Hi!
I've backported 13 commits of mine and one from Richard.
Bootstrapped/regtested on x86_64-linux and i686-linux, committed
to gcc-6-branch.
Jakub
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-06-30 Jakub Jelinek <ja...@redhat.com>
PR target/81225
* config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For
V8FI, V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
of nonimmediate_operand and <store_mask_constraint> instead of m for
the input operand. For V8FI iterator, always split if input is a MEM.
For V16FI and V8SF_256 iterators, don't test if both operands are MEM
if <mask_applied>. For VI4F_256 iterator, use <store_mask_predicate>
instead of register_operand and <store_mask_constraint> instead of v for
the input operand. Make sure both operands aren't MEMs for if not
<mask_applied>.
* gcc.target/i386/pr81225.c: New test.
--- gcc/config/i386/sse.md (revision 250284)
+++ gcc/config/i386/sse.md (revision 250285)
@@ -7230,12 +7230,13 @@ (define_insn "vec_extract_lo_<mode>_mask
(define_insn "vec_extract_lo_<mode><mask_name>"
[(set (match_operand:<ssehalfvecmode> 0 "<store_mask_predicate>"
"=<store_mask_constraint>,v")
(vec_select:<ssehalfvecmode>
- (match_operand:V8FI 1 "nonimmediate_operand" "v,m")
+ (match_operand:V8FI 1 "<store_mask_predicate>"
"v,<store_mask_constraint>")
(parallel [(const_int 0) (const_int 1)
(const_int 2) (const_int 3)])))]
- "TARGET_AVX512F && !(MEM_P (operands[0]) && MEM_P (operands[1]))"
+ "TARGET_AVX512F
+ && (<mask_applied> || !(MEM_P (operands[0]) && MEM_P (operands[1])))"
{
- if (<mask_applied> || !TARGET_AVX512VL)
+ if (<mask_applied> || (!TARGET_AVX512VL && !MEM_P (operands[1])))
return "vextract<shuffletype>64x4\t{$0x0, %1,
%0<mask_operand2>|%0<mask_operand2>, %1, 0x0}";
else
return "#";
@@ -7374,14 +7375,15 @@ (define_expand "avx_vextractf128<mode>"
(define_insn "vec_extract_lo_<mode><mask_name>"
[(set (match_operand:<ssehalfvecmode> 0 "nonimmediate_operand" "=v,m")
(vec_select:<ssehalfvecmode>
- (match_operand:V16FI 1 "nonimmediate_operand" "vm,v")
+ (match_operand:V16FI 1 "<store_mask_predicate>"
+ "<store_mask_constraint>,v")
(parallel [(const_int 0) (const_int 1)
(const_int 2) (const_int 3)
(const_int 4) (const_int 5)
(const_int 6) (const_int 7)])))]
"TARGET_AVX512F
&& <mask_mode512bit_condition>
- && !(MEM_P (operands[0]) && MEM_P (operands[1]))"
+ && (<mask_applied> || !(MEM_P (operands[0]) && MEM_P (operands[1])))"
{
if (<mask_applied>)
return "vextract<shuffletype>32x8\t{$0x0, %1,
%0<mask_operand2>|%0<mask_operand2>, %1, 0x0}";
@@ -7413,11 +7415,12 @@ (define_split
(define_insn "vec_extract_lo_<mode><mask_name>"
[(set (match_operand:<ssehalfvecmode> 0 "<store_mask_predicate>" "=v,m")
(vec_select:<ssehalfvecmode>
- (match_operand:VI8F_256 1 "nonimmediate_operand" "vm,v")
+ (match_operand:VI8F_256 1 "<store_mask_predicate>"
+ "<store_mask_constraint>,v")
(parallel [(const_int 0) (const_int 1)])))]
"TARGET_AVX
&& <mask_avx512vl_condition> && <mask_avx512dq_condition>
- && !(MEM_P (operands[0]) && MEM_P (operands[1]))"
+ && (<mask_applied> || !(MEM_P (operands[0]) && MEM_P (operands[1])))"
{
if (<mask_applied>)
return "vextract<shuffletype>64x2\t{$0x0, %1, %0%{%3%}|%0%{%3%}, %1, 0x0}";
@@ -7493,12 +7496,16 @@ (define_split
(define_insn "vec_extract_lo_<mode><mask_name>"
- [(set (match_operand:<ssehalfvecmode> 0 "<store_mask_predicate>"
"=<store_mask_constraint>")
+ [(set (match_operand:<ssehalfvecmode> 0 "<store_mask_predicate>"
+ "=<store_mask_constraint>,v")
(vec_select:<ssehalfvecmode>
- (match_operand:VI4F_256 1 "register_operand" "v")
+ (match_operand:VI4F_256 1 "<store_mask_predicate>"
+ "v,<store_mask_constraint>")
(parallel [(const_int 0) (const_int 1)
(const_int 2) (const_int 3)])))]
- "TARGET_AVX && <mask_avx512vl_condition> && <mask_avx512dq_condition>"
+ "TARGET_AVX
+ && <mask_avx512vl_condition> && <mask_avx512dq_condition>
+ && (<mask_applied> || !(MEM_P (operands[0]) && MEM_P (operands[1])))"
{
if (<mask_applied>)
return "vextract<shuffletype>32x4\t{$0x0, %1,
%0<mask_operand2>|%0<mask_operand2>, %1, 0x0}";
--- gcc/testsuite/gcc.target/i386/pr81225.c (nonexistent)
+++ gcc/testsuite/gcc.target/i386/pr81225.c (revision 250285)
@@ -0,0 +1,14 @@
+/* PR target/81225 */
+/* { dg-do compile } */
+/* { dg-options "-mavx512ifma -O3 -ffloat-store" } */
+
+long a[24];
+float b[4], c[24];
+int d;
+
+void
+foo ()
+{
+ for (d = 0; d < 24; d++)
+ c[d] = (float) d ? : b[a[d]];
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
PR libquadmath/65757
* math/roundq.c: Cherry-pick upstream glibc 2015-04-28 change.
--- libquadmath/math/roundq.c (revision 250378)
+++ libquadmath/math/roundq.c (revision 250379)
@@ -1,5 +1,5 @@
/* Round __float128 to integer away from zero.
- Copyright (C) 1997, 1999 Free Software Foundation, Inc.
+ Copyright (C) 1997-2017 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Ulrich Drepper <drep...@cygnus.com>, 1997 and
Jakub Jelinek <j...@ultra.linux.cz>, 1999.
@@ -32,7 +32,7 @@ roundq (__float128 x)
GET_FLT128_WORDS64 (i0, i1, x);
j0 = ((i0 >> 48) & 0x7fff) - 0x3fff;
- if (j0 < 31)
+ if (j0 < 48)
{
if (j0 < 0)
{
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-08-08 Richard Biener <rguent...@suse.de>
PR middle-end/81766
* function.c (thread_prologue_and_epilogue_insns): Restore
behavior of always calling find_many_sub_basic_blocks on
the inserted prologue.
* gcc.target/i386/pr81766.c: New testcase.
2017-08-02 Jakub Jelinek <ja...@redhat.com>
PR middle-end/79499
* function.c (thread_prologue_and_epilogue_insns): Determine blocks
for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
of first NONDEBUG_INSN_P in each of the split_prologue_seq and
prologue_seq sequences - if any.
* gcc.dg/pr79499.c: New test.
--- gcc/function.c (revision 250814)
+++ gcc/function.c (revision 250958)
@@ -6068,13 +6068,19 @@ thread_prologue_and_epilogue_insns (void
try_shrink_wrapping (&entry_edge, &bb_flags, prologue_seq);
+ rtx_insn *split_prologue_insn = split_prologue_seq;
if (split_prologue_seq != NULL_RTX)
{
+ while (split_prologue_insn && !NONDEBUG_INSN_P (split_prologue_insn))
+ split_prologue_insn = NEXT_INSN (split_prologue_insn);
insert_insn_on_edge (split_prologue_seq, orig_entry_edge);
inserted = true;
}
+ rtx_insn *prologue_insn = prologue_seq;
if (prologue_seq != NULL_RTX)
{
+ while (prologue_insn && !NONDEBUG_INSN_P (prologue_insn))
+ prologue_insn = NEXT_INSN (prologue_insn);
insert_insn_on_edge (prologue_seq, entry_edge);
inserted = true;
}
@@ -6215,8 +6221,19 @@ epilogue_done:
commit_edge_insertions ();
/* Look for basic blocks within the prologue insns. */
+ if (split_prologue_insn
+ && BLOCK_FOR_INSN (split_prologue_insn) == NULL)
+ split_prologue_insn = NULL;
+ if (prologue_insn
+ && BLOCK_FOR_INSN (prologue_insn) == NULL)
+ prologue_insn = NULL;
blocks = sbitmap_alloc (last_basic_block_for_fn (cfun));
bitmap_clear (blocks);
+ if (split_prologue_insn)
+ bitmap_set_bit (blocks,
+ BLOCK_FOR_INSN (split_prologue_insn)->index);
+ if (prologue_insn)
+ bitmap_set_bit (blocks, BLOCK_FOR_INSN (prologue_insn)->index);
bitmap_set_bit (blocks, entry_edge->dest->index);
bitmap_set_bit (blocks, orig_entry_edge->dest->index);
find_many_sub_basic_blocks (blocks);
--- gcc/testsuite/gcc.dg/pr79499.c (nonexistent)
+++ gcc/testsuite/gcc.dg/pr79499.c (revision 250815)
@@ -0,0 +1,13 @@
+/* PR middle-end/79499 */
+/* { dg-do compile { target split_stack } } */
+/* { dg-options "-O2 -fsplit-stack -fno-omit-frame-pointer" } */
+
+struct S { struct S *a, *b; };
+
+void
+foo (struct S *x)
+{
+ do
+ x->b = x->a;
+ while (x = x->a);
+}
--- gcc/testsuite/gcc.target/i386/pr81766.c (nonexistent)
+++ gcc/testsuite/gcc.target/i386/pr81766.c (revision 250958)
@@ -0,0 +1,5 @@
+/* { dg-do compile } */
+/* { dg-require-effective-target lp64 } */
+/* { dg-options "-O2 -fPIE -mcmodel=large" } */
+
+int main() { return 0; }
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-07-27 Jakub Jelinek <ja...@redhat.com>
PR c/45784
* c-omp.c (c_finish_omp_for): If the condition is wrapped in
rhs of COMPOUND_EXPR(s), skip them and readd their lhs into
new COMPOUND_EXPRs around the rhs of the comparison.
* testsuite/libgomp.c/pr45784.c: New test.
* testsuite/libgomp.c++/pr45784.C: New test.
--- gcc/c-family/c-omp.c (revision 251847)
+++ gcc/c-family/c-omp.c (revision 251848)
@@ -531,6 +531,12 @@ c_finish_omp_for (location_t locus, enum
{
bool cond_ok = false;
+ /* E.g. C sizeof (vla) could add COMPOUND_EXPRs with
+ evaluation of the vla VAR_DECL. We need to readd
+ them to the non-decl operand. See PR45784. */
+ while (TREE_CODE (cond) == COMPOUND_EXPR)
+ cond = TREE_OPERAND (cond, 1);
+
if (EXPR_HAS_LOCATION (cond))
elocus = EXPR_LOCATION (cond);
@@ -605,6 +611,21 @@ c_finish_omp_for (location_t locus, enum
else if (code != CILK_SIMD && code != CILK_FOR)
cond_ok = false;
}
+
+ if (cond_ok && TREE_VEC_ELT (condv, i) != cond)
+ {
+ tree ce = NULL_TREE, *pce = &ce;
+ tree type = TREE_TYPE (TREE_OPERAND (cond, 1));
+ for (tree c = TREE_VEC_ELT (condv, i); c != cond;
+ c = TREE_OPERAND (c, 1))
+ {
+ *pce = build2 (COMPOUND_EXPR, type, TREE_OPERAND (c, 0),
+ TREE_OPERAND (cond, 1));
+ pce = &TREE_OPERAND (*pce, 1);
+ }
+ TREE_OPERAND (cond, 1) = ce;
+ TREE_VEC_ELT (condv, i) = cond;
+ }
}
if (!cond_ok)
--- libgomp/testsuite/libgomp.c/pr45784.c (nonexistent)
+++ libgomp/testsuite/libgomp.c/pr45784.c (revision 251848)
@@ -0,0 +1,41 @@
+/* PR c/45784 */
+/* { dg-do run } */
+
+void
+foo (int n)
+{
+ char *p, vla[2 * n];
+ int i;
+ #pragma omp parallel for
+ for (p = vla; p < vla + (sizeof (vla) / sizeof (vla[0])); p++)
+ *p = ' ';
+ #pragma omp parallel for
+ for (i = 0; i < 2 * n; i++)
+ if (vla[i] != ' ')
+ __builtin_abort ();
+}
+
+void
+bar (int n)
+{
+ char *p, vla1[n], vla2[n * 2], vla3[n * 3], vla4[n * 4];
+ int i;
+ __builtin_memset (vla4, ' ', n * 4);
+ #pragma omp parallel for
+ for (p = vla4 + sizeof (vla1); p < vla4 + sizeof (vla3) - sizeof (vla2) +
sizeof (vla1); p += sizeof (vla4) / sizeof (vla4))
+ p[0] = '!';
+ #pragma omp parallel for
+ for (i = 0; i < n * 4; i++)
+ if (vla4[i] != ((i >= n && i < 2 * n) ? '!' : ' '))
+ __builtin_abort ();
+}
+
+int
+main ()
+{
+ volatile int n;
+ n = 128;
+ foo (n);
+ bar (n);
+ return 0;
+}
--- libgomp/testsuite/libgomp.c++/pr45784.C (nonexistent)
+++ libgomp/testsuite/libgomp.c++/pr45784.C (revision 251848)
@@ -0,0 +1,5 @@
+// PR c/45784
+// { dg-do run }
+
+#include "../libgomp.c/pr45784.c"
+
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-08-03 Jakub Jelinek <ja...@redhat.com>
PR middle-end/81052
* omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
(pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
* c-c++-common/pr81052.c: New test.
--- gcc/omp-low.c (revision 251848)
+++ gcc/omp-low.c (revision 251849)
@@ -9089,7 +9089,7 @@ diagnose_sb_0 (gimple_stmt_iterator *gsi
}
if (kind == NULL)
{
- gcc_checking_assert (flag_openmp);
+ gcc_checking_assert (flag_openmp || flag_openmp_simd);
kind = "OpenMP";
}
@@ -9349,7 +9349,7 @@ public:
/* opt_pass methods: */
virtual bool gate (function *)
{
- return flag_cilkplus || flag_openacc || flag_openmp;
+ return flag_cilkplus || flag_openacc || flag_openmp || flag_openmp_simd;
}
virtual unsigned int execute (function *)
{
--- gcc/testsuite/c-c++-common/pr81052.c (nonexistent)
+++ gcc/testsuite/c-c++-common/pr81052.c (revision 251849)
@@ -0,0 +1,28 @@
+/* PR middle-end/81052 */
+/* { dg-do compile } */
+/* { dg-options "-fopenmp-simd -O2" } */
+
+int
+foo (int x, int y)
+{
+ int i;
+#pragma omp simd
+ for (i = x; i < y; ++i)
+ return 0; /* { dg-error "invalid branch to/from OpenMP
structured block" } */
+ return 1;
+}
+
+#ifdef __cplusplus
+template <typename T>
+T
+bar (T x, T y)
+{
+ T i;
+#pragma omp simd
+ for (i = x; i < y; ++i)
+ return 0; /* { dg-error "invalid branch to/from OpenMP
structured block" "" { target c++ } } */
+ return 1;
+}
+
+int x = bar (1, 7);
+#endif
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-08-03 Jakub Jelinek <ja...@redhat.com>
PR target/81621
* bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
after setting changeable df flags.
* gcc.dg/pr81621.c: New test.
--- gcc/bb-reorder.c (revision 251850)
+++ gcc/bb-reorder.c (revision 251851)
@@ -2881,7 +2881,8 @@ pass_partition_blocks::execute (function
crossing_edges = find_rarely_executed_basic_blocks_and_crossing_edges ();
if (!crossing_edges.exists ())
- return 0;
+ /* Make sure to process deferred rescans and clear changeable df flags. */
+ return TODO_df_finish;
crtl->has_bb_partition = true;
@@ -2947,7 +2948,8 @@ pass_partition_blocks::execute (function
df_analyze ();
}
- return 0;
+ /* Make sure to process deferred rescans and clear changeable df flags. */
+ return TODO_df_finish;
}
} // anon namespace
--- gcc/testsuite/gcc.dg/pr81621.c (nonexistent)
+++ gcc/testsuite/gcc.dg/pr81621.c (revision 251851)
@@ -0,0 +1,5 @@
+/* PR target/81621 */
+/* { dg-do compile { target freorder } } */
+/* { dg-options "-Og -fno-split-wide-types -freorder-blocks-and-partition" } */
+
+#include "graphite/scop-10.c"
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-08-09 Jakub Jelinek <ja...@redhat.com>
PR c/81687
* omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
LABEL_DECLs.
* tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
or DECL_NONLOCAL labels.
(move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
or DECL_NONLOCAL labels here.
* testsuite/libgomp.c/pr81687-1.c: New test.
* testsuite/libgomp.c/pr81687-2.c: New test.
--- gcc/omp-low.c (revision 251852)
+++ gcc/omp-low.c (revision 251853)
@@ -800,6 +800,8 @@ omp_copy_decl (tree var, copy_body_data
if (TREE_CODE (var) == LABEL_DECL)
{
+ if (FORCED_LABEL (var) || DECL_NONLOCAL (var))
+ return var;
new_var = create_artificial_label (DECL_SOURCE_LOCATION (var));
DECL_CONTEXT (new_var) = current_function_decl;
insert_decl_map (&ctx->cb, var, new_var);
--- gcc/tree-cfg.c (revision 251852)
+++ gcc/tree-cfg.c (revision 251853)
@@ -6667,7 +6667,15 @@ move_stmt_op (tree *tp, int *walk_subtre
*tp = t = out->to;
}
- DECL_CONTEXT (t) = p->to_context;
+ /* For FORCED_LABELs we can end up with references from other
+ functions if some SESE regions are outlined. It is UB to
+ jump in between them, but they could be used just for printing
+ addresses etc. In that case, DECL_CONTEXT on the label should
+ be the function containing the glabel stmt with that LABEL_DECL,
+ rather than whatever function a reference to the label was seen
+ last time. */
+ if (!FORCED_LABEL (t) && !DECL_NONLOCAL (t))
+ DECL_CONTEXT (t) = p->to_context;
}
else if (p->remap_decls_p)
{
@@ -6785,6 +6793,21 @@ move_stmt_r (gimple_stmt_iterator *gsi_p
case GIMPLE_OMP_RETURN:
case GIMPLE_OMP_CONTINUE:
break;
+
+ case GIMPLE_LABEL:
+ {
+ /* For FORCED_LABEL, move_stmt_op doesn't adjust DECL_CONTEXT,
+ so that such labels can be referenced from other regions.
+ Make sure to update it when seeing a GIMPLE_LABEL though,
+ that is the owner of the label. */
+ walk_gimple_op (stmt, move_stmt_op, wi);
+ *handled_ops_p = true;
+ tree label = gimple_label_label (as_a <glabel *> (stmt));
+ if (FORCED_LABEL (label) || DECL_NONLOCAL (label))
+ DECL_CONTEXT (label) = p->to_context;
+ }
+ break;
+
default:
if (is_gimple_omp (stmt))
{
--- libgomp/testsuite/libgomp.c/pr81687-1.c (nonexistent)
+++ libgomp/testsuite/libgomp.c/pr81687-1.c (revision 251853)
@@ -0,0 +1,23 @@
+/* PR c/81687 */
+/* { dg-do link } */
+/* { dg-additional-options "-O2" } */
+
+extern int printf (const char *, ...);
+
+int
+main ()
+{
+ #pragma omp parallel
+ {
+ lab1:
+ printf ("lab1=%p\n", (void *)(&&lab1));
+ }
+ lab2:
+ #pragma omp parallel
+ {
+ lab3:
+ printf ("lab2=%p\n", (void *)(&&lab2));
+ }
+ printf ("lab3=%p\n", (void *)(&&lab3));
+ return 0;
+}
--- libgomp/testsuite/libgomp.c/pr81687-2.c (nonexistent)
+++ libgomp/testsuite/libgomp.c/pr81687-2.c (revision 251853)
@@ -0,0 +1,27 @@
+/* PR c/81687 */
+/* { dg-do link } */
+/* { dg-additional-options "-O2" } */
+
+int
+main ()
+{
+ __label__ lab4, lab5, lab6;
+ volatile int l = 0;
+ int m = l;
+ void foo (int x) { if (x == 1) goto lab4; }
+ void bar (int x) { if (x == 2) goto lab5; }
+ void baz (int x) { if (x == 3) goto lab6; }
+ #pragma omp parallel
+ {
+ foo (m + 1);
+ lab4:;
+ }
+ #pragma omp task
+ {
+ bar (m + 2);
+ lab5:;
+ }
+ baz (m + 3);
+ lab6:;
+ return 0;
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-04 Jakub Jelinek <ja...@redhat.com>
* lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
--- gcc/lra-remat.c (revision 251854)
+++ gcc/lra-remat.c (revision 251855)
@@ -684,7 +684,7 @@ reg_overlap_for_remat_p (lra_insn_reg *r
if (regno2 >= FIRST_PSEUDO_REGISTER && reg_renumber[regno2] >= 0)
regno2 = reg_renumber[regno2];
- if (regno >= FIRST_PSEUDO_REGISTER)
+ if (regno2 >= FIRST_PSEUDO_REGISTER)
nregs2 = 1;
else
nregs2 = hard_regno_nregs[regno2][reg->biggest_mode];
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-05 Jakub Jelinek <ja...@redhat.com>
PR middle-end/81768
* omp-low.c (expand_omp_simd): Force second operands of COND_EXPR
into gimple val before gimplification fo the COND_EXPR.
* gcc.dg/gomp/pr81768-1.c: New test.
--- gcc/omp-low.c (revision 251855)
+++ gcc/omp-low.c (revision 251856)
@@ -4725,24 +4725,28 @@ expand_omp_simd (struct omp_region *regi
tree itype2 = TREE_TYPE (fd->loops[i - 1].v);
if (POINTER_TYPE_P (itype2))
itype2 = signed_type_for (itype2);
+ t = fold_convert (itype2, fd->loops[i - 1].step);
+ t = force_gimple_operand_gsi (&gsi, t, true, NULL_TREE, true,
+ GSI_SAME_STMT);
t = build3 (COND_EXPR, itype2,
build2 (fd->loops[i].cond_code, boolean_type_node,
fd->loops[i].v,
fold_convert (itype, fd->loops[i].n2)),
- build_int_cst (itype2, 0),
- fold_convert (itype2, fd->loops[i - 1].step));
+ build_int_cst (itype2, 0), t);
if (POINTER_TYPE_P (TREE_TYPE (fd->loops[i - 1].v)))
t = fold_build_pointer_plus (fd->loops[i - 1].v, t);
else
t = fold_build2 (PLUS_EXPR, itype2, fd->loops[i - 1].v, t);
expand_omp_build_assign (&gsi, fd->loops[i - 1].v, t);
+ t = fold_convert (itype, fd->loops[i].n1);
+ t = force_gimple_operand_gsi (&gsi, t, true, NULL_TREE, true,
+ GSI_SAME_STMT);
t = build3 (COND_EXPR, itype,
build2 (fd->loops[i].cond_code, boolean_type_node,
fd->loops[i].v,
fold_convert (itype, fd->loops[i].n2)),
- fd->loops[i].v,
- fold_convert (itype, fd->loops[i].n1));
+ fd->loops[i].v, t);
expand_omp_build_assign (&gsi, fd->loops[i].v, t);
}
}
--- gcc/testsuite/gcc.dg/gomp/pr81768-1.c (nonexistent)
+++ gcc/testsuite/gcc.dg/gomp/pr81768-1.c (revision 251856)
@@ -0,0 +1,15 @@
+/* PR middle-end/81768 */
+/* { dg-do compile } */
+
+float b[10][15][10];
+
+void
+foo (void)
+{
+ float *i;
+#pragma omp target parallel for simd schedule(static, 32) collapse(3)
+ for (i = &b[0][0][0]; i < &b[0][0][10]; i++)
+ for (float *j = &b[0][15][0]; j > &b[0][0][0]; j -= 10)
+ for (float *k = &b[0][0][10]; k > &b[0][0][0]; --k)
+ b[i - &b[0][0][0]][(j - &b[0][0][0]) / 10 - 1][(k - &b[0][0][0]) - 1]
-= 3.5;
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-05 Jakub Jelinek <ja...@redhat.com>
PR middle-end/81768
* omp-low.c (lower_omp_for): Recompute tree invariant if
gimple_omp_for_initial/final is ADDR_EXPR.
* gcc.dg/gomp/pr81768-2.c: New test.
--- gcc/omp-low.c (revision 251856)
+++ gcc/omp-low.c (revision 251857)
@@ -6927,10 +6927,14 @@ lower_omp_for (gimple_stmt_iterator *gsi
rhs_p = gimple_omp_for_initial_ptr (stmt, i);
if (!is_gimple_min_invariant (*rhs_p))
*rhs_p = get_formal_tmp_var (*rhs_p, &body);
+ else if (TREE_CODE (*rhs_p) == ADDR_EXPR)
+ recompute_tree_invariant_for_addr_expr (*rhs_p);
rhs_p = gimple_omp_for_final_ptr (stmt, i);
if (!is_gimple_min_invariant (*rhs_p))
*rhs_p = get_formal_tmp_var (*rhs_p, &body);
+ else if (TREE_CODE (*rhs_p) == ADDR_EXPR)
+ recompute_tree_invariant_for_addr_expr (*rhs_p);
rhs_p = &TREE_OPERAND (gimple_omp_for_incr (stmt, i), 1);
if (!is_gimple_min_invariant (*rhs_p))
--- gcc/testsuite/gcc.dg/gomp/pr81768-2.c (nonexistent)
+++ gcc/testsuite/gcc.dg/gomp/pr81768-2.c (revision 251857)
@@ -0,0 +1,15 @@
+/* PR middle-end/81768 */
+/* { dg-do compile } */
+
+float b[10][15][10];
+
+void
+foo (void)
+{
+ float *i;
+#pragma omp target parallel for schedule(static, 32) collapse(3)
+ for (i = &b[0][0][0]; i < &b[0][0][10]; i++)
+ for (float *j = &b[0][15][0]; j > &b[0][0][0]; j -= 10)
+ for (float *k = &b[0][0][10]; k > &b[0][0][0]; --k)
+ b[i - &b[0][0][0]][(j - &b[0][0][0]) / 10 - 1][(k - &b[0][0][0]) - 1]
-= 3.5;
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-12 Jakub Jelinek <ja...@redhat.com>
PR target/82112
* c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
assertion check that in the condition.
(get_atomic_generic_size): Likewise. Before testing if parameter
has pointer type, if it has array type, call for C++
default_conversion to perform array-to-pointer conversion.
* c-c++-common/pr82112.c: New test.
* gcc.dg/pr82112.c: New test.
--- gcc/c-family/c-common.c (revision 252802)
+++ gcc/c-family/c-common.c (revision 252803)
@@ -6576,10 +6576,9 @@ sync_resolve_size (tree function, vec<tr
}
argtype = type = TREE_TYPE ((*params)[0]);
- if (TREE_CODE (type) == ARRAY_TYPE)
+ if (TREE_CODE (type) == ARRAY_TYPE && c_dialect_cxx ())
{
/* Force array-to-pointer decay for C++. */
- gcc_assert (c_dialect_cxx());
(*params)[0] = default_conversion ((*params)[0]);
type = TREE_TYPE ((*params)[0]);
}
@@ -6741,10 +6740,9 @@ get_atomic_generic_size (location_t loc,
/* Get type of first parameter, and determine its size. */
type_0 = TREE_TYPE ((*params)[0]);
- if (TREE_CODE (type_0) == ARRAY_TYPE)
+ if (TREE_CODE (type_0) == ARRAY_TYPE && c_dialect_cxx ())
{
/* Force array-to-pointer decay for C++. */
- gcc_assert (c_dialect_cxx());
(*params)[0] = default_conversion ((*params)[0]);
type_0 = TREE_TYPE ((*params)[0]);
}
@@ -6783,6 +6781,12 @@ get_atomic_generic_size (location_t loc,
/* __atomic_compare_exchange has a bool in the 4th position, skip it. */
if (n_param == 6 && x == 3)
continue;
+ if (TREE_CODE (type) == ARRAY_TYPE && c_dialect_cxx ())
+ {
+ /* Force array-to-pointer decay for C++. */
+ (*params)[x] = default_conversion ((*params)[x]);
+ type = TREE_TYPE ((*params)[x]);
+ }
if (!POINTER_TYPE_P (type))
{
error_at (loc, "argument %d of %qE must be a pointer type", x + 1,
--- gcc/testsuite/gcc.dg/pr82112.c (nonexistent)
+++ gcc/testsuite/gcc.dg/pr82112.c (revision 252803)
@@ -0,0 +1,21 @@
+/* PR target/82112 */
+/* { dg-do compile } */
+/* { dg-options "-std=gnu90" } */
+
+struct S { int a[10]; } bar (void);
+int b, c;
+
+void
+foo (void)
+{
+ __atomic_load (bar ().a, &b, __ATOMIC_ACQUIRE); /* { dg-error "argument
1 of .__atomic_load. must be a non-void pointer type" } */
+ __atomic_load (&b, bar ().a, __ATOMIC_ACQUIRE); /* { dg-error "argument
2 of .__atomic_load. must be a pointer type" } */
+ __atomic_store (bar ().a, &b, __ATOMIC_SEQ_CST); /* { dg-error "argument
1 of .__atomic_store. must be a non-void pointer type" } */
+ __atomic_store (&b, bar ().a, __ATOMIC_SEQ_CST); /* { dg-error "argument
2 of .__atomic_store. must be a pointer type" } */
+ __atomic_exchange (bar ().a, &b, &c, __ATOMIC_RELAXED); /* { dg-error
"argument 1 of .__atomic_exchange. must be a non-void pointer type" } */
+ __atomic_exchange (&b, bar ().a, &c, __ATOMIC_RELAXED); /* { dg-error
"argument 2 of .__atomic_exchange. must be a pointer type" } */
+ __atomic_exchange (&b, &c, bar ().a, __ATOMIC_RELAXED); /* { dg-error
"argument 3 of .__atomic_exchange. must be a pointer type" } */
+ __atomic_compare_exchange (bar ().a, &b, &c, 1, __ATOMIC_RELAXED,
__ATOMIC_RELAXED); /* { dg-error "argument 1 of .__atomic_compare_exchange.
must be a non-void pointer type" } */
+ __atomic_compare_exchange (&b, bar ().a, &c, 1, __ATOMIC_RELAXED,
__ATOMIC_RELAXED); /* { dg-error "argument 2 of .__atomic_compare_exchange.
must be a pointer type" } */
+ __atomic_compare_exchange (&b, &c, bar ().a, 1, __ATOMIC_RELAXED,
__ATOMIC_RELAXED); /* { dg-error "argument 3 of .__atomic_compare_exchange.
must be a pointer type" } */
+}
--- gcc/testsuite/c-c++-common/pr82112.c (nonexistent)
+++ gcc/testsuite/c-c++-common/pr82112.c (revision 252803)
@@ -0,0 +1,13 @@
+/* PR target/82112 */
+/* { dg-do compile } */
+
+int c[10], d[10], e[10], f[10], g[10], h[10], i[10], j[10], k[10], l[10];
+
+void
+foo (void)
+{
+ __atomic_load (c, d, __ATOMIC_ACQUIRE);
+ __atomic_store (e, f, __ATOMIC_SEQ_CST);
+ __atomic_exchange (g, h, i, __ATOMIC_RELAXED);
+ __atomic_compare_exchange (j, k, l, 1, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-12 Jakub Jelinek <ja...@redhat.com>
PR target/82112
* gcc.target/powerpc/pr82112.c: New test.
* g++.dg/ext/altivec-18.C: New test.
--- gcc/testsuite/gcc.target/powerpc/pr82112.c (nonexistent)
+++ gcc/testsuite/gcc.target/powerpc/pr82112.c (revision 252804)
@@ -0,0 +1,16 @@
+/* PR target/82112 */
+/* { dg-do compile } */
+/* { dg-require-effective-target powerpc_altivec_ok } */
+/* { dg-options "-maltivec -std=gnu90" } */
+
+#include <altivec.h>
+
+struct __attribute__((aligned (16))) S { unsigned char c[64]; } bar (void);
+vector unsigned char v;
+
+void
+foo (void)
+{
+ vec_ld (0, bar ().c); /* { dg-error "invalid parameter combination
for AltiVec intrinsic" } */
+ vec_st (v, 0, bar ().c); /* { dg-error "invalid parameter combination
for AltiVec intrinsic" } */
+}
--- gcc/testsuite/g++.dg/ext/altivec-18.C (nonexistent)
+++ gcc/testsuite/g++.dg/ext/altivec-18.C (revision 252804)
@@ -0,0 +1,14 @@
+// PR target/82112
+// { dg-do compile { target powerpc*-*-* } }
+// { dg-require-effective-target powerpc_altivec_ok }
+// { dg-options "-maltivec" }
+
+#include <altivec.h>
+
+__attribute__((aligned (16))) extern const unsigned char c[16];
+
+void
+foo (void)
+{
+ vec_ld (0, c);
+}
2017-09-15 Jakub Jelinek <ja...@redhat.com>
Backported from mainline
2017-09-14 Jakub Jelinek <ja...@redhat.com>
PR c++/81314
* cp-gimplify.c (omp_var_to_track): Look through references.
(omp_cxx_notice_variable): Likewise.
* testsuite/libgomp.c++/pr81314.C: New test.
--- gcc/cp/cp-gimplify.c (revision 252805)
+++ gcc/cp/cp-gimplify.c (revision 252806)
@@ -924,6 +924,8 @@ omp_var_to_track (tree decl)
tree type = TREE_TYPE (decl);
if (is_invisiref_parm (decl))
type = TREE_TYPE (type);
+ else if (TREE_CODE (type) == REFERENCE_TYPE)
+ type = TREE_TYPE (type);
while (TREE_CODE (type) == ARRAY_TYPE)
type = TREE_TYPE (type);
if (type == error_mark_node || !CLASS_TYPE_P (type))
@@ -976,6 +978,8 @@ omp_cxx_notice_variable (struct cp_gener
tree type = TREE_TYPE (decl);
if (is_invisiref_parm (decl))
type = TREE_TYPE (type);
+ else if (TREE_CODE (type) == REFERENCE_TYPE)
+ type = TREE_TYPE (type);
while (TREE_CODE (type) == ARRAY_TYPE)
type = TREE_TYPE (type);
get_copy_ctor (type, tf_none);
--- libgomp/testsuite/libgomp.c++/pr81314.C (nonexistent)
+++ libgomp/testsuite/libgomp.c++/pr81314.C (revision 252806)
@@ -0,0 +1,38 @@
+// PR c++/81314
+// { dg-do link }
+
+template <int N>
+struct S {
+ S () { s = 0; }
+ S (const S &x) { s = x.s; }
+ ~S () {}
+ int s;
+};
+
+void
+foo (S<2> &x)
+{
+ #pragma omp taskloop
+ for (int i = 0; i < 100; ++i)
+ x.s++;
+}
+
+void
+bar (S<3> &x)
+{
+ #pragma omp task
+ x.s++;
+}
+
+int
+main ()
+{
+ S<2> s;
+ S<3> t;
+ #pragma omp parallel
+ #pragma omp master
+ {
+ foo (s);
+ bar (t);
+ }
+}
