]> git.ipfire.org Git - thirdparty/gcc.git/commitdiff
Daily bump.
authorGCC Administrator <gccadmin@gcc.gnu.org>
Wed, 10 Aug 2022 00:16:43 +0000 (00:16 +0000)
committerGCC Administrator <gccadmin@gcc.gnu.org>
Wed, 10 Aug 2022 00:16:43 +0000 (00:16 +0000)
gcc/ChangeLog
gcc/DATESTAMP
gcc/analyzer/ChangeLog
gcc/d/ChangeLog
gcc/jit/ChangeLog
gcc/testsuite/ChangeLog

index 36879ec41091144633e8e7a8a7fa35886366b8e9..94508380d6dca9fea2ffa93b1875a4178faee4ca 100644 (file)
@@ -1,3 +1,98 @@
+2022-08-09  Roger Sayle  <roger@nextmovesoftware.com>
+
+       * config/i386/i386-features.cc (scalar_chain::convert_compare):
+       Create new pseudos only when/if needed.  Add support for TEST,
+       i.e. (COMPARE (AND x y) (const_int 0)), using UNSPEC_PTEST.
+       When broadcasting V2DImode and V4SImode use new pseudo register.
+       (timode_scalar_chain::convert_op): Do nothing if operand is
+       already V1TImode.  Avoid generating useless SUBREG conversions,
+       i.e. (SUBREG:V1TImode (REG:V1TImode) 0).  Handle CONST_WIDE_INT
+       in addition to CONST_INT by using CONST_SCALAR_INT_P.
+       (convertible_comparison_p): Use CONST_SCALAR_INT_P to match both
+       CONST_WIDE_INT and CONST_INT.  Recognize new *testti_doubleword
+       pattern as an STV candidate.
+       (timode_scalar_to_vector_candidate_p): Allow CONST_SCALAR_INT_P
+       operands in binary logic operations.
+       * config/i386/i386.cc (ix86_rtx_costs) <case UNSPEC>: Add costs
+       for UNSPEC_PTEST; a PTEST that performs an AND has the same cost
+       as regular PTEST, i.e. cost->sse_op.
+       * config/i386/i386.md (*testti_doubleword): New pre-reload
+       define_insn_and_split that recognizes comparison of TI mode AND
+       against zero.
+       * config/i386/sse.md (*ptest<mode>_and): New pre-reload
+       define_insn_and_split that recognizes UNSPEC_PTEST of identical
+       AND operands.
+
+2022-08-09  Roger Sayle  <roger@nextmovesoftware.com>
+           Richard Biener  <rguenther@suse.de>
+
+       PR middle-end/21137
+       PR tree-optimization/98954
+       * fold-const.cc (fold_binary_loc): Remove optimizations to
+       optimize ((X >> C1) & C2) ==/!= 0.
+       * match.pd (cmp (bit_and (lshift @0 @1) @2) @3): Remove wi::ctz
+       check, and handle all values of INTEGER_CSTs @2 and @3.
+       (cmp (bit_and (rshift @0 @1) @2) @3): Likewise, remove wi::clz
+       checks, and handle all values of INTEGER_CSTs @2 and @3.
+
+2022-08-09  David Malcolm  <dmalcolm@redhat.com>
+
+       * doc/invoke.texi (Static Analyzer Options): Add notes on which
+       functions the analyzer has hardcoded knowledge of.
+
+2022-08-09  Andrew Stubbs  <ams@codesourcery.com>
+
+       * config/gcn/gcn.cc (gcn_function_value): Allow vector return values.
+       (num_arg_regs): Allow vector arguments.
+       (gcn_function_arg): Likewise.
+       (gcn_function_arg_advance): Likewise.
+       (gcn_arg_partial_bytes): Likewise.
+       (gcn_return_in_memory): Likewise.
+       (gcn_expand_epilogue): Get return value from v8.
+       * config/gcn/gcn.h (RETURN_VALUE_REG): Set to v8.
+       (FIRST_PARM_REG): USE FIRST_SGPR_REG for clarity.
+       (FIRST_VPARM_REG): New.
+       (FUNCTION_ARG_REGNO_P): Allow vector parameters.
+       (struct gcn_args): Add vnum field.
+       (LIBCALL_VALUE): All vector return values.
+       * config/gcn/gcn.md (gcn_call_value): Add vector constraints.
+       (gcn_call_value_indirect): Likewise.
+
+2022-08-09  Richard Biener  <rguenther@suse.de>
+
+       * omp-expand.cc (expand_omp_atomic_load): Emit GIMPLE
+       directly.  Avoid update_ssa when in SSA form.
+       (expand_omp_atomic_store): Likewise.
+       (expand_omp_atomic_fetch_op): Avoid update_ssa when in SSA
+       form.
+       (expand_omp_atomic_pipeline): Likewise.
+       (expand_omp_atomic_mutex): Likewise.
+       * tree-parloops.cc (gen_parallel_loop): Use
+       TODO_update_ssa_no_phi after loop_version.
+
+2022-08-09  Richard Biener  <rguenther@suse.de>
+
+       * doc/invoke.texi (max-fsm-thread-length): Remove.
+       * params.opt (max-fsm-thread-length): Likewise.
+       * tree-ssa-threadbackward.cc
+       (back_threader_profitability::profitable_path_p): Do not
+       check max-fsm-thread-length.
+
+2022-08-09  Richard Biener  <rguenther@suse.de>
+
+       PR tree-optimization/106514
+       * params.opt (max-jump-thread-paths): New.
+       * doc/invoke.texi (max-jump-thread-paths): Document.
+       * tree-ssa-threadbackward.cc (back_threader::find_paths_to_names):
+       Honor max-jump-thread-paths, take overall_path argument.
+       (back_threader::find_paths): Pass 1 as initial overall_path.
+
+2022-08-09  Tobias Burnus  <tobias@codesourcery.com>
+
+       PR middle-end/106492
+       * omp-low.cc (lower_rec_input_clauses): Add missing folding
+       to data type of linear-clause list item.
+
 2022-08-08  Andrew MacLeod  <amacleod@redhat.com>
 
        PR tree-optimization/106556
index 8bd1173be90b3ce5239dc9cd334c6a8403b5fcab..9049af15948f12d6ea4eae416b269556f29a54f3 100644 (file)
@@ -1 +1 @@
-20220809
+20220810
index 0b93219e5c3095a9ba37999701d83215dfb105f9..b16971b9eafdae4cd058f7275d0af473c9baf002 100644 (file)
@@ -1,3 +1,9 @@
+2022-08-09  David Malcolm  <dmalcolm@redhat.com>
+
+       PR analyzer/106573
+       * region-model.cc (region_model::on_call_pre): Ensure that we call
+       get_arg_svalue on all arguments.
+
 2022-08-05  David Malcolm  <dmalcolm@redhat.com>
 
        PR analyzer/105947
index 0bb74b134c5bf77da5877d3ab3c463957a73583a..0af724b18a1ec4f9043aa18c6fd8bcb562a43a05 100644 (file)
@@ -1,3 +1,17 @@
+2022-08-09  Iain Buclaw  <ibuclaw@gdcproject.org>
+
+       PR d/106563
+       * decl.cc (DeclVisitor::visit (FuncDeclaration *)): Set semanticRun
+       before generating its symbol.
+       (function_defined_in_root_p): New function.
+       (function_needs_inline_definition_p): New function.
+       (maybe_build_decl_tree): New function.
+       (get_symbol_decl): Call maybe_build_decl_tree before returning symbol.
+       (start_function): Use function_defined_in_root_p instead of inline
+       test for locally defined symbols.
+       (set_linkage_for_decl): Check for inline functions before private or
+       protected symbols.
+
 2022-08-08  Iain Buclaw  <ibuclaw@gdcproject.org>
 
        PR d/106555
index df8bf40b603cf01babf109b30c64f1242a36d48a..314b8310b0ed3b5828fe2320a681be7f8c59d058 100644 (file)
@@ -1,3 +1,8 @@
+2022-08-09  Vibhav Pant  <vibhavp@gmail.com>
+
+       * libgccjit.h (LIBGCCJIT_HAVE_gcc_jit_context_new_bitcast): Move
+       definition out of comment.
+
 2022-07-28  David Malcolm  <dmalcolm@redhat.com>
 
        * docs/internals/index.rst: Remove reference to ".c" extensions
index 19a5303bc1a597ffad0856e7d17b0b18754ea043..9a31d8d23ec85e6b0cc496d8fc805e324848b641 100644 (file)
@@ -1,3 +1,42 @@
+2022-08-09  David Malcolm  <dmalcolm@redhat.com>
+
+       PR analyzer/106573
+       * gcc.dg/analyzer/error-uninit.c: New test.
+       * gcc.dg/analyzer/fd-uninit-1.c: New test.
+       * gcc.dg/analyzer/file-uninit-1.c: New test.
+
+2022-08-09  Roger Sayle  <roger@nextmovesoftware.com>
+
+       * gcc.target/i386/sse4_1-stv-8.c: New test case.
+
+2022-08-09  Roger Sayle  <roger@nextmovesoftware.com>
+           Richard Biener  <rguenther@suse.de>
+
+       PR middle-end/21137
+       PR tree-optimization/98954
+       * gcc.dg/fold-eqandshift-4.c: New test case.
+
+2022-08-09  Iain Buclaw  <ibuclaw@gdcproject.org>
+
+       PR d/106563
+       * gdc.dg/torture/torture.exp (srcdir): New proc.
+       * gdc.dg/torture/imports/pr106563math.d: New test.
+       * gdc.dg/torture/imports/pr106563regex.d: New test.
+       * gdc.dg/torture/imports/pr106563uni.d: New test.
+       * gdc.dg/torture/pr106563.d: New test.
+
+2022-08-09  Richard Biener  <rguenther@suse.de>
+
+       PR tree-optimization/106514
+       * gcc.dg/tree-ssa/ssa-thread-16.c: New testcase.
+       * gcc.dg/tree-ssa/ssa-thread-17.c: Likewise.
+       * gcc.dg/tree-ssa/ssa-dom-thread-7.c: Adjust.
+
+2022-08-09  Tobias Burnus  <tobias@codesourcery.com>
+
+       PR middle-end/106492
+       * g++.dg/gomp/pr106492.C: New test.
+
 2022-08-08  Andrew MacLeod  <amacleod@redhat.com>
 
        PR tree-optimization/106556