From patchwork Fri Nov 30 15:09:58 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Teresa Johnson X-Patchwork-Id: 202968 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) by ozlabs.org (Postfix) with SMTP id B59BD2C0086 for ; Sat, 1 Dec 2012 02:10:18 +1100 (EST) Comment: DKIM? See http://www.dkim.org DKIM-Signature: v=1; a=rsa-sha1; c=relaxed/relaxed; d=gcc.gnu.org; s=default; x=1354893020; h=Comment: DomainKey-Signature:Received:Received:Received:Received:Received: Received:Received:Received:To:Subject:Message-Id:Date:From: Mailing-List:Precedence:List-Id:List-Unsubscribe:List-Archive: List-Post:List-Help:Sender:Delivered-To; bh=87oUMtph63s+yUhln5Ve Hgf18jA=; b=T5z9bHNbjIZgHwZZNcU1utTscNhhGliJOdHzYH8YxmimKjRsonB0 PMCyZZ0AsDOCn9zk/JSIqFNoUJtpw55Hfv2q6D/y/E54T7S1Mu8oBX+S1CPgK5Sv q0mz8yBuWAivogszzozAB0E+p6FFB4+JWWuuphNOELd4WBn36CSx3rY= Comment: DomainKeys? See http://antispam.yahoo.com/domainkeys DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=default; d=gcc.gnu.org; h=Received:Received:X-SWARE-Spam-Status:X-Spam-Check-By:Received:Received:X-Google-DKIM-Signature:Received:Received:Received:Received:To:Subject:Message-Id:Date:From:X-Gm-Message-State:X-IsSubscribed:Mailing-List:Precedence:List-Id:List-Unsubscribe:List-Archive:List-Post:List-Help:Sender:Delivered-To; b=PihBsIdNTYnh6lklqN06AbSvelAB1Cg9Ld5QJvs+3KH1gndpf3uI3+9G3hATO8 ZNnrlliJf/jLmo4JGkr8jryrlNdiRBqXMSvRnnAhEO5/DWW5PToLM5/2Di0izifd 1YlaxQyRVrHQTX7AjIwB2YK+i+Ei4IhvcqXv5DUTBgr6Q=; Received: (qmail 8827 invoked by alias); 30 Nov 2012 15:10:10 -0000 Received: (qmail 8805 invoked by uid 22791); 30 Nov 2012 15:10:08 -0000 X-SWARE-Spam-Status: No, hits=-4.4 required=5.0 tests=AWL, BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, KHOP_RCVD_TRUST, RCVD_IN_DNSWL_LOW, RCVD_IN_HOSTKARMA_YE, RP_MATCHES_RCVD X-Spam-Check-By: sourceware.org Received: from mail-ia0-f201.google.com (HELO mail-ia0-f201.google.com) (209.85.210.201) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Fri, 30 Nov 2012 15:10:00 +0000 Received: by mail-ia0-f201.google.com with SMTP id i18so115852iac.2 for ; Fri, 30 Nov 2012 07:10:00 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20120113; h=to:subject:message-id:date:from:x-gm-message-state; bh=eqn0h1EFKeLt1b1SclYdGtj8He5wFeR3hBLFCSAISmw=; b=IZVkw6UT70e6y/90mPlevv76QiTnlwbECN8By2aCR8S6513314m45APlFS6nSzF6pJ +XNYARE25mBk4KmRtcudvvOIcCfsuKsmYwSLfLpXRHGT9utreQbgS5s/OerRk7RwhJhq VWqhpRJJnaGi6S6Ig9hiQj+/In5tPPur+5oKi4YEp6A489qQyvwaHwx3/37RN4xoqhIW m5sQlzrI/+J84e2OtXeI1E8XlC5wWUQ1tmYP+3ga4jbPB2YB95Jm7/2u5Fdl0PsujgKj 6ERaHhlSq+IdOcGphyAwM3x7775W4iqMUZmPTJjuWPNMxIWJbGc0Ab2mDRuvQ9LQ1qNT HRZQ== Received: by 10.50.5.210 with SMTP id u18mr1436973igu.4.1354288200037; Fri, 30 Nov 2012 07:10:00 -0800 (PST) Received: from wpzn3.hot.corp.google.com (216-239-44-65.google.com [216.239.44.65]) by gmr-mx.google.com with ESMTPS id dx8si1325439igc.1.2012.11.30.07.09.59 (version=TLSv1/SSLv3 cipher=AES128-SHA); Fri, 30 Nov 2012 07:10:00 -0800 (PST) Received: from tjsboxrox.mtv.corp.google.com (tjsboxrox.mtv.corp.google.com [172.17.129.49]) by wpzn3.hot.corp.google.com (Postfix) with ESMTP id E6D6D100047; Fri, 30 Nov 2012 07:09:58 -0800 (PST) Received: by tjsboxrox.mtv.corp.google.com (Postfix, from userid 147431) id 7B61061422; Fri, 30 Nov 2012 07:09:58 -0800 (PST) To: reply@codereview.appspotmail.com,hubicka@ucw.cz,gcc-patches@gcc.gnu.org Subject: [PATCH] Stream profile summary histogram through LTO files (issue6782131) Message-Id: <20121130150958.7B61061422@tjsboxrox.mtv.corp.google.com> Date: Fri, 30 Nov 2012 07:09:58 -0800 (PST) From: tejohnson@google.com (Teresa Johnson) X-Gm-Message-State: ALoCoQn3hFWZyM2/F+PDFkiArQ2njAJ247eFDmDSQIWstHp2mTwz5Fjxev0zRDhTDIOtLFdoemiRa/0e0CyfUWjEfnenqBj2u7N4yHpLPO6JWpulTQqdxaY5NMhUrLZt1vaJdJjfeEtq6LbgElNQC7wvK5wMjmyIoqfpRQudOqlPrgPcyfZeic4Y7hEWv6+/8ssS+oO84AFH X-IsSubscribed: yes Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org Revised patch to ensure that histograms from the profile summary are streamed through the LTO files so that the working set can be computed for use in downstream optimizations. Bootstrapped and tested on x86_64-unknown-linux-gnu. Ok for trunk? Thanks, Teresa 2012-11-29 Teresa Johnson * lto-cgraph.c (output_profile_summary): Stream out sum_all and histogram. (input_profile_summary): Stream in sum_all and histogram. (merge_profile_summaries): Merge sum_all and histogram, and change to use RDIV. (input_symtab): Call compute_working_sets after merging summaries. * gcov-io.c (gcov_histo_index): Make extern for compiler. * gcov-io.h (gcov_histo_index): Ditto. * profile.c (compute_working_sets): Remove static keyword. * profile.h (compute_working_sets): Ditto. * Makefile.in (lto-cgraph.o): Depend on profile.h. --- This patch is available for review at 1 Index: lto-cgraph.c =================================================================== --- lto-cgraph.c (revision 193909) +++ lto-cgraph.c (working copy) @@ -46,6 +46,7 @@ along with GCC; see the file COPYING3. If not see #include "tree-streamer.h" #include "gcov-io.h" #include "tree-pass.h" +#include "profile.h" static void output_cgraph_opt_summary (void); static void input_cgraph_opt_summary (vec nodes); @@ -593,14 +594,39 @@ lto_output_ref (struct lto_simple_output_block *ob static void output_profile_summary (struct lto_simple_output_block *ob) { + unsigned h_ix; + struct bitpack_d bp; + if (profile_info) { - /* We do not output num, sum_all and run_max, they are not used by - GCC profile feedback and they are difficult to merge from multiple - units. */ + /* We do not output num and run_max, they are not used by + GCC profile feedback and they are difficult to merge from multiple + units. */ gcc_assert (profile_info->runs); streamer_write_uhwi_stream (ob->main_stream, profile_info->runs); streamer_write_uhwi_stream (ob->main_stream, profile_info->sum_max); + + /* sum_all is needed for computing the working set with the + histogram. */ + streamer_write_uhwi_stream (ob->main_stream, profile_info->sum_all); + + /* Create and output a bitpack of non-zero histogram entries indices. */ + bp = bitpack_create (ob->main_stream); + for (h_ix = 0; h_ix < GCOV_HISTOGRAM_SIZE; h_ix++) + bp_pack_value (&bp, profile_info->histogram[h_ix].num_counters > 0, 1); + streamer_write_bitpack (&bp); + /* Now stream out only those non-zero entries. */ + for (h_ix = 0; h_ix < GCOV_HISTOGRAM_SIZE; h_ix++) + { + if (!profile_info->histogram[h_ix].num_counters) + continue; + streamer_write_uhwi_stream (ob->main_stream, + profile_info->histogram[h_ix].num_counters); + streamer_write_uhwi_stream (ob->main_stream, + profile_info->histogram[h_ix].min_value); + streamer_write_uhwi_stream (ob->main_stream, + profile_info->histogram[h_ix].cum_value); + } } else streamer_write_uhwi_stream (ob->main_stream, 0); @@ -1227,11 +1253,38 @@ static void input_profile_summary (struct lto_input_block *ib, struct lto_file_decl_data *file_data) { + unsigned h_ix; + struct bitpack_d bp; unsigned int runs = streamer_read_uhwi (ib); if (runs) { file_data->profile_info.runs = runs; file_data->profile_info.sum_max = streamer_read_uhwi (ib); + file_data->profile_info.sum_all = streamer_read_uhwi (ib); + + memset (file_data->profile_info.histogram, 0, + sizeof (gcov_bucket_type) * GCOV_HISTOGRAM_SIZE); + /* Input the bitpack of non-zero histogram indices. */ + bp = streamer_read_bitpack (ib); + /* Read in and unpack the full bitpack, flagging non-zero + histogram entries by setting the num_counters non-zero. */ + for (h_ix = 0; h_ix < GCOV_HISTOGRAM_SIZE; h_ix++) + { + file_data->profile_info.histogram[h_ix].num_counters + = bp_unpack_value (&bp, 1); + } + for (h_ix = 0; h_ix < GCOV_HISTOGRAM_SIZE; h_ix++) + { + if (!file_data->profile_info.histogram[h_ix].num_counters) + continue; + + file_data->profile_info.histogram[h_ix].num_counters + = streamer_read_uhwi (ib); + file_data->profile_info.histogram[h_ix].min_value + = streamer_read_uhwi (ib); + file_data->profile_info.histogram[h_ix].cum_value + = streamer_read_uhwi (ib); + } } } @@ -1242,10 +1295,13 @@ static void merge_profile_summaries (struct lto_file_decl_data **file_data_vec) { struct lto_file_decl_data *file_data; - unsigned int j; + unsigned int j, h_ix; gcov_unsigned_t max_runs = 0; struct cgraph_node *node; struct cgraph_edge *edge; + gcov_type saved_sum_all = 0; + gcov_ctr_summary *saved_profile_info = 0; + int saved_scale = 0; /* Find unit with maximal number of runs. If we ever get serious about roundoff errors, we might also consider computing smallest common @@ -1269,6 +1325,8 @@ merge_profile_summaries (struct lto_file_decl_data profile_info = <o_gcov_summary; lto_gcov_summary.runs = max_runs; lto_gcov_summary.sum_max = 0; + memset (lto_gcov_summary.histogram, 0, + sizeof (gcov_bucket_type) * GCOV_HISTOGRAM_SIZE); /* Rescale all units to the maximal number of runs. sum_max can not be easily merged, as we have no idea what files come from @@ -1276,16 +1334,48 @@ merge_profile_summaries (struct lto_file_decl_data for (j = 0; (file_data = file_data_vec[j]) != NULL; j++) if (file_data->profile_info.runs) { - int scale = ((REG_BR_PROB_BASE * max_runs - + file_data->profile_info.runs / 2) - / file_data->profile_info.runs); + int scale = RDIV (REG_BR_PROB_BASE * max_runs, + file_data->profile_info.runs); lto_gcov_summary.sum_max = MAX (lto_gcov_summary.sum_max, - (file_data->profile_info.sum_max - * scale - + REG_BR_PROB_BASE / 2) - / REG_BR_PROB_BASE); + RDIV (file_data->profile_info.sum_max + * scale, REG_BR_PROB_BASE)); + lto_gcov_summary.sum_all = MAX (lto_gcov_summary.sum_all, + RDIV (file_data->profile_info.sum_all + * scale, REG_BR_PROB_BASE)); + /* Save a pointer to the profile_info with the largest + scaled sum_all and the scale for use in merging the + histogram. */ + if (lto_gcov_summary.sum_all > saved_sum_all) + { + saved_profile_info = &file_data->profile_info; + saved_sum_all = lto_gcov_summary.sum_all; + saved_scale = scale; + } } + gcc_assert (saved_profile_info); + + /* Scale up the histogram from the profile that had the largest + scaled sum_all above. */ + for (h_ix = 0; h_ix < GCOV_HISTOGRAM_SIZE; h_ix++) + { + /* Scale up the min value as we did the corresponding sum_all + above. Use that to find the new histogram index. */ + int scaled_min = RDIV (saved_profile_info->histogram[h_ix].min_value + * saved_scale, REG_BR_PROB_BASE); + unsigned new_ix = gcov_histo_index (scaled_min); + lto_gcov_summary.histogram[new_ix].min_value = scaled_min; + /* Some of the scaled counter values would ostensibly need to be placed + into different (larger) histogram buckets, but we keep things simple + here and place the scaled cumulative counter value in the bucket + corresponding to the scaled minimum counter value. */ + lto_gcov_summary.histogram[new_ix].cum_value + = RDIV (saved_profile_info->histogram[h_ix].cum_value + * saved_scale, REG_BR_PROB_BASE); + lto_gcov_summary.histogram[new_ix].num_counters + = saved_profile_info->histogram[h_ix].num_counters; + } + /* Watch roundoff errors. */ if (lto_gcov_summary.sum_max < max_runs) lto_gcov_summary.sum_max = max_runs; @@ -1303,10 +1393,8 @@ merge_profile_summaries (struct lto_file_decl_data { int scale; - scale = - ((node->count_materialization_scale * max_runs - + node->symbol.lto_file_data->profile_info.runs / 2) - / node->symbol.lto_file_data->profile_info.runs); + scale = RDIV (node->count_materialization_scale * max_runs, + node->symbol.lto_file_data->profile_info.runs); node->count_materialization_scale = scale; if (scale < 0) fatal_error ("Profile information in %s corrupted", @@ -1315,10 +1403,8 @@ merge_profile_summaries (struct lto_file_decl_data if (scale == REG_BR_PROB_BASE) continue; for (edge = node->callees; edge; edge = edge->next_callee) - edge->count = ((edge->count * scale + REG_BR_PROB_BASE / 2) - / REG_BR_PROB_BASE); - node->count = ((node->count * scale + REG_BR_PROB_BASE / 2) - / REG_BR_PROB_BASE); + edge->count = RDIV (edge->count * scale, REG_BR_PROB_BASE); + node->count = RDIV (node->count * scale, REG_BR_PROB_BASE); } } @@ -1365,7 +1451,9 @@ input_symtab (void) } merge_profile_summaries (file_data_vec); + compute_working_sets (); + /* Clear out the aux field that was used to store enough state to tell which nodes should be overwritten. */ FOR_EACH_FUNCTION (node) Index: gcov-io.c =================================================================== --- gcov-io.c (revision 193909) +++ gcov-io.c (working copy) @@ -622,11 +622,15 @@ gcov_time (void) } #endif /* IN_GCOV */ -#if IN_LIBGCOV || !IN_GCOV +#if !IN_GCOV /* Determine the index into histogram for VALUE. */ +#if IN_LIBGCOV static unsigned -gcov_histo_index(gcov_type value) +#else +GCOV_LINKAGE unsigned +#endif +gcov_histo_index (gcov_type value) { gcov_type_unsigned v = (gcov_type_unsigned)value; unsigned r = 0; @@ -664,8 +668,8 @@ static unsigned its entry's original cumulative counter value when computing the new merged cum_value. */ -static void gcov_histogram_merge(gcov_bucket_type *tgt_histo, - gcov_bucket_type *src_histo) +static void gcov_histogram_merge (gcov_bucket_type *tgt_histo, + gcov_bucket_type *src_histo) { int src_i, tgt_i, tmp_i = 0; unsigned src_num, tgt_num, merge_num; @@ -801,4 +805,4 @@ static unsigned /* Finally, copy the merged histogram into tgt_histo. */ memcpy(tgt_histo, tmp_histo, sizeof (gcov_bucket_type) * GCOV_HISTOGRAM_SIZE); } -#endif /* IN_LIBGCOV || !IN_GCOV */ +#endif /* !IN_GCOV */ Index: gcov-io.h =================================================================== --- gcov-io.h (revision 193909) +++ gcov-io.h (working copy) @@ -612,6 +612,7 @@ GCOV_LINKAGE void gcov_write_unsigned (gcov_unsign #if !IN_GCOV && !IN_LIBGCOV /* Available only in compiler */ +GCOV_LINKAGE unsigned gcov_histo_index (gcov_type value); GCOV_LINKAGE void gcov_write_string (const char *); GCOV_LINKAGE gcov_position_t gcov_write_tag (gcov_unsigned_t); GCOV_LINKAGE void gcov_write_length (gcov_position_t /*position*/); Index: profile.c =================================================================== --- profile.c (revision 193909) +++ profile.c (working copy) @@ -207,7 +207,7 @@ instrument_values (histogram_values values) the number of counters required to cover that working set percentage and the minimum counter value in that working set. */ -static void +void compute_working_sets (void) { gcov_type working_set_cum_values[NUM_GCOV_WORKING_SETS]; Index: profile.h =================================================================== --- profile.h (revision 193909) +++ profile.h (working copy) @@ -47,4 +47,6 @@ extern gcov_type sum_edge_counts (vec extern void init_node_map (void); extern void del_node_map (void); +extern void compute_working_sets (void); + #endif /* PROFILE_H */ Index: Makefile.in =================================================================== --- Makefile.in (revision 193909) +++ Makefile.in (working copy) @@ -2163,7 +2163,7 @@ lto-cgraph.o: lto-cgraph.c $(CONFIG_H) $(SYSTEM_H) $(HASHTAB_H) langhooks.h $(BASIC_BLOCK_H) \ $(TREE_FLOW_H) $(CGRAPH_H) $(FUNCTION_H) $(GGC_H) $(DIAGNOSTIC_CORE_H) \ $(EXCEPT_H) $(TIMEVAR_H) pointer-set.h $(LTO_STREAMER_H) \ - $(GCOV_IO_H) $(DATA_STREAMER_H) $(TREE_STREAMER_H) $(TREE_PASS_H) + $(GCOV_IO_H) $(DATA_STREAMER_H) $(TREE_STREAMER_H) $(TREE_PASS_H) profile.h lto-streamer-in.o: lto-streamer-in.c $(CONFIG_H) $(SYSTEM_H) coretypes.h \ $(TM_H) toplev.h $(DIAGNOSTIC_CORE_H) $(EXPR_H) $(FLAGS_H) $(PARAMS_H) \ input.h $(HASHTAB_H) $(BASIC_BLOCK_H) $(TREE_FLOW_H) $(TREE_PASS_H) \