Skip to content

Commit d9450aa

Browse files
author
GCC Administrator
committed
Daily bump.
1 parent 04a7455 commit d9450aa

File tree

10 files changed

+403
-1
lines changed

10 files changed

+403
-1
lines changed

‎ChangeLog

+4
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,7 @@
1+
2022-01-10 Martin Liska <[email protected]>
2+
3+
* MAINTAINERS: Fix obvious issues with sorting.
4+
15
2022-01-07 liuhongt <[email protected]>
26

37
* build.log: New file.

‎gcc/ChangeLog

+128
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,131 @@
1+
2022-01-10 Uroš Bizjak <[email protected]>
2+
3+
PR target/103861
4+
* config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
5+
Handle V2QImode.
6+
* config/i386/mmx.md (<sat_plusminus:insn><mode>3):
7+
Use VI1_16_32 mode iterator.
8+
(*eq<mode>3): Ditto.
9+
(*gt<mode>3): Ditto.
10+
(*xop_maskcmp<mode>3): Ditto.
11+
(*xop_maskcmp_uns<mode>3): Ditto.
12+
(vec_cmp<mode><mode>): Ditto.
13+
(vec_cmpu<mode><mode>): Ditto.
14+
15+
2022-01-10 Richard Biener <[email protected]>
16+
17+
PR tree-optimization/103948
18+
* tree-vect-generic.c (expand_vector_condition): Return true if
19+
all ones vector is returned for true, all zeros vector for false
20+
and the target defines corresponding vec_cmp{,u}MN named RTX pattern.
21+
22+
2022-01-10 Paul A. Clarke <[email protected]>
23+
24+
* config/rs6000/smmintrin.h (_mm_blendv_epi8): Use vec_blendv
25+
when _ARCH_PWR10. Use signed types.
26+
(_mm_blendv_ps): Use vec_blendv when _ARCH_PWR10.
27+
(_mm_blendv_pd): Likewise.
28+
29+
2022-01-10 Andre Vieira <[email protected]>
30+
31+
* tree-vectorizer.c (better_epilogue_loop_than_p): Round factors up for
32+
epilogue costing.
33+
* tree-vect-loop.c (vect_analyze_loop): Re-analyze all modes for
34+
epilogues, unless we are guaranteed that we can't have partial vectors.
35+
* genopinit.c: (partial_vectors_supported): Generate new function.
36+
37+
2022-01-10 Jakub Jelinek <[email protected]>
38+
39+
PR target/102024
40+
* config/i386/i386.c (classify_argument): Add zero_width_bitfields
41+
argument, when seeing DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD bitfields,
42+
always ignore them, when seeing other zero sized bitfields, either
43+
set zero_width_bitfields to 1 and ignore it or if equal to 2 process
44+
it. Pass it to recursive calls. Add wrapper
45+
with old arguments and diagnose ABI differences for C structures
46+
with zero width bitfields. Formatting fixes.
47+
48+
2022-01-10 Richard Sandiford <[email protected]>
49+
50+
PR rtl-optimization/98782
51+
* ira-int.h (ira_soft_conflict): Declare.
52+
* ira-color.c (max_soft_conflict_loop_depth): New constant.
53+
(ira_soft_conflict): New function.
54+
(spill_soft_conflicts): Likewise.
55+
(assign_hard_reg): Use them to handle the case described by
56+
the comment above ira_soft_conflict.
57+
(improve_allocation): Likewise.
58+
* ira.c (check_allocation): Allow allocnos with "soft" conflicts
59+
to share the same register.
60+
61+
2022-01-10 Richard Sandiford <[email protected]>
62+
63+
PR rtl-optimization/98782
64+
* ira-int.h (ira_caller_save_cost): New function.
65+
(ira_caller_save_loop_spill_p): Likewise.
66+
* ira-build.c (ira_propagate_hard_reg_costs): Test whether it is
67+
cheaper to spill a call-clobbered register throughout a loop rather
68+
than spill it around each individual call. If so, treat all
69+
call-clobbered registers as conflicts and...
70+
(propagate_allocno_info): ...do not propagate call information
71+
from the child to the parent.
72+
* ira-color.c (move_spill_restore): Update accordingly.
73+
* ira-costs.c (ira_tune_allocno_costs): Use ira_caller_save_cost.
74+
75+
2022-01-10 Richard Sandiford <[email protected]>
76+
77+
PR rtl-optimization/98782
78+
* ira-int.h (ira_allocno::might_conflict_with_parent_p): New field.
79+
(ALLOCNO_MIGHT_CONFLICT_WITH_PARENT_P): New macro.
80+
(ira_single_region_allocno_p): New function.
81+
(ira_total_conflict_hard_regs): Likewise.
82+
* ira-build.c (ira_create_allocno): Initialize
83+
ALLOCNO_MIGHT_CONFLICT_WITH_PARENT_P.
84+
(ira_propagate_hard_reg_costs): New function.
85+
(propagate_allocno_info): Use it. Try to avoid propagating
86+
hard register conflicts to parent allocnos if we can handle
87+
the conflicts by spilling instead. Limit the propagated
88+
register costs to the cost of spilling throughout the child loop.
89+
* ira-color.c (color_pass): Use ira_single_region_allocno_p to
90+
test whether a child and parent allocno can share the same
91+
register.
92+
(move_spill_restore): Adjust for the new behavior of
93+
propagate_allocno_info.
94+
95+
2022-01-10 Richard Sandiford <[email protected]>
96+
97+
PR rtl-optimization/98782
98+
* ira-int.h (ira_subloop_allocnos_can_differ_p): New function,
99+
extracted from...
100+
* ira-color.c (color_pass): ...here.
101+
102+
2022-01-10 Richard Sandiford <[email protected]>
103+
104+
PR rtl-optimization/98782
105+
* ira-color.c (color_pass): Add comments to describe the spill costs.
106+
(move_spill_restore): Likewise. Fix reversed calculation.
107+
108+
2022-01-10 Richard Sandiford <[email protected]>
109+
110+
PR rtl-optimization/98782
111+
* ira-int.h (ira_loop_border_costs): New class.
112+
* ira-color.c (ira_loop_border_costs::ira_loop_border_costs):
113+
New constructor.
114+
(calculate_allocno_spill_cost): Use ira_loop_border_costs.
115+
(color_pass): Likewise.
116+
(move_spill_restore): Likewise.
117+
118+
2022-01-10 Eric Botcazou <[email protected]>
119+
120+
PR target/103465
121+
* coretypes.h (unwind_info_type): Swap UI_SEH and UI_TARGET.
122+
123+
2022-01-10 Richard Biener <[email protected]>
124+
125+
PR tree-optimization/100359
126+
* tree-ssa-loop-ivcanon.c (try_unroll_loop_completely):
127+
Allow non-growing peeling with !allow_peel and UL_ALL.
128+
1129
2022-01-08 Roger Sayle <[email protected]>
2130

3131
* config/i386/i386-expand.c (ix86_expand_vector_move): Add

‎gcc/DATESTAMP

+1-1
Original file line numberDiff line numberDiff line change
@@ -1 +1 @@
1-
20220110
1+
20220111

‎gcc/ada/ChangeLog

+73
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,76 @@
1+
2022-01-10 Eric Botcazou <[email protected]>
2+
3+
* gcc-interface/trans.c (Identifier_to_gnu): Use correct subtype.
4+
(elaborate_profile): New function.
5+
(Call_to_gnu): Call it on the formals and the result type before
6+
retrieving the translated result type from the subprogram type.
7+
8+
2022-01-10 Eric Botcazou <[email protected]>
9+
10+
* gcc-interface/decl.c (gnat_to_gnu_entity) <E_Record_Type>: Fix
11+
computation of boolean result in the unchecked union case.
12+
(components_to_record): Rename MAYBE_UNUSED parameter to IN_VARIANT
13+
and remove local variable of the same name. Pass NULL recursively
14+
as P_GNU_REP_LIST for nested variants in the unchecked union case.
15+
16+
2022-01-10 Eric Botcazou <[email protected]>
17+
18+
* gcc-interface/trans.c (lvalue_required_p) <N_Pragma>: New case.
19+
<N_Pragma_Argument_Association>: Likewise.
20+
(Pragma_to_gnu) <Pragma_Inspection_Point>: Fetch the corresponding
21+
variable of a constant before marking it as addressable.
22+
23+
2022-01-10 Arnaud Charlet <[email protected]>
24+
25+
* gcc-interface/Make-lang.in (ADA_GENERATED_FILES): Remove
26+
s-casuti.ad?, s-crtl.ad?, s-os_lib.ad?. Update list of object
27+
files accordingly.
28+
29+
2022-01-10 Piotr Trojanek <[email protected]>
30+
31+
* libgnat/s-atopri.ads (Atomic_Compare_Exchange): Replaces
32+
deprecated Sync_Compare_And_Swap.
33+
* libgnat/s-atopri.adb (Lock_Free_Try_Write): Switch from __sync
34+
to __atomic builtins.
35+
36+
2022-01-10 Piotr Trojanek <[email protected]>
37+
38+
* libgnat/s-exponn.adb, libgnat/s-expont.adb,
39+
libgnat/s-exponu.adb, libgnat/s-widthi.adb,
40+
libgnat/s-widthu.adb: Remove CodePeer annotations for pragma
41+
Loop_Variant.
42+
43+
2022-01-10 Piotr Trojanek <[email protected]>
44+
45+
* exp_prag.adb (Expand_Pragma_Loop_Variant): Disable expansion
46+
in CodePeer mode.
47+
48+
2022-01-10 Piotr Trojanek <[email protected]>
49+
50+
* sem_util.adb (Is_Child_Or_Sibling): Fix typo in comment.
51+
52+
2022-01-10 Eric Botcazou <[email protected]>
53+
54+
* exp_pakd.adb (Install_PAT): If the PAT is a scalar type, apply
55+
the canonical adjustment to its alignment.
56+
57+
2022-01-10 Piotr Trojanek <[email protected]>
58+
59+
* libgnat/s-atocou__builtin.adb (Decrement, Increment): Switch
60+
from __sync to __atomic builtins; use 'Address to be consistent
61+
with System.Atomic_Primitives.
62+
63+
2022-01-10 Eric Botcazou <[email protected]>
64+
65+
* exp_pakd.adb (Install_PAT): Do not reset the alignment here.
66+
* layout.adb (Layout_Type): Call Adjust_Esize_Alignment after having
67+
copied the RM_Size onto the Esize when the latter is too small.
68+
69+
2022-01-10 Justin Squirek <[email protected]>
70+
71+
* sem_warn.adb (Check_References): Handle arrays of tasks
72+
similar to task objects.
73+
174
2022-01-07 Patrick Bernardi <[email protected]>
275

376
* libgnat/a-direct.adb (Search_Data): Remove type.

‎gcc/c-family/ChangeLog

+6
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,9 @@
1+
2022-01-10 Richard Biener <[email protected]>
2+
3+
PR middle-end/101530
4+
* c-common.c (c_build_shufflevector): Wrap the BIT_FIELD_REF
5+
in a TARGET_EXPR to force a temporary.
6+
17
2022-01-06 Marek Polacek <[email protected]>
28

39
PR c++/103758

‎gcc/cp/ChangeLog

+23
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,26 @@
1+
2022-01-10 Patrick Palka <[email protected]>
2+
3+
PR c++/103879
4+
* constexpr.c (cxx_fold_indirect_ref): Split out object/offset
5+
canonicalization step into a local lambda. Strengthen it to
6+
absorb more components at position 0. Use it before both calls
7+
to cxx_fold_indirect_ref_1.
8+
9+
2022-01-10 Patrick Palka <[email protected]>
10+
11+
PR c++/103783
12+
* call.c (cand_parms_match): Skip over 'this' when given one
13+
static and one non-static member function.
14+
15+
2022-01-10 Jakub Jelinek <[email protected]>
16+
17+
PR c++/103912
18+
* semantics.c (expand_or_defer_fn): For immediate functions, set
19+
node->body_removed to true and clear analyzed, definition and
20+
force_output.
21+
* decl2.c (c_parse_final_cleanups): Ignore immediate functions for
22+
expand_or_defer_fn.
23+
124
2022-01-08 Jakub Jelinek <[email protected]>
225

326
PR c++/89074

‎gcc/fortran/ChangeLog

+6
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,9 @@
1+
2022-01-10 Paul Thomas <[email protected]>
2+
3+
PR fortran/103366
4+
* trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Allow unlimited
5+
polymorphic actual argument passed to assumed type formal.
6+
17
2022-01-09 Harald Anlauf <[email protected]>
28

39
PR fortran/103777

‎gcc/testsuite/ChangeLog

+69
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,72 @@
1+
2022-01-10 Uroš Bizjak <[email protected]>
2+
3+
PR target/103861
4+
* gcc.target/i386/pr103861-2.c: New test.
5+
6+
2022-01-10 Patrick Palka <[email protected]>
7+
8+
PR c++/103879
9+
* g++.dg/cpp1y/constexpr-base2.C: New test.
10+
* g++.dg/cpp1y/constexpr-base2a.C: New test.
11+
12+
2022-01-10 Patrick Palka <[email protected]>
13+
14+
PR c++/103783
15+
* g++.dg/cpp2a/concepts-memfun2.C: New test.
16+
17+
2022-01-10 Jakub Jelinek <[email protected]>
18+
19+
PR c++/103912
20+
* g++.dg/cpp2a/consteval26.C: New test.
21+
22+
2022-01-10 Paul A. Clarke <[email protected]>
23+
24+
* gcc.target/powerpc/sse4_1-blendvps.c: Copy from gcc.target/i386,
25+
adjust dg directives to suit.
26+
27+
2022-01-10 Andre Vieira <[email protected]>
28+
29+
* gcc.target/aarch64/masked_epilogue.c: New test.
30+
31+
2022-01-10 Paul Thomas <[email protected]>
32+
33+
PR fortran/103366
34+
* gfortran.dg/pr103366.f90: New test.
35+
36+
2022-01-10 Jakub Jelinek <[email protected]>
37+
38+
PR target/102024
39+
* gcc.target/i386/pr102024.c: New test.
40+
* g++.target/i386/pr102024.C: New test.
41+
42+
2022-01-10 Richard Sandiford <[email protected]>
43+
44+
* gcc.target/aarch64/reg-alloc-4.c: New test.
45+
46+
2022-01-10 Richard Sandiford <[email protected]>
47+
48+
* gcc.target/aarch64/reg-alloc-3.c: New test.
49+
50+
2022-01-10 Richard Sandiford <[email protected]>
51+
52+
* gcc.target/aarch64/reg-alloc-2.c: New test.
53+
54+
2022-01-10 Francois-Xavier Coudert <[email protected]>
55+
56+
PR fortran/82207
57+
* gfortran.dg/ieee/signaling_1.f90: New test.
58+
* gfortran.dg/ieee/signaling_1_c.c: New file.
59+
60+
2022-01-10 Richard Biener <[email protected]>
61+
62+
PR middle-end/101530
63+
* c-c++-common/builtin-shufflevector-3.c: New testcase.
64+
65+
2022-01-10 Richard Biener <[email protected]>
66+
67+
PR tree-optimization/100359
68+
* gcc.dg/tree-ssa/pr100359.c: New testcase.
69+
170
2022-01-09 Harald Anlauf <[email protected]>
271

372
PR fortran/103777

‎libgfortran/ChangeLog

+8
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,11 @@
1+
2022-01-10 Francois-Xavier Coudert <[email protected]>
2+
3+
PR fortran/82207
4+
* Makefile.am: Pass -fsignaling-nans for IEEE files.
5+
* Makefile.in: Regenerate.
6+
* ieee/ieee_helper.c: Use issignaling macro to recognized
7+
signaling NaNs.
8+
19

210
Copyright (C) 2022 Free Software Foundation, Inc.
311

0 commit comments

Comments
 (0)