* [PATCH] x86: minor improvements to optimize_imm() (part III)
@ 2022-10-28 12:35 Jan Beulich
2022-10-28 16:03 ` H.J. Lu
0 siblings, 1 reply; 4+ messages in thread
From: Jan Beulich @ 2022-10-28 12:35 UTC (permalink / raw)
To: Binutils; +Cc: H.J. Lu
Earlier tidying still missed an opportunity: There's no need for the
"anyimm" static variable. Instead of using it in the loop to mask
"allowed" (which is necessary to satisfy operand_type_or()'s assertions)
simply use "mask", requiring it to be calculated first. That way the
post-loop masking by "mask" ahead of the operand_type_all_zero() can be
dropped.
--- a/gas/config/tc-i386.c
+++ b/gas/config/tc-i386.c
@@ -1906,7 +1906,6 @@ operand_type_xor (i386_operand_type x, i
static const i386_operand_type disp16_32 = OPERAND_TYPE_DISP16_32;
static const i386_operand_type anydisp = OPERAND_TYPE_ANYDISP;
-static const i386_operand_type anyimm = OPERAND_TYPE_ANYIMM;
static const i386_operand_type regxmm = OPERAND_TYPE_REGXMM;
static const i386_operand_type imm8 = OPERAND_TYPE_IMM8;
static const i386_operand_type imm8s = OPERAND_TYPE_IMM8S;
@@ -5812,13 +5811,6 @@ optimize_imm (void)
const insn_template *t = current_templates->start;
operand_type_set (&mask, 0);
- allowed = t->operand_types[op];
-
- while (++t < current_templates->end)
- {
- allowed = operand_type_and (allowed, anyimm);
- allowed = operand_type_or (allowed, t->operand_types[op]);
- }
switch (guess_suffix)
{
case QWORD_MNEM_SUFFIX:
@@ -5837,7 +5829,14 @@ optimize_imm (void)
default:
break;
}
- allowed = operand_type_and (mask, allowed);
+
+ allowed = operand_type_and (t->operand_types[op], mask);
+ while (++t < current_templates->end)
+ {
+ allowed = operand_type_or (allowed, t->operand_types[op]);
+ allowed = operand_type_and (allowed, mask);
+ }
+
if (!operand_type_all_zero (&allowed))
i.types[op] = operand_type_and (i.types[op], mask);
}
--- a/opcodes/i386-gen.c
+++ b/opcodes/i386-gen.c
@@ -537,8 +537,6 @@ static initializer operand_type_init[] =
"Imm32|Imm32S|Imm64|Disp32" },
{ "OPERAND_TYPE_IMM32_32S_64_DISP32_64",
"Imm32|Imm32S|Imm64|Disp32|Disp64" },
- { "OPERAND_TYPE_ANYIMM",
- "Imm1|Imm8|Imm8S|Imm16|Imm32|Imm32S|Imm64" },
};
typedef struct bitfield
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] x86: minor improvements to optimize_imm() (part III)
2022-10-28 12:35 [PATCH] x86: minor improvements to optimize_imm() (part III) Jan Beulich
@ 2022-10-28 16:03 ` H.J. Lu
2022-10-31 10:11 ` Jan Beulich
0 siblings, 1 reply; 4+ messages in thread
From: H.J. Lu @ 2022-10-28 16:03 UTC (permalink / raw)
To: Jan Beulich; +Cc: Binutils
On Fri, Oct 28, 2022 at 5:35 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> Earlier tidying still missed an opportunity: There's no need for the
> "anyimm" static variable. Instead of using it in the loop to mask
> "allowed" (which is necessary to satisfy operand_type_or()'s assertions)
> simply use "mask", requiring it to be calculated first. That way the
> post-loop masking by "mask" ahead of the operand_type_all_zero() can be
> dropped.
>
> --- a/gas/config/tc-i386.c
> +++ b/gas/config/tc-i386.c
> @@ -1906,7 +1906,6 @@ operand_type_xor (i386_operand_type x, i
>
> static const i386_operand_type disp16_32 = OPERAND_TYPE_DISP16_32;
> static const i386_operand_type anydisp = OPERAND_TYPE_ANYDISP;
> -static const i386_operand_type anyimm = OPERAND_TYPE_ANYIMM;
> static const i386_operand_type regxmm = OPERAND_TYPE_REGXMM;
> static const i386_operand_type imm8 = OPERAND_TYPE_IMM8;
> static const i386_operand_type imm8s = OPERAND_TYPE_IMM8S;
> @@ -5812,13 +5811,6 @@ optimize_imm (void)
> const insn_template *t = current_templates->start;
>
> operand_type_set (&mask, 0);
> - allowed = t->operand_types[op];
> -
> - while (++t < current_templates->end)
> - {
> - allowed = operand_type_and (allowed, anyimm);
> - allowed = operand_type_or (allowed, t->operand_types[op]);
> - }
> switch (guess_suffix)
> {
> case QWORD_MNEM_SUFFIX:
> @@ -5837,7 +5829,14 @@ optimize_imm (void)
> default:
> break;
> }
> - allowed = operand_type_and (mask, allowed);
> +
> + allowed = operand_type_and (t->operand_types[op], mask);
> + while (++t < current_templates->end)
> + {
> + allowed = operand_type_or (allowed, t->operand_types[op]);
> + allowed = operand_type_and (allowed, mask);
mask isn't changed in the loop. Can the AND operation be moved after
the loop?
> + }
> +
> if (!operand_type_all_zero (&allowed))
> i.types[op] = operand_type_and (i.types[op], mask);
> }
> --- a/opcodes/i386-gen.c
> +++ b/opcodes/i386-gen.c
> @@ -537,8 +537,6 @@ static initializer operand_type_init[] =
> "Imm32|Imm32S|Imm64|Disp32" },
> { "OPERAND_TYPE_IMM32_32S_64_DISP32_64",
> "Imm32|Imm32S|Imm64|Disp32|Disp64" },
> - { "OPERAND_TYPE_ANYIMM",
> - "Imm1|Imm8|Imm8S|Imm16|Imm32|Imm32S|Imm64" },
> };
>
> typedef struct bitfield
--
H.J.
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] x86: minor improvements to optimize_imm() (part III)
2022-10-28 16:03 ` H.J. Lu
@ 2022-10-31 10:11 ` Jan Beulich
2022-10-31 16:24 ` H.J. Lu
0 siblings, 1 reply; 4+ messages in thread
From: Jan Beulich @ 2022-10-31 10:11 UTC (permalink / raw)
To: H.J. Lu; +Cc: Binutils
On 28.10.2022 18:03, H.J. Lu wrote:
> On Fri, Oct 28, 2022 at 5:35 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> Earlier tidying still missed an opportunity: There's no need for the
>> "anyimm" static variable. Instead of using it in the loop to mask
>> "allowed" (which is necessary to satisfy operand_type_or()'s assertions)
>> simply use "mask", requiring it to be calculated first. That way the
>> post-loop masking by "mask" ahead of the operand_type_all_zero() can be
>> dropped.
>>
>> --- a/gas/config/tc-i386.c
>> +++ b/gas/config/tc-i386.c
>> @@ -1906,7 +1906,6 @@ operand_type_xor (i386_operand_type x, i
>>
>> static const i386_operand_type disp16_32 = OPERAND_TYPE_DISP16_32;
>> static const i386_operand_type anydisp = OPERAND_TYPE_ANYDISP;
>> -static const i386_operand_type anyimm = OPERAND_TYPE_ANYIMM;
>> static const i386_operand_type regxmm = OPERAND_TYPE_REGXMM;
>> static const i386_operand_type imm8 = OPERAND_TYPE_IMM8;
>> static const i386_operand_type imm8s = OPERAND_TYPE_IMM8S;
>> @@ -5812,13 +5811,6 @@ optimize_imm (void)
>> const insn_template *t = current_templates->start;
>>
>> operand_type_set (&mask, 0);
>> - allowed = t->operand_types[op];
>> -
>> - while (++t < current_templates->end)
>> - {
>> - allowed = operand_type_and (allowed, anyimm);
>> - allowed = operand_type_or (allowed, t->operand_types[op]);
>> - }
>> switch (guess_suffix)
>> {
>> case QWORD_MNEM_SUFFIX:
>> @@ -5837,7 +5829,14 @@ optimize_imm (void)
>> default:
>> break;
>> }
>> - allowed = operand_type_and (mask, allowed);
>> +
>> + allowed = operand_type_and (t->operand_types[op], mask);
>> + while (++t < current_templates->end)
>> + {
>> + allowed = operand_type_or (allowed, t->operand_types[op]);
>> + allowed = operand_type_and (allowed, mask);
>
> mask isn't changed in the loop. Can the AND operation be moved after
> the loop?
Initially I thought it could, yes, but it cannot: As said in the
description the AND-ing "is necessary to satisfy operand_type_or()'s
assertions".
Jan
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] x86: minor improvements to optimize_imm() (part III)
2022-10-31 10:11 ` Jan Beulich
@ 2022-10-31 16:24 ` H.J. Lu
0 siblings, 0 replies; 4+ messages in thread
From: H.J. Lu @ 2022-10-31 16:24 UTC (permalink / raw)
To: Jan Beulich; +Cc: Binutils
On Mon, Oct 31, 2022 at 3:11 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 28.10.2022 18:03, H.J. Lu wrote:
> > On Fri, Oct 28, 2022 at 5:35 AM Jan Beulich <jbeulich@suse.com> wrote:
> >>
> >> Earlier tidying still missed an opportunity: There's no need for the
> >> "anyimm" static variable. Instead of using it in the loop to mask
> >> "allowed" (which is necessary to satisfy operand_type_or()'s assertions)
> >> simply use "mask", requiring it to be calculated first. That way the
> >> post-loop masking by "mask" ahead of the operand_type_all_zero() can be
> >> dropped.
> >>
> >> --- a/gas/config/tc-i386.c
> >> +++ b/gas/config/tc-i386.c
> >> @@ -1906,7 +1906,6 @@ operand_type_xor (i386_operand_type x, i
> >>
> >> static const i386_operand_type disp16_32 = OPERAND_TYPE_DISP16_32;
> >> static const i386_operand_type anydisp = OPERAND_TYPE_ANYDISP;
> >> -static const i386_operand_type anyimm = OPERAND_TYPE_ANYIMM;
> >> static const i386_operand_type regxmm = OPERAND_TYPE_REGXMM;
> >> static const i386_operand_type imm8 = OPERAND_TYPE_IMM8;
> >> static const i386_operand_type imm8s = OPERAND_TYPE_IMM8S;
> >> @@ -5812,13 +5811,6 @@ optimize_imm (void)
> >> const insn_template *t = current_templates->start;
> >>
> >> operand_type_set (&mask, 0);
> >> - allowed = t->operand_types[op];
> >> -
> >> - while (++t < current_templates->end)
> >> - {
> >> - allowed = operand_type_and (allowed, anyimm);
> >> - allowed = operand_type_or (allowed, t->operand_types[op]);
> >> - }
> >> switch (guess_suffix)
> >> {
> >> case QWORD_MNEM_SUFFIX:
> >> @@ -5837,7 +5829,14 @@ optimize_imm (void)
> >> default:
> >> break;
> >> }
> >> - allowed = operand_type_and (mask, allowed);
> >> +
> >> + allowed = operand_type_and (t->operand_types[op], mask);
> >> + while (++t < current_templates->end)
> >> + {
> >> + allowed = operand_type_or (allowed, t->operand_types[op]);
> >> + allowed = operand_type_and (allowed, mask);
> >
> > mask isn't changed in the loop. Can the AND operation be moved after
> > the loop?
>
> Initially I thought it could, yes, but it cannot: As said in the
> description the AND-ing "is necessary to satisfy operand_type_or()'s
> assertions".
>
> Jan
OK.
Thanks.
--
H.J.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2022-10-31 16:25 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-28 12:35 [PATCH] x86: minor improvements to optimize_imm() (part III) Jan Beulich
2022-10-28 16:03 ` H.J. Lu
2022-10-31 10:11 ` Jan Beulich
2022-10-31 16:24 ` H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).