qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Helge Deller <deller@kernel.org>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: qemu-devel@nongnu.org, deller@gmx.de
Subject: Re: [PATCH v2 04/45] target/hppa: Pass displacement to do_dbranch
Date: Tue, 14 May 2024 16:07:50 +0200	[thread overview]
Message-ID: <ZkNwNqGwd9QK9hy6@carbonx1> (raw)
In-Reply-To: <20240513074717.130949-5-richard.henderson@linaro.org>

* Richard Henderson <richard.henderson@linaro.org>:
> Pass a displacement instead of an absolute value.
> 
> In trans_be, remove the user-only do_dbranch case.  The branch we are
> attempting to optimize is to the zero page, which is perforce on a
> different page than the code currently executing, which means that
> we will *not* use a goto_tb.  Use a plain indirect branch instead,
> which is what we got out of the attempted direct branch anyway.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Reviewed-by: Helge Deller <deller@gmx.de>

> ---
>  target/hppa/translate.c | 33 +++++++++------------------------
>  1 file changed, 9 insertions(+), 24 deletions(-)
> 
> diff --git a/target/hppa/translate.c b/target/hppa/translate.c
> index 398803981c..4c42b518c5 100644
> --- a/target/hppa/translate.c
> +++ b/target/hppa/translate.c
> @@ -1766,9 +1766,11 @@ static bool do_fop_dedd(DisasContext *ctx, unsigned rt,
>  
>  /* Emit an unconditional branch to a direct target, which may or may not
>     have already had nullification handled.  */
> -static bool do_dbranch(DisasContext *ctx, uint64_t dest,
> +static bool do_dbranch(DisasContext *ctx, int64_t disp,
>                         unsigned link, bool is_n)
>  {
> +    uint64_t dest = iaoq_dest(ctx, disp);
> +
>      if (ctx->null_cond.c == TCG_COND_NEVER && ctx->null_lab == NULL) {
>          if (link != 0) {
>              copy_iaoq_entry(ctx, cpu_gr[link], ctx->iaoq_n, ctx->iaoq_n_var);
> @@ -1815,10 +1817,7 @@ static bool do_cbranch(DisasContext *ctx, int64_t disp, bool is_n,
>  
>      /* Handle TRUE and NEVER as direct branches.  */
>      if (c == TCG_COND_ALWAYS) {
> -        return do_dbranch(ctx, dest, 0, is_n && disp >= 0);
> -    }
> -    if (c == TCG_COND_NEVER) {
> -        return do_dbranch(ctx, ctx->iaoq_n, 0, is_n && disp < 0);
> +        return do_dbranch(ctx, disp, 0, is_n && disp >= 0);
>      }
>  
>      taken = gen_new_label();
> @@ -3914,22 +3913,6 @@ static bool trans_be(DisasContext *ctx, arg_be *a)
>  {
>      TCGv_i64 tmp;
>  
> -#ifdef CONFIG_USER_ONLY
> -    /* ??? It seems like there should be a good way of using
> -       "be disp(sr2, r0)", the canonical gateway entry mechanism
> -       to our advantage.  But that appears to be inconvenient to
> -       manage along side branch delay slots.  Therefore we handle
> -       entry into the gateway page via absolute address.  */
> -    /* Since we don't implement spaces, just branch.  Do notice the special
> -       case of "be disp(*,r0)" using a direct branch to disp, so that we can
> -       goto_tb to the TB containing the syscall.  */
> -    if (a->b == 0) {
> -        return do_dbranch(ctx, a->disp, a->l, a->n);
> -    }
> -#else
> -    nullify_over(ctx);
> -#endif
> -
>      tmp = tcg_temp_new_i64();
>      tcg_gen_addi_i64(tmp, load_gpr(ctx, a->b), a->disp);
>      tmp = do_ibranch_priv(ctx, tmp);
> @@ -3939,6 +3922,8 @@ static bool trans_be(DisasContext *ctx, arg_be *a)
>  #else
>      TCGv_i64 new_spc = tcg_temp_new_i64();
>  
> +    nullify_over(ctx);
> +
>      load_spr(ctx, new_spc, a->sp);
>      if (a->l) {
>          copy_iaoq_entry(ctx, cpu_gr[31], ctx->iaoq_n, ctx->iaoq_n_var);
> @@ -3968,7 +3953,7 @@ static bool trans_be(DisasContext *ctx, arg_be *a)
>  
>  static bool trans_bl(DisasContext *ctx, arg_bl *a)
>  {
> -    return do_dbranch(ctx, iaoq_dest(ctx, a->disp), a->l, a->n);
> +    return do_dbranch(ctx, a->disp, a->l, a->n);
>  }
>  
>  static bool trans_b_gate(DisasContext *ctx, arg_b_gate *a)
> @@ -4022,7 +4007,7 @@ static bool trans_b_gate(DisasContext *ctx, arg_b_gate *a)
>          save_gpr(ctx, a->l, tmp);
>      }
>  
> -    return do_dbranch(ctx, dest, 0, a->n);
> +    return do_dbranch(ctx, dest - iaoq_dest(ctx, 0), 0, a->n);
>  }
>  
>  static bool trans_blr(DisasContext *ctx, arg_blr *a)
> @@ -4035,7 +4020,7 @@ static bool trans_blr(DisasContext *ctx, arg_blr *a)
>          return do_ibranch(ctx, tmp, a->l, a->n);
>      } else {
>          /* BLR R0,RX is a good way to load PC+8 into RX.  */
> -        return do_dbranch(ctx, ctx->iaoq_f + 8, a->l, a->n);
> +        return do_dbranch(ctx, 0, a->l, a->n);
>      }
>  }
>  
> -- 
> 2.34.1
> 


  reply	other threads:[~2024-05-14 14:08 UTC|newest]

Thread overview: 100+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-05-13  7:46 [PATCH v2 00/45] target/hppa: Misc improvements Richard Henderson
2024-05-13  7:46 ` [PATCH v2 01/45] target/hppa: Move cpu_get_tb_cpu_state out of line Richard Henderson
2024-05-14 13:49   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 02/45] target/hppa: Use hppa_form_gva_psw in hppa_cpu_get_pc Richard Henderson
2024-05-14 14:05   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 03/45] target/hppa: Move constant destination check into use_goto_tb Richard Henderson
2024-05-14 14:06   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 04/45] target/hppa: Pass displacement to do_dbranch Richard Henderson
2024-05-14 14:07   ` Helge Deller [this message]
2024-05-13  7:46 ` [PATCH v2 05/45] target/hppa: Allow prior nullification in do_ibranch Richard Henderson
2024-05-14 14:09   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 06/45] target/hppa: Use CF_BP_PAGE instead of cpu_breakpoint_test Richard Henderson
2024-05-14 14:12   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 07/45] target/hppa: Add install_iaq_entries Richard Henderson
2024-05-14 14:23   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 08/45] target/hppa: Add install_link Richard Henderson
2024-05-14 14:37   ` Helge Deller
2024-05-14 15:43     ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 09/45] target/hppa: Delay computation of IAQ_Next Richard Henderson
2024-05-14 14:39   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 10/45] target/hppa: Skip nullified insns in unconditional dbranch path Richard Henderson
2024-05-14 14:40   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 11/45] target/hppa: Simplify TB end Richard Henderson
2024-05-14 14:43   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 12/45] target/hppa: Add IASQ entries to DisasContext Richard Henderson
2024-05-14 15:23   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 13/45] target/hppa: Add space arguments to install_iaq_entries Richard Henderson
2024-05-14 15:27   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 14/45] target/hppa: Add space argument to do_ibranch Richard Henderson
2024-05-14 15:35   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 15/45] target/hppa: Use umax in do_ibranch_priv Richard Henderson
2024-05-13 11:18   ` Philippe Mathieu-Daudé
2024-05-13 13:23     ` Richard Henderson
2024-05-13 14:15       ` Philippe Mathieu-Daudé
2024-05-13  7:46 ` [PATCH v2 16/45] target/hppa: Always make a copy " Richard Henderson
2024-05-13 10:33   ` Philippe Mathieu-Daudé
2024-05-13  7:46 ` [PATCH v2 17/45] target/hppa: Introduce and use DisasIAQE for branch management Richard Henderson
2024-05-14 15:48   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 18/45] target/hppa: Use displacements in DisasIAQE Richard Henderson
2024-05-14 15:54   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 19/45] target/hppa: Rename cond_make_* helpers Richard Henderson
2024-05-13  9:27   ` Philippe Mathieu-Daudé
2024-05-14 16:17   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 20/45] target/hppa: Use TCG_COND_TST* in do_cond Richard Henderson
2024-05-14 16:23   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 21/45] target/hppa: Use TCG_COND_TST* in do_log_cond Richard Henderson
2024-05-14 16:27   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 22/45] target/hppa: Use TCG_COND_TST* in do_unit_zero_cond Richard Henderson
2024-05-14 16:28   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 23/45] target/hppa: Use TCG_COND_TST* in do_unit_addsub Richard Henderson
2024-05-13  9:37   ` Philippe Mathieu-Daudé
2024-05-14 16:29   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 24/45] target/hppa: Use TCG_COND_TST* in trans_bb_imm Richard Henderson
2024-05-13 10:18   ` Philippe Mathieu-Daudé
2024-05-14 16:36   ` Helge Deller
2024-05-15  7:28     ` Richard Henderson
2024-05-13  7:46 ` [PATCH v2 25/45] target/hppa: Use registerfields.h for FPSR Richard Henderson
2024-05-14 18:03   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 26/45] target/hppa: Use TCG_COND_TST* in trans_ftest Richard Henderson
2024-05-13 10:27   ` Philippe Mathieu-Daudé
2024-05-14 18:05   ` Helge Deller
2024-05-13  7:46 ` [PATCH v2 27/45] target/hppa: Remove cond_free Richard Henderson
2024-05-13 10:24   ` Philippe Mathieu-Daudé
2024-05-14 18:15   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 28/45] target/hppa: Introduce DisasDelayException Richard Henderson
2024-05-14 18:49   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 29/45] target/hppa: Use delay_excp for conditional traps Richard Henderson
2024-05-14 18:55   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 30/45] target/hppa: Use delay_excp for conditional trap on overflow Richard Henderson
2024-05-14 19:00   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 31/45] linux-user/hppa: Force all code addresses to PRIV_USER Richard Henderson
2024-05-14 19:58   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 32/45] target/hppa: Store full iaoq_f and page offset of iaoq_b in TB Richard Henderson
2024-05-14 20:11   ` Helge Deller
2024-05-15  7:35     ` Richard Henderson
2024-05-13  7:47 ` [PATCH v2 33/45] target/hppa: Do not mask in copy_iaoq_entry Richard Henderson
2024-05-14 20:13   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 34/45] target/hppa: Improve hppa_cpu_dump_state Richard Henderson
2024-05-14 20:16   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 35/45] target/hppa: Split PSW X and B into their own field Richard Henderson
2024-05-13 10:30   ` Philippe Mathieu-Daudé
2024-05-14 20:22   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 36/45] target/hppa: Manage PSW_X and PSW_B in translator Richard Henderson
2024-05-14 21:18   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 37/45] target/hppa: Implement PSW_B Richard Henderson
2024-05-14 21:19   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 38/45] target/hppa: Implement PSW_X Richard Henderson
2024-05-14 21:22   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 39/45] target/hppa: Drop tlb_entry return from hppa_get_physical_address Richard Henderson
2024-05-13  9:26   ` Philippe Mathieu-Daudé
2024-05-14 21:23   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 40/45] target/hppa: Adjust priv for B,GATE at runtime Richard Henderson
2024-05-14 21:26   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 41/45] target/hppa: Implement CF_PCREL Richard Henderson
2024-05-14 21:32   ` Helge Deller
2024-05-13  7:47 ` [PATCH v2 42/45] target/hppa: Implement PSW_T Richard Henderson
2024-05-13  7:47 ` [PATCH v2 43/45] target/hppa: Implement PSW_H, PSW_L Richard Henderson
2024-05-13  7:47 ` [PATCH v2 44/45] target/hppa: Log cpu state at interrupt Richard Henderson
2024-05-13  7:47 ` [PATCH v2 45/45] target/hppa: Log cpu state on return-from-interrupt Richard Henderson
2024-05-14 14:02 ` [PATCH v2 00/45] target/hppa: Misc improvements Helge Deller

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=ZkNwNqGwd9QK9hy6@carbonx1 \
    --to=deller@kernel.org \
    --cc=deller@gmx.de \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.org \
    /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).