[gcc]
2014-09-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* config/rs6000/rs6000.c (special_handling_values): Add
SH_EXTRACT.
(rtx_is_swappable_p): Look for patterns with a VEC_SELECT, perhaps
wrapped in a VEC_DUPLICATE, representing an extract. Mark these
as swappable with special handling SH_EXTRACT. Remove
UNSPEC_VSX_XXSPLTW from the list of disallowed unspecs for the
optimization.
(adjust_extract): New function.
(handle_special_swappables): Add default to case statement; add
case for SH_EXTRACT that calls adjust_extract.
(dump_swap_insn_table): Handle SH_EXTRACT.
[gcc/testsuite]
2014-09-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* gcc.target/powerpc/swaps-p8-13.c: New test.
* gcc.target/powerpc/swaps-p8-14.c: New test.
* gcc.target/powerpc/swaps-p8-15.c: New test.
From-SVN: r214903
[gcc]
2014-09-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* config/rs6000/vsx.md (*vsx_extract_<mode>_load): Always match
selection of 0th memory doubleword, regardless of endianness.
[gcc/testsuite]
2014-09-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* gcc.target/powerpc/vsx-extract-1.c: Test 0th doubleword
regardless of endianness.
From-SVN: r214902
* doc/invoke.texi: Document that -Wlogical-not-parentheses is enabled
by -Wall.
c-family/
* c.opt (Wlogical-not-parentheses): Enable by -Wall.
From-SVN: r214893
* config/aarch64/aarch64.c (aarch64_popwb_single_reg): Remove.
(aarch64_popwb_pair_reg): Remove.
(aarch64_set_frame_expr): Remove.
(aarch64_restore_callee_saves): Add CFI_OPS argument; fill it with
the restore ops performed by the insns generated.
(aarch64_expand_epilogue): Attach CFI_OPS to the stack deallocation
insn. Perform the calls_eh_return addition later; do not attempt to
preserve the CFA in that case. Don't use aarch64_set_frame_expr.
(aarch64_expand_prologue): Use REG_CFA_ADJUST_CFA directly, or no
special markup at all. Load cfun->machine->frame.hard_fp_offset
into a local variable.
(aarch64_frame_pointer_required): Don't check calls_alloca.
From-SVN: r214886
2014-09-03 Martin Jambor <mjambor@suse.cz>
PR ipa/61986
* ipa-cp.c (find_aggregate_values_for_callers_subset): Chain
created replacements in ascending order of offsets.
(known_aggs_to_agg_replacement_list): Likewise.
gcc/testsuite/
* gcc.dg/ipa/pr61986.c: New test.
From-SVN: r214877
[NDS32] No need to take padding into consideration in Andes ABI2 because
we can pass arguments in registers for variadic function.
-- By default GCC uses must_pass_in_stack_var_size_or_pad for TARGET_MUST_PASS_IN_STACK.
For Andes ABI2, the data layout in memory will be incorrect when calling variadic function
under big-endian configuration.
* config/nds32/nds32.c (nds32_must_pass_in_stack): New implementation
for TARGET_MUST_PASS_IN_STACK.
From-SVN: r214869
partially in registers and stack.
-- This is needed if we would like to pass variadic arguments via registers
that are pused on the stack by callee. So that va_arg() can access
arguments continously in memory.
* config/nds32/nds32.c (nds32_arg_partial_bytes): New implementation
for TARGET_ARG_PARTIAL_BYTES.
From-SVN: r214868
[NDS32] Emit stack push instructions for varargs at prologue stage and
emit stack adjustment instructions for varargs at epilogue stage.
-- Note that we do not need to handle it in v3push/v3pop prologue/epilogue
expanding because we do not apply v3push/v3pop for variadic function.
* config/nds32/nds32.c (nds32_expand_prologue): Emit stack push
instructions for varargs implementation.
(nds32_expand_epilogue): Emit stack adjustment instructions for
varargs implementation.
From-SVN: r214867
we better not to set cfun->machine->fp_as_gp_p at prologue expanding.
* config/nds32/nds32.c (nds32_expand_prologue): Suppress fp-as-gp
optimization detection.
From-SVN: r214866
1. We have caller prepare nameless arguments in the way of calling a normal function.
e.g. Assume a function: void va_test (int n, ...);
When calling "va_test (6, 11, 22, 33, 44, 55, 66);",
r0 <- 6
r1 <- 11
r2 <- 22
r3 <- 33
r4 <- 44
r5 <- 55
[sp + 0] <- 66
2. Note that we DO NOT handle it for TARGET_HARD_FLOAT case.
The TARGET_HARD_FLOAT should push all the nameless arguments into
stack and GCC is able to take care of them itself.
In addition, we have not implemented hard float on trunk yet.
* config/nds32/nds32.c (nds32_function_arg): Deal with nameless
arguments.
(nds32_function_arg_advance): Deal with nameless arguments.
* config/nds32/nds32.h (NDS32_ARG_PASS_IN_REG_P): Split it into ...
(NDS32_ARG_ENTIRE_IN_GPR_REG_P): ... this one and ...
(NDS32_ARG_PARTIAL_IN_GPR_REG_P): ... this one.
From-SVN: r214865
[NDS32] Since argument pointer must be 8-byte alignment, our first parameter may
require 4-byte offset if pushed varargs is not 8-byte aligned.
* config/nds32/nds32.h (FIRST_PARM_OFFSET): Set proper location
according to the value of crtl->args.pretend_args_size.
From-SVN: r214863
-- The rationale of such avoidance is that the 'pop25' instruction also
contains return behavior and we have no chance to further adjust stack
pointer before returning to caller, which is very important to
variadic function.
-- So we do not setup v3push/v3pop (push25/pop25) information for a
variadic function. It will use normal stack push/pop instead.
* config/nds32/nds32.c (nds32_compute_stack_frame): Do not use
v3push/v3pop for variadic function.
* config/nds32/nds32.md (prologue, epilogue): Likewise.
From-SVN: r214857
-- In fact, we only need to take care of 'push.s' situation,
because we have to push all the nameless arguments into stack.
-- We do not have to use 'pop.s' to restore them back to registers,
because we can just adjust stack pointer to set frame location.
This operation should be done in epilogue expanding.
-- Because 'pop25' instruction also performs return operation, we will
not use v3push/v3pop on isr and vararg functions.
Therefore, their instruction length should be 4 bytes.
* config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push):
Check rtx for varargs implementation.
(nds32_output_stack_pop): Likewise.
* config/nds32/nds32-protos.h: Have a rtx argument for
nds32_output_stack_push and nds32_output_stack_pop.
* config/nds32/nds32.md: Likewise.
From-SVN: r214856
* config/nds32/nds32-isr.c (nds32_isr_function_p): Define new function
to check if FUNC is an interrupt service routine.
* config/nds32/nds32-protos.h (nds32_isr_function_p): Declaration.
From-SVN: r214855
-- The varargs area may also need padding bytes so that we can ensure
that argument pointer is 8-byte alignment.
* config/nds32/nds32.h (machine_function): Add some fields for variadic
arguments implementation.
From-SVN: r214854
* config/nds32/nds32.c (nds32_gen_stack_v3push): Rename to ...
(nds32_emit_stack_v3push): ... this.
(nds32_gen_stack_v3pop): Rename to ...
(nds32_emit_stack_v3pop): ... this and consider CFA restore
information.
From-SVN: r214852
* config/nds32/nds32.c (nds32_gen_stack_push_multiple): Rename to ...
(nds32_emit_stack_push_multiple): ... this.
(nds32_gen_stack_pop_multiple): Rename to ...
(nds32_emit_stack_pop_multiple): ... this and consider CFA restore
information.
From-SVN: r214851
2014-09-03 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
PR target/61078
* config/s390/s390.md ("*negdi2_31"): Add s390_split_ok_p check
and add a second splitter to handle the remaining cases.
2014-09-03 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
PR target/61078
* gcc.target/s390/pr61078.c: New testcase.
From-SVN: r214850
Bill Schmidt
Jan-Benedict Glaw
Markus Trippelsdorf
Alan Modra
Matthew Fortune
GCC Administrator
Chris Manghane
Marek Polacek
Fritz Reese
Richard Sandiford
Richard Henderson
Richard Biener
Martin Jambor
Martin Liska
Kaz Kojima
Chung-Ju Wu
Ilya Tocar
Andreas Krebbel
Uros Bizjak
Max Ostapenko
Tobias Burnus
Trevor Saunders
Nathan Sidwell