qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <f4bug@amsat.org>
To: qemu-devel@nongnu.org, Aurelien Jarno <aurelien@aurel32.net>,
	Richard Henderson <rth@twiddle.net>,
	Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>,
	Eric Blake <eblake@redhat.com>,
	Markus Armbruster <armbru@redhat.com>,
	Laurent Vivier <lvivier@redhat.com>,
	Michael Tokarev <mjt@tls.msk.ru>,
	Eduardo Habkost <ehabkost@redhat.com>,
	Paolo Bonzini <pbonzini@redhat.com>
Cc: "Philippe Mathieu-Daudé" <f4bug@amsat.org>
Subject: [Qemu-devel] [RFC PATCH v2] coccinelle: add a script to optimize tcg op using tcg_gen_extract()
Date: Wed, 10 May 2017 20:52:04 -0300	[thread overview]
Message-ID: <20170510235204.9657-1-f4bug@amsat.org> (raw)
In-Reply-To: <0f1f5bcd-815a-c070-5ba7-281f5a0ee146@twiddle.net>

Apply this script using:

$ docker run -v `pwd`:`pwd` -w `pwd` petersenna/coccinelle \
    --sp-file scripts/coccinelle/tcg_gen_extract.cocci \
    --macro-file scripts/cocci-macro-file.h \
    --dir target \
    --in-place

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
---

This is a new version of the coccinelle script addressing Richard comments and
trying to do it correctly. Also changed license to GPLv2+.

The first rule matches, it calls a python2 script that basically checks the
target_ulong is not overflowed: (msk << ofs) >> sizeof(target_ulong) == 0 

I think this is enough but if I'm wrong I can try to do as Richard suggested
(counting the low bits).

If this rule fails, there is no further processing, else the 3rd rule doing the
patch is called.
I couldn't figure still how to reuse args from rule2 in rule3 so it is mostly a
copy of rule1.

I also don't know yet how to process target-generic functions (ending _tl
instead of _32 or _64) eventually the correct --include-option may resolve this.

Applied it verified the ARM/M68K/PPC patches (2,3,5,7 from serie v1).
The other archs weren't patched since they use _tl generic functions.

I apologize for my python!

Regards,

Phil.

 scripts/coccinelle/tcg_gen_extract.cocci | 61 ++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
 create mode 100644 scripts/coccinelle/tcg_gen_extract.cocci

diff --git a/scripts/coccinelle/tcg_gen_extract.cocci b/scripts/coccinelle/tcg_gen_extract.cocci
new file mode 100644
index 0000000000..d75ef7d98c
--- /dev/null
+++ b/scripts/coccinelle/tcg_gen_extract.cocci
@@ -0,0 +1,61 @@
+// optimize TCG using extract op
+//
+// Copyright: (C) 2017 Philippe Mathieu-Daudé. GPLv2+.
+// Confidence: Low
+// Options: --macro-file scripts/cocci-macro-file.h
+
+@match@ // match shri*+andi* pattern, calls script check_len_correct
+identifier ret, arg;
+constant ofs, len;
+identifier tcg_gen_shri =~ "^tcg_gen_shri_";
+identifier tcg_gen_andi =~ "^tcg_gen_andi_";
+position p;
+@@
+(
+tcg_gen_shri(ret, arg, ofs);
+tcg_gen_andi(ret, ret, len);@p
+)
+
+@script:python check_len_correct@
+ret_s << match.ret;
+arg_s << match.arg;
+ofs_s << match.ofs;
+msk_s << match.len;
+shr_s << match.tcg_gen_shri;
+and_s << match.tcg_gen_andi;
+@@
+try:
+    # only eval integer, no #define like 'SR_M'
+    ofs = long(ofs_s, 0)
+    msk = long(msk_s, 0)
+    # get op_size: 32/64
+    shr_sz = int(shr_s[-2:])
+    and_sz = int(and_s[-2:])
+    # op_size shr<and allowed
+    # check overflow
+    if shr_sz > and_sz or (msk << ofs) >> and_sz:
+        cocci.include_match(False) # no further process
+except ValueError: # op_size: "tl" not handled yet
+    cocci.include_match(False) # no further process
+
+@use_extract depends on check_len_correct@ // ofs/len are valid, we can replace
+identifier ret, arg;
+constant ofs, len;
+@@
+(
+// from Nikunj A Dadhania comment:
+// http://lists.nongnu.org/archive/html/qemu-devel/2017-02/msg05211.html
+-tcg_gen_shri_tl(ret, arg, ofs);
+-tcg_gen_andi_tl(ret, ret, len);
++tcg_gen_extract_tl(ret, arg, ofs, len);
+|
+// from Aurelien Jarno comment:
+// http://lists.nongnu.org/archive/html/qemu-devel/2017-05/msg01466.html
+-tcg_gen_shri_i32(ret, arg, ofs);
+-tcg_gen_andi_i32(ret, ret, len);
++tcg_gen_extract_i32(ret, arg, ofs, len);
+|
+-tcg_gen_shri_i64(ret, arg, ofs);
+-tcg_gen_andi_i64(ret, ret, len);
++tcg_gen_extract_i64(ret, arg, ofs, len);
+)
-- 
2.11.0

  parent reply	other threads:[~2017-05-10 23:52 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-05-10 20:05 [Qemu-devel] [PATCH 0/8] optimize various tcg_gen() functions using extract op Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 1/8] coccinelle: add a script to optimize tcg op using tcg_gen_extract() Philippe Mathieu-Daudé
2017-05-10 20:12   ` Eric Blake
2017-05-10 20:23     ` Philippe Mathieu-Daudé
2017-05-10 20:19   ` Richard Henderson
2017-05-10 20:23     ` Eric Blake
2017-05-10 21:27     ` Philippe Mathieu-Daudé
2017-05-10 23:52     ` Philippe Mathieu-Daudé [this message]
2017-05-11  0:13       ` [Qemu-devel] [RFC PATCH v2] " Philippe Mathieu-Daudé
2017-05-11  9:03         ` Markus Armbruster
2017-05-12  2:01           ` Philippe Mathieu-Daudé
2017-05-12  2:04             ` Philippe Mathieu-Daudé
2017-05-15  7:04             ` Markus Armbruster
2017-05-10 20:05 ` [Qemu-devel] [PATCH 2/8] target/arm: optimize smul_dual() and neon_trn_u8() using extract op Philippe Mathieu-Daudé
2017-05-10 20:15   ` Eric Blake
2017-05-12  2:49     ` Philippe Mathieu-Daudé
2017-05-10 20:20   ` Richard Henderson
2017-05-10 20:32     ` Philippe Mathieu-Daudé
2017-05-12  1:31       ` Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 3/8] target/arm: optimize rev16() " Philippe Mathieu-Daudé
2017-05-12  1:54   ` Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 4/8] target/cris: optimize gen_swapb() " Philippe Mathieu-Daudé
2017-05-12  1:29   ` Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 5/8] target/m68k: optimize bcd_flags() " Philippe Mathieu-Daudé
2017-05-11  8:41   ` Laurent Vivier
2017-05-12  1:52     ` Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 6/8] target/mips: optimize bshfl() " Philippe Mathieu-Daudé
2017-05-12  2:13   ` Philippe Mathieu-Daudé
2017-05-10 20:05 ` [Qemu-devel] [PATCH 7/8] target/ppc: optimize various functions " Philippe Mathieu-Daudé
2017-05-11  0:41   ` David Gibson
2017-05-11  8:46     ` [Qemu-devel] [Qemu-ppc] " Laurent Vivier
2017-05-12  5:13       ` David Gibson
2017-05-11  4:54   ` [Qemu-devel] " Nikunj A Dadhania
2017-05-12  1:48     ` Philippe Mathieu-Daudé
2017-05-12  5:16       ` David Gibson
2017-05-10 20:05 ` [Qemu-devel] [PATCH 8/8] target/sparc: " Philippe Mathieu-Daudé
2017-05-12  1:50   ` Philippe Mathieu-Daudé
2017-05-10 20:20 ` [Qemu-devel] [PATCH 0/8] optimize various tcg_gen() " no-reply
2017-05-12  2:34   ` Philippe Mathieu-Daudé

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=20170510235204.9657-1-f4bug@amsat.org \
    --to=f4bug@amsat.org \
    --cc=armbru@redhat.com \
    --cc=aurelien@aurel32.net \
    --cc=eblake@redhat.com \
    --cc=ehabkost@redhat.com \
    --cc=lvivier@redhat.com \
    --cc=mjt@tls.msk.ru \
    --cc=nikunj@linux.vnet.ibm.com \
    --cc=pbonzini@redhat.com \
    --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).