From: "H.J. Lu" <hjl.tools@gmail.com>
To: Jan Beulich <jbeulich@suse.com>
Cc: Binutils <binutils@sourceware.org>
Subject: Re: [PATCH 7/7] x86: introduce a state stack for .arch
Date: Thu, 30 Jun 2022 16:15:41 -0700 [thread overview]
Message-ID: <CAMe9rOq2+cL2-XJQPAdHrkZ9mC52yw6ZOG3ta9udCXDw47Ayog@mail.gmail.com> (raw)
In-Reply-To: <53e719d3-482e-e1bc-663c-367fe0de7277@suse.com>
On Thu, Jun 30, 2022 at 5:55 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> When using just slightly non-trivial combinations of .arch, it can be
> quite useful to be able to go back to prior state without needing to
> re-invoke perhaps many earlier directives and without needing to invoke
> perhaps many "negative" ones. Like some other architectures allow
> saving (pushing) and restoring (popping) present/prior state.
>
> For now require the same .code<N> to be in effect for ".arch pop" that
> was in effect for the corresponding ".arch push".
>
> Also change the global "no_cond_jump_promotion" to be bool, to match the
> new struct field.
>
> --- a/gas/config/tc-i386.c
> +++ b/gas/config/tc-i386.c
> @@ -788,7 +788,7 @@ i386_cpu_flags cpu_arch_isa_flags;
>
> /* If set, conditional jumps are not automatically promoted to handle
> larger than a byte offset. */
> -static unsigned int no_cond_jump_promotion = 0;
> +static bool no_cond_jump_promotion = false;
>
> /* Encode SSE instructions with VEX prefix. */
> static unsigned int sse2avx;
> @@ -2663,6 +2663,20 @@ extend_cpu_sub_arch_name (const char *na
> static void
> set_cpu_arch (int dummy ATTRIBUTE_UNUSED)
> {
> + typedef struct arch_stack_entry
> + {
> + const struct arch_stack_entry *prev;
> + const char *name;
> + char *sub_name;
> + i386_cpu_flags flags;
> + i386_cpu_flags isa_flags;
> + enum processor_type isa;
> + enum flag_code flag_code;
> + char stackop_size;
> + bool no_cond_jump_promotion;
> + } arch_stack_entry;
> + static const arch_stack_entry *arch_stack_top;
> +
> SKIP_WHITESPACE ();
>
> if (!is_end_of_line[(unsigned char) *input_line_pointer])
> @@ -2706,6 +2720,67 @@ set_cpu_arch (int dummy ATTRIBUTE_UNUSED
> j = ARRAY_SIZE (cpu_arch) + 1;
> }
> }
> + else if (strcmp (string, "push") == 0)
> + {
> + arch_stack_entry *top = XNEW (arch_stack_entry);
> +
> + top->name = cpu_arch_name;
> + if (cpu_sub_arch_name)
> + top->sub_name = xstrdup (cpu_sub_arch_name);
> + else
> + top->sub_name = NULL;
> + top->flags = cpu_arch_flags;
> + top->isa = cpu_arch_isa;
> + top->isa_flags = cpu_arch_isa_flags;
> + top->flag_code = flag_code;
> + top->stackop_size = stackop_size;
> + top->no_cond_jump_promotion = no_cond_jump_promotion;
> +
> + top->prev = arch_stack_top;
> + arch_stack_top = top;
> +
> + (void) restore_line_pointer (e);
> + demand_empty_rest_of_line ();
> + return;
> + }
> + else if (strcmp (string, "pop") == 0)
> + {
> + const arch_stack_entry *top = arch_stack_top;
> +
> + if (!top)
> + as_bad (_(".arch stack is empty"));
> + else if (top->flag_code != flag_code
> + || top->stackop_size != stackop_size)
> + {
> + static const unsigned int bits[] = {
> + [CODE_16BIT] = 16,
> + [CODE_32BIT] = 32,
> + [CODE_64BIT] = 64,
> + };
> +
> + as_bad (_("this `.arch pop' requires `.code%u%s' to be in effect"),
> + bits[top->flag_code],
> + top->stackop_size == LONG_MNEM_SUFFIX ? "gcc" : "");
> + }
> + else
> + {
> + arch_stack_top = top->prev;
> +
> + cpu_arch_name = top->name;
> + xfree (cpu_sub_arch_name);
Can we just use free?
> + cpu_sub_arch_name = top->sub_name;
> + cpu_arch_flags = top->flags;
> + cpu_arch_isa = top->isa;
> + cpu_arch_isa_flags = top->isa_flags;
> + no_cond_jump_promotion = top->no_cond_jump_promotion;
> +
> + XDELETE (top);
> + }
> +
> + (void) restore_line_pointer (e);
> + demand_empty_rest_of_line ();
> + return;
> + }
>
> for (; j < ARRAY_SIZE (cpu_arch); j++)
> {
> @@ -13685,6 +13760,10 @@ show_arch (FILE *stream, int ext, int ch
> {
> p = output_message (stream, p, message, start, &left,
> STRING_COMMA_LEN ("default"));
> + p = output_message (stream, p, message, start, &left,
> + STRING_COMMA_LEN ("push"));
> + p = output_message (stream, p, message, start, &left,
> + STRING_COMMA_LEN ("pop"));
> }
>
> for (j = 0; j < ARRAY_SIZE (cpu_arch); j++)
> --- a/gas/doc/c-i386.texi
> +++ b/gas/doc/c-i386.texi
> @@ -1504,7 +1504,7 @@ directive enables a warning when gas det
> supported on the CPU specified. The choices for @var{cpu_type} are:
>
> @multitable @columnfractions .20 .20 .20 .20
> -@item @samp{default}
> +@item @samp{default} @tab @samp{push} @tab @samp{pop}
> @item @samp{i8086} @tab @samp{i186} @tab @samp{i286} @tab @samp{i386}
> @item @samp{i486} @tab @samp{i586} @tab @samp{i686} @tab @samp{pentium}
> @item @samp{pentiumpro} @tab @samp{pentiumii} @tab @samp{pentiumiii} @tab @samp{pentium4}
> --- /dev/null
> +++ b/gas/testsuite/gas/i386/arch-stk.l
> @@ -0,0 +1,43 @@
> +.*: Assembler messages:
> +.*:3: Error:.*`cmovl'.*
> +.*:10: Error:.*`cmovg'.*
> +.*:17: Error:.*`cmovz'.*
> +.*:21: Error:.*`\.arch pop'.*`\.code32'.*
> +.*:28: Error:.*`\.arch pop'.*`\.code16gcc'.*
> +.*:32: Error:.*\.arch.*empty.*
> +GAS LISTING .*
> +
> +
> +[ ]*[0-9]*[ ]+\.text
> +[ ]*[0-9]*[ ]+start:
> +[ ]*[0-9]*[ ]+cmovl %eax, %ecx
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch push
> +[ ]*[0-9]*[ ]+\.arch default
> +[ ]*[0-9]*[ ]+\?\?\?\? 0F4DC8[ ]+cmovnl %eax, %ecx
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]+cmovg %eax, %ecx
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch push
> +[ ]*[0-9]*[ ]+\.arch \.cmov
> +[ ]*[0-9]*[ ]+\?\?\?\? 0F4EC8[ ]+cmovng %eax, %ecx
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]+cmovz %eax, %ecx
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch push
> +[ ]*[0-9]*[ ]+\.code16
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]+\.code32
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.code16gcc
> +[ ]*[0-9]*[ ]+\.arch push
> +[ ]*[0-9]*[ ]+\.code32
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]+\.code16gcc
> +[ ]*[0-9]*[ ]+\.arch pop
> +[ ]*[0-9]*[ ]*
> +[ ]*[0-9]*[ ]+\.arch pop
> +#pass
> --- /dev/null
> +++ b/gas/testsuite/gas/i386/arch-stk.s
> @@ -0,0 +1,34 @@
> + .text
> +start:
> + cmovl %eax, %ecx
> +
> + .arch push
> + .arch default
> + cmovnl %eax, %ecx
> +
> + .arch pop
> + cmovg %eax, %ecx
> +
> + .arch push
> + .arch .cmov
> + cmovng %eax, %ecx
> +
> + .arch pop
> + cmovz %eax, %ecx
> +
> + .arch push
> + .code16
> + .arch pop
> + .code32
> + .arch pop
> +
> + .code16gcc
> + .arch push
> + .code32
> + .arch pop
> + .code16gcc
> + .arch pop
> +
> + .arch pop
> +
> + .end
> --- a/gas/testsuite/gas/i386/i386.exp
> +++ b/gas/testsuite/gas/i386/i386.exp
> @@ -206,6 +206,7 @@ if [gas_32_check] then {
> run_dump_test "arch-13"
> run_dump_test "arch-14"
> run_list_test "arch-dflt" "-march=generic32 -al"
> + run_list_test "arch-stk" "-march=generic32 -al"
> run_dump_test "8087"
> run_dump_test "287"
> run_dump_test "387"
>
--
H.J.
prev parent reply other threads:[~2022-06-30 23:16 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-30 12:52 [PATCH] gas/x86: .arch / -march= enhancements Jan Beulich
2022-06-30 12:53 ` [PATCH 1/7] x86: don't leak sub-architecture accumulated strings Jan Beulich
2022-06-30 22:54 ` H.J. Lu
2022-07-01 10:18 ` Jan Beulich
2022-07-01 19:44 ` H.J. Lu
2022-06-30 12:53 ` [PATCH 2/7] x86: de-duplicate sub-architecture strings accumulation Jan Beulich
2022-06-30 22:55 ` H.J. Lu
2022-06-30 12:53 ` [PATCH 3/7] x86: permit "default" with .arch Jan Beulich
2022-06-30 22:58 ` H.J. Lu
2022-06-30 12:54 ` [PATCH 4/7] x86: macro-ize cpu_arch[] entries Jan Beulich
2022-06-30 23:00 ` H.J. Lu
2022-06-30 12:54 ` [PATCH 5/7] x86: introduce fake processor type to mark sub-arch entries in cpu_arch[] Jan Beulich
2022-06-30 23:03 ` H.J. Lu
2022-06-30 12:55 ` [PATCH 6/7] x86: generalize disabling of sub-architectures Jan Beulich
2022-06-30 23:11 ` H.J. Lu
2022-06-30 12:55 ` [PATCH 7/7] x86: introduce a state stack for .arch Jan Beulich
2022-06-30 23:15 ` H.J. Lu [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAMe9rOq2+cL2-XJQPAdHrkZ9mC52yw6ZOG3ta9udCXDw47Ayog@mail.gmail.com \
--to=hjl.tools@gmail.com \
--cc=binutils@sourceware.org \
--cc=jbeulich@suse.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).