qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Aurelien Jarno <aurelien@aurel32.net>
To: Richard Henderson <rth@twiddle.net>
Cc: mark.cave-ayland@ilande.co.uk, qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH v2 11/14] tcg: Implement indirect memory registers
Date: Thu, 31 Dec 2015 12:54:30 +0100	[thread overview]
Message-ID: <20151231115430.GF360@aurel32.net> (raw)
In-Reply-To: <1450382429-6019-1-git-send-email-rth@twiddle.net>

On 2015-12-17 12:00, Richard Henderson wrote:
> That is, global_mem registers whose base is another global_mem
> register, rather than a fixed register.
> 
> Signed-off-by: Richard Henderson <rth@twiddle.net>
> ---
>  tcg/tcg.c | 95 ++++++++++++++++++++++++++++++++++++++++++++-------------------
>  tcg/tcg.h |  2 ++
>  2 files changed, 68 insertions(+), 29 deletions(-)
> 
> diff --git a/tcg/tcg.c b/tcg/tcg.c
> index c51e0ec..7150a3f 100644
> --- a/tcg/tcg.c
> +++ b/tcg/tcg.c
> @@ -509,17 +509,23 @@ int tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
>      TCGContext *s = &tcg_ctx;
>      TCGTemp *base_ts = &s->temps[GET_TCGV_PTR(base)];
>      TCGTemp *ts = tcg_global_alloc(s);
> -    int bigendian = 0;
> +    int indirect_reg = 0, bigendian = 0;
>  #ifdef HOST_WORDS_BIGENDIAN
>      bigendian = 1;
>  #endif
>  
> +    if (!base_ts->fixed_reg) {
> +        indirect_reg = 1;
> +        base_ts->indirect_base = 1;
> +    }
> +
>      if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
>          TCGTemp *ts2 = tcg_global_alloc(s);
>          char buf[64];
>  
>          ts->base_type = TCG_TYPE_I64;
>          ts->type = TCG_TYPE_I32;
> +        ts->indirect_reg = indirect_reg;

Do we really need to add this new bit, while we can simply test
ts->mem_base->fixed_reg? This means one more derefence, but anyway it
has to be done later when calling temp_load?

>          ts->mem_allocated = 1;
>          ts->mem_base = base_ts;
>          ts->mem_offset = offset + bigendian * 4;

> @@ -1781,13 +1803,16 @@ static inline void temp_save(TCGContext *s, TCGTemp *ts,
>                               TCGRegSet allocated_regs)
>  {
>  #ifdef USE_LIVENESS_ANALYSIS
> -    /* The liveness analysis already ensures that globals are back
> -       in memory. Keep an assert for safety. */
> -    tcg_debug_assert(ts->val_type == TEMP_VAL_MEM || ts->fixed_reg);
> -#else
> +    /* ??? Liveness does not yet incorporate indirect bases.  */
> +    if (!ts->indirect_base) {
> +        /* The liveness analysis already ensures that globals are back
> +           in memory. Keep an assert for safety. */
> +        tcg_debug_assert(ts->val_type == TEMP_VAL_MEM || ts->fixed_reg);
> +        return;
> +    }

This basically disables the assert. What does it mean in practice? Can
it generates bad code?

> +#endif
>      temp_sync(s, ts, allocated_regs);
>      temp_dead(s, ts);
> -#endif
>  }
>  
>  /* save globals to their canonical location and assume they can be
> @@ -1812,12 +1837,15 @@ static void sync_globals(TCGContext *s, TCGRegSet allocated_regs)
>      for (i = 0; i < s->nb_globals; i++) {
>          TCGTemp *ts = &s->temps[i];
>  #ifdef USE_LIVENESS_ANALYSIS
> -        tcg_debug_assert(ts->val_type != TEMP_VAL_REG
> -                         || ts->fixed_reg
> -                         || ts->mem_coherent);
> -#else
> -        temp_sync(s, ts, allocated_regs);
> +        /* ??? Liveness does not yet incorporate indirect bases.  */
> +        if (!ts->indirect_base) {
> +            tcg_debug_assert(ts->val_type != TEMP_VAL_REG
> +                             || ts->fixed_reg
> +                             || ts->mem_coherent);
> +            continue;
> +        }

Same here.

>  #endif
> +        temp_sync(s, ts, allocated_regs);
>      }
>  }
>  
> @@ -1833,12 +1861,15 @@ static void tcg_reg_alloc_bb_end(TCGContext *s, TCGRegSet allocated_regs)
>              temp_save(s, ts, allocated_regs);
>          } else {
>  #ifdef USE_LIVENESS_ANALYSIS
> -            /* The liveness analysis already ensures that temps are dead.
> -               Keep an assert for safety. */
> -            assert(ts->val_type == TEMP_VAL_DEAD);
> -#else
> -            temp_dead(s, ts);
> +            /* ??? Liveness does not yet incorporate indirect bases.  */
> +            if (!ts->indirect_base) {
> +                /* The liveness analysis already ensures that temps are dead.
> +                   Keep an assert for safety. */
> +                assert(ts->val_type == TEMP_VAL_DEAD);
> +                continue;
> +            }

And here.

>  #endif
> +            temp_dead(s, ts);
>          }
>      }
>  

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                 http://www.aurel32.net

  reply	other threads:[~2015-12-31 11:59 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-12-17 19:58 [Qemu-devel] [PATCH v2 00/14] tcg: Improve sparc register windows Richard Henderson
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 02/14] tcg: Change ts->mem_reg to ts->mem_base Richard Henderson
2015-12-31 10:52   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 03/14] tcg: Tidy temporary allocation Richard Henderson
2015-12-31 11:13   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 04/14] tcg: More use of TCGReg where appropriate Richard Henderson
2015-12-31 11:14   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 05/14] tcg: Remove tcg_get_arg_str_i32/64 Richard Henderson
2015-12-31 11:15   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 06/14] tcg: Change reg_to_temp to TCGTemp pointer Richard Henderson
2015-12-31 11:19   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 07/14] tcg: Change temp_dead argument to TCGTemp Richard Henderson
2015-12-31 11:24   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 08/14] tcg: Change temp_sync " Richard Henderson
2015-12-31 11:24   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 09/14] tcg: Change temp_save " Richard Henderson
2015-12-31 11:24   ` Aurelien Jarno
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 10/14] tcg: Introduce temp_load Richard Henderson
2015-12-31 11:33   ` Aurelien Jarno
2016-01-15 22:08     ` Richard Henderson
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 11/14] tcg: Implement indirect memory registers Richard Henderson
2015-12-31 11:54   ` Aurelien Jarno [this message]
2016-01-15 22:31     ` Richard Henderson
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 12/14] tcg: Allocate indirect_base temporaries in a different order Richard Henderson
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 13/14] target-sparc: Tidy global register initialization Richard Henderson
2015-12-17 20:00 ` [Qemu-devel] [PATCH v2 14/14] target-sparc: Use global registers for the register window Richard Henderson

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=20151231115430.GF360@aurel32.net \
    --to=aurelien@aurel32.net \
    --cc=mark.cave-ayland@ilande.co.uk \
    --cc=qemu-devel@nongnu.org \
    --cc=rth@twiddle.net \
    /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 NNTP newsgroup(s).