qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: richard.henderson@linaro.org
Subject: [PATCH 07/14] target/i386: optimize computation of ZF from CC_OP_DYNAMIC
Date: Sun, 20 Oct 2024 17:53:17 +0200	[thread overview]
Message-ID: <20241020155324.35273-8-pbonzini@redhat.com> (raw)
In-Reply-To: <20241020155324.35273-1-pbonzini@redhat.com>

Most uses of CC_OP_DYNAMIC are for CMP/JB/JE or similar sequences.
We can optimize many of them to avoid computation of the flags.
This eliminates both TCG ops to set up the new cc_op, and helper
instructions because evaluating just ZF is much cheaper.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/helper.h        |  1 +
 target/i386/tcg/cc_helper.c | 20 ++++++++++++++++++++
 target/i386/tcg/translate.c | 10 +++++++---
 3 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/target/i386/helper.h b/target/i386/helper.h
index eeb8df56eaa..3f67098f11f 100644
--- a/target/i386/helper.h
+++ b/target/i386/helper.h
@@ -1,5 +1,6 @@
 DEF_HELPER_FLAGS_4(cc_compute_all, TCG_CALL_NO_RWG_SE, tl, tl, tl, tl, int)
 DEF_HELPER_FLAGS_4(cc_compute_c, TCG_CALL_NO_RWG_SE, tl, tl, tl, tl, int)
+DEF_HELPER_FLAGS_3(cc_compute_nz, TCG_CALL_NO_RWG_SE, tl, tl, tl, int)
 
 DEF_HELPER_3(write_eflags, void, env, tl, i32)
 DEF_HELPER_1(read_eflags, tl, env)
diff --git a/target/i386/tcg/cc_helper.c b/target/i386/tcg/cc_helper.c
index 40583c04cf9..c24e6a14c07 100644
--- a/target/i386/tcg/cc_helper.c
+++ b/target/i386/tcg/cc_helper.c
@@ -95,6 +95,26 @@ static target_ulong compute_all_adcox(target_ulong dst, target_ulong src1,
     return (src1 & ~(CC_C | CC_O)) | (dst * CC_C) | (src2 * CC_O);
 }
 
+target_ulong helper_cc_compute_nz(target_ulong dst, target_ulong src1,
+                                  int op)
+{
+    target_ulong mask;
+
+    if (CC_OP_HAS_EFLAGS(op)) {
+        return ~src1 & CC_Z;
+    } else {
+        MemOp size = cc_op_size(op);
+
+        if (size == MO_TL) {
+            /* Avoid shift count overflow when computing the mask below.  */
+            return dst;
+        }
+
+        mask = (1ull << (8 << size)) - 1;
+        return dst & mask;
+    }
+}
+
 target_ulong helper_cc_compute_all(target_ulong dst, target_ulong src1,
                                    target_ulong src2, int op)
 {
diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index 1a9a2fe709e..5e326ab1aff 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -1008,15 +1008,19 @@ static CCPrepare gen_prepare_eflags_o(DisasContext *s, TCGv reg)
 static CCPrepare gen_prepare_eflags_z(DisasContext *s, TCGv reg)
 {
     switch (s->cc_op) {
-    case CC_OP_DYNAMIC:
-        gen_compute_eflags(s);
-        /* FALLTHRU */
     case CC_OP_EFLAGS:
     case CC_OP_ADCX:
     case CC_OP_ADOX:
     case CC_OP_ADCOX:
         return (CCPrepare) { .cond = TCG_COND_TSTNE, .reg = cpu_cc_src,
                              .imm = CC_Z };
+    case CC_OP_DYNAMIC:
+        gen_update_cc_op(s);
+        if (!reg) {
+            reg = tcg_temp_new();
+        }
+        gen_helper_cc_compute_nz(reg, cpu_cc_dst, cpu_cc_src, cpu_cc_op);
+        return (CCPrepare) { .cond = TCG_COND_EQ, .reg = reg, .imm = 0 };
     default:
         {
             MemOp size = cc_op_size(s->cc_op);
-- 
2.46.2



  parent reply	other threads:[~2024-10-20 15:54 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-20 15:53 [PATCH 00/14] target/i386: miscellaneous flags improvements Paolo Bonzini
2024-10-20 15:53 ` [PATCH 01/14] target/i386: use tcg_gen_ext_tl when applicable Paolo Bonzini
2024-10-20 19:50   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 02/14] target/i386: Tidy cc_op_str usage Paolo Bonzini
2024-10-20 15:53 ` [PATCH 03/14] target/i386: remove CC_OP_CLR Paolo Bonzini
2024-10-20 19:55   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 04/14] target/i386: Rearrange CCOp Paolo Bonzini
2024-10-20 15:53 ` [PATCH 05/14] target/i386: Introduce cc_op_size Paolo Bonzini
2024-10-20 15:53 ` [PATCH 06/14] target/i386: Wrap cc_op_live with a validity check Paolo Bonzini
2024-10-20 20:05   ` Richard Henderson
2024-10-20 15:53 ` Paolo Bonzini [this message]
2024-10-20 20:18   ` [PATCH 07/14] target/i386: optimize computation of ZF from CC_OP_DYNAMIC Richard Henderson
2024-10-20 15:53 ` [PATCH 08/14] target/i386: optimize TEST+Jxx sequences Paolo Bonzini
2024-10-20 20:19   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 09/14] target/i386: add a few more trivial CCPrepare cases Paolo Bonzini
2024-10-20 20:23   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 10/14] target/i386: add a note about gen_jcc1 Paolo Bonzini
2024-10-20 15:53 ` [PATCH 11/14] target/i386: make flag variables unsigned Paolo Bonzini
2024-10-20 20:26   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 12/14] target/i386: use builtin popcnt or parity to compute PF, if available Paolo Bonzini
2024-10-20 20:40   ` Richard Henderson
2024-10-20 15:53 ` [PATCH 13/14] target/i386: use higher-precision arithmetic to compute CF Paolo Bonzini
2024-10-20 15:53 ` [PATCH 14/14] target/i386: use + to put flags together Paolo Bonzini
  -- strict thread matches above, loose matches on Subject: below --
2024-10-28 15:18 [PATCH v2 00/14] target/i386: miscellaneous flags improvements Paolo Bonzini
2024-10-28 15:18 ` [PATCH 07/14] target/i386: optimize computation of ZF from CC_OP_DYNAMIC Paolo Bonzini
2024-10-29 15:43   ` 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=20241020155324.35273-8-pbonzini@redhat.com \
    --to=pbonzini@redhat.com \
    --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).