FreeChainXenon/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

i386.c (pentium4_cost, [...]): Update 32bit memcpy/memset decriptors.

Browse source 
* config/i386/i386.c (pentium4_cost, nocona_cost): Update
	32bit memcpy/memset decriptors.
	(decide_alg): With -minline-all-stringops and sizes that are best
	to be copied via libcall still work hard enough to pick non-libcall
	strategy.

Co-Authored-By: Uros Bizjak <ubizjak@gmail.com>

From-SVN: r119445
This commit is contained in:
Jan Hubicka 2006-12-02 14:16:27 +01:00 • committed by Jan Hubicka
parent 9c134b65cd
commit cc0faf9dc9
2 changed files with 24 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
gcc/ChangeLog
View file

@ -1,3 +1,12 @@
2006-11-30 Jan Hubicka <jh@suse.cz>
Uros Bizjak <ubizjak@gmail.com>
* config/i386/i386.c (pentium4_cost, nocona_cost): Update
32bit memcpy/memset decriptors.
(decide_alg): With -minline-all-stringops and sizes that are best
to be copied via libcall still work hard enough to pick non-libcall
strategy.
2006-12-02 Kazu Hirata <kazu@codesourcery.com> 2006-12-02 Kazu Hirata <kazu@codesourcery.com>
* doc/extend.texi, doc/invoke.texi, doc/md.texi: Fix typos. * doc/extend.texi, doc/invoke.texi, doc/md.texi: Fix typos.

24
gcc/config/i386/i386.c
View file

@ -655,10 +655,11 @@ struct processor_costs pentium4_cost = {
COSTS_N_INSNS (2), /* cost of FABS instruction. */ COSTS_N_INSNS (2), /* cost of FABS instruction. */
COSTS_N_INSNS (2), /* cost of FCHS instruction. */ COSTS_N_INSNS (2), /* cost of FCHS instruction. */
COSTS_N_INSNS (43), /* cost of FSQRT instruction. */ COSTS_N_INSNS (43), /* cost of FSQRT instruction. */
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}, {{libcall, {{12, loop_1_byte}, {64, loop}, {-1, rep_prefix_4_byte}}},
{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}}, DUMMY_STRINGOP_ALGS},
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}, {{libcall, {{6, loop_1_byte}, {64, loop}, {20480, rep_prefix_4_byte},
{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}} {-1, libcall}}},
DUMMY_STRINGOP_ALGS},
}; };
static const static const
@ -712,10 +713,11 @@ struct processor_costs nocona_cost = {
COSTS_N_INSNS (3), /* cost of FABS instruction. */ COSTS_N_INSNS (3), /* cost of FABS instruction. */
COSTS_N_INSNS (3), /* cost of FCHS instruction. */ COSTS_N_INSNS (3), /* cost of FCHS instruction. */
COSTS_N_INSNS (44), /* cost of FSQRT instruction. */ COSTS_N_INSNS (44), /* cost of FSQRT instruction. */
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}, {{libcall, {{12, loop_1_byte}, {64, loop}, {-1, rep_prefix_4_byte}}},
{libcall, {{32, loop}, {20000, rep_prefix_8_byte}, {libcall, {{32, loop}, {20000, rep_prefix_8_byte},
{100000, unrolled_loop}, {-1, libcall}}}}, {100000, unrolled_loop}, {-1, libcall}}}},
{{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}}, {{libcall, {{6, loop_1_byte}, {64, loop}, {20480, rep_prefix_4_byte},
{-1, libcall}}},
{libcall, {{24, loop}, {64, unrolled_loop}, {libcall, {{24, loop}, {64, unrolled_loop},
{8192, rep_prefix_8_byte}, {-1, libcall}}}} {8192, rep_prefix_8_byte}, {-1, libcall}}}}
}; };
@ -13507,14 +13509,18 @@ decide_alg (HOST_WIDE_INT count, HOST_WIDE_INT expected_size, bool memset,
last non-libcall inline algorithm. */ last non-libcall inline algorithm. */
if (TARGET_INLINE_ALL_STRINGOPS) if (TARGET_INLINE_ALL_STRINGOPS)
{ {
gcc_assert (alg != libcall); /* When the current size is best to be copied by a libcall,
return alg; but we are still forced to inline, run the heuristic bellow
that will pick code for medium sized blocks. */
if (alg != libcall)
return alg;
break;
} }
else else
return algs->size[i].alg; return algs->size[i].alg;
} }
} }
gcc_unreachable (); gcc_assert (TARGET_INLINE_ALL_STRINGOPS);
} }
/* When asked to inline the call anyway, try to pick meaningful choice. /* When asked to inline the call anyway, try to pick meaningful choice.
We look for maximal size of block that is faster to copy by hand and We look for maximal size of block that is faster to copy by hand and