Re: [Qemu-devel] qemu-system-ppc TCG assert with git master

[Bastian Koppelmann <kbastian@mail.uni-paderborn.de>]

[-- Attachment #1.1: Type: text/plain, Size: 2393 bytes --]

Hi Mark, sorry I forgot to cc qemu-devel :/

On 03/12/2015 08:41 AM, Mark Cave-Ayland wrote:
> Hi all,
>
> Whilst testing git master in preparation for some OpenBIOS updates, I'm
> seeing the following TCG assert in one of my older test images:
>
>
> $ gdb --args ./qemu-system-ppc -cdrom
> /home/build/src/qemu/image/ppc/ubuntu-5.10-live-powerpc.iso -boot d -g
> 800x600x8
> GNU gdb (GDB) 7.4.1-debian
> Copyright (C) 2012 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later
> <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.  Type "show 
> copying"
> and "show warranty" for details.
> This GDB was configured as "x86_64-linux-gnu".
> For bug reporting instructions, please see:
> <http://www.gnu.org/software/gdb/bugs/>...
> Reading symbols from /home/build/rel-qemu-git/bin/qemu-system-ppc...done.
> (gdb) r
> Starting program: /home/build/rel-qemu-git/bin/qemu-system-ppc -cdrom
> /home/build/src/qemu/image/ppc/ubuntu-5.10-live-powerpc.iso -boot d -g
> 800x600x8
> [Thread debugging using libthread_db enabled]
> Using host libthread_db library 
> "/lib/x86_64-linux-gnu/libthread_db.so.1".
> [New Thread 0x7fffe9ef3700 (LWP 11437)]
> [New Thread 0x7fffe5bf1700 (LWP 11438)]
> [New Thread 0x7fffe53f0700 (LWP 11439)]
> [Thread 0x7fffe5bf1700 (LWP 11438) exited]
> [New Thread 0x7fffe5bf1700 (LWP 11443)]
> [Thread 0x7fffe5bf1700 (LWP 11443) exited]
> qemu-system-ppc: /home/build/src/qemu/git/qemu/tcg/optimize.c:212:
> tcg_opt_gen_mov: Assertion `temps[src].state != TCG_TEMP_CONST' failed.

does this fix it?

diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index 96adf9a..4c8ff70 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -396,7 +396,11 @@ static inline void tcg_gen_and_i32(TCGv_i32 ret, 
TCGv_i32 arg1, TCGv_i32 arg2)

  static inline void tcg_gen_or_i32(TCGv_i32 ret, TCGv_i32 arg1, 
TCGv_i32 arg2)
  {
-    tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
+    if (TCGV_EQUAL_I32(arg1, arg2)) {
+        tcg_gen_mov_i32(ret, arg1);
+    } else {
+        tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
+    }
  }

  static inline void tcg_gen_xor_i32(TCGv_i32 ret, TCGv_i32 arg1, 
TCGv_i32 arg2)

The patch is also attached. It is just a revert of tcg_gen_or_i32. I'll 
investigate some more why the optimizer fails here.

Cheers,
Bastian


[-- Attachment #1.2: Type: text/html, Size: 4098 bytes --]

[-- Attachment #2: 0001-tcg-tcg_gen_or_i32-causing-in-some-cases-tcg_opt_gen.patch --]
[-- Type: text/x-patch, Size: 1059 bytes --]

>From 64c6c9e8a312ae6030794ae1c0fb8e45ce0d77a3 Mon Sep 17 00:00:00 2001
From: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Date: Thu, 12 Mar 2015 14:47:29 +0000
Subject: [PATCH] tcg: tcg_gen_or_i32 causing in some cases tcg_opt_gen_mov to
 fail

A first try to fix this is to emit the mov in tcg_or_i32.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
---
 tcg/tcg-op.h | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index 96adf9a..4c8ff70 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -396,7 +396,11 @@ static inline void tcg_gen_and_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
 
 static inline void tcg_gen_or_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
 {
-    tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
+    if (TCGV_EQUAL_I32(arg1, arg2)) {
+        tcg_gen_mov_i32(ret, arg1);
+    } else {
+        tcg_gen_op3_i32(INDEX_op_or_i32, ret, arg1, arg2);
+    }
 }
 
 static inline void tcg_gen_xor_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
-- 
2.3.2