This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] allow zero_extract combines - checked in
- From: Eric Christopher <echristo at redhat dot com>
- To: gcc-patches at gcc dot gnu dot org
- Date: 31 Mar 2003 14:13:34 -0800
- Subject: [PATCH] allow zero_extract combines - checked in
After some debate I've checked this patch into mainline.
I've tested on every platform and simulator I could find, but if anyone
runs into problems please let me know since this is touching code older
than the repository. :)
2003-03-31 Eric Christopher <echristo at redhat dot com>
* combine.c (can_combine_p): Allow ZERO_EXTRACT and STRICT_LOW_PART.
(combinable_i3pat): Remove call to expand_field_assignment and
#if 0'd code.
--
o/~ got caught stealing fire from the sky o/~
Index: combine.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/combine.c,v
retrieving revision 1.352
diff -u -p -w -r1.352 combine.c
--- combine.c 24 Mar 2003 11:29:09 -0000 1.352
+++ combine.c 31 Mar 2003 22:11:50 -0000
@@ -1065,8 +1065,6 @@ can_combine_p (insn, i3, pred, succ, pde
/* Don't eliminate a store in the stack pointer. */
if (dest == stack_pointer_rtx
- /* If we couldn't eliminate a field assignment, we can't combine. */
- || GET_CODE (dest) == ZERO_EXTRACT || GET_CODE (dest) == STRICT_LOW_PART
/* Don't combine with an insn that sets a register to itself if it has
a REG_EQUAL note. This may be part of a REG_NO_CONFLICT sequence. */
|| (rtx_equal_p (src, dest) && find_reg_note (insn, REG_EQUAL, NULL_RTX))
@@ -1319,46 +1317,16 @@ combinable_i3pat (i3, loc, i2dest, i1des
if (GET_CODE (x) == SET)
{
- rtx set = expand_field_assignment (x);
+ rtx set = x ;
rtx dest = SET_DEST (set);
rtx src = SET_SRC (set);
rtx inner_dest = dest;
-#if 0
- rtx inner_src = src;
-#endif
-
- SUBST (*loc, set);
-
while (GET_CODE (inner_dest) == STRICT_LOW_PART
|| GET_CODE (inner_dest) == SUBREG
|| GET_CODE (inner_dest) == ZERO_EXTRACT)
inner_dest = XEXP (inner_dest, 0);
- /* We probably don't need this any more now that LIMIT_RELOAD_CLASS
- was added. */
-#if 0
- while (GET_CODE (inner_src) == STRICT_LOW_PART
- || GET_CODE (inner_src) == SUBREG
- || GET_CODE (inner_src) == ZERO_EXTRACT)
- inner_src = XEXP (inner_src, 0);
-
- /* If it is better that two different modes keep two different pseudos,
- avoid combining them. This avoids producing the following pattern
- on a 386:
- (set (subreg:SI (reg/v:QI 21) 0)
- (lshiftrt:SI (reg/v:SI 20)
- (const_int 24)))
- If that were made, reload could not handle the pair of
- reg 20/21, since it would try to get any GENERAL_REGS
- but some of them don't handle QImode. */
-
- if (rtx_equal_p (inner_src, i2dest)
- && GET_CODE (inner_dest) == REG
- && ! MODES_TIEABLE_P (GET_MODE (i2dest), GET_MODE (inner_dest)))
- return 0;
-#endif
-
/* Check for the case where I3 modifies its output, as
discussed above. */
if ((inner_dest != dest