From patchwork Mon May 13 09:27:02 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tom de Vries X-Patchwork-Id: 32656 Received: (qmail 87093 invoked by alias); 13 May 2019 09:27:07 -0000 Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org Delivered-To: mailing list gdb-patches@sourceware.org Received: (qmail 87085 invoked by uid 89); 13 May 2019 09:27:07 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-25.2 required=5.0 tests=AWL, BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, SPF_PASS autolearn=ham version=3.3.1 spammy=reverting X-HELO: mx1.suse.de Received: from mx2.suse.de (HELO mx1.suse.de) (195.135.220.15) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Mon, 13 May 2019 09:27:06 +0000 Received: from relay2.suse.de (unknown [195.135.220.254]) by mx1.suse.de (Postfix) with ESMTP id 332E6ADAB for ; Mon, 13 May 2019 09:27:04 +0000 (UTC) Date: Mon, 13 May 2019 11:27:02 +0200 From: Tom de Vries To: gdb-patches@sourceware.org Subject: [PATCH][gdb/symtab] Fix symbol loading performance regression Message-ID: <20190513092700.GA9733@delia> MIME-Version: 1.0 Content-Disposition: inline User-Agent: Mutt/1.10.1 (2018-07-13) X-IsSubscribed: yes Hi, The commit "[gdb/symtab] Fix language of duplicate static minimal symbol" introduces a performance regression, when loading a cc1 executable build with -O0 -g and gcc 7.4.0. The performance regression, measured in 'real' time is about 175%. The slower execution comes from the fact that the fix in symbol_set_names makes the call to symbol_find_demangled_name unconditional. Fix this by reverting the commit, and redoing the fix as follows. Recapturing the original problem, the first time symbol_set_names is called with gsymbol.language == lang_auto and linkage_name == "_ZL3foov", the name is not present in the per_bfd->demangled_names_hash hash table, so symbol_find_demangled_name is called to demangle the name, after which the mangled/demangled pair is added to the hashtable. The call to symbol_find_demangled_name also sets gsymbol.language to lang_cplus. The second time symbol_set_names is called with gsymbol.language == lang_auto and linkage_name == "_ZL3foov", the name is present in the hash table, so the demangled name from the hash table is used. However, the language of the symbol remains lang_auto. Fix this by adding a field language in struct demangled_name_entry, and using the field in symbol_set_names to set the language of gsymbol, if necessary. Tested on x86_64-linux. OK for trunk? Thanks, - Tom [gdb/symtab] Fix symbol loading performance regression gdb/ChangeLog: 2019-05-11 Tom de Vries PR symtab/24545 * symtab.c (struct demangled_name_entry): Add language field. (symbol_set_names): Revert "[gdb/symtab] Fix language of duplicate static minimal symbol". Set and use language field. --- gdb/symtab.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) diff --git a/gdb/symtab.c b/gdb/symtab.c index 130d5cd48f..44964533ee 100644 --- a/gdb/symtab.c +++ b/gdb/symtab.c @@ -713,6 +713,7 @@ symbol_set_language (struct general_symbol_info *gsymbol, struct demangled_name_entry { const char *mangled; + ENUM_BITFIELD(language) language : LANGUAGE_BITS; char demangled[1]; }; @@ -853,11 +854,6 @@ symbol_set_names (struct general_symbol_info *gsymbol, else linkage_name_copy = linkage_name; - /* Set the symbol language. */ - char *demangled_name_ptr - = symbol_find_demangled_name (gsymbol, linkage_name_copy); - gdb::unique_xmalloc_ptr demangled_name (demangled_name_ptr); - entry.mangled = linkage_name_copy; slot = ((struct demangled_name_entry **) htab_find_slot (per_bfd->demangled_names_hash.get (), @@ -870,7 +866,9 @@ symbol_set_names (struct general_symbol_info *gsymbol, || (gsymbol->language == language_go && (*slot)->demangled[0] == '\0')) { - int demangled_len = demangled_name ? strlen (demangled_name.get ()) : 0; + char *demangled_name = symbol_find_demangled_name (gsymbol, + linkage_name_copy); + int demangled_len = demangled_name ? strlen (demangled_name) : 0; /* Suppose we have demangled_name==NULL, copy_name==0, and linkage_name_copy==linkage_name. In this case, we already have the @@ -906,12 +904,19 @@ symbol_set_names (struct general_symbol_info *gsymbol, strcpy (mangled_ptr, linkage_name_copy); (*slot)->mangled = mangled_ptr; } + (*slot)->language = gsymbol->language; if (demangled_name != NULL) - strcpy ((*slot)->demangled, demangled_name.get()); + { + strcpy ((*slot)->demangled, demangled_name); + xfree (demangled_name); + } else (*slot)->demangled[0] = '\0'; } + else if (gsymbol->language == language_unknown + || gsymbol->language == language_auto) + gsymbol->language = (*slot)->language; gsymbol->name = (*slot)->mangled; if ((*slot)->demangled[0] != '\0')