* config/i386/i386.md (*movti_internal): Substitute Ye constraint
with Yd constraint. Set "preferred_for_speed" attribute from
TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
with Yd constraint.
(*movdi_internal): Ditto.
(movti_interunit splitters): Remove
TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
(movdi_interunit splitters): Ditto.
* config/i386/constraints.md (Ye): Remove.
(Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
From-SVN: r259701
PR tree-optimization/85529
* tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
zero extension or masking of the MSB bit.
(optimize_range_tests): Add FIRST_BB argument, pass it through
to optimize_range_tests_var_bound.
(maybe_optimize_range_tests, reassociate_bb): Adjust
optimize_range_tests callers.
* gcc.c-torture/execute/pr85529-1.c: New test.
* gcc.c-torture/execute/pr85529-2.c: New test.
* gcc.dg/pr85529.c: New test.
From-SVN: r259696
With the soft-float n64 ABI and the data-sym-pool.c test case code like
below is produced:
.file 1 "data-sym-pool.c"
.section .mdebug.abi64
.previous
.nan legacy
.module softfloat
.module oddspreg
.abicalls
.option pic0
.text
.align 2
.globl frob
.set mips16
.set nomicromips
.ent frob
.type frob, @function
frob:
.frame $17,16,$31 # vars= 0, regs= 1/0, args= 0, gp= 0
.mask 0x00020000,-8
.fmask 0x00000000,0
daddiu $sp,-16
sd $17,8($sp)
move $17,$sp
ld $2,.L3
move $sp,$17
ld $17,8($sp)
daddiu $sp,16
jr $31
.type __pool_frob_3, @object
__pool_frob_3:
.align 3
.L3:
.dword 305419896
.type __pend_frob_3, @function
__pend_frob_3:
.insn
.end frob
.size frob, .-frob
.ident "GCC: (GNU) 8.0.1 20180410 (experimental)"
(we have no support for hard-float n64 MIPS16 code generation), which
means that the test case will fail, as the regular expression pattern
expects `lw' and `.word' rather than `ld' and `.dword' respectively to
appear in assembly code generation. Correct the pattern in an obvious
way then making it accept both intructions and pseudo-ops.
gcc/testsuite/
* gcc.target/mips/data-sym-pool.c (dg-options): Match `ld' and
`.dword' in addition to `lw' and `.word'.
From-SVN: r259691
With GCC configurations using the SVR4 rather than the PLT dynamic
executable model and the o32 ABI with the data-sym-pool.c test case code
like below is produced:
.file 1 "data-sym-pool.c"
.section .mdebug.abi32
.previous
.nan legacy
.module fp=xx
.module nooddspreg
.abicalls
.text
.align 2
.globl frob
.set mips16
.set nomicromips
.ent frob
.type frob, @function
frob:
.frame $17,8,$31 # vars= 0, regs= 1/0, args= 0, gp= 0
.mask 0x00020000,-4
.fmask 0x00000000,0
save 8,$17
move $17,$sp
lw $2,$L4
move $sp,$17
restore 8,$17
jr $31
.type __pool_frob_3, @object
__pool_frob_3:
.align 2
$L3:
.word __gnu_local_gp
$L4:
.word 305419896
.type __pend_frob_3, @function
__pend_frob_3:
.insn
.end frob
.size frob, .-frob
.ident "GCC: (GNU) 8.0.1 20180410 (experimental)"
causing a failure due to the unexpected `__gnu_local_gp' entry in the
constant pool, even though there is nothing wrong with it as far as the
annotation being examined is concerned.
Given that the SVR4 vs PLT code model consideration is irrelevant for
this test case rather than rewriting the regular expression to match
this variant of code just enforce the PLT model by using the `-mplt'
option. It is safe to use this option unconditionally as it is silently
ignored with configurations that do not support this model, e.g. bare
metal ELF.
gcc/testsuite/
* gcc.target/mips/data-sym-pool.c (dg-options): Add `-mplt'.
From-SVN: r259690
2018-04-26 Tom de Vries <tom@codesourcery.com>
PR target/84952
* config/nvptx/nvptx.c (verify_neutering_jumps)
(verify_neutering_labels): New function
(nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
From-SVN: r259677
2018-04-26 Tom de Vries <tom@codesourcery.com>
PR target/84025
* config/nvptx/nvptx.c (needs_neutering_p): New function.
(nvptx_single): Use needs_neutering_p to skip over insns that do not
need neutering.
From-SVN: r259676
2018-04-26 Richard Biener <rguenther@suse.de>
Tom de Vries <tom@codesourcery.com>
PR lto/85422
* lto-streamer-out.c (output_function): Fixup loops if required to match
discovery done in the reader.
* testsuite/libgomp.oacc-c-c++-common/pr85422.c: New test.
Co-Authored-By: Tom de Vries <tom@codesourcery.com>
From-SVN: r259675
2018-04-26 Richard Biener <rguenther@suse.de>
PR tree-optimization/85116
* tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
have a loop exit from the single latch predecessor. Remove
case of header with just condition.
(ch_base::copy_headers): Exclude infinite loops from any
processing.
(pass_ch::execute): Record exits.
* gcc.dg/tree-ssa/copy-headers-2.c: New testcase.
* gcc.dg/tree-ssa/copy-headers-3.c: Likewise.
* gcc.dg/tree-ssa/copy-headers-4.c: Likewise.
* gcc.dg/tree-ssa/loadpre6.c: Adjust.
From-SVN: r259672
2018-04-26 Richard Biener <rguenther@suse.de>
* tree-vect-data-refs.c (vect_get_data_access_cost): Get
prologue cost vector and pass it to vect_get_load_cost.
(vect_get_peeling_costs_all_drs): Likewise.
(vect_peeling_hash_get_lowest_cost): Likewise.
(vect_enhance_data_refs_alignment): Likewise.
From-SVN: r259668
* config/mips/mips.c (mips_asan_shadow_offset): New function.
(TARGET_ASAN_SHADOW_OFFSET): Define.
* config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
true for -fsanitize=address.
Co-Authored-By: Jean Lee <xiaoyur347@gmail.com>
From-SVN: r259666
If someone has access to a 64-bit mips-linux system to test
this (with the obvious edit), that'd be really nice. Until
then, best to not introduce possible build failures.
* configure.tgt <mips*-*-linux*>: Enable build, excluding
mips*64*-*-linux*.
From-SVN: r259665
As mentioned in <http://gcc.gnu.org/ml/gcc/2018-03/msg00133.html>
the bogus adjustment to 160 from 144 (which is reverted here),
is a single-token commit in upstream r301307, an attempt to
correct a failed build due to an upstream change to compile the
runtime with D_FILE_OFFSET_BITS=64. The correct fix is here:
just use the right include. Yes, user-struct-stat64-as-stat is
actually 160 for MIPS o32 and I hear user-struct-stat is also
160 for n32. There are additional fields appended for
user-struct-stat! I guess for MIPS it's as bad as it gets for
mixing up kernel and user struct stat. The context of the patch
doesn't show that in the #else there's the correct include, the
one for <asm/stat.h> to get the kernel-struct-stat. If you
can't compile it, IMHO the kernel headers are just too old; 3.2
is fine for example.
* sanitizer_common/sanitizer_platform_limits_linux.cc: Do not
take the shortcut to #include <sys/stat.h> for MIPS instead of
the kernel <asm/stat.h>. Explain why sys/stat.h is misleading
or wrong to get the kernel struct stat.
* sanitizer_common/sanitizer_platform_limits_posix.h [__mips__]:
Correct the value for 32-bit non-android struct_kernel_stat_sz.
From-SVN: r259664
This appears to be present in compiler-rt upstream, but as part
of more intrusive changes. For gcc, the lack of this results in
a fatal warning (-Werror) at build-time.
* sanitizer_common/sanitizer_atomic_clang_other.h [_MIPS_SIM
&& _MIPS_SIM == _ABIO32] (lock): Add initializer for .pad member.
From-SVN: r259663
When gcc dwarf2out generates the .debug_line table itself (for example
when generating one for a split DWARF .dwo) it uses natural sorting for
the directory table. Shorter directory paths come before longer directory
paths with the same prefix. This causes the files in the line table to
pick the shorter dir. Creating slightly ineffecient line tables because
the longer directory paths will never be used.
Fix this by changing file_info_cmp () to pick longer directory prefixes
before shorter ones. We still sort files (the compilation unit) without
any directory path before all entries with a directory path, so they
will still use dir entry 0 (the working directory).
A hello.c program would get the following dir and line table before:
Directory table:
/opt/local/install/gcc/lib/gcc/x86_64-pc-linux-gnu/8.0.1/include
/usr/include
/usr/include/bits
File name table:
Entry Dir Time Size Name
1 0 0 0 hello.c
2 1 0 0 stddef.h
3 2 0 0 bits/types.h
4 2 0 0 libio.h
5 2 0 0 stdio.h
6 2 0 0 bits/sys_errlist.h
Note that the last directory table entry is never used.
After this patch it looks as follows:
Directory table:
/opt/local/install/gcc/lib/gcc/x86_64-pc-linux-gnu/8.0.1/include
/usr/include/bits
/usr/include
File name table:
Entry Dir Time Size Name
1 0 0 0 hello.c
2 1 0 0 stddef.h
3 2 0 0 types.h
4 3 0 0 libio.h
5 3 0 0 stdio.h
6 2 0 0 sys_errlist.h
Which is similar to what gas would output.
gcc/ChangeLog:
* dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
shorter ones.
From-SVN: r259655
* config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
than "alu", remove explicit "memory" and "imm_disp" attributes.
(*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
From-SVN: r259650
PR middle-end/85414
* simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
gen_lowpart_no_emit.
From-SVN: r259649
PR sanitizer/84307
* c-decl.c (build_compound_literal): Call pushdecl (decl) even when
it is not TREE_STATIC.
* c-typeck.c (c_mark_addressable) <case COMPOUND_LITERAL_EXPR>: Mark
not just the COMPOUND_LITERAL_EXPR node itself addressable, but also
its COMPOUND_LITERAL_EXPR_DECL.
From-SVN: r259641
