Fix test fails on powerpc-darwin.

The current implementation of “speculation_barrier”
and “group_end_nop” insns emit hard-wired register
names which causes tests using them to fail on Darwin,
at least, which uses “rNN” instead of “NN”.

The patch makes the register names for these insns use
the operand output mechanism to substitute the
appropriate variant when needed.

gcc/

2019-04-21  Iain Sandoe  <iain@sandoe.co.uk>

	* config/rs6000/rs6000.md (group_end_nop): Emit insn register
	names using operand format, rather than hard-wired.
	(speculation_barrier): Likewise.

From-SVN: r270480
This commit is contained in:
Iain Sandoe 2019-04-21 08:25:44 +00:00 committed by Iain Sandoe
parent 375eb99a55
commit befa871199
2 changed files with 13 additions and 4 deletions

View file

@ -1,3 +1,9 @@
2019-04-21 Iain Sandoe <iain@sandoe.co.uk>
* config/rs6000/rs6000.md (group_end_nop): Emit insn register
names using operand format, rather than hard-wired.
(speculation_barrier): Likewise.
2019-04-19 Segher Boessenkool <segher@kernel.crashing.org>
PR tree-optimization/88055

View file

@ -12445,15 +12445,18 @@
[(unspec [(const_int 0)] UNSPEC_GRP_END_NOP)]
""
{
if (rs6000_tune == PROCESSOR_POWER6)
return "ori 1,1,0";
return "ori 2,2,0";
operands[0] = gen_rtx_REG (Pmode,
rs6000_tune == PROCESSOR_POWER6 ? 1 : 2);
return "ori %0,%0,0";
})
(define_insn "speculation_barrier"
[(unspec_volatile:BLK [(const_int 0)] UNSPECV_SPEC_BARRIER)]
""
"ori 31,31,0")
{
operands[0] = gen_rtx_REG (Pmode, 31);
return "ori %0,%0,0";
})
;; Define the subtract-one-and-jump insns, starting with the template
;; so loop.c knows what to generate.