16d
+	{rex}	movd	%xmm1, %r16d
+	{rex2}	movd	%xmm1, %r16d
+
 	.intel_syntax noprefix
 # Tests for op mem64
 	ldmxcsr DWORD PTR [rcx]
--- a/opcodes/i386-opc.tbl
+++ b/opcodes/i386-opc.tbl
@@ -1071,10 +1071,40 @@  pause, 0xf390, i186, NoSuf, {}
     $avx:AVX:66:Vex128|VexVVVV|VexW0|SSE2AVX:RegXMM:Xmmword, +
     $sse:SSE2:66::RegXMM:Xmmword, +
     $mmx:MMX:::RegMMX:Qword>
+// As above, but also allowing AVX512 (EVEX) encoding, to transform
+// in particular insns using eGPR-s.
+<MMX:cpu:pfx:attr:reg:mem, +
+    $avx:AVX|AVX512VL:66:Vex128|EVex128|VexVVVV|VexW0|Disp8MemShift=4|SSE2AVX:RegXMM:Xmmword, +
+    $sse:SSE2:66::RegXMM:Xmmword, +
+    $mmx:MMX:::RegMMX:Qword>
+<MMXdq:opc:cpu:pfx:attr:reg:mem, +
+    d:0:AVX|AVX512VL:66:Vex128|EVex128|VexVVVV|VexW0|Disp8MemShift=4|SSE2AVX:RegXMM:Xmmword, +
+    d:0:SSE2:66::RegXMM:Xmmword, +
+    d:0:MMX:::RegMMX:Qword, +
+    q:1:AVX:66:Vex128|VexVVVV|VexW0|SSE2AVX:RegXMM:Xmmword, +
+    q:1:AVX512VL:66:EVex128|VexVVVV|VexW1|Disp8MemShift=4|SSE2AVX:RegXMM:Xmmword, +
+    q:1:SSE2:66::RegXMM:Xmmword, +
+    q:1:MMX:::RegMMX:Qword>
+<MMXBW:cpu:pfx:attr:reg:mem, +
+    $avx:AVX:66:Vex128|VexVVVV|VexW0|SSE2AVX:RegXMM:Xmmword, +
+    $apx:AVX512BW&AVX512VL:66:EVex128|VexVVVV|VexW0|Disp8MemShift=4|SSE2AVX:RegXMM:Xmmword, +
+    $sse:SSE2:66::RegXMM:Xmmword, +
+    $mmx:MMX:::RegMMX:Qword>
 
 <sse2:cpu:attr:scal:vvvv, +
     $avx:AVX:Vex128|VexW0|SSE2AVX:VexLIG|VexW0|SSE2AVX:VexVVVV, +
     $sse:SSE2:::>
+<SSE2BW:cpu:attr:vvvv, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512BW&AVX512VL:EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexVVVV, +
+    $sse:SSE2::>
+<SSE2D:cpu:attr:scal:vvvv, +
+    $avx:AVX|AVX512VL:Vex128|EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexLIG|EVexLIG|VexW0|Disp8MemShift=2|SSE2AVX:VexVVVV, +
+    $sse:SSE2:::>
+<SSE2Q:cpu:attr:scal:vvvv, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexLIG|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512VL:EVex128|VexW1|Disp8MemShift=4|SSE2AVX:EVexLIG|VexW1|Disp8MemShift=3|SSE2AVX:VexVVVV, +
+    $sse:SSE2:::>
 
 <bw:opc:vexw:elem:kcpu:kpfx:cpubmi, +
     b:0:VexW0:Byte:AVX512DQ:66:AVX512VBMI, +
@@ -1089,8 +1119,8 @@  emms, 0xf77, MMX, NoSuf, {}
 // copying between Reg64/Mem64 and RegXMM/RegMMX, as is mandated by Intel's
 // spec). AMD's spec, having been in existence for much longer, failed to
 // recognize that and specified movd for 32- and 64-bit operations.
-movd, 0x666e, AVX, D|Modrm|Vex128|Space0F|VexW0|NoSuf|SSE2AVX, { Reg32|Unspecified|BaseIndex, RegXMM }
-movd, 0x666e, AVX&x64, D|Modrm|Vex=1|Space0F|VexW1|NoSuf|Size64|SSE2AVX, { Reg64|BaseIndex, RegXMM }
+movd, 0x666e, AVX|AVX512F, D|Modrm|Vex128|EVex128|Space0F|VexW0|Disp8MemShift=2|NoSuf|SSE2AVX, { Reg32|Unspecified|BaseIndex, RegXMM }
+movd, 0x666e, x64&(AVX|AVX512F), D|Modrm|Vex128|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|Size64|SSE2AVX, { Reg64|BaseIndex, RegXMM }
 movd, 0x660f6e, SSE2, D|Modrm|IgnoreSize|NoSuf, { Reg32|Unspecified|BaseIndex, RegXMM }
 movd, 0x660f6e, SSE2&x64, D|Modrm|NoSuf|Size64, { Reg64|BaseIndex, RegXMM }
 // The MMX templates have to remain after at least the SSE2AVX ones.
@@ -1098,247 +1128,272 @@  movd, 0xf6e, MMX, D|Modrm|IgnoreSize|NoS
 movd, 0xf6e, MMX&x64, D|Modrm|NoSuf|Size64, { Reg64|BaseIndex, RegMMX }
 movq, 0xf37e, AVX, Load|Modrm|Vex128|Space0F|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
 movq, 0x66d6, AVX, Modrm|Vex128|Space0F|VexW0|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex|RegXMM }
-movq, 0x666e, AVX&x64, D|Modrm|Vex=1|Space0F|VexW1|NoSuf|Size64|SSE2AVX, { Reg64|Unspecified|BaseIndex, RegXMM }
+movq, 0xf37e, AVX512F, Load|Modrm|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+movq, 0x66d6, AVX512F, Modrm|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex|RegXMM }
+movq, 0x666e, x64&(AVX|AVX512F), D|Modrm|Vex128|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|Size64|SSE2AVX, { Reg64|Unspecified|BaseIndex, RegXMM }
 movq, 0xf30f7e, SSE2, Load|Modrm|NoSuf, { Unspecified|Qword|BaseIndex|RegXMM, RegXMM }
 movq, 0x660fd6, SSE2, Modrm|NoSuf, { RegXMM, Unspecified|Qword|BaseIndex|RegXMM }
 movq, 0x660f6e, SSE2&x64, D|Modrm|NoSuf|Size64, { Reg64|Unspecified|BaseIndex, RegXMM }
 // The MMX templates have to remain after at least the SSE2AVX ones.
 movq, 0xf6f, MMX, D|Modrm|NoSuf, { Unspecified|Qword|BaseIndex|RegMMX, RegMMX }
 movq, 0xf6e, MMX&x64, D|Modrm|NoSuf|Size64, { Reg64|Unspecified|BaseIndex, RegMMX }
-packssdw<mmx>, 0x<mmx:pfx>0f6b, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-packsswb<mmx>, 0x<mmx:pfx>0f63, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-packuswb<mmx>, 0x<mmx:pfx>0f67, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-padd<bw><mmx>, 0x<mmx:pfx>0ffc | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-paddd<mmx>, 0x<mmx:pfx>0ffe, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-paddq<sse2>, 0x660fd4, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+packssdw<MMXBW>, 0x<MMXBW:pfx>0f6b, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+packsswb<MMXBW>, 0x<MMXBW:pfx>0f63, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+packuswb<MMXBW>, 0x<MMXBW:pfx>0f67, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+padd<bw><MMXBW>, 0x<MMXBW:pfx>0ffc | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+paddd<MMX>, 0x<MMX:pfx>0ffe, <MMX:cpu>, Modrm|<MMX:attr>|C|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+paddq<SSE2Q>, 0x660fd4, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 paddq, 0xfd4, SSE2, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-padds<bw><mmx>, 0x<mmx:pfx>0fec | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-paddus<bw><mmx>, 0x<mmx:pfx>0fdc | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-pand<mmx>, 0x<mmx:pfx>0fdb, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-pandn<mmx>, 0x<mmx:pfx>0fdf, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
+padds<bw><MMXBW>, 0x<MMXBW:pfx>0fec | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+paddus<bw><MMXBW>, 0x<MMXBW:pfx>0fdc | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+pand<MMX>, 0x<MMX:pfx>0fdb, <MMX:cpu>, Modrm|<MMX:attr>|C|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+pandn<MMX>, 0x<MMX:pfx>0fdf, <MMX:cpu>, Modrm|<MMX:attr>|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
 pcmpeq<bw><mmx>, 0x<mmx:pfx>0f74 | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
 pcmpeqd<mmx>, 0x<mmx:pfx>0f76, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
 pcmpgt<bw><mmx>, 0x<mmx:pfx>0f64 | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf|Optimize, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
 pcmpgtd<mmx>, 0x<mmx:pfx>0f66, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf|Optimize, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-pmaddwd<mmx>, 0x<mmx:pfx>0ff5, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-pmulhw<mmx>, 0x<mmx:pfx>0fe5, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-pmullw<mmx>, 0x<mmx:pfx>0fd5, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-por<mmx>, 0x<mmx:pfx>0feb, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psllw<mmx>, 0x<mmx:pfx>0ff1, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psllw<mmx>, 0x<mmx:pfx>0f71/6, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psll<dq><mmx>, 0x<mmx:pfx>0ff2 | <dq:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psll<dq><mmx>, 0x<mmx:pfx>0f72 | <dq:opc>/6, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psraw<mmx>, 0x<mmx:pfx>0fe1, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psraw<mmx>, 0x<mmx:pfx>0f71/4, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psrad<mmx>, 0x<mmx:pfx>0fe2, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psrad<mmx>, 0x<mmx:pfx>0f72/4, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psrlw<mmx>, 0x<mmx:pfx>0fd1, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psrlw<mmx>, 0x<mmx:pfx>0f71/2, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psrl<dq><mmx>, 0x<mmx:pfx>0fd2 | <dq:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psrl<dq><mmx>, 0x<mmx:pfx>0f72 | <dq:opc>/2, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { Imm8, <mmx:reg> }
-psub<bw><mmx>, 0x<mmx:pfx>0ff8 | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psubd<mmx>, 0x<mmx:pfx>0ffa, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psubq<sse2>, 0x660ffb, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaddwd<MMXBW>, 0x<MMXBW:pfx>0ff5, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+pmulhw<MMXBW>, 0x<MMXBW:pfx>0fe5, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+pmullw<MMXBW>, 0x<MMXBW:pfx>0fd5, <MMXBW:cpu>, Modrm|<MMXBW:attr>|C|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+por<MMX>, 0x<MMX:pfx>0feb, <MMX:cpu>, Modrm|<MMX:attr>|C|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+psllw<MMXBW>, 0x<MMXBW:pfx>0ff1, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+psllw<MMXBW>, 0x<MMXBW:pfx>0f71/6, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { Imm8, <MMXBW:reg> }
+psll<MMXdq>, 0x<MMXdq:pfx>0ff2 | <MMXdq:opc>, <MMXdq:cpu>, Modrm|<MMXdq:attr>|NoSuf, { <MMXdq:reg>|<MMXdq:mem>|Unspecified|BaseIndex, <MMXdq:reg> }
+psll<MMXdq>, 0x<MMXdq:pfx>0f72 | <MMXdq:opc>/6, <MMXdq:cpu>, Modrm|<MMXdq:attr>|NoSuf, { Imm8, <MMXdq:reg> }
+psraw<MMXBW>, 0x<MMXBW:pfx>0fe1, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+psraw<MMXBW>, 0x<MMXBW:pfx>0f71/4, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { Imm8, <MMXBW:reg> }
+psrad<MMX>, 0x<MMX:pfx>0fe2, <MMX:cpu>, Modrm|<MMX:attr>|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+psrad<MMX>, 0x<MMX:pfx>0f72/4, <MMX:cpu>, Modrm|<MMX:attr>|NoSuf, { Imm8, <MMX:reg> }
+psrlw<MMXBW>, 0x<MMXBW:pfx>0fd1, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+psrlw<MMXBW>, 0x<MMXBW:pfx>0f71/2, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { Imm8, <MMXBW:reg> }
+psrl<MMXdq>, 0x<MMXdq:pfx>0fd2 | <MMXdq:opc>, <MMXdq:cpu>, Modrm|<MMXdq:attr>|NoSuf, { <MMXdq:reg>|<MMXdq:mem>|Unspecified|BaseIndex, <MMXdq:reg> }
+psrl<MMXdq>, 0x<MMXdq:pfx>0f72 | <MMXdq:opc>/2, <MMXdq:cpu>, Modrm|<MMXdq:attr>|NoSuf, { Imm8, <MMXdq:reg> }
+psub<bw><MMXBW>, 0x<MMXBW:pfx>0ff8 | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+psubd<MMX>, 0x<MMX:pfx>0ffa, <MMX:cpu>, Modrm|<MMX:attr>|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+psubq<SSE2Q>, 0x660ffb, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 psubq, 0xffb, SSE2, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-psubs<bw><mmx>, 0x<mmx:pfx>0fe8 | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-psubus<bw><mmx>, 0x<mmx:pfx>0fd8 | <bw:opc>, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-punpckhbw<mmx>, 0x<mmx:pfx>0f68, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-punpckhwd<mmx>, 0x<mmx:pfx>0f69, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-punpckhdq<mmx>, 0x<mmx:pfx>0f6a, <mmx:cpu>, Modrm|<mmx:attr>|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
-punpcklbw<sse2>, 0x660f60, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+psubs<bw><MMXBW>, 0x<MMXBW:pfx>0fe8 | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+psubus<bw><MMXBW>, 0x<MMXBW:pfx>0fd8 | <bw:opc>, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+punpckhbw<MMXBW>, 0x<MMXBW:pfx>0f68, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+punpckhwd<MMXBW>, 0x<MMXBW:pfx>0f69, <MMXBW:cpu>, Modrm|<MMXBW:attr>|NoSuf, { <MMXBW:reg>|<MMXBW:mem>|Unspecified|BaseIndex, <MMXBW:reg> }
+punpckhdq<MMX>, 0x<MMX:pfx>0f6a, <MMX:cpu>, Modrm|<MMX:attr>|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
+punpcklbw<SSE2BW>, 0x660f60, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 punpcklbw, 0xf60, MMX, Modrm|NoSuf, { Dword|Unspecified|BaseIndex|RegMMX, RegMMX }
-punpcklwd<sse2>, 0x660f61, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+punpcklwd<SSE2BW>, 0x660f61, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 punpcklwd, 0xf61, MMX, Modrm|NoSuf, { Dword|Unspecified|BaseIndex|RegMMX, RegMMX }
-punpckldq<sse2>, 0x660f62, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+punpckldq<SSE2D>, 0x660f62, <SSE2D:cpu>, Modrm|<SSE2D:attr>|<SSE2D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 punpckldq, 0xf62, MMX, Modrm|NoSuf, { Dword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pxor<mmx>, 0x<mmx:pfx>0fef, <mmx:cpu>, Modrm|<mmx:attr>|C|NoSuf, { <mmx:reg>|<mmx:mem>|Unspecified|BaseIndex, <mmx:reg> }
+pxor<MMX>, 0x<MMX:pfx>0fef, <MMX:cpu>, Modrm|<MMX:attr>|C|NoSuf, { <MMX:reg>|<MMX:mem>|Unspecified|BaseIndex, <MMX:reg> }
 
 // SSE instructions.
 
 <sse:cpu:attr:scal:vvvv, +
     $avx:AVX:Vex128|VexW0|SSE2AVX:VexLIG|VexW0|SSE2AVX:VexVVVV, +
     $sse:SSE:::>
+<SSE:cpu:attr:scal:vvvv, +
+    $avx:AVX|AVX512VL:Vex128|EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexLIG|EVexLIG|VexW0|Disp8MemShift=2|SSE2AVX:VexVVVV, +
+    $sse:SSE:::>
+<SSEDQ:cpu:attr, +
+    $avx:AVX:Vex128|VexW0|VexVVVV|SSE2AVX, +
+    $apx:AVX512DQ&AVX512VL:EVex128|VexW0|VexVVVV|Disp8MemShift=4|SSE2AVX, +
+    $sse:SSE:>
 <frel:imm:comm, eq:0:C, lt:1:, le:2:, unord:3:C, neq:4:C, nlt:5:, nle:6:, ord:7:C>
 
-addps<sse>, 0x0f58, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-addss<sse>, 0xf30f58, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-andnps<sse>, 0x0f55, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-andps<sse>, 0x0f54, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+addps<SSE>, 0x0f58, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+addss<SSE>, 0xf30f58, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+andnps<SSEDQ>, 0x0f55, <SSEDQ:cpu>, Modrm|<SSEDQ:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+andps<SSEDQ>, 0x0f54, <SSEDQ:cpu>, Modrm|<SSEDQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 cmp<frel>ps<sse>, 0x0fc2/<frel:imm>, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|<frel:comm>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
 cmp<frel>ss<sse>, 0xf30fc2/<frel:imm>, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|<frel:comm>|NoSuf|ImmExt, { RegXMM|Dword|Unspecified|BaseIndex, RegXMM }
 cmpps<sse>, 0x0fc2, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
 cmpss<sse>, 0xf30fc2, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Imm8, Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-comiss<sse>, 0x0f2f, <sse:cpu>, Modrm|<sse:scal>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+comiss<SSE>, 0x0f2f, <SSE:cpu>, Modrm|<SSE:scal>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 cvtpi2ps, 0xf2a, SSE, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegXMM }
 cvtps2pi, 0xf2d, SSE, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegMMX }
 cvtsi2ss<sse>, 0xf30f2a, <sse:cpu>&No64, Modrm|<sse:scal>|<sse:vvvv>|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|No_qSuf, { Reg32|Unspecified|BaseIndex, RegXMM }
-cvtsi2ss, 0xf32a, AVX&x64, Modrm|Vex=3|Space0F|VexVVVV|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
-cvtsi2ss, 0xf32a, AVX&x64, Modrm|Vex=3|Space0F|VexVVVV|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
+cvtsi2ss, 0xf32a, x64&(AVX|AVX512F), Modrm|VexLIG|EVexLIG|Space0F|VexVVVV|IgnoreSize|Disp8ShiftVL|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
+cvtsi2ss, 0xf32a, x64&(AVX|AVX512F), Modrm|VexLIG|EVexLIG|Space0F|VexVVVV|Disp8ShiftVL|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
 cvtsi2ss, 0xf30f2a, SSE&x64, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
 cvtsi2ss, 0xf30f2a, SSE&x64, Modrm|No_bSuf|No_wSuf|No_sSuf|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
-cvtss2si, 0xf32d, AVX, Modrm|VexLIG|Space0F|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
+cvtss2si, 0xf32d, AVX|AVX512F, Modrm|VexLIG|EVexLIG|Space0F|Disp8MemShift=2|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
 cvtss2si, 0xf30f2d, SSE, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
 cvttps2pi, 0xf2c, SSE, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegMMX }
-cvttss2si, 0xf32c, AVX, Modrm|VexLIG|Space0F|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
+cvttss2si, 0xf32c, AVX|AVX512F, Modrm|VexLIG|EVexLIG|Space0F|Disp8MemShift=2|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
 cvttss2si, 0xf30f2c, SSE, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf, { Dword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
-divps<sse>, 0x0f5e, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-divss<sse>, 0xf30f5e, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+divps<SSE>, 0x0f5e, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+divss<SSE>, 0xf30f5e, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 ldmxcsr<sse>, 0x0fae/2, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { Dword|Unspecified|BaseIndex }
 maskmovq, 0xff7, SSE|3dnowA, Modrm|NoSuf, { RegMMX, RegMMX }
-maxps<sse>, 0x0f5f, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-maxss<sse>, 0xf30f5f, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-minps<sse>, 0x0f5d, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-minss<sse>, 0xf30f5d, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movaps<sse>, 0x0f28, <sse:cpu>, D|Modrm|<sse:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+maxps<SSE>, 0x0f5f, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+maxss<SSE>, 0xf30f5f, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+minps<SSE>, 0x0f5d, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+minss<SSE>, 0xf30f5d, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+movaps<SSE>, 0x0f28, <SSE:cpu>, D|Modrm|<SSE:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 movhlps<sse>, 0x0f12, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM, RegXMM }
-movhps, 0x16, AVX, Modrm|Vex|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
-movhps, 0x17, AVX, Modrm|Vex|Space0F|VexW0|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
+movhps, 0x16, AVX|AVX512F, Modrm|Vex128|EVex128|Space0F|VexVVVV|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+movhps, 0x17, AVX|AVX512F, Modrm|Vex128|EVex128|Space0F|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
 movhps, 0xf16, SSE, D|Modrm|NoSuf, { Qword|Unspecified|BaseIndex, RegXMM }
 movlhps<sse>, 0x0f16, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM, RegXMM }
-movlps, 0x12, AVX, Modrm|Vex|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
-movlps, 0x13, AVX, Modrm|Vex|Space0F|VexW0|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
+movlps, 0x12, AVX|AVX512F, Modrm|Vex128|EVex128|Space0F|VexVVVV|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+movlps, 0x13, AVX|AVX512F, Modrm|Vex128|EVex128|Space0F|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
 movlps, 0xf12, SSE, D|Modrm|NoSuf, { Qword|Unspecified|BaseIndex, RegXMM }
 movmskps<sse>, 0x0f50, <sse:cpu>, Modrm|<sse:attr>|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { RegXMM, Reg32|Reg64 }
-movntps<sse>, 0x0f2b, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
+movntps<SSE>, 0x0f2b, <SSE:cpu>, Modrm|<SSE:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
 movntq, 0xfe7, SSE|3dnowA, Modrm|NoSuf, { RegMMX, Qword|Unspecified|BaseIndex }
-movntdq<sse2>, 0x660fe7, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
-movss, 0xf310, AVX, D|Modrm|VexLIG|Space0F|VexW0|NoSuf|SSE2AVX, { Dword|Unspecified|BaseIndex, RegXMM }
+movntdq<SSE2D>, 0x660fe7, <SSE2D:cpu>, Modrm|<SSE2D:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
+movss, 0xf310, AVX|AVX512F, D|Modrm|VexLIG|EVexLIG|Space0F|VexW0|Disp8MemShift=2|NoSuf|SSE2AVX, { Dword|Unspecified|BaseIndex, RegXMM }
 movss, 0xf310, AVX, D|Modrm|VexLIG|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { RegXMM, RegXMM }
 movss, 0xf30f10, SSE, D|Modrm|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movups<sse>, 0x0f10, <sse:cpu>, D|Modrm|<sse:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-mulps<sse>, 0x0f59, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-mulss<sse>, 0xf30f59, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-orps<sse>, 0x0f56, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+movups<SSE>, 0x0f10, <SSE:cpu>, D|Modrm|<SSE:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+mulps<SSE>, 0x0f59, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+mulss<SSE>, 0xf30f59, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+orps<SSEDQ>, 0x0f56, <SSEDQ:cpu>, Modrm|<SSEDQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pavg<bw>, 0xfe0 | (3 * <bw:opc>), SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pavg<bw><sse2>, 0x660fe0 | (3 * <bw:opc>), <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pextrw<sse2>, 0x660fc5, <sse2:cpu>, Load|Modrm|<sse2:attr>|No_bSuf|No_wSuf|No_sSuf|IgnoreSize|NoRex64, { Imm8, RegXMM, Reg32|Reg64 }
+pavg<bw><SSE2BW>, 0x660fe0 | (3 * <bw:opc>), <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pextrw<SSE2D>, 0x660fc5, <SSE2D:cpu>, Load|Modrm|<SSE2D:attr>|No_bSuf|No_wSuf|No_sSuf|IgnoreSize|NoRex64, { Imm8, RegXMM, Reg32|Reg64 }
 pextrw, 0xfc5, SSE|3dnowA, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { Imm8, RegMMX, Reg32|Reg64 }
-pinsrw<sse2>, 0x660fc4, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|No_bSuf|No_wSuf|No_sSuf|IgnoreSize|NoRex64, { Imm8, Reg32|Reg64, RegXMM }
-pinsrw<sse2>, 0x660fc4, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { Imm8, Word|Unspecified|BaseIndex, RegXMM }
+pinsrw<SSE2D>, 0x660fc4, <SSE2D:cpu>, Modrm|<SSE2D:attr>|<SSE2D:vvvv>|No_bSuf|No_wSuf|No_sSuf|IgnoreSize|NoRex64, { Imm8, Reg32|Reg64, RegXMM }
+pinsrw<SSE2D>, 0x660fc4, <SSE2D:cpu>, Modrm|<SSE2D:attr>|<SSE2D:vvvv>|Disp8MemShift|NoSuf, { Imm8, Word|Unspecified|BaseIndex, RegXMM }
 pinsrw, 0xfc4, SSE|3dnowA, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { Imm8, Reg32|Reg64, RegMMX }
 pinsrw, 0xfc4, SSE|3dnowA, Modrm|NoSuf, { Imm8, Word|Unspecified|BaseIndex, RegMMX }
-pmaxsw<sse2>, 0x660fee, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxsw<SSE2BW>, 0x660fee, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pmaxsw, 0xfee, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pmaxub<sse2>, 0x660fde, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxub<SSE2BW>, 0x660fde, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pmaxub, 0xfde, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pminsw<sse2>, 0x660fea, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminsw<SSE2BW>, 0x660fea, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pminsw, 0xfea, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pminub<sse2>, 0x660fda, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminub<SSE2BW>, 0x660fda, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pminub, 0xfda, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
 pmovmskb<sse2>, 0x660fd7, <sse2:cpu>, Modrm|<sse2:attr>|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { RegXMM, Reg32|Reg64 }
 pmovmskb, 0xfd7, SSE|3dnowA, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { RegMMX, Reg32|Reg64 }
-pmulhuw<sse2>, 0x660fe4, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmulhuw<SSE2BW>, 0x660fe4, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pmulhuw, 0xfe4, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
 prefetchnta, 0xf18/0, SSE|3dnowA, Modrm|Anysize|IgnoreSize|NoSuf, { BaseIndex }
 prefetcht0, 0xf18/1, SSE|3dnowA, Modrm|Anysize|IgnoreSize|NoSuf, { BaseIndex }
 prefetcht1, 0xf18/2, SSE|3dnowA, Modrm|Anysize|IgnoreSize|NoSuf, { BaseIndex }
 prefetcht2, 0xf18/3, SSE|3dnowA, Modrm|Anysize|IgnoreSize|NoSuf, { BaseIndex }
 psadbw, 0xff6, SSE|3dnowA, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-psadbw<sse2>, 0x660ff6, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+psadbw<SSE2BW>, 0x660ff6, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|<SSE2BW:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pshufw, 0xf70, SSE|3dnowA, Modrm|NoSuf, { Imm8|Imm8S, Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
 rcpps<sse>, 0x0f53, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 rcpss<sse>, 0xf30f53, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 rsqrtps<sse>, 0x0f52, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 rsqrtss<sse>, 0xf30f52, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 sfence, 0xfaef8, SSE|3dnowA, NoSuf, {}
-shufps<sse>, 0x0fc6, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-sqrtps<sse>, 0x0f51, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-sqrtss<sse>, 0xf30f51, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+shufps<SSE>, 0x0fc6, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+sqrtps<SSE>, 0x0f51, <SSE:cpu>, Modrm|<SSE:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+sqrtss<SSE>, 0xf30f51, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 stmxcsr<sse>, 0x0fae/3, <sse:cpu>, Modrm|<sse:attr>|NoSuf, { Dword|Unspecified|BaseIndex }
-subps<sse>, 0x0f5c, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-subss<sse>, 0xf30f5c, <sse:cpu>, Modrm|<sse:scal>|<sse:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-ucomiss<sse>, 0x0f2e, <sse:cpu>, Modrm|<sse:scal>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-unpckhps<sse>, 0x0f15, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-unpcklps<sse>, 0x0f14, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-xorps<sse>, 0x0f57, <sse:cpu>, Modrm|<sse:attr>|<sse:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+subps<SSE>, 0x0f5c, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+subss<SSE>, 0xf30f5c, <SSE:cpu>, Modrm|<SSE:scal>|<SSE:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+ucomiss<SSE>, 0x0f2e, <SSE:cpu>, Modrm|<SSE:scal>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+unpckhps<SSE>, 0x0f15, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+unpcklps<SSE>, 0x0f14, <SSE:cpu>, Modrm|<SSE:attr>|<SSE:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+xorps<SSEDQ>, 0x0f57, <SSEDQ:cpu>, Modrm|<SSEDQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 
 // SSE2 instructions.
 
-addpd<sse2>, 0x660f58, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-addsd<sse2>, 0xf20f58, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-andnpd<sse2>, 0x660f55, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-andpd<sse2>, 0x660f54, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+<SSE2DQ:cpu:attr, +
+    $avx:AVX:Vex128|VexW0|VexVVVV|SSE2AVX, +
+    $apx:AVX512DQ&AVX512VL:EVex128|VexW1|VexVVVV|Disp8MemShift=4|SSE2AVX, +
+    $sse:SSE2:>
+
+addpd<SSE2Q>, 0x660f58, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+addsd<SSE2Q>, 0xf20f58, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+andnpd<SSE2DQ>, 0x660f55, <SSE2DQ:cpu>, Modrm|<SSE2DQ:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+andpd<SSE2DQ>, 0x660f54, <SSE2DQ:cpu>, Modrm|<SSE2DQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 cmp<frel>pd<sse2>, 0x660fc2/<frel:imm>, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|<frel:comm>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
 cmp<frel>sd<sse2>, 0xf20fc2/<frel:imm>, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|<frel:comm>|NoSuf|ImmExt, { RegXMM|Qword|Unspecified|BaseIndex, RegXMM }
 cmppd<sse2>, 0x660fc2, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
 cmpsd<sse2>, 0xf20fc2, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Imm8, Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-comisd<sse2>, 0x660f2f, <sse2:cpu>, Modrm|<sse2:scal>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+comisd<SSE2Q>, 0x660f2f, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
 cvtpi2pd, 0x660f2a, SSE2, Modrm|NoSuf, { RegMMX, RegXMM }
-cvtpi2pd, 0xf3e6, AVX, Modrm|Vex|Space0F|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+cvtpi2pd, 0xf3e6, AVX|AVX512VL, Modrm|Vex128|EVex128|Space0F|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
 cvtpi2pd, 0x660f2a, SSE2, Modrm|NoSuf, { Qword|Unspecified|BaseIndex, RegXMM }
 cvtsi2sd<sse2>, 0xf20f2a, <sse2:cpu>&No64, Modrm|IgnoreSize|<sse2:scal>|<sse2:vvvv>|No_bSuf|No_wSuf|No_sSuf|No_qSuf, { Reg32|Unspecified|BaseIndex, RegXMM }
-cvtsi2sd, 0xf22a, AVX&x64, Modrm|Vex=3|Space0F|VexVVVV|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
-cvtsi2sd, 0xf22a, AVX&x64, Modrm|Vex=3|Space0F|VexVVVV|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
+cvtsi2sd, 0xf22a, x64&(AVX|AVX512F), Modrm|VexLIG|EVexLIG|Space0F|VexVVVV|Disp8ShiftVL|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
+cvtsi2sd, 0xf22a, x64&(AVX|AVX512F), Modrm|VexLIG|EVexLIG|Space0F|VexVVVV|Disp8ShiftVL|No_bSuf|No_wSuf|No_sSuf|SSE2AVX|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
 cvtsi2sd, 0xf20f2a, SSE2&x64, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|ATTSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
 cvtsi2sd, 0xf20f2a, SSE2&x64, Modrm|No_bSuf|No_wSuf|No_sSuf|IntelSyntax, { Reg32|Reg64|Unspecified|BaseIndex, RegXMM }
-divpd<sse2>, 0x660f5e, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-divsd<sse2>, 0xf20f5e, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-maxpd<sse2>, 0x660f5f, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-maxsd<sse2>, 0xf20f5f, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-minpd<sse2>, 0x660f5d, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-minsd<sse2>, 0xf20f5d, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movapd<sse2>, 0x660f28, <sse2:cpu>, D|Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+divpd<SSE2Q>, 0x660f5e, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+divsd<SSE2Q>, 0xf20f5e, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+maxpd<SSE2Q>, 0x660f5f, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+maxsd<SSE2Q>, 0xf20f5f, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+minpd<SSE2Q>, 0x660f5d, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+minsd<SSE2Q>, 0xf20f5d, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+movapd<SSE2Q>, 0x660f28, <SSE2Q:cpu>, D|Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 movhpd, 0x6616, AVX, Modrm|Vex|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+movhpd, 0x6616, AVX512F, Modrm|EVex128|Space0F|VexVVVV|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
 movhpd, 0x6617, AVX, Modrm|Vex|Space0F|VexW0|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
+movhpd, 0x6617, AVX512F, Modrm|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
 movhpd, 0x660f16, SSE2, D|Modrm|NoSuf, { Qword|Unspecified|BaseIndex, RegXMM }
 movlpd, 0x6612, AVX, Modrm|Vex|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+movlpd, 0x6612, AVX512F, Modrm|EVex128|Space0F|VexVVVV|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
 movlpd, 0x6613, AVX, Modrm|Vex|Space0F|VexW0|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
+movlpd, 0x6613, AVX512F, Modrm|EVex128|Space0F|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { RegXMM, Qword|Unspecified|BaseIndex }
 movlpd, 0x660f12, SSE2, D|Modrm|NoSuf, { Qword|Unspecified|BaseIndex, RegXMM }
 movmskpd<sse2>, 0x660f50, <sse2:cpu>, Modrm|<sse2:attr>|IgnoreSize|No_bSuf|No_wSuf|No_sSuf|NoRex64, { RegXMM, Reg32|Reg64 }
-movntpd<sse2>, 0x660f2b, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
+movntpd<SSE2Q>, 0x660f2b, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|NoSuf, { RegXMM, Xmmword|Unspecified|BaseIndex }
 movsd, 0xf210, AVX, D|Modrm|VexLIG|Space0F|VexW0|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
+movsd, 0xf210, AVX512F, D|Modrm|VexLIG|EVexLIG|Space0F|VexW0|Disp8MemShift=3|NoSuf|SSE2AVX, { Qword|Unspecified|BaseIndex, RegXMM }
 movsd, 0xf210, AVX, D|Modrm|VexLIG|Space0F|VexVVVV|VexW0|NoSuf|SSE2AVX, { RegXMM, RegXMM }
 movsd, 0xf20f10, SSE2, D|Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movupd<sse2>, 0x660f10, <sse2:cpu>, D|Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-mulpd<sse2>, 0x660f59, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-mulsd<sse2>, 0xf20f59, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-orpd<sse2>, 0x660f56, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-shufpd<sse2>, 0x660fc6, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-sqrtpd<sse2>, 0x660f51, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-sqrtsd<sse2>, 0xf20f51, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-subpd<sse2>, 0x660f5c, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-subsd<sse2>, 0xf20f5c, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-ucomisd<sse2>, 0x660f2e, <sse2:cpu>, Modrm|<sse2:scal>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-unpckhpd<sse2>, 0x660f15, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-unpcklpd<sse2>, 0x660f14, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-xorpd<sse2>, 0x660f57, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-cvtdq2pd<sse2>, 0xf30fe6, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-cvtpd2dq<sse2>, 0xf20fe6, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-cvtdq2ps<sse2>, 0x0f5b, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+movupd<SSE2Q>, 0x660f10, <SSE2Q:cpu>, D|Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+mulpd<SSE2Q>, 0x660f59, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+mulsd<SSE2Q>, 0xf20f59, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+orpd<SSE2DQ>, 0x660f56, <SSE2DQ:cpu>, Modrm|<SSE2DQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+shufpd<SSE2Q>, 0x660fc6, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+sqrtpd<SSE2Q>, 0x660f51, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+sqrtsd<SSE2Q>, 0xf20f51, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+subpd<SSE2Q>, 0x660f5c, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+subsd<SSE2Q>, 0xf20f5c, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+ucomisd<SSE2Q>, 0x660f2e, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+unpckhpd<SSE2Q>, 0x660f15, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+unpcklpd<SSE2Q>, 0x660f14, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+xorpd<SSE2DQ>, 0x660f57, <SSE2DQ:cpu>, Modrm|<SSE2DQ:attr>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvtdq2pd<SSE2D>, 0xf30fe6, <SSE2D:cpu>, Modrm|<SSE2D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+cvtpd2dq<SSE2Q>, 0xf20fe6, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvtdq2ps<SSE2D>, 0x0f5b, <SSE2D:cpu>, Modrm|<SSE2D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 cvtpd2pi, 0x660f2d, SSE2, Modrm|NoSuf, { RegXMM|Unspecified|BaseIndex, RegMMX }
-cvtpd2ps<sse2>, 0x660f5a, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-cvtps2pd<sse2>, 0x0f5a, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-cvtps2dq<sse2>, 0x660f5b, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-cvtsd2si, 0xf22d, AVX, Modrm|VexLIG|Space0F|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
+cvtpd2ps<SSE2Q>, 0x660f5a, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvtps2pd<SSE2D>, 0x0f5a, <SSE2D:cpu>, Modrm|<SSE2D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+cvtps2dq<SSE2D>, 0x660f5b, <SSE2D:cpu>, Modrm|<SSE2D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvtsd2si, 0xf22d, AVX|AVX512F, Modrm|VexLIG|EVexLIG|Space0F|Disp8MemShift=3|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
 cvtsd2si, 0xf20f2d, SSE2, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
-cvtsd2ss<sse2>, 0xf20f5a, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-cvtss2sd<sse2>, 0xf30f5a, <sse2:cpu>, Modrm|<sse2:scal>|<sse2:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-
+cvtsd2ss<SSE2Q>, 0xf20f5a, <SSE2Q:cpu>, Modrm|<SSE2Q:scal>|<SSE2Q:vvvv>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+cvtss2sd<SSE2D>, 0xf30f5a, <SSE2D:cpu>, Modrm|<SSE2D:scal>|<SSE2D:vvvv>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 cvttpd2pi, 0x660f2c, SSE2, Modrm|NoSuf, { RegXMM|Unspecified|BaseIndex, RegMMX }
-cvttsd2si, 0xf22c, AVX, Modrm|VexLIG|Space0F|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
+cvttsd2si, 0xf22c, AVX|AVX512F, Modrm|VexLIG|EVexLIG|Space0F|Disp8MemShift=3|No_bSuf|No_wSuf|No_sSuf|SSE2AVX, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
 cvttsd2si, 0xf20f2c, SSE2, Modrm|IgnoreSize|No_bSuf|No_wSuf|No_sSuf, { Qword|Unspecified|BaseIndex|RegXMM, Reg32|Reg64 }
-cvttpd2dq<sse2>, 0x660fe6, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-cvttps2dq<sse2>, 0xf30f5b, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvttpd2dq<SSE2Q>, 0x660fe6, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+cvttps2dq<SSE2D>, 0xf30f5b, <SSE2D:cpu>, Modrm|<SSE2D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 maskmovdqu<sse2>, 0x660ff7, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { RegXMM, RegXMM }
-movdqa<sse2>, 0x660f6f, <sse2:cpu>, D|Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-movdqu<sse2>, 0xf30f6f, <sse2:cpu>, D|Modrm|<sse2:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+movdqa<SSE2D>, 0x660f6f, <SSE2D:cpu>, D|Modrm|<SSE2D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+movdqu<SSE2D>, 0xf30f6f, <SSE2D:cpu>, D|Modrm|<SSE2D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 movdq2q, 0xf20fd6, SSE2, Modrm|NoSuf, { RegXMM, RegMMX }
 movq2dq, 0xf30fd6, SSE2, Modrm|NoSuf, { RegMMX, RegXMM }
-pmuludq<sse2>, 0x660ff4, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmuludq<SSE2Q>, 0x660ff4, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|C|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pmuludq, 0xff4, SSE2, Modrm|NoSuf, { Qword|Unspecified|BaseIndex|RegMMX, RegMMX }
-pshufd<sse2>, 0x660f70, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-pshufhw<sse2>, 0xf30f70, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-pshuflw<sse2>, 0xf20f70, <sse2:cpu>, Modrm|<sse2:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+pshufd<SSE2D>, 0x660f70, <SSE2D:cpu>, Modrm|<SSE2D:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+pshufhw<SSE2BW>, 0xf30f70, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+pshuflw<SSE2BW>, 0xf20f70, <SSE2BW:cpu>, Modrm|<SSE2BW:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
 pslldq<sse2>, 0x660f73/7, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { Imm8, RegXMM }
 psrldq<sse2>, 0x660f73/3, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { Imm8, RegXMM }
-punpckhqdq<sse2>, 0x660f6d, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-punpcklqdq<sse2>, 0x660f6c, <sse2:cpu>, Modrm|<sse2:attr>|<sse2:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+punpckhqdq<SSE2Q>, 0x660f6d, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+punpcklqdq<SSE2Q>, 0x660f6c, <SSE2Q:cpu>, Modrm|<SSE2Q:attr>|<SSE2Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 
 <frel>
 
 // SSE3 instructions.
 
 <sse3:cpu:attr:vvvv, $avx:AVX:Vex128|VexW0|SSE2AVX:VexVVVV, $sse:SSE3::>
+<SSE3D:cpu:attr, +
+    $avx:AVX|AVX512VL:Vex128|EVex128|VexW0|Disp8MemShift=4|SSE2AVX, +
+    $sse:SSE3:>
+<SSE3Q:cpu:attr, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512VL:EVex128|VexW1|Disp8MemShift=3|SSE2AVX, +
+    $sse:SSE3:>
 
 addsubpd<sse3>, 0x660fd0, <sse3:cpu>, Modrm|<sse3:attr>|<sse3:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 addsubps<sse3>, 0xf20fd0, <sse3:cpu>, Modrm|<sse3:attr>|<sse3:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
@@ -1346,10 +1401,13 @@  haddpd<sse3>, 0x660f7c, <sse3:cpu>, Modr
 haddps<sse3>, 0xf20f7c, <sse3:cpu>, Modrm|<sse3:attr>|<sse3:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 hsubpd<sse3>, 0x660f7d, <sse3:cpu>, Modrm|<sse3:attr>|<sse3:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 hsubps<sse3>, 0xf20f7d, <sse3:cpu>, Modrm|<sse3:attr>|<sse3:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-lddqu<sse3>, 0xf20ff0, <sse3:cpu>, Modrm|<sse3:attr>|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
-movddup<sse3>, 0xf20f12, <sse3:cpu>, Modrm|<sse3:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movshdup<sse3>, 0xf30f16, <sse3:cpu>, Modrm|<sse3:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-movsldup<sse3>, 0xf30f12, <sse3:cpu>, Modrm|<sse3:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+lddqu, 0xf20ff0, AVX, Modrm|Vex128|VexW0|SSE2AVX|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
+// For use with eGPR-s in the memory operand, utilize VMOVDQU32.
+lddqu, 0xf30f6f, AVX512VL, Modrm|EVex128|VexW0|Disp8MemShift=4|SSE2AVX|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
+lddqu, 0xf20ff0, SSE3, Modrm|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
+movddup<SSE3Q>, 0xf20f12, <SSE3Q:cpu>, Modrm|<SSE3Q:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+movshdup<SSE3D>, 0xf30f16, <SSE3D:cpu>, Modrm|<SSE3D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+movsldup<SSE3D>, 0xf30f12, <SSE3D:cpu>, Modrm|<SSE3D:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 
 // FPU instructions also covered by SSE3 CPUID flag.
 
@@ -1419,6 +1477,15 @@  invpcid, 0xf3f2, INVPCID&APX_F, Modrm|No
     $avx:AVX:66:Vex128|VexW0|SSE2AVX:VexVVVV:RegXMM:Xmmword, +
     $sse:SSSE3:66:::RegXMM:Xmmword, +
     $mmx:SSSE3::::RegMMX:Qword>
+<SSSE3BW:cpu:pfx:attr:vvvv:reg:mem, +
+    $avx:AVX:66:Vex128|VexW0|SSE2AVX:VexVVVV:RegXMM:Xmmword, +
+    $apx:AVX512BW&AVX512VL:66:EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexVVVV:RegXMM:Xmmword, +
+    $sse:SSSE3:66:::RegXMM:Xmmword, +
+    $mmx:SSSE3::::RegMMX:Qword>
+<SSSE3D:cpu:pfx:attr:reg:mem, +
+    $avx:AVX|AVX512VL:66:Vex128|EVex128|VexW0|Disp8MemShift=4|SSE2AVX:RegXMM:Xmmword, +
+    $sse:SSSE3:66::RegXMM:Xmmword, +
+    $mmx:SSSE3:::RegMMX:Qword>
 
 phaddw<ssse3>, 0x<ssse3:pfx>0f3801, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
 phaddd<ssse3>, 0x<ssse3:pfx>0f3802, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
@@ -1426,18 +1493,34 @@  phaddsw<ssse3>, 0x<ssse3:pfx>0f3803, <ss
 phsubw<ssse3>, 0x<ssse3:pfx>0f3805, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
 phsubd<ssse3>, 0x<ssse3:pfx>0f3806, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
 phsubsw<ssse3>, 0x<ssse3:pfx>0f3807, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-pmaddubsw<ssse3>, 0x<ssse3:pfx>0f3804, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-pmulhrsw<ssse3>, 0x<ssse3:pfx>0f380b, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-pshufb<ssse3>, 0x<ssse3:pfx>0f3800, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
+pmaddubsw<SSSE3BW>, 0x<SSSE3BW:pfx>0f3804, <SSSE3BW:cpu>, Modrm|<SSSE3BW:attr>|<SSSE3BW:vvvv>|NoSuf, { <SSSE3BW:reg>|<SSSE3BW:mem>|Unspecified|BaseIndex, <SSSE3BW:reg> }
+pmulhrsw<SSSE3BW>, 0x<SSSE3BW:pfx>0f380b, <SSSE3BW:cpu>, Modrm|<SSSE3BW:attr>|<SSSE3BW:vvvv>|NoSuf, { <SSSE3BW:reg>|<SSSE3BW:mem>|Unspecified|BaseIndex, <SSSE3BW:reg> }
+pshufb<SSSE3BW>, 0x<SSSE3BW:pfx>0f3800, <SSSE3BW:cpu>, Modrm|<SSSE3BW:attr>|<SSSE3BW:vvvv>|NoSuf, { <SSSE3BW:reg>|<SSSE3BW:mem>|Unspecified|BaseIndex, <SSSE3BW:reg> }
 psign<bw><ssse3>, 0x<ssse3:pfx>0f3808 | <bw:opc>, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
 psignd<ssse3>, 0x<ssse3:pfx>0f380a, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-palignr<ssse3>, 0x<ssse3:pfx>0f3a0f, <ssse3:cpu>, Modrm|<ssse3:attr>|<ssse3:vvvv>|NoSuf, { Imm8, <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-pabs<bw><ssse3>, 0x<ssse3:pfx>0f381c | <bw:opc>, <ssse3:cpu>, Modrm|<ssse3:attr>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
-pabsd<ssse3>, 0x<ssse3:pfx>0f381e, <ssse3:cpu>, Modrm|<ssse3:attr>|NoSuf, { <ssse3:reg>|<ssse3:mem>|Unspecified|BaseIndex, <ssse3:reg> }
+palignr<SSSE3BW>, 0x<SSSE3BW:pfx>0f3a0f, <SSSE3BW:cpu>, Modrm|<SSSE3BW:attr>|<SSSE3BW:vvvv>|NoSuf, { Imm8, <SSSE3BW:reg>|<SSSE3BW:mem>|Unspecified|BaseIndex, <SSSE3BW:reg> }
+pabs<bw><SSSE3BW>, 0x<SSSE3BW:pfx>0f381c | <bw:opc>, <SSSE3BW:cpu>, Modrm|<SSSE3BW:attr>|NoSuf, { <SSSE3BW:reg>|<SSSE3BW:mem>|Unspecified|BaseIndex, <SSSE3BW:reg> }
+pabsd<SSSE3D>, 0x<SSSE3D:pfx>0f381e, <SSSE3D:cpu>, Modrm|<SSSE3D:attr>|NoSuf, { <SSSE3D:reg>|<SSSE3D:mem>|Unspecified|BaseIndex, <SSSE3D:reg> }
 
 // SSE4.1 instructions.
 
 <sse41:cpu:attr:scal:vvvv, $avx:AVX:Vex128|VexW0|SSE2AVX:VexLIG|VexW0|SSE2AVX:VexVVVV, $sse:SSE4_1:::>
+<SSE41BW:cpu:attr:vvvv, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512BW&AVX512VL:EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexVVVV, +
+    $sse:SSE4_1::>
+<SSE41DQ:cpu:attr:vvvv, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512DQ&AVX512VL:EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexVVVV, +
+    $sse:SSE4_1::>
+<SSE41D:cpu:attr:scal:vvvv, +
+    $avx:AVX|AVX512VL:Vex128|EVex128|VexW0|Disp8MemShift=4|SSE2AVX:VexLIG|EVexLIG|VexW0|Disp8MemShift=2|SSE2AVX:VexVVVV, +
+    $sse:SSE4_1:::>
+<SSE41Q:cpu:attr:scal:vvvv, +
+    $avx:AVX:Vex128|VexW0|SSE2AVX:VexLIG|VexW0|SSE2AVX:VexVVVV, +
+    $apx:AVX512VL:EVex128|VexW1|Disp8MemShift=4|SSE2AVX:EVexLIG|VexW1|Disp8MemShift=3|SSE2AVX:VexVVVV, +
+    $sse:SSE4_1:::>
+
 <sd:ppfx:spfx:opc:vexw:elem, s::f3:0:VexW0:Dword, d:66:f2:1:VexW1:Qword>
 
 blendp<sd><sse41>, 0x660f3a0c | <sd:opc>, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
@@ -1446,56 +1529,60 @@  blendvp<sd>, 0x664a | <sd:opc>, AVX, Mod
 blendvp<sd>, 0x660f3814 | <sd:opc>, SSE4_1, Modrm|NoSuf, { Acc|Xmmword, RegXMM|Unspecified|BaseIndex, RegXMM }
 blendvp<sd>, 0x660f3814 | <sd:opc>, SSE4_1, Modrm|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 dpp<sd><sse41>, 0x660f3a40 | <sd:opc>, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-extractps, 0x6617, AVX, Modrm|Vex128|Space0F3A|VexW0|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg32|Unspecified|BaseIndex }
-extractps, 0x6617, AVX&x64, RegMem|Vex128|Space0F3A|VexW1|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg64 }
+extractps, 0x6617, AVX|AVX512F, Modrm|Vex128|EVex128|Space0F3A|VexW0|Disp8MemShift=2|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg32|Unspecified|BaseIndex }
+extractps, 0x6617, x64&(AVX|AVX512F), RegMem|Vex128|EVex128|Space0F3A|VexW1|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg64 }
 extractps, 0x660f3a17, SSE4_1, Modrm|IgnoreSize|NoSuf, { Imm8, RegXMM, Reg32|Unspecified|BaseIndex }
 extractps, 0x660f3a17, SSE4_1&x64, RegMem|NoSuf|NoRex64, { Imm8, RegXMM, Reg64 }
-insertps<sse41>, 0x660f3a21, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8, Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-movntdqa<sse41>, 0x660f382a, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
+insertps<SSE41D>, 0x660f3a21, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|Disp8MemShift|NoSuf, { Imm8, Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+movntdqa<SSE41D>, 0x660f382a, <SSE41D:cpu>, Modrm|<SSE41D:attr>|NoSuf, { Xmmword|Unspecified|BaseIndex, RegXMM }
 mpsadbw<sse41>, 0x660f3a42, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-packusdw<sse41>, 0x660f382b, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+packusdw<SSE41BW>, 0x660f382b, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pblendvb, 0x664c, AVX, Modrm|Vex128|Space0F3A|VexVVVV|VexW0|NoSuf|SSE2AVX, { Acc|Xmmword, RegXMM|Unspecified|BaseIndex, RegXMM }
 pblendvb, 0x664c, AVX, Modrm|Vex128|Space0F3A|VexVVVV|VexW0|NoSuf|Implicit1stXmm0|SSE2AVX, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pblendvb, 0x660f3810, SSE4_1, Modrm|NoSuf, { Acc|Xmmword, RegXMM|Unspecified|BaseIndex, RegXMM }
 pblendvb, 0x660f3810, SSE4_1, Modrm|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 pblendw<sse41>, 0x660f3a0e, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
 pcmpeqq<sse41>, 0x660f3829, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf|Optimize, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pextr<bw><sse41>, 0x660f3a14 | <bw:opc>, <sse41:cpu>, RegMem|<sse41:attr>|NoSuf|IgnoreSize|NoRex64, { Imm8, RegXMM, Reg32|Reg64 }
-pextr<bw><sse41>, 0x660f3a14 | <bw:opc>, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Imm8, RegXMM, <bw:elem>|Unspecified|BaseIndex }
-pextrd<sse41>, 0x660f3a16, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf|IgnoreSize, { Imm8, RegXMM, Reg32|Unspecified|BaseIndex }
+pextr<bw><SSE41BW>, 0x660f3a14 | <bw:opc>, <SSE41BW:cpu>, RegMem|<SSE41BW:attr>|NoSuf|IgnoreSize|NoRex64, { Imm8, RegXMM, Reg32|Reg64 }
+pextr<bw><SSE41BW>, 0x660f3a14 | <bw:opc>, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|Disp8MemShift|NoSuf, { Imm8, RegXMM, <bw:elem>|Unspecified|BaseIndex }
+pextrd<SSE41DQ>, 0x660f3a16, <SSE41DQ:cpu>, Modrm|<SSE41DQ:attr>|Disp8MemShift|NoSuf|IgnoreSize, { Imm8, RegXMM, Reg32|Unspecified|BaseIndex }
 pextrq, 0x6616, AVX&x64, Modrm|Vex|Space0F3A|VexW1|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg64|Unspecified|BaseIndex }
+pextrq, 0x6616, AVX512DQ&AVX512VL&x64, Modrm|EVex128|Space0F3A|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { Imm8, RegXMM, Reg64|Unspecified|BaseIndex }
 pextrq, 0x660f3a16, SSE4_1&x64, Modrm|Size64|NoSuf, { Imm8, RegXMM, Reg64|Unspecified|BaseIndex }
 phminposuw<sse41>, 0x660f3841, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pinsrb<sse41>, 0x660f3a20, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf|IgnoreSize|NoRex64, { Imm8, Reg32|Reg64, RegXMM }
-pinsrb<sse41>, 0x660f3a20, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { Imm8, Byte|Unspecified|BaseIndex, RegXMM }
-pinsrd<sse41>, 0x660f3a22, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf|IgnoreSize, { Imm8, Reg32|Unspecified|BaseIndex, RegXMM }
+pinsrb<SSE41BW>, 0x660f3a20, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf|IgnoreSize|NoRex64, { Imm8, Reg32|Reg64, RegXMM }
+pinsrb<SSE41BW>, 0x660f3a20, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|Disp8MemShift|NoSuf, { Imm8, Byte|Unspecified|BaseIndex, RegXMM }
+pinsrd<SSE41DQ>, 0x660f3a22, <SSE41DQ:cpu>, Modrm|<SSE41DQ:attr>|<SSE41DQ:vvvv>|Disp8MemShift|NoSuf|IgnoreSize, { Imm8, Reg32|Unspecified|BaseIndex, RegXMM }
 pinsrq, 0x6622, AVX&x64, Modrm|Vex|Space0F3A|VexVVVV|VexW1|NoSuf|SSE2AVX, { Imm8, Reg64|Unspecified|BaseIndex, RegXMM }
+pinsrq, 0x6622, AVX512DQ&AVX512VL&AVX&x64, Modrm|EVex128|Space0F3A|VexVVVV|VexW1|Disp8MemShift=3|NoSuf|SSE2AVX, { Imm8, Reg64|Unspecified|BaseIndex, RegXMM }
 pinsrq, 0x660f3a22, SSE4_1&x64, Modrm|Size64|NoSuf, { Imm8, Reg64|Unspecified|BaseIndex, RegXMM }
-pmaxsb<sse41>, 0x660f383c, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pmaxsd<sse41>, 0x660f383d, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pmaxud<sse41>, 0x660f383f, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pmaxuw<sse41>, 0x660f383e, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pminsb<sse41>, 0x660f3838, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pminsd<sse41>, 0x660f3839, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pminud<sse41>, 0x660f383b, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pminuw<sse41>, 0x660f383a, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pmovsxbw<sse41>, 0x660f3820, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovsxbd<sse41>, 0x660f3821, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovsxbq<sse41>, 0x660f3822, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Word|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovsxwd<sse41>, 0x660f3823, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovsxwq<sse41>, 0x660f3824, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovsxdq<sse41>, 0x660f3825, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxbw<sse41>, 0x660f3830, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxbd<sse41>, 0x660f3831, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxbq<sse41>, 0x660f3832, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Word|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxwd<sse41>, 0x660f3833, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxwq<sse41>, 0x660f3834, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmovzxdq<sse41>, 0x660f3835, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
-pmuldq<sse41>, 0x660f3828, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pmulld<sse41>, 0x660f3840, <sse41:cpu>, Modrm|<sse41:attr>|<sse41:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxsb<SSE41BW>, 0x660f383c, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxsd<SSE41D>, 0x660f383d, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxud<SSE41D>, 0x660f383f, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmaxuw<SSE41BW>, 0x660f383e, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminsb<SSE41BW>, 0x660f3838, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminsd<SSE41D>, 0x660f3839, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminud<SSE41D>, 0x660f383b, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pminuw<SSE41BW>, 0x660f383a, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|<SSE41BW:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmovsxbw<SSE41BW>, 0x660f3820, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovsxbd<SSE41D>, 0x660f3821, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovsxbq<SSE41D>, 0x660f3822, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Word|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovsxwd<SSE41D>, 0x660f3823, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovsxwq<SSE41D>, 0x660f3824, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovsxdq<SSE41D>, 0x660f3825, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxbw<SSE41BW>, 0x660f3830, <SSE41BW:cpu>, Modrm|<SSE41BW:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxbd<SSE41D>, 0x660f3831, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxbq<SSE41D>, 0x660f3832, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Word|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxwd<SSE41D>, 0x660f3833, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxwq<SSE41D>, 0x660f3834, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmovzxdq<SSE41D>, 0x660f3835, <SSE41D:cpu>, Modrm|<SSE41D:attr>|Disp8MemShift|NoSuf, { Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+pmuldq<SSE41Q>, 0x660f3828, <SSE41Q:cpu>, Modrm|<SSE41Q:attr>|<SSE41Q:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pmulld<SSE41D>, 0x660f3840, <SSE41D:cpu>, Modrm|<SSE41D:attr>|<SSE41D:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 ptest<sse41>, 0x660f3817, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-roundp<sd><sse41>, 0x660f3a08 | <sd:opc>, <sse41:cpu>, Modrm|<sse41:attr>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
-rounds<sd><sse41>, 0x660f3a0a | <sd:opc>, <sse41:cpu>, Modrm|<sse41:scal>|<sse41:vvvv>|NoSuf, { Imm8, <sd:elem>|Unspecified|BaseIndex|RegXMM, RegXMM }
+roundpd<SSE41Q>, 0x660f3a09, <SSE41Q:cpu>, Modrm|<SSE41Q:attr>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
+roundps<SSE41D>, 0x660f3a08, <SSE41D:cpu>, Modrm|<SSE41D:attr>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
+roundsd<SSE41Q>, 0x660f3a0b, <SSE41Q:cpu>, Modrm|<SSE41Q:scal>|<SSE41Q:vvvv>|NoSuf, { Imm8, Qword|Unspecified|BaseIndex|RegXMM, RegXMM }
+roundss<SSE41D>, 0x660f3a0a, <SSE41D:cpu>, Modrm|<SSE41D:scal>|<SSE41D:vvvv>|NoSuf, { Imm8, Dword|Unspecified|BaseIndex|RegXMM, RegXMM }
 
 // SSE4.2 instructions.
 
@@ -1532,31 +1619,38 @@  xsaveopt64, 0xfae/6, Xsaveopt&x64, Modrm
 // AES instructions.
 
 <aes:cpu:attr:vvvv, $avx:AVX&:Vex128|VexW0|SSE2AVX:VexVVVV, $sse:::>
-
-aesdec<aes>, 0x660f38de, <aes:cpu>AES, Modrm|<aes:attr>|<aes:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-aesdeclast<aes>, 0x660f38df, <aes:cpu>AES, Modrm|<aes:attr>|<aes:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-aesenc<aes>, 0x660f38dc, <aes:cpu>AES, Modrm|<aes:attr>|<aes:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
-aesenclast<aes>, 0x660f38dd, <aes:cpu>AES, Modrm|<aes:attr>|<aes:vvvv>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+<AES:cpu:attr, +
+    $avx:&(AVX|AVX512VL):Vex128|EVex128|VexW0|VexVVVV|Disp8MemShift=4|SSE2AVX, +
+    $sse::>
+
+aesdec<AES>, 0x660f38de, AES<AES:cpu>, Modrm|<AES:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+aesdeclast<AES>, 0x660f38df, AES<AES:cpu>, Modrm|<AES:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+aesenc<AES>, 0x660f38dc, AES<AES:cpu>, Modrm|<AES:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+aesenclast<AES>, 0x660f38dd, AES<AES:cpu>, Modrm|<AES:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 aesimc<aes>, 0x660f38db, <aes:cpu>AES, Modrm|<aes:attr>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 aeskeygenassist<aes>, 0x660f3adf, <aes:cpu>AES, Modrm|<aes:attr>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
 
 // PCLMULQDQ
 
-<pclmul:cpu:attr, $avx:AVX&:Vex128|VexW0|SSE2AVX|VexVVVV, $sse::>
-
-pclmulqdq<pclmul>, 0x660f3a44, <pclmul:cpu>PCLMULQDQ, Modrm|<pclmul:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
-pclmullqlqdq<pclmul>, 0x660f3a44/0x00, <pclmul:cpu>PCLMULQDQ, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pclmulhqlqdq<pclmul>, 0x660f3a44/0x01, <pclmul:cpu>PCLMULQDQ, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pclmullqhqdq<pclmul>, 0x660f3a44/0x10, <pclmul:cpu>PCLMULQDQ, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
-pclmulhqhqdq<pclmul>, 0x660f3a44/0x11, <pclmul:cpu>PCLMULQDQ, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
+<pclmul:cpu:attr, +
+    $avx:&(AVX|AVX512VL):Vex128|EVex128|VexW0|VexVVVV|Disp8MemShift=4|SSE2AVX, +
+    $sse::>
+
+pclmulqdq<pclmul>, 0x660f3a44, PCLMULQDQ<pclmul:cpu>, Modrm|<pclmul:attr>|NoSuf, { Imm8|Imm8S, RegXMM|Unspecified|BaseIndex, RegXMM }
+pclmullqlqdq<pclmul>, 0x660f3a44/0x00, PCLMULQDQ<pclmul:cpu>, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pclmulhqlqdq<pclmul>, 0x660f3a44/0x01, PCLMULQDQ<pclmul:cpu>, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pclmullqhqdq<pclmul>, 0x660f3a44/0x10, PCLMULQDQ<pclmul:cpu>, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
+pclmulhqhqdq<pclmul>, 0x660f3a44/0x11, PCLMULQDQ<pclmul:cpu>, Modrm|<pclmul:attr>|NoSuf|ImmExt, { RegXMM|Unspecified|BaseIndex, RegXMM }
 
 // GFNI
 
-<gfni:cpu:w0:w1, $avx:AVX&:Vex128|VexW0|SSE2AVX|VexVVVV:Vex128|VexW1|SSE2AVX|VexVVVV, $sse:::>
-
-gf2p8affineqb<gfni>, 0x660f3ace, <gfni:cpu>GFNI, Modrm|<gfni:w1>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
-gf2p8affineinvqb<gfni>, 0x660f3acf, <gfni:cpu>GFNI, Modrm|<gfni:w1>|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
-gf2p8mulb<gfni>, 0x660f38cf, <gfni:cpu>GFNI, Modrm|<gfni:w0>|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
+<gfni:cpu:attr, +
+    $avx:&(AVX|AVX512VL):Vex128|EVex128|VexVVVV|Disp8MemShift=4|SSE2AVX, +
+    $sse::>
+
+gf2p8affineqb<gfni>, 0x660f3ace, GFNI<gfni:cpu>, Modrm|<gfni:attr>|VexW1|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
+gf2p8affineinvqb<gfni>, 0x660f3acf, GFNI<gfni:cpu>, Modrm|<gfni:attr>|VexW1|NoSuf, { Imm8, RegXMM|Unspecified|BaseIndex, RegXMM }
+gf2p8mulb<gfni>, 0x660f38cf, GFNI<gfni:cpu>, Modrm|<gfni:attr>|VexW0|NoSuf, { RegXMM|Unspecified|BaseIndex, RegXMM }
 
 // AVX instructions.