From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1MNf4T-0001a7-NB for qemu-devel@nongnu.org; Sun, 05 Jul 2009 23:40:49 -0400 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1MNf4O-0001Zv-M0 for qemu-devel@nongnu.org; Sun, 05 Jul 2009 23:40:48 -0400 Received: from [199.232.76.173] (port=51036 helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1MNf4O-0001Zs-HW for qemu-devel@nongnu.org; Sun, 05 Jul 2009 23:40:44 -0400 Received: from mail-yx0-f188.google.com ([209.85.210.188]:47735) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1MNf4O-00062Z-6B for qemu-devel@nongnu.org; Sun, 05 Jul 2009 23:40:44 -0400 Received: by yxe26 with SMTP id 26so5136467yxe.4 for ; Sun, 05 Jul 2009 20:40:42 -0700 (PDT) MIME-Version: 1.0 Date: Mon, 6 Jul 2009 11:40:42 +0800 Message-ID: <488ae4ca0907052040u6b0ac7cdg8a45d50e559c1e97@mail.gmail.com> From: =?GB2312?B?yO6wsrDu?= Content-Type: multipart/alternative; boundary=001636af022e421249046e014500 Subject: [Qemu-devel] cpu_exit every time for a jump tb (containing only the "b" instruction) List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org --001636af022e421249046e014500 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Hi, list I am currently profiling qemu-system-arm. I wrote a simple ARM assembly only to add r0 infinitely: * add_nums: add r0, r0, r1 b add_nums* I forced a tb to contain only one instruction, and I got cpu_exit every time the tb for "*b add_nums"* was executed. It exited the cpu_exec loop every time, and then reentered it and executed it without cpu_exit. Following are the executing sequences. I dumped the "in_asm" for a tb every time it was executed. And wrapped the *tcg_qemu_tb_exec*() with qemu_log(). *IN: 0x8001000c: b 0x80010008 before tcg_qemu_tb_exec cpu_exit after tcg_qemu_tb_exec* *IN: 0x8001000c: b 0x80010008 before tcg_qemu_tb_exec after tcg_qemu_tb_exec IN: 0x80010008: add r0, r0, r1 before tcg_qemu_tb_exec after tcg_qemu_tb_exec IN: 0x8001000c: b 0x80010008 before tcg_qemu_tb_exec cpu_exit after tcg_qemu_tb_exec * It seems like the cpu_exit was called when * tcg_qemu_tb_exec *was executing. However, * tcg_qemu_tb_exec* is only to execute the translated native code (am i right?), how can it call the qemu func (cpu_exit)? Or it is called by other func executing simultaneously with * tcg_qemu_tb_exec*?* * And why cpu_exit is called everytime, it really comsumes lots of ticks.~ Any suggestions? Great thanks! Regards, Ryan --001636af022e421249046e014500 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Hi, list

=A0=A0=A0=A0=A0=A0 I am currently profiling qemu-system-arm= . I wrote a simple ARM assembly only to add r0 infinitely:

=A0=A0= =A0 add_nums:
=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 add r0, r0, r1=A0=A0=A0= =A0=A0
=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 b add_nums

=A0=A0= =A0=A0=A0=A0 I forced a tb to contain only one instruction, and I got cpu_e= xit every time the tb for "b add_nums" was executed. It ex= ited the cpu_exec loop every time, and then reentered it and executed it wi= thout cpu_exit.

=A0=A0=A0=A0=A0=A0 Following are the executing sequences. I dumped the = "in_asm" for a tb every time it was executed. And wrapped the = tcg_qemu_tb_exec() with qemu_log().


IN:
0x8001000c:=A0 b=A0=A0=A0 0x80010008

before tcg_qemu_tb_e= xec
cpu_exit
after tcg_qemu_tb_exec

IN:
0x8001000c:=A0 b=A0=A0=A0 0x= 80010008

before tcg_qemu_tb_exec
after tcg_qemu_tb_exec

IN:
0x80010008= :=A0 add=A0=A0=A0 r0, r0, r1

before tcg_qemu_tb_exec
after tcg_qe= mu_tb_exec


IN:
0x8001000c:=A0 b=A0=A0=A0 0x80010008

b= efore tcg_qemu_tb_exec
cpu_exit
after tcg_qemu_tb_exec

=A0=A0=A0=A0=A0 It seem= s like the cpu_exit was called when tcg_qemu_tb_exec was executing.= However,=A0 tcg_qemu_tb_exec is only to execute the translated nat= ive code (am i right?), how can it call the qemu func (cpu_exit)? Or it is = called by other func executing simultaneously with=A0 tcg_qemu_tb_exec<= /i>? And why cpu_exit is called everytime, it really comsumes lots = of ticks.~

=A0=A0=A0=A0=A0 Any suggestions? Great thanks!


Regards,
R= yan
--001636af022e421249046e014500--