From issues-return-7893-archive-asf-public=cust-asf.ponee.io@systemml.apache.org Sat Feb 24 06:19:08 2018 Return-Path: X-Original-To: archive-asf-public@cust-asf.ponee.io Delivered-To: archive-asf-public@cust-asf.ponee.io Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by mx-eu-01.ponee.io (Postfix) with SMTP id 10BAD18067E for ; Sat, 24 Feb 2018 06:19:07 +0100 (CET) Received: (qmail 79534 invoked by uid 500); 24 Feb 2018 05:19:06 -0000 Mailing-List: contact issues-help@systemml.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@systemml.apache.org Delivered-To: mailing list issues@systemml.apache.org Received: (qmail 79525 invoked by uid 99); 24 Feb 2018 05:19:06 -0000 Received: from pnap-us-west-generic-nat.apache.org (HELO spamd2-us-west.apache.org) (209.188.14.142) by apache.org (qpsmtpd/0.29) with ESMTP; Sat, 24 Feb 2018 05:19:06 +0000 Received: from localhost (localhost [127.0.0.1]) by spamd2-us-west.apache.org (ASF Mail Server at spamd2-us-west.apache.org) with ESMTP id 705011A06D2 for ; Sat, 24 Feb 2018 05:19:06 +0000 (UTC) X-Virus-Scanned: Debian amavisd-new at spamd2-us-west.apache.org X-Spam-Flag: NO X-Spam-Score: -101.261 X-Spam-Level: X-Spam-Status: No, score=-101.261 tagged_above=-999 required=6.31 tests=[KAM_ASCII_DIVIDERS=0.8, KAM_LOTSOFHASH=0.25, RCVD_IN_DNSWL_MED=-2.3, SPF_PASS=-0.001, T_RP_MATCHES_RCVD=-0.01, USER_IN_WHITELIST=-100] autolearn=disabled Received: from mx1-lw-us.apache.org ([10.40.0.8]) by localhost (spamd2-us-west.apache.org [10.40.0.9]) (amavisd-new, port 10024) with ESMTP id prNYBUDkHc79 for ; Sat, 24 Feb 2018 05:19:05 +0000 (UTC) Received: from mailrelay1-us-west.apache.org (mailrelay1-us-west.apache.org [209.188.14.139]) by mx1-lw-us.apache.org (ASF Mail Server at mx1-lw-us.apache.org) with ESMTP id 8DBA95F504 for ; Sat, 24 Feb 2018 05:19:04 +0000 (UTC) Received: from jira-lw-us.apache.org (unknown [207.244.88.139]) by mailrelay1-us-west.apache.org (ASF Mail Server at mailrelay1-us-west.apache.org) with ESMTP id 85ADDE091C for ; Sat, 24 Feb 2018 05:19:02 +0000 (UTC) Received: from jira-lw-us.apache.org (localhost [127.0.0.1]) by jira-lw-us.apache.org (ASF Mail Server at jira-lw-us.apache.org) with ESMTP id 328A4218FA for ; Sat, 24 Feb 2018 05:19:00 +0000 (UTC) Date: Sat, 24 Feb 2018 05:19:00 +0000 (UTC) From: "Janardhan (JIRA)" To: issues@systemml.apache.org Message-ID: In-Reply-To: References: Subject: [jira] [Updated] (SYSTEMML-1140) Sparse/Caching performance bugs related to deep learning scripts MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable X-JIRA-FingerPrint: 30527f35849b9dde25b450d4833f0394 [ https://issues.apache.org/jira/browse/SYSTEMML-1140?page=3Dcom.atlas= sian.jira.plugin.system.issuetabpanels:all-tabpanel ] Janardhan updated SYSTEMML-1140: -------------------------------- Description:=20 We have identified two performance bugs that frequently occurs in deep lear= ning script. First, we repeatedly perform unnecessary conversion to sparse format. Also,= the operations such as matrix multiplication (including BLAS and CuBLAS) a= re optimized for dense. Second, even with large memory budget, we sometimes spend almost 20-30% tim= e in caching. [~mboehm7] [~reinwald] [~mwdusenb@us.ibm.com] I am labeling this bug as blo= cker for SystemML 1.0. Please feel free to assign this issue to yourself. *Improvements so far:* 1. Disabled sparse conversions & caching,=C2=A0 by [commit|https://github.c= om/apache/systemml/commit/caaaec90b61e529e50021d89f9f108230fa307a8] 2. binary sparse-dense mult/div, preallocation by=C2=A0[commit=C2=A0|https:= //github.com/apache/systemml/commit/4f86485939d4777d2799a697b2cbc23ea93ee7e= 4] 3. For `conv_2d_bias_add`, the `elementWiseInPlaceTransposedAddition` metho= d - first, aggreates partial blocks w/o transpose. secondly, does a cache c= onscious transpose to output. by [commit|https://github.com/apache/systemml= /commit/de1e119de0b2fc2a6c6a2c57bf64c4172a26890d] 4. serialization overhead of sparse matrices(in MCSR) on bufferpool write, = by using inMemorySize of cache block. by [commit=C2=A0|https://github.com/a= pache/systemml/commit/a68648ded00b0dc2510cd16ae8a0e5fa7ae822c3] 5. removeEmpty(rows) or order perfomance improved by ,=C2=A0shallow copy of= sparse rows, exploiting the fact that removeEmpty(rows) and order do not m= odify the actual sparse rows. by [commit|https://github.com/apache/systemml= /commit/4288da9eec006891025379b9fc723cf0e72adeaf] was: We have identified two performance bugs that frequently occurs in deep lear= ning script. First, we repeatedly perform unnecessary conversion to sparse format. Also,= the operations such as matrix multiplication (including BLAS and CuBLAS) a= re optimized for dense. Second, even with large memory budget, we sometimes spend almost 20-30% tim= e in caching. [~mboehm7] [~reinwald] [~mwdusenb@us.ibm.com] I am labeling this bug as blo= cker for SystemML 1.0. Please feel free to assign this issue to yourself. *Improvements so far:* 1. Disabled sparse conversions & caching,=C2=A0 by [commit|https://github.c= om/apache/systemml/commit/caaaec90b61e529e50021d89f9f108230fa307a8] 2. binary sparse-dense mult/div, preallocation by=C2=A0[commit=C2=A0|https:= //github.com/apache/systemml/commit/4f86485939d4777d2799a697b2cbc23ea93ee7e= 4] 3. For `conv_2d_bias_add`, the `elementWiseInPlaceTransposedAddition` metho= d - first, aggreates partial blocks w/o transpose. secondly, does a cache c= onscious transpose to output. by [commit|https://github.com/apache/systemml= /commit/de1e119de0b2fc2a6c6a2c57bf64c4172a26890d] 4. serialization overhead of sparse matrices(in MCSR) on bufferpool write, = by using inMemorySize of cache block. by [commit=C2=A0|https://github.com/a= pache/systemml/commit/a68648ded00b0dc2510cd16ae8a0e5fa7ae822c3] 5. removeEmpty(rows) or order perfomance improved by ,=C2=A0shallow copy of= sparse rows, exploiting the fact that removeEmpty(rows) and order do not m= odify the actual sparse rows. by commit > Sparse/Caching performance bugs related to deep learning scripts > ---------------------------------------------------------------- > > Key: SYSTEMML-1140 > URL: https://issues.apache.org/jira/browse/SYSTEMML-1140 > Project: SystemML > Issue Type: Bug > Affects Versions: SystemML 1.0.0, SystemML 1.1 > Reporter: Niketan Pansare > Priority: Blocker > > We have identified two performance bugs that frequently occurs in deep le= arning script. > First, we repeatedly perform unnecessary conversion to sparse format. Als= o, the operations such as matrix multiplication (including BLAS and CuBLAS)= are optimized for dense. > Second, even with large memory budget, we sometimes spend almost 20-30% t= ime in caching. > [~mboehm7] [~reinwald] [~mwdusenb@us.ibm.com] I am labeling this bug as b= locker for SystemML 1.0. Please feel free to assign this issue to yourself. > *Improvements so far:* > 1. Disabled sparse conversions & caching,=C2=A0 by [commit|https://github= .com/apache/systemml/commit/caaaec90b61e529e50021d89f9f108230fa307a8] > 2. binary sparse-dense mult/div, preallocation by=C2=A0[commit=C2=A0|http= s://github.com/apache/systemml/commit/4f86485939d4777d2799a697b2cbc23ea93ee= 7e4] > 3. For `conv_2d_bias_add`, the `elementWiseInPlaceTransposedAddition` met= hod - first, aggreates partial blocks w/o transpose. secondly, does a cache= conscious transpose to output. by [commit|https://github.com/apache/system= ml/commit/de1e119de0b2fc2a6c6a2c57bf64c4172a26890d] > 4. serialization overhead of sparse matrices(in MCSR) on bufferpool write= , by using inMemorySize of cache block. by [commit=C2=A0|https://github.com= /apache/systemml/commit/a68648ded00b0dc2510cd16ae8a0e5fa7ae822c3] > 5. removeEmpty(rows) or order perfomance improved by ,=C2=A0shallow copy = of sparse rows, exploiting the fact that removeEmpty(rows) and order do not= modify the actual sparse rows. by [commit|https://github.com/apache/system= ml/commit/4288da9eec006891025379b9fc723cf0e72adeaf] -- This message was sent by Atlassian JIRA (v7.6.3#76005)