This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Comment update for find_reloads_address
- To: law at cygnus dot com, gcc-patches at gcc dot gnu dot org
- Subject: Comment update for find_reloads_address
- From: Joern Rennecke <amylaar at cygnus dot co dot uk>
- Date: Wed, 10 Nov 1999 19:27:37 +0000 (GMT)
- Cc: amylaar at cygnus dot co dot uk
find_reloads_address still has a comment that became outdated when
regsiter elimination was changed so that no pseudos are replaced with
MEMs at that stage.
Index: reload.c
===================================================================
RCS file: /cvs/gcc/egcs/gcc/reload.c,v
retrieving revision 1.89
diff -p -r1.89 reload.c
*** reload.c 1999/11/03 12:43:35 1.89
--- reload.c 1999/11/10 19:26:21
*************** find_reloads_address (mode, memrefloc, a
*** 4699,4712 ****
that the index needs a reload and find_reloads_address_1 will take care
of it.
- There is still a case when we might generate an extra reload,
- however. In certain cases eliminate_regs will return a MEM for a REG
- (see the code there for details). In those cases, memory_address_p
- applied to our address will return 0 so we will think that our offset
- must be too large. But it might indeed be valid and the only problem
- is that a MEM is present where a REG should be. This case should be
- very rare and there doesn't seem to be any way to avoid it.
-
If we decide to do something here, it must be that
`double_reg_address_ok' is true and that this address rtl was made by
eliminate_regs. We generate a reload of the fp/sp/ap + constant and
--- 4699,4704 ----