On 06/21/12 18:49, Pedro F. Giffuni wrote:
Author: pfg Date: Thu Jun 21 16:49:20 2012 New Revision: 237406 URL: http://svn.freebsd.org/changeset/base/237406Log: Bring a couple of fixes for gcc optimizations. The GCC4.3 branch contains some optimization fixes that were not considered regressions and therefore were never backported. We are bringing a couple of them that are under GPLv2 since they were made before the license switch upstream. While here, add the GCC revision numbers in the log. Discussed with: jkim MFC after: 1 week Modified: head/contrib/gcc/ChangeLog.gcc43 head/contrib/gcc/fold-const.c head/contrib/gcc/gimplify.c head/contrib/gcc/tree-ssa-ccp.c head/contrib/gcc/tree-ssa-pre.c Modified: head/contrib/gcc/ChangeLog.gcc43 ============================================================================== --- head/contrib/gcc/ChangeLog.gcc43 Thu Jun 21 16:48:56 2012 (r237405) +++ head/contrib/gcc/ChangeLog.gcc43 Thu Jun 21 16:49:20 2012 (r237406) @@ -5,6 +5,18 @@ with SSE3 instruction set support. * doc/invoke.texi: Likewise. +2007-04-12 Richard Guenther<rguent...@suse.de> (r123736) + + PR tree-optimization/24689 + PR tree-optimization/31307 + * fold-const.c (operand_equal_p): Compare INTEGER_CST array + indices by value. + * gimplify.c (canonicalize_addr_expr): To be consistent with + gimplify_compound_lval only set operands two and three of + ARRAY_REFs if they are not gimple_min_invariant. This makes + it never at this place. + * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Likewise. + 2007-04-07 H.J. Lu<hongjiu...@intel.com> (r123639) * config/i386/i386.c (ix86_handle_option): Handle SSSE3. @@ -96,7 +108,7 @@ * doc/invoke.texi: Add entry about geode processor. -2006-10-24 Richard Guenther<rguent...@suse.de> +2006-10-24 Richard Guenther<rguent...@suse.de> (r118001) PR middle-end/28796 * builtins.c (fold_builtin_classify): Use HONOR_INFINITIES @@ -170,7 +182,13 @@ * doc/invoke.texi: Document -mssse3/-mno-ssse3 switches. -2006-10-21 Richard Guenther<rguent...@suse.de> +2006-10-21 Richard Guenther<rguent...@suse.de> (r117932) + + PR tree-optimization/3511 + * tree-ssa-pre.c (phi_translate): Fold CALL_EXPRs that + got new invariant arguments during PHI translation. + +2006-10-21 Richard Guenther<rguent...@suse.de> (r117929) * builtins.c (fold_builtin_classify): Fix typo. Modified: head/contrib/gcc/fold-const.c ============================================================================== --- head/contrib/gcc/fold-const.c Thu Jun 21 16:48:56 2012 (r237405) +++ head/contrib/gcc/fold-const.c Thu Jun 21 16:49:20 2012 (r237406) @@ -2802,9 +2802,13 @@ operand_equal_p (tree arg0, tree arg1, u case ARRAY_REF: case ARRAY_RANGE_REF: - /* Operands 2 and 3 may be null. */ + /* Operands 2 and 3 may be null. + Compare the array index by value if it is constant first as we + may have different types but same value here. */ return (OP_SAME (0) - && OP_SAME (1) + && (tree_int_cst_equal (TREE_OPERAND (arg0, 1), + TREE_OPERAND (arg1, 1)) + || OP_SAME (1)) && OP_SAME_WITH_NULL (2) && OP_SAME_WITH_NULL (3)); Modified: head/contrib/gcc/gimplify.c ============================================================================== --- head/contrib/gcc/gimplify.c Thu Jun 21 16:48:56 2012 (r237405) +++ head/contrib/gcc/gimplify.c Thu Jun 21 16:49:20 2012 (r237406) @@ -1600,9 +1600,7 @@ canonicalize_addr_expr (tree *expr_p) /* All checks succeeded. Build a new node to merge the cast. */ *expr_p = build4 (ARRAY_REF, dctype, obj_expr, TYPE_MIN_VALUE (TYPE_DOMAIN (datype)), - TYPE_MIN_VALUE (TYPE_DOMAIN (datype)), - size_binop (EXACT_DIV_EXPR, TYPE_SIZE_UNIT (dctype), - size_int (TYPE_ALIGN_UNIT (dctype)))); + NULL_TREE, NULL_TREE); *expr_p = build1 (ADDR_EXPR, ctype, *expr_p); } Modified: head/contrib/gcc/tree-ssa-ccp.c ============================================================================== --- head/contrib/gcc/tree-ssa-ccp.c Thu Jun 21 16:48:56 2012 (r237405) +++ head/contrib/gcc/tree-ssa-ccp.c Thu Jun 21 16:49:20 2012 (r237406) @@ -1621,9 +1621,7 @@ maybe_fold_offset_to_array_ref (tree bas if (!integer_zerop (elt_offset)) idx = int_const_binop (PLUS_EXPR, idx, elt_offset, 0); - return build4 (ARRAY_REF, orig_type, base, idx, min_idx, - size_int (tree_low_cst (elt_size, 1) - / (TYPE_ALIGN_UNIT (elt_type)))); + return build4 (ARRAY_REF, orig_type, base, idx, NULL_TREE, NULL_TREE); } Modified: head/contrib/gcc/tree-ssa-pre.c ============================================================================== --- head/contrib/gcc/tree-ssa-pre.c Thu Jun 21 16:48:56 2012 (r237405) +++ head/contrib/gcc/tree-ssa-pre.c Thu Jun 21 16:49:20 2012 (r237406) @@ -1076,6 +1076,7 @@ phi_translate (tree expr, value_set_t se tree newexpr; tree vh = get_value_handle (expr); bool listchanged = false; + bool invariantarg = false; VEC (tree, gc) *vuses = VALUE_HANDLE_VUSES (vh); VEC (tree, gc) *tvuses; @@ -1134,10 +1135,26 @@ phi_translate (tree expr, value_set_t se if (newval != oldval) { listchanged = true; + invariantarg |= is_gimple_min_invariant (newval); TREE_VALUE (newwalker) = get_value_handle (newval); } } } + + /* In case of new invariant args we might try to fold the call + again. */ + if (invariantarg) + { + tree tmp = fold_ternary (CALL_EXPR, TREE_TYPE (expr), + newop0, newarglist, newop2); + if (tmp) + { + STRIP_TYPE_NOPS (tmp); + if (is_gimple_min_invariant (tmp)) + return tmp; + } + } + if (listchanged) vn_lookup_or_add (newarglist, NULL);
Could this fix the issues with mozilla ports not compiling on -O2? I will test later, but I'm just checking if you have any insight... Regards! -- Niclas Zeising _______________________________________________ svn-src-head@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-head To unsubscribe, send any mail to "svn-src-head-unsubscr...@freebsd.org"
