qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Richard Henderson <rth@twiddle.net>
To: qemu-devel@nongnu.org
Subject: [Qemu-devel] [PATCH 49/65] tcg/i386: Rely on undefined/undocumented behaviour of BSF/BSR
Date: Fri, 23 Dec 2016 20:00:26 -0800	[thread overview]
Message-ID: <20161224040042.12654-50-rth@twiddle.net> (raw)
In-Reply-To: <20161224040042.12654-1-rth@twiddle.net>

The ISA manual documents the output is undefined if the input was zero.

However, we document in target-i386 that the behavior of real silicon
is to preserve the contents of the output register.  We also mention
that there are real applications that depend on this.  That this is
baked into silicon is mentioned as a potential cause for some false
sharing behaviour wrt lzcnt/tzcnt.

Taking advantage of this allows us to save 2 insns in the normal case,
and 4 insns for i686 emulating a 64-bit clz.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/i386/tcg-target.inc.c | 35 ++++++++++++++++++++++-------------
 1 file changed, 22 insertions(+), 13 deletions(-)

diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c
index 3ed8cd1..3650340 100644
--- a/tcg/i386/tcg-target.inc.c
+++ b/tcg/i386/tcg-target.inc.c
@@ -1146,9 +1146,12 @@ static void tcg_out_ctz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1,
         tcg_debug_assert(arg2 == (rexw ? 64 : 32));
         tcg_out_modrm(s, OPC_TZCNT + rexw, dest, arg1);
     } else {
-        tcg_debug_assert(dest != arg2);
+        /* ??? The manual says that the output is undefined when the
+           input is zero, but real hardware leaves it unchanged.  As
+           noted in target-i386/translate.c, real programs depend on
+           this -- now we are one more of those.  */
+        tcg_debug_assert(dest == arg2);
         tcg_out_modrm(s, OPC_BSF + rexw, dest, arg1);
-        tcg_out_cmov(s, TCG_COND_EQ, rexw, dest, arg2);
     }
 }
 
@@ -1161,20 +1164,26 @@ static void tcg_out_clz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1,
             tcg_debug_assert(arg2 == (rexw ? 64 : 32));
         } else {
             tcg_debug_assert(dest != arg2);
+            /* LZCNT sets C if the input was zero.  */
             tcg_out_cmov(s, TCG_COND_LTU, rexw, dest, arg2);
         }
     } else {
-        tcg_debug_assert(!const_a2);
-        tcg_debug_assert(dest != arg1);
-        tcg_debug_assert(dest != arg2);
+        TCGType type = rexw ? TCG_TYPE_I64: TCG_TYPE_I32;
+        TCGArg rev = rexw ? 63 : 31;
 
-        /* Recall that the output of BSR is the index not the count.  */
+        /* Recall that the output of BSR is the index not the count.
+           Therefore we must adjust the result by ^ (SIZE-1).  In some
+           cases below, we prefer an extra XOR to a JMP.  */
+        /* ??? See the comment in tcg_out_ctz re BSF.  */
+        if (const_a2) {
+            tcg_debug_assert(dest != arg1);
+            tcg_out_movi(s, type, dest, arg2 ^ rev);
+        } else {
+            tcg_debug_assert(dest == arg2);
+            tgen_arithi(s, ARITH_XOR + rexw, dest, rev, 0);
+        }
         tcg_out_modrm(s, OPC_BSR + rexw, dest, arg1);
-        tgen_arithi(s, ARITH_XOR + rexw, dest, rexw ? 63 : 31, 0);
-
-        /* Since we have destroyed the flags from BSR, we have to re-test.  */
-        tcg_out_cmp(s, arg1, 0, 1, rexw);
-        tcg_out_cmov(s, TCG_COND_EQ, rexw, dest, arg2);
+        tgen_arithi(s, ARITH_XOR + rexw, dest, rev, 0);
     }
 }
 
@@ -2443,7 +2452,7 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op)
     case INDEX_op_ctz_i64:
         {
             static const TCGTargetOpDef ctz[2] = {
-                { .args_ct_str = { "&r", "r", "r" } },
+                { .args_ct_str = { "r", "r", "0" } },
                 { .args_ct_str = { "&r", "r", "rW" } },
             };
             return &ctz[have_bmi1];
@@ -2452,7 +2461,7 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op)
     case INDEX_op_clz_i64:
         {
             static const TCGTargetOpDef clz[2] = {
-                { .args_ct_str = { "&r", "r", "r" } },
+                { .args_ct_str = { "&r", "r", "0i" } },
                 { .args_ct_str = { "&r", "r", "rW" } },
             };
             return &clz[have_lzcnt];
-- 
2.9.3

  parent reply	other threads:[~2016-12-24  4:01 UTC|newest]

Thread overview: 69+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-12-24  3:59 [Qemu-devel] [PATCH v5 00/65] tcg 2.9 patch queue Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 01/65] tcg: Add field extraction primitives Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 02/65] tcg: Minor adjustments to deposit expanders Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 03/65] tcg: Add deposit_z expander Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 04/65] tcg/aarch64: Implement field extraction opcodes Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 05/65] tcg/arm: Move isa detection to tcg-target.h Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 06/65] tcg/arm: Implement field extraction opcodes Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 07/65] tcg/i386: " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 08/65] tcg/mips: " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 09/65] tcg/ppc: " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 10/65] tcg/s390: Expose host facilities to tcg-target.h Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 11/65] tcg/s390: Implement field extraction opcodes Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 12/65] tcg/s390: Support deposit into zero Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 13/65] target-alpha: Use deposit and extract ops Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 14/65] target-arm: Use new " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 15/65] target-i386: " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 16/65] target-mips: Use the new extract op Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 17/65] target-ppc: Use the new deposit and extract ops Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 18/65] target-s390x: " Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 19/65] tcg/optimize: Fold movcond 0/1 into setcond Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 20/65] tcg: Add markup for output requires new register Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 21/65] tcg: Transition flat op_defs array to a target callback Richard Henderson
2016-12-24  3:59 ` [Qemu-devel] [PATCH 22/65] tcg: Pass the opcode width to target_parse_constraint Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 23/65] tcg: Allow an operand to be matching or a constant Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 24/65] tcg: Add clz and ctz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 25/65] disas/i386.c: Handle tzcnt Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 26/65] disas/ppc: Handle popcnt and cnttz Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 27/65] target-alpha: Use the ctz and clz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 28/65] target-cris: Use clz opcode Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 29/65] target-microblaze: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 30/65] target-mips: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 31/65] target-openrisc: Use clz and ctz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 32/65] target-ppc: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 33/65] target-s390x: Use clz opcode Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 34/65] target-tilegx: Use clz and ctz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 35/65] target-tricore: Use clz opcode Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 36/65] target-unicore32: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 37/65] target-xtensa: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 38/65] target-arm: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 39/65] target-i386: Use clz and ctz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 40/65] tcg/ppc: Handle ctz and clz opcodes Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 41/65] tcg/aarch64: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 42/65] tcg/arm: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 43/65] tcg/mips: Handle clz opcode Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 44/65] tcg/s390: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 45/65] tcg/i386: Fuly convert tcg_target_op_def Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 46/65] tcg/i386: Hoist common arguments in tcg_out_op Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 47/65] tcg/i386: Allow bmi2 shiftx to have non-matching operands Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 48/65] tcg/i386: Handle ctz and clz opcodes Richard Henderson
2016-12-24  4:00 ` Richard Henderson [this message]
2017-01-16 19:19   ` [Qemu-devel] [PATCH 49/65] tcg/i386: Rely on undefined/undocumented behaviour of BSF/BSR Eduardo Habkost
2017-01-16 19:35     ` Eduardo Habkost
2016-12-24  4:00 ` [Qemu-devel] [PATCH 50/65] tcg: Add helpers for clrsb Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 51/65] target-arm: Use clrsb helper Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 52/65] target-tricore: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 53/65] target-xtensa: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 54/65] tcg: Add opcode for ctpop Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 55/65] target-alpha: Use ctpop helper Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 56/65] target-ppc: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 57/65] target-s390x: Avoid a loop for popcnt Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 58/65] target-sparc: Use ctpop helper Richard Henderson
2016-12-30 18:25   ` Mark Cave-Ayland
2016-12-24  4:00 ` [Qemu-devel] [PATCH 59/65] target-tilegx: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 60/65] target-i386: " Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 61/65] qemu/host-utils.h: Reduce the operation count in the fallback ctpop Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 62/65] tests: New test-bitcnt Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 63/65] tcg: Use ctpop to generate ctz if needed Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 64/65] tcg/ppc: Handle ctpop opcode Richard Henderson
2016-12-24  4:00 ` [Qemu-devel] [PATCH 65/65] tcg/i386: " 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=20161224040042.12654-50-rth@twiddle.net \
    --to=rth@twiddle.net \
    --cc=qemu-devel@nongnu.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).