Merge branch 'staging' into master

See merge request !171

Merge branch 'staging' into master
See merge request !171
912338a6 · Swann Perarnau · bfc976af · 83e7c188 · 912338a6 · 912338a6
Commit 912338a6 authored Dec 09, 2020 by Swann Perarnau
20 changed files
--- a/.clang-format
+++ b/.clang-format
+# Copyright 2019 UChicago Argonne, LLC.
+# (c.f. AUTHORS, LICENSE)
+#
+# This file is part of the AML project.
+# For more info, see https://xgitlab.cels.anl.gov/argo/aml
+#
+# SPDX-License-Identifier: BSD-3-Clause
+---
+Language:        Cpp
+AccessModifierOffset: -2
+AlignAfterOpenBracket: Align
+AlignConsecutiveAssignments: false
+AlignConsecutiveDeclarations: false
+AlignEscapedNewlines: Right
+AlignOperands:   true
+AlignTrailingComments: false
+AllowAllParametersOfDeclarationOnNextLine: true
+AllowShortBlocksOnASingleLine: false
+AllowShortCaseLabelsOnASingleLine: false
+AllowShortFunctionsOnASingleLine: None
+AllowShortIfStatementsOnASingleLine: false
+AllowShortLoopsOnASingleLine: false
+AlwaysBreakAfterDefinitionReturnType: None
+AlwaysBreakAfterReturnType: None
+AlwaysBreakBeforeMultilineStrings: false
+BinPackArguments: true
+BinPackParameters: false
+BreakBeforeBinaryOperators: None
+BreakBeforeBraces: Linux
+BreakBeforeInheritanceComma: false
+BreakInheritanceList: BeforeColon
+BreakBeforeTernaryOperators: false
+BreakConstructorInitializersBeforeComma: false
+BreakConstructorInitializers: BeforeComma
+BreakAfterJavaFieldAnnotations: false
+BreakStringLiterals: false
+ColumnLimit:     80
+CommentPragmas:  '^ IWYU pragma:'
+CompactNamespaces: false
+ConstructorInitializerAllOnOneLineOrOnePerLine: false
+ConstructorInitializerIndentWidth: 8
+ContinuationIndentWidth: 8
+Cpp11BracedListStyle: true
+DerivePointerAlignment: false
+DisableFormat:   false
+ExperimentalAutoDetectBinPacking: false
+FixNamespaceComments: false
+IncludeBlocks:   Regroup
+IncludeCategories:
+  - Regex:           'config\.h'
+    Priority:        1
+  - Regex:           'test.*'
+    Priority:        1
+  - Regex:           'aml\.h'
+    Priority:        3
+  - Regex:           'aml/.*'
+    Priority:        4
+  - Regex:           '^<.*\.h>'
+    Priority:        2
+  - Regex:           '.*'
+    Priority:        5
+IncludeIsMainRegex: 'aml\.h'
+IndentCaseLabels: false
+IndentWidth:     8
+IndentWrappedFunctionNames: false
+JavaScriptQuotes: Leave
+JavaScriptWrapImports: true
+KeepEmptyLinesAtTheStartOfBlocks: true
+MacroBlockBegin: ''
+MacroBlockEnd:   ''
+MaxEmptyLinesToKeep: 1
+NamespaceIndentation: Inner
+PenaltyBreakAssignment: 10
+PenaltyBreakBeforeFirstCallParameter: 30
+PenaltyBreakComment: 10
+PenaltyBreakFirstLessLess: 0
+PenaltyBreakString: 10
+PenaltyBreakTemplateDeclaration: 10
+PenaltyExcessCharacter: 100
+PenaltyReturnTypeOnItsOwnLine: 60
+PointerAlignment: Right
+ReflowComments:  true
+SortIncludes:    true
+SortUsingDeclarations: true
+SpaceAfterCStyleCast: false
+SpaceAfterTemplateKeyword: true
+SpaceBeforeAssignmentOperators: true
+SpaceBeforeCpp11BracedList: false
+SpaceBeforeCtorInitializerColon: true
+SpaceBeforeInheritanceColon: true
+SpaceBeforeParens: ControlStatements
+SpaceBeforeRangeBasedForLoopColon: true
+SpaceInEmptyParentheses: false
+SpacesBeforeTrailingComments: 1
+SpacesInAngles:  false
+SpacesInContainerLiterals: false
+SpacesInCStyleCastParentheses: false
+SpacesInParentheses: false
+SpacesInSquareBrackets: false
+Standard:        Cpp03
+TabWidth:        8
+UseTab:          ForIndentation
+...
+
--- a/.ecp-ci.yml
+++ b/.ecp-ci.yml
@@ -2,6 +2,7 @@ variables:
  NMC_FE1_SLURM_PARAMETERS: "--nodes=1 --partition=ecp-p9-4v100"
  ANL_THETA_PROJECT_SERVICE_USER: "cscstss"
  ANL_THETA_SCHEDULER_PARAMETERS: "-A CSC250STPR19 -n 1 -t 20 -q debug-flat-quad"
+  GIT_SUBMODULE_STRATEGY: "normal"

 stages:
  - build
@@ -86,21 +87,20 @@ make:theta-batch:
      - staging
    variables:
      - $ECP_MIRROR == "THETA"
+  variables:
+    CC: "icc"
+    CFLAGS: "-xKNL"
+    LDFLAGS: "-qopenmp-link=static"
  script:
-    - |
-      cat > aml-ci-script.sh << EOF
-      #!/bin/bash
-      module list
-      ./autogen.sh
-      ./configure
-      make -j63
-      make check VERBOSE=1
-      EOF
-    - cat aml-ci-script.sh
-    - chmod +x aml-ci-script.sh
-    - aprun aml-ci-script.sh
+    - ./autogen.sh
+    - ./configure --host=x86_64
+    - make
+    - make check-programs
+    - aprun make check VERBOSE=1
  artifacts:
    when: on_failure
    paths:
      - tests/*.log
+      - doc/tutorials/*.log
+      - benchmarks/*.log
      - config.log
--- a/.gitignore
+++ b/.gitignore
@@ -52,6 +52,7 @@ stamp-h1
 /missing
 /stamp-h1
 version.h
+features.h
 /m4/libtool.m4
 /m4/ltmain.sh
 /m4/ltoptions.m4
@@ -60,9 +61,8 @@ version.h
 /m4/lt~obsolete.m4

 # autotest
-tests/**/*.trs
-tests/**/*.log
-*.test.*
+*.trs
+*.log

 # libtool
 /libtool

--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
 variables:
  ARGOPKGS : "https://xgitlab.cels.anl.gov/argo/argopkgs/-/archive/master/argopkgs-master.tar.gz"
+  DEFAULT_MERGE_TARGET: "origin/staging"
+  GIT_SUBMODULE_STRATEGY: "normal"

 stages:
  - style
  - build
  - docs
-  - release
+  - distribution

 repoquality:
  stage: style
@@ -13,7 +15,7 @@ repoquality:
    - /^wip.*/
    - /^WIP.*/
  script:
-    - git ls-files *.c *.h | grep -v -e benchmarks > .repoquality
+    - git ls-files '*.c' '*.h' > .repoquality
    - nix run -f "$ARGOPKGS" repoquality --command repoquality
  tags:
    - nix
@@ -23,7 +25,7 @@ repoquality:
    paths:
      - .repoquality

-checkpatch:
+clang-format:
  stage: style
  except:
    - /^wip.*/
@@ -32,8 +34,16 @@ checkpatch:
    - nix
    - kvm
  script:
-    - git ls-files *.c *.h | grep -v -e benchmarks >> .checkpatch.conf
-    - nix run -f "$ARGOPKGS" checkpatch --command checkpatch.pl
+    - rm -f clang-format-diff
+    - target=$(git rev-parse $DEFAULT_MERGE_TARGET)
+    - nix-shell --run "git-clang-format --quiet --diff $target > clang-format-diff"
+    - lint=$(grep -v --color=never "no modified files to format" clang-format-diff || true)
+    - if [ ! -z "$lint" ]; then echo "format errors, inspect the clang-format-diff artifact for info"; exit 1; else exit 0; fi
+  artifacts:
+    when: on_failure
+    paths:
+      - ./clang-format-diff
+    expire_in: 1 week

 style:docs:
  stage: style
@@ -45,12 +55,12 @@ style:docs:
    - kvm
  script:
    - |
-      nix-shell "$ARGOPKGS" -A aml-dist --arg aml-src ./. --run bash << EOF
-      ./autogen.sh
-      mkdir build
-      ./configure --prefix=`pwd`/build --enable-docs
-      make install-data
-      EOF
+      nix-shell --run bash <<< '
+        ./autogen.sh
+        mkdir build
+        ./configure --prefix=`pwd`/build --enable-docs
+        make install-data
+      '
  artifacts:
    when: on_failure
    paths:
@@ -58,20 +68,24 @@ style:docs:

 make:generic:
  tags:
-    - knl
+    - nix
+    - kvm
  stage: build
  except:
    - /^wip.*/
    - /^WIP.*/
  variables:
-    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter"
+    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter -Wno-builtin-declaration-mismatch"
  script:
-    - ./autogen.sh
-    - mkdir build
-    - ./configure --prefix=`pwd`/build
-    - make
-    - make check
-    - make install
+    - |
+      nix-shell --run bash <<< '
+        ./autogen.sh
+        mkdir build
+        ./configure --prefix=`pwd`/build
+        make
+        make check
+        make install
+      '
  artifacts:
    when: on_failure
    paths:
@@ -80,21 +94,52 @@ make:generic:

 make:out-of-tree:
  tags:
-    - knl
+    - nix
+    - kvm
  stage: build
  except:
    - /^wip.*/
    - /^WIP.*/
  variables:
-    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter"
+    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter -Wno-builtin-declaration-mismatch"
  script:
-    - ./autogen.sh
-    - mkdir out
-    - cd out
-    - mkdir build
-    - ../configure --prefix=`pwd`/build
-    - make
-    - make check
+    - |
+      nix-shell --run bash <<< '
+        ./autogen.sh
+        mkdir out
+        cd out
+        mkdir build
+        ../configure --prefix=`pwd`/build
+        make
+        make check
+      '
+  artifacts:
+    when: on_failure
+    paths:
+      - config.log
+      - tests/*.log
+
+make:valgrind:
+  tags:
+    - nix
+    - kvm
+  stage: build
+  except:
+    - /^wip.*/
+    - /^WIP.*/
+  variables:
+    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter -Wno-builtin-declaration-mismatch"
+    VERBOSE: "1"
+    OMP_NUM_THREADS: "1"
+  script:
+    - |
+      nix-shell --run bash <<< '
+        ./autogen.sh
+        mkdir build
+        ./configure --prefix=`pwd`/build --enable-valgrind
+        make
+        make check-valgrind
+      '
  artifacts:
    when: on_failure
    paths:
@@ -106,15 +151,18 @@ make:knl:
  except:
    - /^wip.*/
    - /^WIP.*/
+  variables:
+    CFLAGS: "-mkl -xhost"
+    CC: "icc"
  tags:
    - knl
  script:
    - source /opt/intel/compilers_and_libraries/linux/bin/compilervars.sh intel64
    - ./autogen.sh
    - mkdir build
-    - CC=icc CFLAGS="-mkl -xhost" ./configure --prefix=`pwd`/build --enable-benchmarks
+    - ./configure --prefix=`pwd`/build
    - make -j64
-    - make check
+    - make -C tests check
    - make install
  artifacts:
    when: on_failure
@@ -122,6 +170,25 @@ make:knl:
      - config.log
      - tests/*.log

+make:chimera:
+  stage: build
+  except:
+    - /^wip.*/
+    - /^WIP.*/
+  tags:
+    - cuda
+  script:
+    - ./autogen.sh
+    - mkdir build
+    - ./configure --prefix=`pwd`/build --with-cuda=yes
+    - make -j
+    - make check
+  artifacts:
+    when: on_failure
+    paths:
+      - config.log
+      - tests/*.log
+
 readthedocs:
  stage: docs
  when: on_success
@@ -138,3 +205,27 @@ readthedocs:
  script:
    - nix run nixpkgs.curl -c curl -X POST -d "branches=$CI_COMMIT_REF_NAME" -d "token=$READTHEDOCS_TOKEN" https://readthedocs.org/api/v2/webhook/argo-aml/83161/

+distcheck:
+  tags:
+    - nix
+    - kvm
+  stage: distribution
+  except:
+    - /^wip.*/
+    - /^WIP.*/
+  variables:
+    CFLAGS: "-std=c99 -pedantic -Wall -Wextra -Werror -Wno-unused-but-set-parameter -Wno-builtin-declaration-mismatch"
+  script:
+    - |
+      nix-shell --run bash <<< '
+        ./autogen.sh
+        mkdir build
+        ./configure --prefix=`pwd`/build
+        make distcheck
+        chmod -R +w aml-*
+      '
+  artifacts:
+    when: on_failure
+    paths:
+      - config.log
+      - tests/*.log
--- a/.gitmodules
+++ b/.gitmodules
+[submodule "excit"]
+	path = excit
+	url = https://xgitlab.cels.anl.gov/argo/excit.git
--- a/AUTHORS
+++ b/AUTHORS
@@ -7,3 +7,4 @@ Clement Foyer <cfoyer@cray.com>
 Brice Videau <bvideau@anl.gov>
 Aleksandr Danilin <danilin96@gmail.com>
 Kyle Shaver <kshaver@anl.gov>
+Florence Monna <fmonna@anl.gov>
--- a/Makefile.am
+++ b/Makefile.am
 ACLOCAL_AMFLAGS = -I m4
-SUBDIRS = src include tests doc
-
-#if ADD_BENCHMARKS
-#SUBDIRS += benchmarks
-#endif
+SUBDIRS = src include tests doc benchmarks excit

 pkgconfigdir = $(libdir)/pkgconfig
 pkgconfig_DATA = aml.pc

--- a/aml.pc.in
+++ b/aml.pc.in
@@ -6,5 +6,5 @@ includedir=@includedir@
 Name: libaml
 Description: AML: Building Blocks for Memory Management
 Version: @PACKAGE_VERSION@
-Libs: -L${libdir} -laml
-Cflags: -I${includedir}
+Libs: -L${libdir} -laml -lexcit @HWLOC_LIBS@ @CUDA_LIBS@
+Cflags: -I${includedir} @HWLOC_CFLAGS@ @CUDA_CFLAGS@
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
-AM_CFLAGS = -I$(top_srcdir)/include -I. $(OPENMP_CFLAGS)
-AM_LDFLAGS = ../src/libaml.la $(OPENMP_CFLAGS)
-
-noinst_LIBRARIES = libutils.a
-libutils_a_SOURCES = utils.c utils.h
-LDADD = libutils.a
-
-noinst_PROGRAMS = stream_add_pth_st \
-		  stream_add_omp_st \
-		  stream_add_omp_mt \
-		  dgemm_vanilla \
-		  dgemm_mkl \
-		  dgemm_prefetch \
-		  dgemm_noprefetch
+AM_COLOR_TESTS = yes
+
+AM_CFLAGS = -I$(top_srcdir)/include -I$(top_srcdir)/excit/src $(PTHREAD_CFLAGS) $(OPENMP_CFLAGS)
+AM_LDFLAGS = ../src/libaml.la $(top_builddir)/excit/src/libexcit.la $(PTHREAD_LIBS) $(OPENMP_CFLAGS)
+
+if HAVE_CUDA
+# LIBS is used instead of AM_LDFLAGS on purpose
+# AM_LDFLAGS appends flags before libraries added before LDADD.
+# Thus, when linking with libaml.la, linking with cuda is not done.
+LIBS += $(CUDA_CFLAGS)
+LIBS += $(CUDA_LIBS)
+AM_CFLAGS += $(CUDA_CFLAGS)
+AM_LDFLAGS += $(CUDA_LIBS)
+endif
+
+if HAVE_HWLOC
+AM_CFLAGS += $(HWLOC_CFLAGS)
+AM_LDFLAGS += $(HWLOC_LIBS)
+endif
+
+# valgrind support
+@VALGRIND_CHECK_RULES@
+
+noinst_LIBRARIES = libkernel.a
+libkernel_a_SOURCES = utils.c utils.h blas_l1_kernel.c blas_l1_kernel.h verify_blas_l1.c verify_blas_l1.h
+LDADD = libkernel.a
+
+NOPREFETCH = noprefetch/flat_blas_l1 \
+	     noprefetch/tiled_blas_l1
+
+BENCHMARKS = $(NOPREFETCH)
+
+# all tests
+check_PROGRAMS = $(BENCHMARKS)
+TESTS = $(BENCHMARKS)
+
+# phony target to allow us to compile the check programs without running the
+# tests. Required for any environment where we are cross-compiling
+check-programs-local: $(BENCHMARKS)
--- a/benchmarks/blas_l1_kernel.c
+++ b/benchmarks/blas_l1_kernel.c
+/*******************************************************************************
+ * Copyright 2019 UChicago Argonne, LLC.
+ * (c.f. AUTHORS, LICENSE)
+ *
+ * This file is part of the AML project.
+ * For more info, see https://xgitlab.cels.anl.gov/argo/aml
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ ******************************************************************************/
+/*
+ * This is a benchmark for the BLAS Level 1 operations for AML.
+ */
+
+#include "blas_l1_kernel.h"
+
+/* Look into another way to define these */
+#define sign(a) ((a > 0) ? 1 : ((a < 0) ? -1 : 0))
+
+double dasum(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*b;
+	(void)*c;
+	(void)scalar;
+	size_t i;
+	double dasum = 0;
+
+	for (i = 0; i < n; i++) {
+		dasum = dasum + fabs(a[i]);
+	}
+	return dasum;
+}
+
+double daxpy(size_t n, double *a, double *b, double *c, double scalar)
+{
+	size_t i;
+#pragma omp parallel for
+	for (i = 0; i < n; i++)
+		c[i] = b[i] + scalar * a[i];
+	return 1;
+}
+
+double dcopy(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*c;
+	(void)scalar;
+	size_t i;
+
+#pragma omp parallel for
+	for (i = 0; i < n; i++)
+		b[i] = a[i];
+	return 1;
+}
+
+double ddot(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*c;
+	(void)scalar;
+	size_t i;
+	long double dot = 0.0;
+
+#pragma omp parallel for reduction(+ : dot)
+	for (i = 0; i < n; i++) {
+		long double temp;
+		temp = a[i] * b[i];
+		dot += temp;
+	}
+	return (double)dot;
+}
+
+double dnrm2(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*b;
+	(void)*c;
+	(void)scalar;
+	size_t i;
+	double scale, ssq, temp;
+
+	scale = 0.0;
+	ssq = 1.0;
+	for (i = 0; i < n; i++) {
+		if (a[i] != 0.0) {
+			temp = fabs(a[i]);
+			if (scale < temp) {
+				ssq = 1.0 + ssq * pow(scale / temp, 2);
+				scale = temp;
+			} else
+				ssq = ssq + pow(temp / scale, 2);
+		}
+	}
+	return scale * sqrt(ssq);
+}
+
+double dscal(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*c;
+	size_t i;
+
+#pragma omp parallel for
+	for (i = 0; i < n; i++)
+		b[i] = scalar * a[i];
+	return 1;
+}
+
+double dswap(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*c;
+	(void)scalar;
+	size_t i;
+	double temp;
+
+#pragma omp parallel for
+	for (i = 0; i < n; i++) {
+		temp = a[i];
+		a[i] = b[i];
+		b[i] = temp;
+	}
+	return 1;
+}
+
+double idmax(size_t n, double *a, double *b, double *c, double scalar)
+{
+	(void)*b;
+	(void)*c;
+	(void)scalar;
+
+	if (n == 1)
+		return 0;
+
+	size_t i;
+	double max;
+	size_t id_max = 0;
+
+	max = a[0];
+	for (i = 1; i < n; i++) {
+		if (fabs(a[i]) > max) {
+			id_max = i;
+			max = fabs(a[i]);
+		}
+	}
+	return id_max;
+}
+
+/* The rotations. Not included in the array of functions because of their
+   parameters */
+/* Plane rotation */
+void drot(size_t n, double *a, double *b, double x, double y)
+{
+	double temp;
+	size_t i;
+
+#pragma omp parallel for