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

Powerpc bootstrap failure due to duplicate case value

Browse source 
PR target/79098
	* config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
	use a switch.

From-SVN: r244489
This commit is contained in:
Alan Modra 2017-01-16 21:42:57 +10:30 • committed by Alan Modra
parent 1bcd0192bf
commit 7cefdfd5aa
2 changed files with 41 additions and 34 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
gcc/ChangeLog
View file

@ -1,3 +1,9 @@
2017-01-16 Alan Modra <amodra@gmail.com>
PR target/79098
* config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
use a switch.
2017-01-16 Georg-Johann Lay <avr@gjlay.de>
* config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.

69
gcc/config/rs6000/rs6000.c
View file

@ -9085,40 +9085,41 @@ rs6000_const_not_ok_for_debug_p (rtx x)
static bool
rs6000_legitimate_combined_insn (rtx_insn *insn)
{
switch (INSN_CODE (insn))
{
/* Reject creating doloop insns. Combine should not be allowed
to create these for a number of reasons:
1) In a nested loop, if combine creates one of these in an
outer loop and the register allocator happens to allocate ctr
to the outer loop insn, then the inner loop can't use ctr.
Inner loops ought to be more highly optimized.
2) Combine often wants to create one of these from what was
originally a three insn sequence, first combining the three
insns to two, then to ctrsi/ctrdi. When ctrsi/ctrdi is not
allocated ctr, the splitter takes use back to the three insn
sequence. It's better to stop combine at the two insn
sequence.
3) Faced with not being able to allocate ctr for ctrsi/crtdi
insns, the register allocator sometimes uses floating point
or vector registers for the pseudo. Since ctrsi/ctrdi is a
jump insn and output reloads are not implemented for jumps,
the ctrsi/ctrdi splitters need to handle all possible cases.
That's a pain, and it gets to be seriously difficult when a
splitter that runs after reload needs memory to transfer from
a gpr to fpr. See PR70098 and PR71763 which are not fixed
for the difficult case. It's better to not create problems
in the first place. */
case CODE_FOR_ctrsi_internal1:
case CODE_FOR_ctrdi_internal1:
case CODE_FOR_ctrsi_internal2:
case CODE_FOR_ctrdi_internal2:
case CODE_FOR_ctrsi_internal3:
case CODE_FOR_ctrdi_internal3:
case CODE_FOR_ctrsi_internal4:
case CODE_FOR_ctrdi_internal4:
return false;
}
int icode = INSN_CODE (insn);
/* Reject creating doloop insns. Combine should not be allowed
to create these for a number of reasons:
1) In a nested loop, if combine creates one of these in an
outer loop and the register allocator happens to allocate ctr
to the outer loop insn, then the inner loop can't use ctr.
Inner loops ought to be more highly optimized.
2) Combine often wants to create one of these from what was
originally a three insn sequence, first combining the three
insns to two, then to ctrsi/ctrdi. When ctrsi/ctrdi is not
allocated ctr, the splitter takes use back to the three insn
sequence. It's better to stop combine at the two insn
sequence.
3) Faced with not being able to allocate ctr for ctrsi/crtdi
insns, the register allocator sometimes uses floating point
or vector registers for the pseudo. Since ctrsi/ctrdi is a
jump insn and output reloads are not implemented for jumps,
the ctrsi/ctrdi splitters need to handle all possible cases.
That's a pain, and it gets to be seriously difficult when a
splitter that runs after reload needs memory to transfer from
a gpr to fpr. See PR70098 and PR71763 which are not fixed
for the difficult case. It's better to not create problems
in the first place. */
if (icode != CODE_FOR_nothing
&& (icode == CODE_FOR_ctrsi_internal1
|| icode == CODE_FOR_ctrdi_internal1
|| icode == CODE_FOR_ctrsi_internal2
|| icode == CODE_FOR_ctrdi_internal2
|| icode == CODE_FOR_ctrsi_internal3
|| icode == CODE_FOR_ctrdi_internal3
|| icode == CODE_FOR_ctrsi_internal4
|| icode == CODE_FOR_ctrdi_internal4))
return false;
return true;
}