[0/2] x86: further template encoding adjustments

Message ID 7e63e19e-ac01-5ce6-d403-c41c8aaa7ca1@suse.com
Headers show
Series
  • x86: further template encoding adjustments
Related show

Message

Jan Beulich July 4, 2019, 8:49 a.m.
Especially for operands (of which there may be up to 5 right now,
and the number is only going to grow) it is quite wasteful to have
unnecessarily many attribute bits. The long term goal is to get this
count down to no more than 32. The two patches here are another two
small steps towards this.

1: fold SReg{2,3}
2: make RegMem an opcode modifier

Jan

Comments

Jan Beulich July 15, 2019, 6:37 a.m. | #1
On 04.07.2019 10:49, Jan Beulich wrote:
> Especially for operands (of which there may be up to 5 right now,

> and the number is only going to grow) it is quite wasteful to have

> unnecessarily many attribute bits. The long term goal is to get this

> count down to no more than 32. The two patches here are another two

> small steps towards this.

> 

> 1: fold SReg{2,3}

> 2: make RegMem an opcode modifier


Gentle ping, also for the two other ones sent a day later.

Thanks, Jan
H.J. Lu July 15, 2019, 8 p.m. | #2
On Thu, Jul 4, 2019 at 1:55 AM Jan Beulich <JBeulich@suse.com> wrote:
>

> Especially for operands (of which there may be up to 5 right now,

> and the number is only going to grow) it is quite wasteful to have

> unnecessarily many attribute bits. The long term goal is to get this

> count down to no more than 32. The two patches here are another two

> small steps towards this.

>

> 1: fold SReg{2,3}

> 2: make RegMem an opcode modifier

>


OK.

Thanks.

-- 
H.J.
Jan Beulich July 16, 2019, 7:35 a.m. | #3
On 15.07.2019 22:00,  H.J. Lu  wrote:
> On Thu, Jul 4, 2019 at 1:55 AM Jan Beulich <JBeulich@suse.com> wrote:

>>

>> Especially for operands (of which there may be up to 5 right now,

>> and the number is only going to grow) it is quite wasteful to have

>> unnecessarily many attribute bits. The long term goal is to get this

>> count down to no more than 32. The two patches here are another two

>> small steps towards this.

>>

>> 1: fold SReg{2,3}

>> 2: make RegMem an opcode modifier

>>

> 

> OK.


Thanks, committed. What about the other two (sent on the 5th)?

Jan