Thx very much.
-----邮件原件-----
发件人: Sun Chan [mailto:sun.c...@gmail.com]
发送时间: 2012年8月24日 7:46
收件人: shenrfen
主题: Re: Please review patch for bug1004(IPA dst merge)
looks good to me
Sun
On Wed, Aug 22, 2012 at 11:08 AM, shenrfen <shenr...@gmail.com> wrote:
> hi, all.
>
>
>
>
> Testcase
>
> int foo(int a)
>
> {
>
> if (a == 0)
>
> return 1;
>
> else
>
> return foo(a - 1) + 10;
>
> }
>
> int main()
>
> {
>
> return foo(10);
>
> }
>
>
>
>
> CMD
>
> opencc a.c -ipa -keep -show
>
>
> Error
>
> ### Assertion failure at line 310 of
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/com
> mon/co
> m/dwarf_DST.cxx:
>
> ### Compiler Error during Inline processing phase:
>
> ### Illegal attempt to append DST child
>
> IPA processing aborted
>
>
>
> Or
>
>
>
> opencc INTERNAL ERROR:
> /home/rfshen/workspace/trunk/build_trunk_x8664/x8664_bin/lib/gcc-lib/x
> 86_64- open64-linux/5.0/ipa_link died due to signal 11
>
>
> Part of backtrace
>
> #2 0xf7e8e110 in DST_enter_inlined_subroutine (parent={byte_idx = 0,
> block_idx = 12830}, inl_routine=
>
> {byte_idx = 144, block_idx = 12829}, begin_label=80215,
> end_label=80215, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:847
>
> #3 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 0,
> block_idx = 12830}, inl_routine=
>
> {byte_idx = 56, block_idx = 12829}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #4 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx = 56,
> block_idx = 12829}, inl_routine=
>
> {byte_idx = 56, block_idx = 12829}, begin_label=82978,
> end_label=81442, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #5 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 56,
> block_idx = 12829}, inl_routine=
>
> {byte_idx = 88, block_idx = 12828}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #6 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx = 88,
> block_idx = 12828}, inl_routine=
>
> {byte_idx = 88, block_idx = 12828}, begin_label=82973,
> end_label=81437, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #7 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 88,
> block_idx = 12828}, inl_routine=
>
> {byte_idx = 144, block_idx = 12827}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #8 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx =
> 144, block_idx = 12827}, inl_routine=
>
> {byte_idx = 144, block_idx = 12827}, begin_label=82968,
> end_label=81432, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #9 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 144,
> block_idx = 12827}, inl_routine=
>
> {byte_idx = 176, block_idx = 12826}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
>
>
>
>
> For this case,
>
> Edge: Foo <0,3> inline foo <0,3>
>
>
> DST_enter_inlined_subroutine (parent={byte_idx = 0, block_idx = 3},
> inl_routine=
>
> {byte_idx = 0, block_idx = 3},..)
>
> {
>
> ….
>
>
>
>
>
>
>
>
>
>
>
>
> idx = DST_mk_inlined_subroutine (low_pc, high_pc,
>
> abstract_origin,
> abstract_file_dst);
>
> Idx = <0, 4> is a copy of inl_routine<0,3>
>
>
>
>
> DST_append_child(parent, idx);
>
> Parent <0,3> append child idx<0,4>
>
>
> DST_enter_cloned_childs (idx, inl_routine, symtab,
>
> caller_file_dst, callee_file_dst,
>
> file_index, TRUE, caller_file_m,
>
> callee_file_m);
>
>
>
> DST_enter_cloned_childs call DST_enter_inlined_subroutine, append
> all child of inl_routine to idx
>
>
>
> The child of Inl_routine<0,3> is <0,4>
>
>
>
> There is be dead recursion between(DST_enter_cloned_childs and
> DST_enter_inlined_subroutine).
>
>
>
> <0,3>à<0,4>à<0,3>
>
>
> ...
>
> }
>
>
>
>
>
> There is a patch for bug 1004.
>
> Could you please give a review, Thx very much.
>
>
>
>
> Index: osprey/be/com/clone_DST_utils.cxx
>
> ===================================================================
>
> --- osprey/be/com/clone_DST_utils.cxx (revision 4002)
>
> +++ osprey/be/com/clone_DST_utils.cxx (working copy)
>
> @@ -866,12 +866,12 @@
>
> idx = DST_mk_inlined_subroutine (low_pc, high_pc,
>
> abstract_origin,
> abstract_file_dst);
>
>
>
> - DST_append_child(parent, idx);
>
> -
>
> DST_enter_cloned_childs (idx, inl_routine, symtab,
>
> caller_file_dst, callee_file_dst,
>
> file_index, TRUE, caller_file_m,
>
> callee_file_m);
>
> +
>
> + DST_append_child(parent, idx);
>
> }
>
> #if (!defined(_LEGO_CLONER))
>
> else {
>
>
>
>
>
> PS:
>
> Recursive edge is treated as common edge if fully inline.
>
> fooàfoo
>
>
> Before inline
>
> int foo(int a)
>
> {
>
> if (a == 0)
>
> return 1;
>
> else
>
> return foo(a - 1) + 10;
>
> }
>
>
>
>
> After inline
>
> int foo(int a)
>
> {
>
> if (a == 0)
>
> return 1;
>
> else
>
> {
>
> if ((a-1) == 0)
>
> t = 1;
>
> else
>
> t = foo(a �C 1 - 1) + 10;
>
> return t+ 10;
>
> }
>
> }
>
>
>
>
>
>
>
>
>
> 发件人: shenrfen [mailto:shenr...@gmail.com]
> 发送时间: 2012年8月17日 11:34
> 收件人: 'shenrfen'; 'Xiaojing Zhang'; 'Jian-Xin Lai'; 'Sun Chan'
> 抄送: open64-devel@lists.sourceforge.net
> 主题: One quesion about bug1004(should directly recursion edged be
> inlined)
>
>
>
> Hi , all
>
> There is a small case for bug1004
(http://bugs.open64.net/show_bug.
> cgi?id=1004)
>
>
> Testcase
>
> int foo(int a)
>
> {
>
> if (a == 0)
>
> return 1;
>
> else
>
> return foo(a - 1) + 10;
>
> }
>
> int main()
>
> {
>
> return foo(10);
>
> }
>
>
>
>
> CMD
>
> opencc a.c -ipa -keep -show
>
>
> Error
>
> ### Assertion failure at line 310 of
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/com
> mon/co
> m/dwarf_DST.cxx:
>
> ### Compiler Error during Inline processing phase:
>
> ### Illegal attempt to append DST child
>
> IPA processing aborted
>
>
>
> Or
>
>
>
> opencc INTERNAL ERROR:
> /home/rfshen/workspace/trunk/build_trunk_x8664/x8664_bin/lib/gcc-lib/x
> 86_64- open64-linux/5.0/ipa_link died due to signal 11
>
>
> Part of backtrace
>
> #2 0xf7e8e110 in DST_enter_inlined_subroutine (parent={byte_idx = 0,
> block_idx = 12830}, inl_routine=
>
> {byte_idx = 144, block_idx = 12829}, begin_label=80215,
> end_label=80215, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:847
>
> #3 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 0,
> block_idx = 12830}, inl_routine=
>
> {byte_idx = 56, block_idx = 12829}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #4 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx = 56,
> block_idx = 12829}, inl_routine=
>
> {byte_idx = 56, block_idx = 12829}, begin_label=82978,
> end_label=81442, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #5 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 56,
> block_idx = 12829}, inl_routine=
>
> {byte_idx = 88, block_idx = 12828}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #6 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx = 88,
> block_idx = 12828}, inl_routine=
>
> {byte_idx = 88, block_idx = 12828}, begin_label=82973,
> end_label=81437, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #7 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 88,
> block_idx = 12828}, inl_routine=
>
> {byte_idx = 144, block_idx = 12827}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:461
>
> #8 0xf7e8e233 in DST_enter_inlined_subroutine (parent={byte_idx =
> 144, block_idx = 12827}, inl_routine=
>
> {byte_idx = 144, block_idx = 12827}, begin_label=82968,
> end_label=81432, caller_file_dst=0x8338c58,
>
> callee_file_dst=0x8338c58, symtab=0x832cf88,
> caller_file_m=0x88b3100, callee_file_m=0x88b3100, cross_file_id=0)
>
> at
> /home/rfshen/workspace/trunk/build_trunk_x8664/osprey/../../osprey/be/
> com/cl
> one_DST_utils.cxx:874
>
> #9 0xf7e8ed14 in DST_enter_cloned_childs (parent={byte_idx = 144,
> block_idx = 12827}, inl_routine=
>
> {byte_idx = 176, block_idx = 12826}, symtab=0x832cf88,
> caller_file_dst=0x8338c58, callee_file_dst=0x8338c58,
>
> parent_file_index=0, inlined=1, caller_file_m=0x88b3100,
> callee_file_m=0x88b3100)
>
>
>
>
>
> For this case, there will be dead recursion
> between(DST_enter_cloned_childs and DST_enter_inlined_subroutine).
>
> DST merge is at the end of IPA:Process
>
> I don’t know why the directly recursion edge is judged to be inlined
> before.
>
>
>
> Could you please help me?
>
> Thx very much.
>
>
>
>
>
> 发件人: shenrfen [mailto:shenr...@gmail.com]
> 发送时间: 2012年8月16日 15:18
> 收件人: 'Xiaojing Zhang'; 'Jian-Xin Lai'
> 主题: 答复: 答复: regression related with your check-in rev3993
>
>
>
> 好的，谢谢！
>
> 如果是这样的话，估计应该是前面判断的 地方存在问题。
>
> 你认为呢？
>
>
>
>
>
> 发件人: Xiaojing Zhang [mailto:xiaojing70...@gmail.com]
> 发送时间: 2012年8月16日 13:49
> 收件人: shenrfen
> 主题: Re: 答复: regression related with your check-in rev3993
>
>
>
> Hi, shenrfen,
>
> 我看了其它case,也都是直接递归的。
>
>
>
> 2012/8/15 Xiaojing Zhang <xiaojing70...@gmail.com>
>
> Hi, shenrfen,
>
> 我这里有一些failed 的case, 我先minimize 一下看看再给你
>
> 对直接递归应该不会被inline吧？
>
>
>
> 2012/8/15 shenrfen <shenr...@gmail.com>
>
> Hi ,jing & jianxin.
>
>
>
> 我发现只有直接递归的时候会出先 bug1004。
>
> 写了个简单的间接递归的case，是正确的。
>
> 不知道你那里有没有更多failure的case。
>
>
>
>
>
> 我在想，IPA阶段是先判断一条边是否可以被inline，然后再执行inline这
个
> 动作的。
>
> 对于一个直接递归这样一条边，该不该 inline呢？
>
>
>
> 发件人: shenrfen [mailto:shenr...@gmail.com]
> 发送时间: 2012年8月13日 11:01
> 收件人: 'Xiaojing Zhang'
> 主题: 答复: regression related with your check-in rev3993
>
>
>
> Hi, jing
>
> I did not see this mail until now as I was on vacation in
> last two weeks.
>
> I will check bug1004 now.
>
> Thx.
>
>
>
> 发件人: Xiaojing Zhang [mailto:xiaojing70...@gmail.com]
> 发送时间: 2012年8月6日 16:23
> 收件人: shenrfen
> 主题: Re: regression related with your check-in rev3993
>
>
>
> Hi, shenrfen,
>
>
>
> With the "DST_get_dirname" declaration problem solved, there still
> another regression persists, I have filed bug1004 for this problem,
> please have a look ASAP, thank you.
>
>
>
>
>
> 2012/8/2 Xiaojing Zhang <xiaojing70...@gmail.com>
>
> Hi, shenrfen,
>
>
>
> There is a regression caused by your check-in of rev 3993. could you
> please have a look?
>
> For the small case list of bug1003, it will got link failure.
>
> ===========
>
> Test case:
>
> 1.c
>
> extern void foo(int a);
>
> int main()
>
> {
>
> foo(135);
>
> }
>
> =======
>
> 2.c
>
> extern void goo(int a);
>
> void foo(int a)
>
> {
>
> too(a);
>
> }
>
> ========
>
> 3.c
>
>
>
> int goo(int a)
>
> {
>
> //printf("a = %d\n", a);
>
> too(a);
>
> }
>
>
>
> void too(int a)
>
> {
>
> printf("a = %d\n", a);
>
> }
>
> with command:
>
> opencc 1.c 2.c 3.c �Cipa
>
> Error:
>
> ../lib/gcc-lib/x86_64-open64-linux/5.0/ipa_link: symbol lookup error:
> .../lib/gcc-lib/x86_64-open64-linux/5.0/ipa.so: undefined symbol:
> _Z15DST_get_dirnamet
>
> openCC INTERNAL ERROR:
> .../lib/gcc-lib/x86_64-open64-linux/5.0/ipa_link
> returned non-zero status 127
>
>
>
> In osprey/common/com/dwarf_DST_producer.cxx: the function
> DST_get_dirname is declared as “static char*”, However, I find in
> osprey/be/com/clone_DST_utils.cxx, there is an “extern char*
> DST_get_dirname”, a static function can’t be “extern”, and removing
> the "static" is able to solve this problem, please make it sure, thank
you.
>
>
>
> Regards,
>
> zhang, xiao-jing
>
>
>
>
>
>
>
>
>
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Open64-devel mailing list
Open64-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/open64-devel
