[{"id":1769077,"web_url":"http://patchwork.ozlabs.org/comment/1769077/","msgid":"","list_archive_url":null,"date":"2017-09-15T09:31:19","subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","submitter":{"id":1765,"url":"http://patchwork.ozlabs.org/api/people/1765/","name":"Richard Biener","email":"richard.guenther@gmail.com"},"content":"On Fri, Sep 15, 2017 at 3:27 AM, Kugan Vivekanandarajah\n wrote:\n> This patch adds separate params for rtl unroller so that they can be\n> tunned accordingly. Default values I have are based on some testing on\n> aarch64. I am happy to leave it as the current value and set them in\n> the back-end.\n\nPARAM_MAX_AVERAGE_UNROLLED_INSNS is only used by the RTL\nunroller. Why should we separate PARAM_MAX_UNROLL_TIMES?\n\nPARAM_MAX_UNROLLED_INSNS is only used by gimple passes\nthat perform unrolling. Since GIMPLE is three-address it should\nmatch RTL reasonably well -- but I'd be ok in having a separate param\nfor those. But I wouldn't name those 'partial'.\n\nThat said, those are magic numbers and I expect we can find some\nthat work well on RTL and GIMPLE.\n\nRichard.\n\n>\n> Thanks,\n> Kugan\n>\n>\n> gcc/ChangeLog:\n>\n> 2017-09-12 Kugan Vivekanandarajah \n>\n> * loop-unroll.c (decide_unroll_constant_iterations): Use new params.\n> (decide_unroll_runtime_iterations): Likewise.\n> (decide_unroll_stupid): Likewise.\n> * params.def (DEFPARAM): Separate and add new params for rtl unroller.","headers":{"Return-Path":"","X-Original-To":"incoming@patchwork.ozlabs.org","Delivered-To":["patchwork-incoming@bilbo.ozlabs.org","mailing list gcc-patches@gcc.gnu.org"],"Authentication-Results":["ozlabs.org;\n\tspf=pass (mailfrom) smtp.mailfrom=gcc.gnu.org\n\t(client-ip=209.132.180.131; helo=sourceware.org;\n\tenvelope-from=gcc-patches-return-462210-incoming=patchwork.ozlabs.org@gcc.gnu.org;\n\treceiver=)","ozlabs.org; dkim=pass (1024-bit key;\n\tunprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org\n\theader.b=\"cYko8pRA\"; dkim-atps=neutral","sourceware.org; auth=none"],"Received":["from sourceware.org (server1.sourceware.org [209.132.180.131])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256\n\tbits)) (No client certificate requested)\n\tby ozlabs.org (Postfix) with ESMTPS id 3xtqt05rlBz9s5L\n\tfor ;\n\tFri, 15 Sep 2017 19:31:51 +1000 (AEST)","(qmail 85195 invoked by alias); 15 Sep 2017 09:31:24 -0000","(qmail 85101 invoked by uid 89); 15 Sep 2017 09:31:24 -0000","from mail-wm0-f42.google.com (HELO mail-wm0-f42.google.com)\n\t(74.125.82.42) by sourceware.org\n\t(qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP;\n\tFri, 15 Sep 2017 09:31:22 +0000","by mail-wm0-f42.google.com with SMTP id v142so6656975wmv.5 for\n\t; Fri, 15 Sep 2017 02:31:21 -0700 (PDT)","by 10.80.180.250 with HTTP; Fri, 15 Sep 2017 02:31:19 -0700 (PDT)"],"DomainKey-Signature":"a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; q=dns; s=default; b=XWmIaRgQGBmPQnQ\n\tBQe/UwcCJlOFik64R+712YY4zfC6n5z9vGNvdZVaBSAV1nigDKm+gEjLpd8HLlvh\n\tnWkApJgl9OZ40/Wvu0zIbzAk+yX/Gr6w/EQ6Otks05oyGAaH+D0tj/QpmG9+uSiN\n\ts+V97rUUWpvzH3s2Hq4A0Tuutd18=","DKIM-Signature":"v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; s=default; bh=tCe88nZqgsdWrljk5UioT\n\tPAd/EM=; b=cYko8pRAuYyNgHG4oqGxJn4OGS3D85dm7AEBIQfU4BoeKebaBv2uJ\n\tRLPzmILhmNXLoU+f1UEyi3t2hbacryPp52jspWyWXOfZUqSeiHc+qEgP1p2w4EdL\n\t5m/OShkuO1tzre6lZ+fm95BAOfSPEn6PdlyMcwCiwA9WLHybJZHRnY=","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","X-Virus-Found":"No","X-Spam-SWARE-Status":"No, score=-1.5 required=5.0 tests=AWL, BAYES_00,\n\tFREEMAIL_FROM, RCVD_IN_DNSWL_NONE, RCVD_IN_SORBS_SPAM,\n\tSPF_PASS autolearn=no version=3.3.2 spammy=Hx-languages-length:1181","X-HELO":"mail-wm0-f42.google.com","X-Google-DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net;\n\ts=20161025;\n\th=x-gm-message-state:mime-version:in-reply-to:references:from:date\n\t:message-id:subject:to:cc;\n\tbh=tmxPDFFxnBimXH9PZLwlQRlYseGAfY9+x3vGIIEa6lI=;\n\tb=ISQsmmDeOzgcQN+Yd6aLXhW2CSFfLqbn2qKScSD3R/kINJvbghvSNgyV/b/qqqPNZ3\n\toDhVvAiLs6d9Nn2cFRtguFhwVskKfA3JbKQ2VuSQoBjTNecdeZD58+M2/aje8fSgEItL\n\tLOw3pRWYN+FoovqURNLUfMEecb19SPOffMCC4SFCM0VkyqxE5DGUh5oqgGETNpX2F0xE\n\tW58di8Ebj6lFSHbGQ6vvhn6VwkJueY/VYOwATJf1YBFVUauJ+wUCEDN3WfY/Z+qpphOQ\n\tMZDHDVAps008eJv83idttD+lSAprNxaAB62d+lN+vpHTEoqu0AbJzGapnjiv3SVNFqkP\n\t98OA==","X-Gm-Message-State":"AHPjjUhAsH12UV1NBj53f8YYY3qNKKBW8ffVR/DvZTkNf8feh7DoyXEs\t8fnpIegfzU5TyrVt/y239QB1ngiOWWnUPAzTlfPLVSBM","X-Google-Smtp-Source":"AOwi7QDVJsR3+ziODJHkI0AYrLjOx4doUaLMQyLq0RkkK/vhjZEnPVBPnGcBLk+k0znnzIy+43FQG7dkp0L0GJdsL48=","X-Received":"by 10.80.194.217 with SMTP id u25mr6783804edf.260.1505467879565;\n\tFri, 15 Sep 2017 02:31:19 -0700 (PDT)","MIME-Version":"1.0","In-Reply-To":"","References":"","From":"Richard Biener ","Date":"Fri, 15 Sep 2017 11:31:19 +0200","Message-ID":"","Subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","To":"Kugan Vivekanandarajah ","Cc":"\"gcc-patches@gcc.gnu.org\" ","Content-Type":"text/plain; charset=\"UTF-8\"","X-IsSubscribed":"yes"}},{"id":1769833,"web_url":"http://patchwork.ozlabs.org/comment/1769833/","msgid":"","list_archive_url":null,"date":"2017-09-18T01:36:40","subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","submitter":{"id":25768,"url":"http://patchwork.ozlabs.org/api/people/25768/","name":"Kugan Vivekanandarajah","email":"kugan.vivekanandarajah@linaro.org"},"content":"Hi Richard,\n\nOn 15 September 2017 at 19:31, Richard Biener\n wrote:\n> On Fri, Sep 15, 2017 at 3:27 AM, Kugan Vivekanandarajah\n> wrote:\n>> This patch adds separate params for rtl unroller so that they can be\n>> tunned accordingly. Default values I have are based on some testing on\n>> aarch64. I am happy to leave it as the current value and set them in\n>> the back-end.\n>\n> PARAM_MAX_AVERAGE_UNROLLED_INSNS is only used by the RTL\n> unroller. Why should we separate PARAM_MAX_UNROLL_TIMES?\n>\n> PARAM_MAX_UNROLLED_INSNS is only used by gimple passes\n> that perform unrolling. Since GIMPLE is three-address it should\n> match RTL reasonably well -- but I'd be ok in having a separate param\n> for those. But I wouldn't name those 'partial'.\n>\n> That said, those are magic numbers and I expect we can find some\n> that work well on RTL and GIMPLE.\n\nThanks for the review. I am mostly interested in having separate\nparams for RTL runtime unrolling as this is different to what GIMPLE\nunroller does. May be I should have separate params only for the\nruntime unrolling (or the partial unroller) and let RTL/GIMPLE share\nthe other. Any preference here ? Any preference on the name ?\n\nI am suspecting that RTL unroller which does the same as GIMPLE is\nkind of obsolete now?\n\nThanks,\nKugan\n\n\n\n\n\n> Richard.\n>\n>>\n>> Thanks,\n>> Kugan\n>>\n>>\n>> gcc/ChangeLog:\n>>\n>> 2017-09-12 Kugan Vivekanandarajah \n>>\n>> * loop-unroll.c (decide_unroll_constant_iterations): Use new params.\n>> (decide_unroll_runtime_iterations): Likewise.\n>> (decide_unroll_stupid): Likewise.\n>> * params.def (DEFPARAM): Separate and add new params for rtl unroller.","headers":{"Return-Path":"","X-Original-To":"incoming@patchwork.ozlabs.org","Delivered-To":["patchwork-incoming@bilbo.ozlabs.org","mailing list gcc-patches@gcc.gnu.org"],"Authentication-Results":["ozlabs.org;\n\tspf=pass (mailfrom) smtp.mailfrom=gcc.gnu.org\n\t(client-ip=209.132.180.131; helo=sourceware.org;\n\tenvelope-from=gcc-patches-return-462346-incoming=patchwork.ozlabs.org@gcc.gnu.org;\n\treceiver=)","ozlabs.org; dkim=pass (1024-bit key;\n\tunprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org\n\theader.b=\"oJx+iZ6x\"; dkim-atps=neutral","sourceware.org; auth=none"],"Received":["from sourceware.org (server1.sourceware.org [209.132.180.131])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256\n\tbits)) (No client certificate requested)\n\tby ozlabs.org (Postfix) with ESMTPS id 3xwTBZ4ffVz9s7G\n\tfor ;\n\tMon, 18 Sep 2017 11:36:52 +1000 (AEST)","(qmail 21022 invoked by alias); 18 Sep 2017 01:36:44 -0000","(qmail 21005 invoked by uid 89); 18 Sep 2017 01:36:43 -0000","from mail-qk0-f174.google.com (HELO mail-qk0-f174.google.com)\n\t(209.85.220.174) by sourceware.org\n\t(qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP;\n\tMon, 18 Sep 2017 01:36:42 +0000","by mail-qk0-f174.google.com with SMTP id d70so6040119qkc.11 for\n\t; Sun, 17 Sep 2017 18:36:42 -0700 (PDT)","by 10.237.37.211 with HTTP; Sun, 17 Sep 2017 18:36:40 -0700 (PDT)"],"DomainKey-Signature":"a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; q=dns; s=default; b=XugFV+IXhV5xzzp\n\txCoyt92AtBH/mjLgjMJcW9boeC7d8kcZ6GI1bMCKExARZFvqVeJz+3nveOvQ3WCY\n\tfiQMen/Qo/EFPU6pJsBFx48kYhqw3wzqPwcj+DU7Vf/f+XObqBErS5rwKLoYeHuI\n\t4x2OMM3Y36q/afuTmGoAphBeg7sc=","DKIM-Signature":"v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; s=default; bh=burCptRAnWr5lHGBJejw0\n\t08/KWw=; b=oJx+iZ6xqBHOLJNJpdVd5kzsz5VDV3G/vDSF2c5V3SpQ/hGjohoL5\n\t5+lvM0fPsVy8EN4pZlenPsdAG7lVTFSDTp9Ssb7I89yYG0C6XBA2H3UVcsewFwWX\n\t4fzi3sjbrgCsojWwnuhXTxzLuROAEY/0bQ/qOUMX/RITxhGlG6WYVA=","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","X-Virus-Found":"No","X-Spam-SWARE-Status":"No, score=-1.6 required=5.0 tests=AWL, BAYES_00,\n\tRCVD_IN_DNSWL_NONE, RCVD_IN_SORBS_SPAM,\n\tSPF_PASS autolearn=no version=3.3.2 spammy=Hx-languages-length:1755","X-HELO":"mail-qk0-f174.google.com","X-Google-DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net;\n\ts=20161025;\n\th=x-gm-message-state:mime-version:in-reply-to:references:from:date\n\t:message-id:subject:to:cc;\n\tbh=DvzBcXgB+ZN22ORa2SlVI/HHoJzGcykAI52hOSLWwHM=;\n\tb=aFKk1/6A3BljGyZ2aL6hcviFe1K99vhcIVaZmNvePPP56tIXBdfUvIsdP8nbr6iCV2\n\tvhgJTqY2gZyikY2vFF5ryId8Y3pDFRmBV10lM9n37QZlt0kTdBfIjj8/CwPw6QBO3fwq\n\t46hd7C99TLh/qbSDCo6p9tWWTuNEj/p3ixGbQV5Ud9PQDm+bxyAXYFti2O4mrwbfiFba\n\tvD57cfWoK6HHQ4ndxeSnJ0gbOiUemQvEIpcVRNhfMXQ1aNDilpcDVNvqiqDV8qOYVv9z\n\tNUCk9+wjdS6Gp9V7htjEwPVVK0RwTzBYfB2UQs0RLM1U0WrIxQYJ0D660xA4EisEy/X9\n\tMalw==","X-Gm-Message-State":"AHPjjUi+SWjtkzuhXbl1tVu4smx663YhHUwLTxzxlRn6aWb5yHbJr1M7\trRRUPKMGScUWyElSL/gmfm+chh1wL9NzBwP3SiPcRcpsBx0=","X-Google-Smtp-Source":"AOwi7QAf5Wfm0TBHRqQcCYb2E718biDBfQuOc2TEvBV0ML9B/wY49UAknHepzvjTLQg8dBmLEwOtbQwO0evGk0asNBI=","X-Received":"by 10.55.104.131 with SMTP id d125mr17042886qkc.37.1505698600670;\n\tSun, 17 Sep 2017 18:36:40 -0700 (PDT)","MIME-Version":"1.0","In-Reply-To":"","References":"\n\t","From":"Kugan Vivekanandarajah ","Date":"Mon, 18 Sep 2017 11:36:40 +1000","Message-ID":"","Subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","To":"Richard Biener ","Cc":"\"gcc-patches@gcc.gnu.org\" ","Content-Type":"text/plain; charset=\"UTF-8\"","X-IsSubscribed":"yes"}},{"id":1769933,"web_url":"http://patchwork.ozlabs.org/comment/1769933/","msgid":"","list_archive_url":null,"date":"2017-09-18T07:50:32","subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","submitter":{"id":1765,"url":"http://patchwork.ozlabs.org/api/people/1765/","name":"Richard Biener","email":"richard.guenther@gmail.com"},"content":"On Mon, Sep 18, 2017 at 3:36 AM, Kugan Vivekanandarajah\n wrote:\n> Hi Richard,\n>\n> On 15 September 2017 at 19:31, Richard Biener\n> wrote:\n>> On Fri, Sep 15, 2017 at 3:27 AM, Kugan Vivekanandarajah\n>> wrote:\n>>> This patch adds separate params for rtl unroller so that they can be\n>>> tunned accordingly. Default values I have are based on some testing on\n>>> aarch64. I am happy to leave it as the current value and set them in\n>>> the back-end.\n>>\n>> PARAM_MAX_AVERAGE_UNROLLED_INSNS is only used by the RTL\n>> unroller. Why should we separate PARAM_MAX_UNROLL_TIMES?\n>>\n>> PARAM_MAX_UNROLLED_INSNS is only used by gimple passes\n>> that perform unrolling. Since GIMPLE is three-address it should\n>> match RTL reasonably well -- but I'd be ok in having a separate param\n>> for those. But I wouldn't name those 'partial'.\n>>\n>> That said, those are magic numbers and I expect we can find some\n>> that work well on RTL and GIMPLE.\n>\n> Thanks for the review. I am mostly interested in having separate\n> params for RTL runtime unrolling as this is different to what GIMPLE\n> unroller does.\n\nWhy? Do you just want to have more magic knobs to machine-auto-tune?\n\n> May be I should have separate params only for the\n> runtime unrolling (or the partial unroller) and let RTL/GIMPLE share\n> the other. Any preference here ? Any preference on the name ?\n>\n> I am suspecting that RTL unroller which does the same as GIMPLE is\n> kind of obsolete now?\n\nWe do not have a GIMPLE loop unroller pass. On GIMPLE we only do\ncomplete peeling as a separate pass and several passes perform\nunrolling as part of their transform.\n\nRichard.\n\n> Thanks,\n> Kugan\n>\n>\n>\n>\n>\n>> Richard.\n>>\n>>>\n>>> Thanks,\n>>> Kugan\n>>>\n>>>\n>>> gcc/ChangeLog:\n>>>\n>>> 2017-09-12 Kugan Vivekanandarajah \n>>>\n>>> * loop-unroll.c (decide_unroll_constant_iterations): Use new params.\n>>> (decide_unroll_runtime_iterations): Likewise.\n>>> (decide_unroll_stupid): Likewise.\n>>> * params.def (DEFPARAM): Separate and add new params for rtl unroller.","headers":{"Return-Path":"","X-Original-To":"incoming@patchwork.ozlabs.org","Delivered-To":["patchwork-incoming@bilbo.ozlabs.org","mailing list gcc-patches@gcc.gnu.org"],"Authentication-Results":["ozlabs.org;\n\tspf=pass (mailfrom) smtp.mailfrom=gcc.gnu.org\n\t(client-ip=209.132.180.131; helo=sourceware.org;\n\tenvelope-from=gcc-patches-return-462354-incoming=patchwork.ozlabs.org@gcc.gnu.org;\n\treceiver=)","ozlabs.org; dkim=pass (1024-bit key;\n\tunprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org\n\theader.b=\"K0ouDyJ4\"; dkim-atps=neutral","sourceware.org; auth=none"],"Received":["from sourceware.org (server1.sourceware.org [209.132.180.131])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256\n\tbits)) (No client certificate requested)\n\tby ozlabs.org (Postfix) with ESMTPS id 3xwdV11kgnz9s3w\n\tfor ;\n\tMon, 18 Sep 2017 17:50:44 +1000 (AEST)","(qmail 98409 invoked by alias); 18 Sep 2017 07:50:36 -0000","(qmail 98032 invoked by uid 89); 18 Sep 2017 07:50:35 -0000","from mail-wm0-f50.google.com (HELO mail-wm0-f50.google.com)\n\t(74.125.82.50) by sourceware.org\n\t(qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP;\n\tMon, 18 Sep 2017 07:50:34 +0000","by mail-wm0-f50.google.com with SMTP id a137so11745411wma.0 for\n\t; Mon, 18 Sep 2017 00:50:34 -0700 (PDT)","by 10.80.180.250 with HTTP; Mon, 18 Sep 2017 00:50:32 -0700 (PDT)"],"DomainKey-Signature":"a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; q=dns; s=default; b=UpU+Lih8zq7P99R\n\tfWMMxM5WAuxcFup4uFqeqaSyyT1hfBAbZOlFElDFbrYqh/NQfLlE7jaZIPd4mVKb\n\tXpkCWlYRX6nhXvDtvDUST6qFamWm0yzxTYg26ibX3raEyRbtMPSW7AGV8bj/Vmu+\n\tV2bjz5DAMase47vKZEdx2GuSL6Zs=","DKIM-Signature":"v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; s=default; bh=ARCSY/F9p34AcgexT+zDv\n\td+/3xI=; b=K0ouDyJ4S2M0zQB6o5/0D+49wiHka3nMUe5RsxPZLcBGvM792idaQ\n\tEOlntI9rfzS6JNEJmTGbLWg9I4mXgG6Yu3L1rDJv8j/nEyu2va+edtoTLJMOAk0t\n\t57I++IoAuMnCA72X3TyqNryKQc0kzNRCvNQyOyaok2gw1p80aZ8nSw=","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","X-Virus-Found":"No","X-Spam-SWARE-Status":"No, score=-1.5 required=5.0 tests=AWL, BAYES_00,\n\tFREEMAIL_FROM, RCVD_IN_DNSWL_NONE, RCVD_IN_SORBS_SPAM,\n\tSPF_PASS autolearn=no version=3.3.2 spammy=","X-HELO":"mail-wm0-f50.google.com","X-Google-DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net;\n\ts=20161025;\n\th=x-gm-message-state:mime-version:in-reply-to:references:from:date\n\t:message-id:subject:to:cc;\n\tbh=v2p/umXPT99rTfS2/iaVlLnAPrtkig0MYKQ5oGyyUTQ=;\n\tb=ZDIGk3iTShW3YFAjVmTkRLQ9pf+iyCZZ7VgfCPS3DGWyaEoJbuC6sTEP3QA4r1AVek\n\tO2fHBJEi167A9r7gjuHW01jcSHivvbc8w57IPF6qR/Q7g52NJ7V2v4mwb/EERUpEkhhb\n\tKR0iUAxoaPD8D/QaHIPdUMuWj8zRz6fUAd4aA3RMiRcEJ72D6WPPT7jnen99MahFE2Zp\n\t8TmogkzUNDBgKeyEQzJvINoELF32VVY9+OyEOofA9PN9UhI6oNjD881s2yuJ63iP9utC\n\tZUbynqKQYkfNs2379yUXwDBuA5rxjh/TNpvcoxStPEkCAXbGC1Wi/yAwYqBDL3v2zBix\n\t3MBg==","X-Gm-Message-State":"AHPjjUik28sns6Jus9xfpVUyPU6Iqi35afmSV10XAZb2tBbZU17E79uC\trlhdB37V/KwGYqo/GV0e50heQItyK8WTI4gWuHE=","X-Google-Smtp-Source":"AOwi7QBFtYTdTHeMSfNF0CmpuFFIHz/fM75uWQe+2Wk7UY/l+dEinXmfFW1GR7mWdGY3rfThT7O+5E7jsl9oE9p6h1Y=","X-Received":"by 10.80.175.228 with SMTP id h91mr6811602edd.292.1505721032438;\n\tMon, 18 Sep 2017 00:50:32 -0700 (PDT)","MIME-Version":"1.0","In-Reply-To":"","References":"\n\t\n\t","From":"Richard Biener ","Date":"Mon, 18 Sep 2017 09:50:32 +0200","Message-ID":"","Subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","To":"Kugan Vivekanandarajah ","Cc":"\"gcc-patches@gcc.gnu.org\" ","Content-Type":"text/plain; charset=\"UTF-8\"","X-IsSubscribed":"yes"}},{"id":1770677,"web_url":"http://patchwork.ozlabs.org/comment/1770677/","msgid":"","list_archive_url":null,"date":"2017-09-19T06:25:25","subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","submitter":{"id":25768,"url":"http://patchwork.ozlabs.org/api/people/25768/","name":"Kugan Vivekanandarajah","email":"kugan.vivekanandarajah@linaro.org"},"content":"Hi Richard,\n\nOn 18 September 2017 at 17:50, Richard Biener\n wrote:\n> On Mon, Sep 18, 2017 at 3:36 AM, Kugan Vivekanandarajah\n> wrote:\n>> Hi Richard,\n>>\n>> On 15 September 2017 at 19:31, Richard Biener\n>> wrote:\n>>> On Fri, Sep 15, 2017 at 3:27 AM, Kugan Vivekanandarajah\n>>> wrote:\n>>>> This patch adds separate params for rtl unroller so that they can be\n>>>> tunned accordingly. Default values I have are based on some testing on\n>>>> aarch64. I am happy to leave it as the current value and set them in\n>>>> the back-end.\n>>>\n>>> PARAM_MAX_AVERAGE_UNROLLED_INSNS is only used by the RTL\n>>> unroller. Why should we separate PARAM_MAX_UNROLL_TIMES?\n>>>\n>>> PARAM_MAX_UNROLLED_INSNS is only used by gimple passes\n>>> that perform unrolling. Since GIMPLE is three-address it should\n>>> match RTL reasonably well -- but I'd be ok in having a separate param\n>>> for those. But I wouldn't name those 'partial'.\n>>>\n>>> That said, those are magic numbers and I expect we can find some\n>>> that work well on RTL and GIMPLE.\n>>\n>> Thanks for the review. I am mostly interested in having separate\n>> params for RTL runtime unrolling as this is different to what GIMPLE\n>> unroller does.\n>\n> Why? Do you just want to have more magic knobs to machine-auto-tune?\n>\n>> May be I should have separate params only for the\n>> runtime unrolling (or the partial unroller) and let RTL/GIMPLE share\n>> the other. Any preference here ? Any preference on the name ?\n>>\n>> I am suspecting that RTL unroller which does the same as GIMPLE is\n>> kind of obsolete now?\n>\n> We do not have a GIMPLE loop unroller pass. On GIMPLE we only do\n> complete peeling as a separate pass and several passes perform\n> unrolling as part of their transform.\n\nSorry for my terminology. I was referring to complete unrolling of\nloops in tree-sea-loop-ivcanon.c. In any case, what I am interested in\nis partial unrolling of loops in loop-unroll.c. For Falkor at least,\npartially unrolling of small loops that satisfies certain condition\nseems to improve performance. Since this is not enabled by default and\nsome of the params are also shared by other optimisations (as you also\nhave pointed out), I thought that having separate params for\nloop-unroll.c will limit the interference of changing this params on\nother optimisations. My initial patch was wrong and this should only\nseparate params that are shared.\n\nAnd also, is anyone working on a partial unroller for Gimple? I would\nlike to give it a try if no one is working on this. Is there any\npreferences here.\n\nThanks,\nKugan\n\n\n\n\n> Richard.\n>\n>> Thanks,\n>> Kugan\n>>\n>>\n>>\n>>\n>>\n>>> Richard.\n>>>\n>>>>\n>>>> Thanks,\n>>>> Kugan\n>>>>\n>>>>\n>>>> gcc/ChangeLog:\n>>>>\n>>>> 2017-09-12 Kugan Vivekanandarajah \n>>>>\n>>>> * loop-unroll.c (decide_unroll_constant_iterations): Use new params.\n>>>> (decide_unroll_runtime_iterations): Likewise.\n>>>> (decide_unroll_stupid): Likewise.\n>>>> * params.def (DEFPARAM): Separate and add new params for rtl unroller.","headers":{"Return-Path":"","X-Original-To":"incoming@patchwork.ozlabs.org","Delivered-To":["patchwork-incoming@bilbo.ozlabs.org","mailing list gcc-patches@gcc.gnu.org"],"Authentication-Results":["ozlabs.org;\n\tspf=pass (mailfrom) smtp.mailfrom=gcc.gnu.org\n\t(client-ip=209.132.180.131; helo=sourceware.org;\n\tenvelope-from=gcc-patches-return-462454-incoming=patchwork.ozlabs.org@gcc.gnu.org;\n\treceiver=)","ozlabs.org; dkim=pass (1024-bit key;\n\tunprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org\n\theader.b=\"SCnAME0L\"; dkim-atps=neutral","sourceware.org; auth=none"],"Received":["from sourceware.org (server1.sourceware.org [209.132.180.131])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256\n\tbits)) (No client certificate requested)\n\tby ozlabs.org (Postfix) with ESMTPS id 3xxCYK1H5qz9s7h\n\tfor ;\n\tTue, 19 Sep 2017 16:25:39 +1000 (AEST)","(qmail 28009 invoked by alias); 19 Sep 2017 06:25:30 -0000","(qmail 27987 invoked by uid 89); 19 Sep 2017 06:25:29 -0000","from mail-qk0-f178.google.com (HELO mail-qk0-f178.google.com)\n\t(209.85.220.178) by sourceware.org\n\t(qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP;\n\tTue, 19 Sep 2017 06:25:28 +0000","by mail-qk0-f178.google.com with SMTP id s132so2673542qke.7 for\n\t; Mon, 18 Sep 2017 23:25:28 -0700 (PDT)","by 10.237.37.211 with HTTP; Mon, 18 Sep 2017 23:25:25 -0700 (PDT)"],"DomainKey-Signature":"a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; q=dns; s=default; b=BIVzZ5pURu/e0iz\n\tWmUV+1hASkVEx20XfxX5sSTHB31pXl1iDI4NNAoXboE5PvgUpZFjboace+scnvJc\n\teSefUHTyAbStZKrpk0at518UUOwh7UdeE9+T43CaCP3eLekFRxyqhSvumoQRn5an\n\tA9Z0gFErGcbUkwlDJvfH6nhOKrwA=","DKIM-Signature":"v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; s=default; bh=msaJGzUhwOqbcGbIehbPb\n\tTtzjho=; b=SCnAME0LNdNV22XXlo95oByhUnjyaM/HYtwa9BK49kqcBrPlV72la\n\tLbjZVL3CAUCIm1wbBs5c+irnglKGrZ+08EsCNTnkjI6GfWnsJfCJSi0+VuglB0i3\n\tI1RqlhqrjyIK3ZuASgWPATqm1kfXzgvEcI9PWpMVd6OqEnMNnDzONk=","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","X-Virus-Found":"No","X-Spam-SWARE-Status":"No, score=-1.6 required=5.0 tests=AWL, BAYES_00,\n\tRCVD_IN_DNSWL_NONE, RCVD_IN_SORBS_SPAM,\n\tSPF_PASS autolearn=no version=3.3.2 spammy=preferences","X-HELO":"mail-qk0-f178.google.com","X-Google-DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net;\n\ts=20161025;\n\th=x-gm-message-state:mime-version:in-reply-to:references:from:date\n\t:message-id:subject:to:cc;\n\tbh=W+I8cbodAUpiCf8OL56/xORl+UwWwMOJIf1SR39bzKk=;\n\tb=VgaU64K2FyZ27osbLtGaBp1RmQPyf2tI6XRQt+AKSqxNh1CuMaMqcFHF8C1GdczqGR\n\tUwV0S11DPMxbukaUJDSCiBKduD8PWol26qDKLy7YLjPkEppWesODOBtAbE55SzhA0kpS\n\tkrY+egyW6v4AyfMACJ8l7ttxY+0Ei0Aw0HfZk9JXYOVHOFWP/Gr7osrW4lqwz6WLxpr7\n\tfS0XiVZjZ3weiAy1oevH+pwR5fJieBSSxajuIptXlr6k1Cd+Gj4hxTzNQVoQVdXLbMlY\n\tH0C4177/kdL4LJRbegKEoga2lz2OfkKQjwmmZ9H1qMmmKdETbaZlsva2kDuOcVenGsBP\n\tctRg==","X-Gm-Message-State":"AHPjjUikIJtdEQome1m+imLokeUi3/HeUKy6l8L2qJMvjRHMQZ3pZ8d2\tJo9HI6OfytNab00sRfFEzHFXBtempfXj1e/ZDnnYYA==","X-Google-Smtp-Source":"AOwi7QBgeUfYMyOXY/8QR3LvSGa5oNLKqvPb9r0p5SiaPBcSAcmG4KU8f4LQ9bWhr22aRCSpRqJk3skOd8neBj3mpyI=","X-Received":"by 10.55.151.135 with SMTP id z129mr469938qkd.119.1505802326506;\n\tMon, 18 Sep 2017 23:25:26 -0700 (PDT)","MIME-Version":"1.0","In-Reply-To":"","References":"\n\t\n\t\n\t","From":"Kugan Vivekanandarajah ","Date":"Tue, 19 Sep 2017 16:25:25 +1000","Message-ID":"","Subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","To":"Richard Biener ","Cc":"\"gcc-patches@gcc.gnu.org\" ","Content-Type":"text/plain; charset=\"UTF-8\"","X-IsSubscribed":"yes"}},{"id":1770897,"web_url":"http://patchwork.ozlabs.org/comment/1770897/","msgid":"","list_archive_url":null,"date":"2017-09-19T11:45:32","subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","submitter":{"id":1765,"url":"http://patchwork.ozlabs.org/api/people/1765/","name":"Richard Biener","email":"richard.guenther@gmail.com"},"content":"On Tue, Sep 19, 2017 at 8:25 AM, Kugan Vivekanandarajah\n wrote:\n> Hi Richard,\n>\n> On 18 September 2017 at 17:50, Richard Biener\n> wrote:\n>> On Mon, Sep 18, 2017 at 3:36 AM, Kugan Vivekanandarajah\n>> wrote:\n>>> Hi Richard,\n>>>\n>>> On 15 September 2017 at 19:31, Richard Biener\n>>> wrote:\n>>>> On Fri, Sep 15, 2017 at 3:27 AM, Kugan Vivekanandarajah\n>>>> wrote:\n>>>>> This patch adds separate params for rtl unroller so that they can be\n>>>>> tunned accordingly. Default values I have are based on some testing on\n>>>>> aarch64. I am happy to leave it as the current value and set them in\n>>>>> the back-end.\n>>>>\n>>>> PARAM_MAX_AVERAGE_UNROLLED_INSNS is only used by the RTL\n>>>> unroller. Why should we separate PARAM_MAX_UNROLL_TIMES?\n>>>>\n>>>> PARAM_MAX_UNROLLED_INSNS is only used by gimple passes\n>>>> that perform unrolling. Since GIMPLE is three-address it should\n>>>> match RTL reasonably well -- but I'd be ok in having a separate param\n>>>> for those. But I wouldn't name those 'partial'.\n>>>>\n>>>> That said, those are magic numbers and I expect we can find some\n>>>> that work well on RTL and GIMPLE.\n>>>\n>>> Thanks for the review. I am mostly interested in having separate\n>>> params for RTL runtime unrolling as this is different to what GIMPLE\n>>> unroller does.\n>>\n>> Why? Do you just want to have more magic knobs to machine-auto-tune?\n>>\n>>> May be I should have separate params only for the\n>>> runtime unrolling (or the partial unroller) and let RTL/GIMPLE share\n>>> the other. Any preference here ? Any preference on the name ?\n>>>\n>>> I am suspecting that RTL unroller which does the same as GIMPLE is\n>>> kind of obsolete now?\n>>\n>> We do not have a GIMPLE loop unroller pass. On GIMPLE we only do\n>> complete peeling as a separate pass and several passes perform\n>> unrolling as part of their transform.\n>\n> Sorry for my terminology. I was referring to complete unrolling of\n> loops in tree-sea-loop-ivcanon.c. In any case, what I am interested in\n> is partial unrolling of loops in loop-unroll.c. For Falkor at least,\n> partially unrolling of small loops that satisfies certain condition\n> seems to improve performance. Since this is not enabled by default and\n> some of the params are also shared by other optimisations (as you also\n> have pointed out), I thought that having separate params for\n> loop-unroll.c will limit the interference of changing this params on\n> other optimisations. My initial patch was wrong and this should only\n> separate params that are shared.\n\nWell, as the complete peeling doesn't use those params but only\npasses that perform \"partial\" unrolling I see nothing wrong with sharing\nthe parameters.\n\n> And also, is anyone working on a partial unroller for Gimple? I would\n> like to give it a try if no one is working on this. Is there any\n> preferences here.\n\nPeople have talked about it but I'm not aware of actual implementations.\nThough an implementation is quite trivial with the interesting piece being\nthe costing (as usual).\n\nRichard.\n\n> Thanks,\n> Kugan\n>\n>\n>\n>\n>> Richard.\n>>\n>>> Thanks,\n>>> Kugan\n>>>\n>>>\n>>>\n>>>\n>>>\n>>>> Richard.\n>>>>\n>>>>>\n>>>>> Thanks,\n>>>>> Kugan\n>>>>>\n>>>>>\n>>>>> gcc/ChangeLog:\n>>>>>\n>>>>> 2017-09-12 Kugan Vivekanandarajah \n>>>>>\n>>>>> * loop-unroll.c (decide_unroll_constant_iterations): Use new params.\n>>>>> (decide_unroll_runtime_iterations): Likewise.\n>>>>> (decide_unroll_stupid): Likewise.\n>>>>> * params.def (DEFPARAM): Separate and add new params for rtl unroller.","headers":{"Return-Path":"","X-Original-To":"incoming@patchwork.ozlabs.org","Delivered-To":["patchwork-incoming@bilbo.ozlabs.org","mailing list gcc-patches@gcc.gnu.org"],"Authentication-Results":["ozlabs.org;\n\tspf=pass (mailfrom) smtp.mailfrom=gcc.gnu.org\n\t(client-ip=209.132.180.131; helo=sourceware.org;\n\tenvelope-from=gcc-patches-return-462469-incoming=patchwork.ozlabs.org@gcc.gnu.org;\n\treceiver=)","ozlabs.org; dkim=pass (1024-bit key;\n\tunprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org\n\theader.b=\"becqPCty\"; dkim-atps=neutral","sourceware.org; auth=none"],"Received":["from sourceware.org (server1.sourceware.org [209.132.180.131])\n\t(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256\n\tbits)) (No client certificate requested)\n\tby ozlabs.org (Postfix) with ESMTPS id 3xxLff1Knmz9s7h\n\tfor ;\n\tTue, 19 Sep 2017 21:45:44 +1000 (AEST)","(qmail 118895 invoked by alias); 19 Sep 2017 11:45:36 -0000","(qmail 118883 invoked by uid 89); 19 Sep 2017 11:45:36 -0000","from mail-wm0-f52.google.com (HELO mail-wm0-f52.google.com)\n\t(74.125.82.52) by sourceware.org\n\t(qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP;\n\tTue, 19 Sep 2017 11:45:34 +0000","by mail-wm0-f52.google.com with SMTP id i189so4431352wmf.1 for\n\t; Tue, 19 Sep 2017 04:45:34 -0700 (PDT)","by 10.80.180.250 with HTTP; Tue, 19 Sep 2017 04:45:32 -0700 (PDT)"],"DomainKey-Signature":"a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; q=dns; s=default; b=KqCC6SNFQqMRd4r\n\tN7B8x3RwzGJU4/e1kWw7Lma6MsKPC7hdWnLm5SiS00231UlaMZELIdmE7A6RGSiX\n\t7DDNmAu3V3DXsY25JSKnw+BQYQthTPMBcgduKT2wwbAX81DOG3zi9r7+5v0cRiHR\n\t3Ob46naRoMNXYSYkp9A7Un+nQTrw=","DKIM-Signature":"v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id\n\t:list-unsubscribe:list-archive:list-post:list-help:sender\n\t:mime-version:in-reply-to:references:from:date:message-id\n\t:subject:to:cc:content-type; s=default; bh=eAUaztG+5qYFeB1b4N0FE\n\tbtv7/4=; b=becqPCtyKgFZWq0lg1OToBdD+vmNoMQ7wiswFHLNAJaEqMjUJRutT\n\tV58pcHLkJl2NhlFz488Dhu5LZhiraBrKGE0a39eZzveQSX53m7ZGFvXYOWLKQAel\n\tw8rX6AVGvPK3W990Kz91RRBZgxTGfz/tHiLvcEicMvZk0+Zq5BODaw=","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","X-Virus-Found":"No","X-Spam-SWARE-Status":"No, score=-2.0 required=5.0 tests=AWL, BAYES_00,\n\tFREEMAIL_FROM, RCVD_IN_DNSWL_NONE,\n\tSPF_PASS autolearn=ham version=3.3.2 spammy=H*f:sk:CAFiYyc,\n\tH*i:sk:CAELXzT, preferences","X-HELO":"mail-wm0-f52.google.com","X-Google-DKIM-Signature":"v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net;\n\ts=20161025;\n\th=x-gm-message-state:mime-version:in-reply-to:references:from:date\n\t:message-id:subject:to:cc;\n\tbh=O1eEHVtHId6Guro5pWCtrKMEIkzx4AC1BhfjA+DoUlo=;\n\tb=T9XP8q2AvRAzWC/2QTxbVaBOjZeoszLlfqK4BfqO++RoOFoInw8ZDGPWkeVZOeCDCz\n\txp4fInxsoJVSx+iYmyYz41F6nrcLhnCbpKBQw6sj4mm/wBWaU0CteIEKMBvszs1hNr6E\n\tUBcbGHa8Ll/IL4AszqPGM097lxuiZrwSPLWxnBT8WHLf5E4JDd3YplJRz7hmDew0MF6S\n\tEvkq1Wkm8ySPymS/75kCMKPPJItBhFRIk2x/kez0/xoz+dHB9i9L8+AVOu5Bnx596K+E\n\tdH7WD2eaIP8ShovYMety35ipcctKOnNmAPC65R7yUGjVKWEdOUAsVIH6qHwGEs2CswWi\n\t71Cg==","X-Gm-Message-State":"AHPjjUjFavrvytbZdmbAaJcH9aBIueiuE3TNtlxTecnI4iAMSn+M/gy8\tBqOA749v7EUBfCg8BaD7jZf+amhlextNx/eHiuo=","X-Google-Smtp-Source":"AOwi7QC/kHFa1I/eUgtEbcADeuEQAxPUhjRKgIH6ztKH6FRo5AiED9CDSmjuaNbhl0SDMatVqM96oTn2mejcporrRsA=","X-Received":"by 10.80.175.228 with SMTP id h91mr1472397edd.292.1505821532647;\n\tTue, 19 Sep 2017 04:45:32 -0700 (PDT)","MIME-Version":"1.0","In-Reply-To":"","References":"\n\t\n\t\n\t\n\t","From":"Richard Biener ","Date":"Tue, 19 Sep 2017 13:45:32 +0200","Message-ID":"","Subject":"Re: [RFC][PATCH 1/5] Add separate parms for rtl unroller","To":"Kugan Vivekanandarajah ","Cc":"\"gcc-patches@gcc.gnu.org\" ","Content-Type":"text/plain; charset=\"UTF-8\"","X-IsSubscribed":"yes"}}]