[Patch i386]: btver2 pipeline descriptions.

Kumar, Venkataramanan Venkataramanan.Kumar@amd.com
Thu Jan 17 13:18:00 GMT 2013


Hi Honza,

There is no change when I find top 10 symbols using nm -t d --size-sort ./cc1 | c++filt | tail. 

Before and after the patch is same.

(snip)
0000000000085808 B default_target_expmed
0000000000093960 r bdver3_fp_transitions
0000000000098957 T c_common_nodes_and_builtins()
0000000000107792 B default_target_ira_int
0000000000109648 R cl_options
0000000000118413 t recog_69(rtx_def*, rtx_def*, int*)
0000000000122680 T recog(rtx_def*, rtx_def*, int*)
0000000000140789 T fold_binary_loc(unsigned int, tree_code, tree_node*, tree_node*, tree_node*)
0000000000155680 R insn_data
0000000000176304 r operand_data
(snip)

When I grep btver2 I get the below symbols.
 
nm -t d --size-sort ./cc1 | c++filt | grep btver2
(Snip)
0000000000000002 r btver2_min_issue_delay
0000000000000012 r btver2_transitions
0000000000000023 r get_attr_btver2_decode(rtx_def*)::__FUNCTION__
0000000000000025 r get_attr_btver2_sse_attr(rtx_def*)::__FUNCTION__
0000000000000028 r insn_default_latency_btver2(rtx_def*)::__FUNCTION__
0000000000000030 r internal_dfa_insn_code_btver2(rtx_def*)::__FUNCTION__
0000000000000195 r btver2_agu_min_issue_delay
0000000000000199 T get_attr_btver2_sse_attr(rtx_def*)
0000000000000256 r btver2_int_base
0000000000000390 r btver2_agu_transitions
0000000000000480 D btver2_cost
0000000000000589 r btver2_int_check
0000000000000589 r btver2_int_transitions
0000000000000664 T get_attr_btver2_decode(rtx_def*)
0000000000000898 r btver2_translate
0000000000000898 r btver2_fp_translate
0000000000000898 r btver2_agu_translate
0000000000000898 r btver2_int_translate
0000000000001920 r btver2_int_min_issue_delay
0000000000001990 r btver2_fp_base
0000000000008101 T insn_default_latency_btver2(rtx_def*)
0000000000009810 T internal_dfa_insn_code_btver2(rtx_def*)
0000000000012788 r btver2_fp_check
0000000000012788 r btver2_fp_transitions
0000000000053730 r btver2_fp_min_issue_delay
(Snip)

Regards,
Venkat.

-----Original Message-----
From: Jan Hubicka [mailto:hubicka@ucw.cz] 
Sent: Thursday, January 17, 2013 6:02 PM
To: Uros Bizjak
Cc: Kumar, Venkataramanan; gcc-patches@gcc.gnu.org
Subject: Re: [Patch i386]: btver2 pipeline descriptions.

> On Thu, Jan 17, 2013 at 8:17 AM, Kumar, Venkataramanan 
> <Venkataramanan.Kumar@amd.com> wrote:
> > Hi Maintainers,
> >
> > This patch adds pipeline descriptions for -march=btver2.
> >
> > Completed bootstrap and  gcc regression test passes with  r194705
> >
> > Is this ok for trunk?
> >
> >
> > Change log
> > --------------
> > 2013-01-17  Venkataramanan Kumar  <venkataramanan.kumar@amd.com>
> >
> >         btver2 pipeline descriptions
> >         * config/i386/i386.c : Enable CPU_BTVER2 to use btver2 pipeline descriptions.
> >         * config/i386/i386.md (btver2_decode) : New type attributes.
> >         * config/i386/sse.md (btver2_decode, btver2_sse_attr): New type attributes.
> >         * config/i386/btver2.md: New file describing btver2 pipelines.
> 
> As agreed previously, tuning patch looks safe even this late in the game.

Just a quick question, what are the effect on the generated automaton sizes?

Honza




More information about the Gcc-patches mailing list