OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Do not bias REG_N_REFS by loop_depth when optimising for size.

Browse Source 
From-SVN: r33833
This commit is contained in:
Nick Clifton 2000-05-10 22:29:27 +00:00 committed by Nick Clifton
parent 402823c45e
commit ebfe71a8d1
2 changed files with 13 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
gcc/ChangeLog
View File

@ -1,3 +1,10 @@
2000-05-10 Nick Clifton <nickc@redhat.com>
* flow.c (find_auto_inc): Do not bias REG_N_REFS by loop_depth
when optimising for size.
(mark_used_regs:) Ditto.
(try_pre_increment_1): Ditto.
2000-05-10 Zack Weinberg <zack@wolery.cumb.org>
* cppexp.c (lex): Use %.*s to print non-NUL-terminated string.

9
gcc/flow.c
View File

@ -4854,7 +4854,8 @@ find_auto_inc (pbi, x, insn)
/* Count an extra reference to the reg. When a reg is
incremented, spilling it is worse, so we want to make
that less likely. */
REG_N_REFS (regno) += pbi->bb->loop_depth + 1;
REG_N_REFS (regno) += (optimize_size ? 1
: pbi->bb->loop_depth + 1);
/* Count the increment as a setting of the register,
even though it isn't a SET in rtl. */
@ -4935,7 +4936,8 @@ mark_used_reg (pbi, reg, cond, insn)
REG_BASIC_BLOCK (regno) = REG_BLOCK_GLOBAL;
/* Count (weighted) number of uses of each reg. */
REG_N_REFS (regno) += pbi->bb->loop_depth + 1;
REG_N_REFS (regno) += (optimize_size ? 1
: pbi->bb->loop_depth + 1);
}
}
@ -5353,7 +5355,8 @@ try_pre_increment_1 (pbi, insn)
less likely. */
if (regno >= FIRST_PSEUDO_REGISTER)
{
REG_N_REFS (regno) += pbi->bb->loop_depth + 1;
REG_N_REFS (regno) += (optimize_size ? 1
: pbi->bb->loop_depth + 1);
REG_N_SETS (regno)++;
}
return 1;