Alexandre Oliva [Tue, 3 Nov 2015 03:57:07 +0000 (03:57 +0000)]
defer mark_addressable calls during expand till the end of expand
for gcc/ChangeLog
* gimple-expr.c: Include hash-set.h and rtl.h.
(mark_addressable_queue): New var.
(mark_addressable): Factor actual marking into...
(mark_addressable_1): ... this. Queue it up during expand.
(mark_addressable_2): New.
(flush_mark_addressable_queue): New.
* gimple-expr.h (flush_mark_addressable_queue): Declare.
* cfgexpand.c: Include gimple-expr.h.
(pass_expand::execute): Flush mark_addressable queue.
Alexandre Oliva [Tue, 3 Nov 2015 00:30:07 +0000 (00:30 +0000)]
[PR68083] don't introduce undefined behavior in ifcombine
The ifcombine pass may move a conditional access to an uninitialized
value before the condition that ensures it is always well-defined,
thus introducing undefined behavior. Stop it from doing so.
for gcc/ChangeLog
PR tree-optimization/68083
* tree-ssa-ifcombine.c: Include tree-ssa.h.
(bb_no_side_effects_p): Test for undefined uses too.
* tree-ssa.c (gimple_uses_undefined_value_p): New.
* tree-ssa.h (gimple_uses_undefined_value_p): Declare.
for gcc/testsuite/ChangeLog
PR tree-optimization/68083
* gcc.dg/torture/pr68083.c: New. From Zhendong Su.
David Malcolm [Mon, 2 Nov 2015 19:47:03 +0000 (19:47 +0000)]
libcpp: Improvements to description of source_location in line-map.h
libcpp/ChangeLog:
* include/line-map.h (source_location): In the table in the
descriptive comment, show UNKNOWN_LOCATION, BUILTINS_LOCATION,
LINE_MAP_MAX_LOCATION_WITH_COLS, LINE_MAP_MAX_SOURCE_LOCATION.
Add notes about ad-hoc values.
Move constant folds for maths functions to new file
The new routines operate on the built-in enum rather than on tree decls.
The idea is to extend this to handle internal functions too, with a
combined enum for both.
The patch also moves fold_fma too, with the same prototype. The long-term
plan is to replace FMA_EXPR with an internal function, for consistency
with the way that things like SQRT will be handled.
Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.
gcc/
* builtins.h (fold_fma): Move to fold-const-call.h.
* builtins.c: Include fold-const-call.h.
(mathfn_built_in_2): New function, split out from...
(mathfn_built_in_1): ...here.
(do_real_to_int_conversion, fold_const_builtin_pow)
(fold_const_builtin_logb, fold_const_builtin_significand)
(fold_const_builtin_load_exponent, do_mpfr_arg1, do_mpfr_arg2)
(do_mpfr_arg3, do_mpfr_sincos, do_mpfr_bessel_n, do_mpc_arg1): Delete.
(fold_builtin_sincos): Use fold_const_call to handle constants.
(fold_builtin_1, fold_builtin_2, fold_builtin_3): Add explicit
checks for ERROR_MARK. Use fold_const_call to handle constant
folds for math functions.
(fold_fma): Move to fold-const-call.c.
* fold-const.c: Include fold-const-call.h.
* Makefile.in (OBJS): Add fold-const-call.o.
(PLUGIN_HEADERS): Add fold-const-call.h.
* realmpfr.h (real_from_mpfr): Allow the format to be specified
directly.
* realmpfr.c (real_from_mpfr): Likewise.
* fold-const-call.h, fold-const-call.c: New files.
re PR go/67968 (go1: internal compiler error: in write_specific_type_functions, at go/gofrontend/types.cc:1812)
PR go/67968
compiler: Traverse types of call expressions.
https://gcc.gnu.org/PR67968 provides a test case that causes a gccgo
crash on valid code. The compiler failed to build the hash and equality
functions required for a type descriptor. The descriptor is for an
unnamed type that is being returned by a function imported from a
different package. The unnamed type is being implicitly converted to an
interface type by a return statement. The fix is to ensure that the
type of a call expression is always traversed.
Test case sent out for the master testsuite as
https://golang.org/cl/16532 .
Ian Lance Taylor [Sat, 31 Oct 2015 23:48:19 +0000 (23:48 +0000)]
runtime: If no split stacks, allocate stacks using mmap on 64-bit systems.
When not using split stacks, libgo allocate large stacks for each
goroutine. On a 64-bit system, libgo allocates a maximum of 128G for
the Go heap, and allocates 4M for each stack. When the stacks are
allocated from the Go heap, the result is that a program can only create
32K goroutines, which is not enough for an active Go server. This patch
changes libgo to allocate the stacks using mmap directly, rather than
allocating them out of the Go heap. This change is only done for 64-bit
systems when not using split stacks. When using split stacks, the
stacks are allocated using mmap directly anyhow. On a 32-bit system,
there is no maximum size for the Go heap, or, rather, the maximum size
is the available address space anyhow.
This one for TARGET_TOC && TARGET_MINIMAL_TOC. Without it, r30 is not
saved in the prologue for functions that do not use r30, but the register
is set later in the prologue anyhow. This made all java tests fail.
* config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Use the
same condition for testing whether RS6000_PIC_OFFSET_TABLE_REGNUM is
live as for using it elsewhere, for TARGET_MINIMAL_TOC.
Jason Merrill [Sat, 31 Oct 2015 16:19:55 +0000 (12:19 -0400)]
Streamline for_each_template_parm.
* pt.c (for_each_template_parm_r): Use WALK_SUBTREE.
Return a meaningful value rather than error_mark_node.
(for_each_template_parm): Return a tree.
(uses_template_parms_level): Return bool.
* cp-tree.h: Adjust.
Tom de Vries [Sat, 31 Oct 2015 08:08:39 +0000 (08:08 +0000)]
Add initial constraints in create_function_info_for
2015-10-31 Tom de Vries <tom@codesourcery.com>
* tree-ssa-structalias.c (ipa_pta_execute): Add extra arg to call to
create_function_info_for. Dump constraints generated during
create_function_info_for. Move intra_create_variable_infos call and
function-return-values-escape bit to ...
(create_function_info_for): ... here, and merge
intra_create_variable_infos call with argument loop. Add and handle
nonlocal_p parameter.
Tom de Vries [Sat, 31 Oct 2015 08:08:28 +0000 (08:08 +0000)]
Improve readability and structure of create_function_info_for
2015-10-31 Tom de Vries <tom@codesourcery.com>
* tree-ssa-structalias.c (create_function_info_for): Make sure prev_vi
updating is alap, and seperated from preceding code. Make sure
insert_vi_for_tree is seperated from surrounding code.
gfortran.h (gfc_omp_namespace): Add locus where member.
gcc/fortran/
* gfortran.h (gfc_omp_namespace): Add locus where member.
* openmp.c (gfc_match_omp_variable_list): Set where for each list
item found.
(resolve_omp_clauses): Remove where argument and use the where
gfc_omp_namespace member when reporting errors.
(resolve_omp_do): Update call to resolve_omp_clauses.
(resolve_oacc_loop): Likewise.
(gfc_resolve_oacc_directive): Likewise.
(gfc_resolve_omp_directive): Likewise.
(gfc_resolve_omp_declare_simd): Likewise.
Steven G. Kargl [Fri, 30 Oct 2015 18:13:50 +0000 (18:13 +0000)]
re PR fortran/51993 (Erroneous type component initialization leads to internal compiler error)
2015-10-15 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/51993
* decl.c (gfc_set_constant_character_len): Convert gcc_assert into an
if-statement causing an early return leads to valid error message.
2015-10-15 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/51993
* gfortran.dg/pr51993.f90: New test.
sincos has always freed dominators at the end, but AFAICT they should
still be up-to-date. (In particular, gimple_purge_dead_eh_edges
updates the information.)
Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.
Allow real_format to be passed to more real.h functions
Most real.h routines used machine modes to specify the format of
an operation and converted that to a float_format * internally.
Some also had alternative versions that accepted a float_format *.
In an upcoming patch it seemed more convenient for the callers
I was adding to use float_format directly, since the callers need
to examine the format themselves for other reasons. This patch
therefore replaces the machine_mode arguments with a new class that
allows both machine modes and float_format pointers to be used.
Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.
fixed-value.c was passing a fixed-point mode to the floating-point
real_2expN routine. That didn't cause a problem in practice because
all real_2expN did was check for decimal float modes, but it triggered
a failure with an upcoming patch.
Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.
Make sure that REAL_MODE_FORMAT aborts if it is passed an invalid mode,
rather than stepping beyond the bounds of an array. It turned out that
some code was passing non-float modes to the real.h routines.
Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.
gcc/
* real.h (REAL_MODE_FORMAT): Abort if the mode isn't a
SCALAR_FLOAT_MODE_P.
Richard Biener [Fri, 30 Oct 2015 12:18:34 +0000 (12:18 +0000)]
gimple-fold.c (fold_gimple_assign): Do not dispatch to fold () on single RHSs.
2015-10-30 Richard Biener <rguenther@suse.de>
* gimple-fold.c (fold_gimple_assign): Do not dispatch to
fold () on single RHSs. Allow CONSTRUCTORS with trailing
zeros to be folded to VECTOR_CSTs.
* tree.c (build_vector_from_ctor): Handle VECTOR_CST elements.
* fold-const.c (fold): Use build_vector_from_ctor.
* config/aarch64/aarch64.md (*movhf_aarch64): Change the type of
"mov %0.h[0], %1.h[0] to "neon_move".
(*movtf_aarch64): Change the type of "fmov %s0, wzr" to "f_mcr".
(*cmov<mode>_insn): Change the types of "mov %<w>0, {-1,1}" to
"mov_imm".
(*cmovsi_insn_uxtw): Likewise.
Tom de Vries [Fri, 30 Oct 2015 08:38:24 +0000 (08:38 +0000)]
Cleanup fipa-pta constraint dumping
2015-10-30 Tom de Vries <tom@codesourcery.com>
* tree-ssa-structalias.c (ipa_pta_execute): Declare variable from as
unsigned, and initialize, and use initial value instead of hardcoded
constant. Add generic constraints dumping section. Don't dump global
initializers constraints dumping section if empty. Don't update
variable from if unused.
Jeff Law [Thu, 29 Oct 2015 20:37:01 +0000 (14:37 -0600)]
[PATCH] Remove unnecessary invalidation support in threading
2015-10-29 Jeff Law <law@redhat.com>
* tree-ssa-scopedtables.h (const_and_copies): Remove invalidate
method.
* tree-ssa-scopedtables.h (const_and_copies::invalidate): Remove.
* tree-ssa-threadedge.c
(record_temporary_equivalences_from_stmts_at_dest): Remove
backedge_seen argument and associated code which invalidated
equivalences based on the value of that argument.
(thread_through_normal_block): Corresponding changes.
The PIC reg (r30) needs to be saved whenever the prologue sets it up,
not just if it is used elsewhere in the function. Without this patch
the prologue for such a function will modify r30 without saving it
first, leading to disaster back in its callers. This happened in the
testsuite for -m32 libgfortran and libstdc++, bootstrapped with -mlra,
many hundred times.
* config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Move this
function earlier in the file.
(first_reg_to_save): Use rs6000_reg_live_or_pic_offset_p instead of
df_regs_ever_live_p.
lra: Don't remove the scratch in (mem:BLK (scratch))
LRA wants to replace SCRATCH registers with real registers. It should
not do that with (mem:BLK (scratch)), which is special, not really a
scratch register.
...which was failing on *-elf. This test is making sure that we narrow
maths calls, e.g. sinh -> sinhf, but we can only do that if we know that
the target has a full c99 libm or if we have other proof that the function
we want to generate is available.
Tested on aarch64-none-elf (where it's skipped) and aarch64-linux-gnu
(where it isn't).
Michael Meissner [Thu, 29 Oct 2015 18:21:44 +0000 (18:21 +0000)]
rs6000.c (rs6000_init_libfuncs): Split libfunc setup into 3 functions...
[gcc]
2015-10-29 Michael Meissner <meissner@linux.vnet.ibm.com>
* config/rs6000/rs6000.c (rs6000_init_libfuncs): Split libfunc
setup into 3 functions: init_float128_ibm, init_float128_ieee, and
rs6000_init_libfuncs. If -mfloat128, add IFmode functions for all
of the traditional names that TFmode uses for handling IEEE
extended double. If -mfloat128, add KFmode functions for all of
the emulation functions. If -mabi=ieeelongdouble and -mfloat128,
make TFmode use the same emulation functions as KFmode.
(init_float128_ibm): Likewise.
(init_float128_ieee): Likewise.
(rs6000_generate_compare): For IEEE 128-bit floating point
comparisons, call the unordered comparison function instead of the
ordered comparison function.
(rs6000_expand_float128_convert): Deal with operands that are
memory operands. Restructure the code to use a switch statement on
the mode. Add support for TFmode defaulting to either IBM extended
double or IEEE 128-bit floating point. If the underlying types are
the same, use a move instead of a conversion function.
(TARGET_C_MODE_FOR_SUFFIX): Define 'q' and 'Q' as the suffix to
use for IEEE 128-bit floating point constants with -mfloat128.
(rs6000_c_mode_for_suffix): Likewise.
(TARGET_INVALID_BINARY_OP): Do not allow inter-mixing of IEEE
128-bit floating point with IBM extended double floating point.
(rs6000_invalid_binary_op): Likewise.
(rs6000_gen_le_vsx_permute): On little endian systems generate a
ROTATE insn instead of VEC_SELECT for IEEE 128-bit floating point
types that can go in vector registers.
(chain_contains_only_swaps): Properly swap IEEE 128-bit floating
point types that can go in vector registers on little endian
PowerPC systems.
(mark_swaps_for_removal): Likewise.
(rs6000_analyze_swaps): Likewise.
(rs6000_mangle_type): Use U10__float128 for IEEE 128-bit floating point.
* config/rs6000/rs6000.md (FLOAT128_SFDFTF): Delete iterator,
rework IEEE 128-bit floating point insns to deal with TFmode being
either IBM extended double or IEEE 128-bit floating point.
(IFKF): Likewise.
(IBM128): Update iterator to add condition that the mode is IBM
extended double.
(IEEE128): New iterator for IEEE 128-bit floating point.
(TFIFKF): Rename TFIFKF iterator to FLOAT128.
(FLOAT128): Likewise.
(signbit<mode>2): FLOAT128_IBM_P condition test moved into IBM128
iterator.
(neg<mode>2): Replace TFIFKF iterator with FLOAT128. Add support
for TFmode being IEEE 128-bit floating point. Use IEEE128 iterator
instead of hard coding TFmode or KFmode.
(negtf2_internal): Likewise.
(neg<mode>2_internal): Likewise.
(abs<mode>2): Likewise.
(abstf2_internal): Likewise.
(abs<mode>2_internal): Likewise.
(ieee_128bit_neg<mode>2): Likewise.
(ieee_128bit_neg<mode>2_internal): Likewise.
(ieee_128bit_abs<mode>2): Likewise.
(ieee_128bit_abs<mode>2_internal): Likewise.
(ieee_128bit_nabs<mode>2): Likewise.
(ieee_128bit_nabs<mode>2_internal): Likewise.
(extendiftf2): Add explicit conversions between 128-bit floating
point types. Drop the old conversions that had become unwieldy.
(extend<FLOAT128_SFDFTF:mode><IFKF:mode>2): Likewise.
(extendifkf2): Likewise.
(trunc<IFKF:mode><FLOAT128_SFDFTF:mode>2): Likewise.
(extendtfkf2): Likewise.
(fix_trunc<IFKF:mode><SDI:mode>2): Likewise.
(trunciftf2): Likewise.
(fixuns_trunc<IFKF:mode><SDI:mode>2): Likewise.
(truncifkf2): Likewise.
(float<SDI:mode><IFKF:mode>2): Likewise.
(trunckftf2): Likewise.
(floatuns<SDI:mode><IFKF:mode>2): Likewise.
(trunctfif2): Likewise.
(FP iterator): Allow TFmode to be IEEE 128-bit floating point.
(extenddftf2): Rework 128-bit floating point conversions to
properly handle -mabi=ieeelongdouble. Merge IFmode, TFmode, and
KFmode expanders into one function.
(extenddf<mode>2): Likewise.
(extenddftf2_fprs): Likewise.
(extenddf<mode>2_fprs): Likewise.
(extenddftf2_vsx): Likewise.
(extenddf<mode>2_vsx): Likewise.
(extendsftf2): Likewise.
(extendsf<mode>2): Likewise.
(trunctfdf2): Likewise.
(trunc<mode>df2): Likewise.
(trunctfdf2_internal1): Likewise.
(trunc<mode>df2_internal1): Likewise.
(trunctfdf2_internal2): Likewise.
(trunc<mode>df2_internal2): Likewise.
(trunctfsf2): Likewise.
(trunc<mode>sf2): Likewise.
(trunctfsf2_fprs): Likewise.
(trunc<mode>sf2_fprs): Likewise.
(floatsit2f): Likewise.
(floatsi<mode>2): Likewise.
(fix_trunc_helper): Likewise.
(fix_trunc_helper<mode>): Likewise.
(fix_trunctfsi2): Likewise.
(fix_trunc<mode>si2): Likewise.
(fix_trunctfsi2_fprs): Likewise.
(fix_trunc<mode>si2_fprs): Likewise.
(fix_trunctfsi2_internal): Likewise.
(fix_trunc<mode>si2_internal): Likewise.
(fix_trunctfdi2): Likewise.
(fix_trunc<mode>di2): Likewise.
(fixuns_trunctf<mode>2): Likewise.
(fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
(floatditf2): Likewise.
(floatdi<mode>2): Likewise.
(floatuns<mode>tf2): Likewise.
(floatuns<SDI:mode><IEEE128:mode>): Likewise.
(cmptf_internal1): Use a mode iterator to add support for both
types (IFmode, TFmode) that support IBM extended double.
(cmp<mode>_internal1): Likewise.
(cmptf_internal2): Likewise.
(cmp<mode>_internal2): Likewise.
* doc/extend.texi (Floating Types): Document __ibm128 and
__float128 on PowerPC.
* doc/invoke.texi (RS/6000 and PowerPC Options): Document
-mfloat128 and -mno-float128.
[gcc/testsuite]
2015-10-29 Michael Meissner <meissner@linux.vnet.ibm.com>
* gcc.target/powerpc/float128-call.c: New test for -mfloat128 on
PowerPC.
* gcc.target/powerpc/float128-mix.c: Likewise.
Ian Lance Taylor [Thu, 29 Oct 2015 18:14:50 +0000 (18:14 +0000)]
compiler, reflect, runtime: remove zero field from type descriptor
Type descriptors picked up a zero field because the gc map
implementation used it. However, it's since been dropped by the gc
library. It was never used by gccgo. Drop it now in preparation for
upgrading to the Go 1.5 library.
projects / thirdparty / gcc.git / log
