What's new in AOCL 3.1 (#27988)

* What's new in AOCL 3.1

1) AMD BLIS:
    1.a) Supports Dynamic Dispatch and AOCL Dynamic feature
    1.b) Improvements in DGEMM, ZGEMM, DTRSM, DSYRK, xGEMV, and DOTV
2) AMD libFLAME:
    2.a) Supports LAPACK 3.10.0 specification
    2.b) Optimized factorization and ZGEEV routines
3) AMD FFTW:
    3.a) Features like 'AMD application optimization layer', 'Fast MPI transpose algorithm' and 'Top N planner' are added
4) AMD LibM:
    4.a) Optimized exp2, log2 (Single and Double precision) scalar and vector
    4.b) Optimized log10f (scalar and vector) and powf vector variants to support WRF4.1.2 benchmark
5) AOCL-Sparse:
    5.a) New API for sparse matrix and dense matrix multiplication
6) AMD ScaLAPACK:
    6.a) ILP64 support has been enabled
7) AOCL enabled MUMPS library:
    7.a) CMake based build system on Windows for AOCL enabled MUMPS sparse solver library will be available shortly on GitHub
        7.a.i)	Refer https://github.com/amd/mumps-build
    7.b) Spack-based recipe on Linux for AOCL enabled MUMPS sparse solver library will be enabled shortly

* Fix invalid version range error

* Incorporated review comments

1) Restore to previous url value
2) Instead of if else statements, used spack's enable_or_disable feature

* Incorporated following review comments:

1. Use of satisfies() for spec checks
2. Seperate conflict statements to check for minimum and maximum GCC versions
3. Used CMakePackage helpers
4. Code rearrangement to have the directives listed before methods

7 files changed, 259 insertions, 53 deletions

diff --git a/var/spack/repos/builtin/packages/amdblis/package.py b/var/spack/repos/builtin/packages/amdblis/package.py
index 22fa9165eb..7e3ebba99f 100644
--- a/var/spack/repos/builtin/packages/amdblis/package.py
+++ b/var/spack/repos/builtin/packages/amdblis/package.py
@@ -22,17 +22,50 @@ class Amdblis(BlisBase):
 
     maintainers = ['amd-toolchain-support']
 
+    version('3.1', sha256='2891948925b9db99eec02a1917d9887a7bee9ad2afc5421c9ba58602a620f2bf')
+    version('3.0.1', sha256='dff643e6ef946846e91e8f81b75ff8fe21f1f2d227599aecd654d184d9beff3e')
     version('3.0', sha256='ac848c040cd6c3550fe49148dbdf109216cad72d3235763ee7ee8134e1528517')
     version('2.2', sha256='e1feb60ac919cf6d233c43c424f6a8a11eab2c62c2c6e3f2652c15ee9063c0c9')
 
+    variant(
+        'ilp64',
+        default=False,
+        description='Build with ILP64 support')
+
+    conflicts(
+        '+ilp64',
+        when='@:3.0.0',
+        msg='ilp64 is supported from amdblis 3.0.1 version onwards')
+
     def configure_args(self):
         spec = self.spec
         args = super(Amdblis, self).configure_args()
 
-        if spec.satisfies('@3.0 %aocc'):
-            """ To enabled Fortran to C calling convention for
-            complex types when compiling with aocc flang"""
+        if spec.satisfies('+ilp64'):
+            args.append('--blas-int-size=64')
+
+        """ To enable Fortran to C calling convention for
+        complex types when compiling with aocc flang"""
+        if self.spec.satisfies("@3.0 %aocc"):
             args.append('CFLAGS={0}'.format("-DAOCL_F2C"))
             args.append('CXXFLAGS={0}'.format("-DAOCL_F2C"))
+        elif self.spec.satisfies("@3.0.1: %aocc"):
+            args.append('--complex-return=intel')
+
+        if self.spec.satisfies("@3.1:"):
+            args.append('--disable-aocl-dynamic')
 
         return args
+
+    def configure(self, spec, prefix):
+        config_args = self.configure_args()
+
+        # "amdzen" - A fat binary or multiarchitecture binary
+        # support for 3.1 release onwards
+        if self.spec.satisfies("@3.1:"):
+            config_args.append("amdzen")
+        else:
+            config_args.append("auto")
+
+        configure("--prefix=" + prefix,
+                  *config_args)
diff --git a/var/spack/repos/builtin/packages/amdfftw/package.py b/var/spack/repos/builtin/packages/amdfftw/package.py
index a43baa1094..8e77b17199 100644
--- a/var/spack/repos/builtin/packages/amdfftw/package.py
+++ b/var/spack/repos/builtin/packages/amdfftw/package.py
@@ -31,36 +31,150 @@ class Amdfftw(FftwBase):
 
     maintainers = ['amd-toolchain-support']
 
+    version('3.1', sha256='3e777f3acef13fa1910db097e818b1d0d03a6a36ef41186247c6ab1ab0afc132')
+    version('3.0.1', sha256='87030c6bbb9c710f0a64f4f306ba6aa91dc4b182bb804c9022b35aef274d1a4c')
     version('3.0', sha256='a69deaf45478a59a69f77c4f7e9872967f1cfe996592dd12beb6318f18ea0bcd')
     version('2.2', sha256='de9d777236fb290c335860b458131678f75aa0799c641490c644c843f0e246f8')
 
-    variant('shared', default=True, description="Builds a shared version of the library")
-    variant('openmp', default=True, description="Enable OpenMP support")
-    variant('threads', default=False, description="Enable SMP threads support")
-    variant('debug', default=False, description="Builds a debug version of the library")
+    variant('shared', default=True,
+            description='Builds a shared version of the library')
+    variant('openmp', default=True,
+            description='Enable OpenMP support')
+    variant('threads', default=False,
+            description='Enable SMP threads support')
+    variant('debug', default=False,
+            description='Builds a debug version of the library')
     variant(
         'amd-fast-planner',
         default=False,
-        description="Option to reduce the planning time without much"
-                    "tradeoff in the performance. It is supported for"
-                    "Float and double precisions only.")
+        description='Option to reduce the planning time without much'
+                    'tradeoff in the performance. It is supported for'
+                    'Float and double precisions only.')
+    variant(
+        'amd-top-n-planner',
+        default=False,
+        description='Build with amd-top-n-planner support')
+    variant(
+        'amd-mpi-vader-limit',
+        default=False,
+        description='Build with amd-mpi-vader-limit support')
+    variant(
+        'static',
+        default=False,
+        description='Build with static suppport')
+    variant(
+        'amd-trans',
+        default=False,
+        description='Build with amd-trans suppport')
+    variant(
+        'amd-app-opt',
+        default=False,
+        description='Build with amd-app-opt suppport')
 
     depends_on('texinfo')
 
     provides('fftw-api@3', when='@2:')
 
-    conflicts('precision=quad', when='@2.2 %aocc', msg="AOCC clang doesn't support quad precision")
-    conflicts('+debug', when='@2.2 %aocc', msg="AOCC clang doesn't support debug")
-    conflicts('%gcc@:7.2', when="@2.2:", msg="Required GCC version above 7.2 for AMDFFTW")
-    conflicts('+amd-fast-planner', when="@2.2", msg="amd-fast-planner is supported from 3.0 onwards")
+    conflicts(
+        'precision=quad',
+        when='@2.2 %aocc',
+        msg='Quad precision is not supported by AOCC clang version 2.2')
+    conflicts(
+        '+debug',
+        when='@2.2 %aocc',
+        msg='debug mode is not supported by AOCC clang version 2.2')
+    conflicts(
+        '%gcc@:7.2',
+        when='@2.2:',
+        msg='GCC version above 7.2 is required for AMDFFTW')
+    conflicts(
+        '+amd-fast-planner ',
+        when='+mpi',
+        msg='mpi thread is not supported with amd-fast-planner')
+    conflicts(
+        '+amd-fast-planner',
+        when='@2.2',
+        msg='amd-fast-planner is supported from 3.0 onwards')
     conflicts(
         '+amd-fast-planner',
         when='precision=quad',
-        msg="amd-fast-planner doesn't support quad precision")
+        msg='Quad precision is not supported with amd-fast-planner')
     conflicts(
         '+amd-fast-planner',
         when='precision=long_double',
-        msg="amd-fast-planner doesn't support long_double precision")
+        msg='long_double precision is not supported with amd-fast-planner')
+    conflicts(
+        '+amd-top-n-planner',
+        when='@:3.0.0',
+        msg='amd-top-n-planner is supported from 3.0.1 onwards')
+    conflicts(
+        '+amd-top-n-planner',
+        when='precision=long_double',
+        msg='long_double precision is not supported with amd-top-n-planner')
+    conflicts(
+        '+amd-top-n-planner',
+        when='precision=quad',
+        msg='Quad precision is not supported with amd-top-n-planner')
+    conflicts(
+        '+amd-top-n-planner',
+        when='+amd-fast-planner',
+        msg='amd-top-n-planner cannot be used with amd-fast-planner')
+    conflicts(
+        '+amd-top-n-planner',
+        when='+threads',
+        msg='amd-top-n-planner works only for single thread')
+    conflicts(
+        '+amd-top-n-planner',
+        when='+mpi',
+        msg='mpi thread is not supported with amd-top-n-planner')
+    conflicts(
+        '+amd-top-n-planner',
+        when='+openmp',
+        msg='openmp thread is not supported with amd-top-n-planner')
+    conflicts(
+        '+amd-mpi-vader-limit',
+        when='@:3.0.0',
+        msg='amd-mpi-vader-limit is supported from 3.0.1 onwards')
+    conflicts(
+        '+amd-mpi-vader-limit',
+        when='precision=quad',
+        msg='Quad precision is not supported with amd-mpi-vader-limit')
+    conflicts(
+        '+amd-trans',
+        when='+threads',
+        msg='amd-trans works only for single thread')
+    conflicts(
+        '+amd-trans',
+        when='+mpi',
+        msg='mpi thread is not supported with amd-trans')
+    conflicts(
+        '+amd-trans',
+        when='+openmp',
+        msg='openmp thread is not supported with amd-trans')
+    conflicts(
+        '+amd-trans',
+        when='precision=long_double',
+        msg='long_double precision is not supported with amd-trans')
+    conflicts(
+        '+amd-trans',
+        when='precision=quad',
+        msg='Quad precision is not supported with amd-trans')
+    conflicts(
+        '+amd-app-opt',
+        when='@:3.0.1',
+        msg='amd-app-opt is supported from 3.1 onwards')
+    conflicts(
+        '+amd-app-opt',
+        when='+mpi',
+        msg='mpi thread is not supported with amd-app-opt')
+    conflicts(
+        '+amd-app-opt',
+        when='precision=long_double',
+        msg='long_double precision is not supported with amd-app-opt')
+    conflicts(
+        '+amd-app-opt',
+        when='precision=quad',
+        msg='Quad precision is not supported with amd-app-opt')
 
     def configure(self, spec, prefix):
         """Configure function"""
@@ -72,28 +186,13 @@ class Amdfftw(FftwBase):
 
         # Check if compiler is AOCC
         if '%aocc' in spec:
-            options.append("CC={0}".format(os.path.basename(spack_cc)))
-            options.append("FC={0}".format(os.path.basename(spack_fc)))
-            options.append("F77={0}".format(os.path.basename(spack_fc)))
-
-        if '+shared' in spec:
-            options.append('--enable-shared')
-        else:
-            options.append('--disable-shared')
+            options.append('CC={0}'.format(os.path.basename(spack_cc)))
+            options.append('FC={0}'.format(os.path.basename(spack_fc)))
+            options.append('F77={0}'.format(os.path.basename(spack_fc)))
 
         if '+debug' in spec:
             options.append('--enable-debug')
 
-        if '+openmp' in spec:
-            options.append('--enable-openmp')
-        else:
-            options.append('--disable-openmp')
-
-        if '+threads' in spec:
-            options.append('--enable-threads')
-        else:
-            options.append('--disable-threads')
-
         if '+mpi' in spec:
             options.append('--enable-mpi')
             options.append('--enable-amd-mpifft')
@@ -101,13 +200,18 @@ class Amdfftw(FftwBase):
             options.append('--disable-mpi')
             options.append('--disable-amd-mpifft')
 
-        if '+amd-fast-planner' in spec:
-            options.append('--enable-amd-fast-planner')
-        else:
-            options.append('--disable-amd-fast-planner')
+        options.extend(self.enable_or_disable('shared'))
+        options.extend(self.enable_or_disable('openmp'))
+        options.extend(self.enable_or_disable('threads'))
+        options.extend(self.enable_or_disable('amd-fast-planner'))
+        options.extend(self.enable_or_disable('amd-top-n-planner'))
+        options.extend(self.enable_or_disable('amd-mpi-vader-limit'))
+        options.extend(self.enable_or_disable('static'))
+        options.extend(self.enable_or_disable('amd-trans'))
+        options.extend(self.enable_or_disable('amd-app-opt'))
 
         if not self.compiler.f77 or not self.compiler.fc:
-            options.append("--disable-fortran")
+            options.append('--disable-fortran')
 
         # Cross compilation is supported in amd-fftw by making use of target
         # variable to set AMD_ARCH configure option.
@@ -115,9 +219,9 @@ class Amdfftw(FftwBase):
         # use target variable to set appropriate -march option in AMD_ARCH.
         arch = spec.architecture
         options.append(
-            "AMD_ARCH={0}".format(
+            'AMD_ARCH={0}'.format(
                 arch.target.optimization_flags(
-                    spec.compiler).split("=")[-1]))
+                    spec.compiler).split('=')[-1]))
 
         # Specific SIMD support.
         # float and double precisions are supported
diff --git a/var/spack/repos/builtin/packages/amdlibflame/package.py b/var/spack/repos/builtin/packages/amdlibflame/package.py
index 56e97e5894..fc27de6d52 100644
--- a/var/spack/repos/builtin/packages/amdlibflame/package.py
+++ b/var/spack/repos/builtin/packages/amdlibflame/package.py
@@ -36,11 +36,22 @@ class Amdlibflame(LibflameBase):
 
     maintainers = ['amd-toolchain-support']
 
+    version('3.1', sha256='97c74086306fa6dea9233a3730407c400c196b55f4461d4861364b1ac131ca42')
+    version('3.0.1', sha256='5859e7b39ffbe73115dd598b035f212d36310462cf3a45e555a5087301710776')
     version('3.0', sha256='d94e08b688539748571e6d4c1ec1ce42732eac18bd75de989234983c33f01ced')
     version('2.2', sha256='12b9c1f92d2c2fa637305aaa15cf706652406f210eaa5cbc17aaea9fcfa576dc')
 
+    variant('ilp64', default=False, description='Build with ILP64 support')
+
+    conflicts('+ilp64', when="@:3.0.0",
+              msg="ILP64 is supported from 3.0.1 onwards")
+    conflicts('threads=pthreads',
+              msg='pthread is not supported')
+    conflicts('threads=openmp',
+              msg='openmp is not supported')
+
     patch('aocc-2.2.0.patch', when="@:2", level=1)
-    patch('cray-compiler-wrapper.patch')
+    patch('cray-compiler-wrapper.patch', when="@:3.0.0", level=1)
 
     provides('flame@5.2', when='@2:')
 
@@ -64,8 +75,29 @@ class Amdlibflame(LibflameBase):
         if "@3.0: %aocc" in self.spec:
             args.append("--enable-f2c-dotc")
 
+        if "@3.0.1: +ilp64" in self.spec:
+            args.append("--enable-ilp64")
+
+        if "@3.1: %aocc" in self.spec:
+            args.append("--enable-void-return-complex")
+
+        if "@3.1: " in self.spec:
+            args.append("--enable-blas-ext-gemmt")
+
         return args
 
+    @run_after('build')
+    @on_package_attributes(run_tests=True)
+    def check(self):
+        """make check for single and multithread"""
+        blas_flags = self.spec['blas'].libs.ld_flags
+        if self.spec.variants['threads'].value != 'none':
+            make('check',
+                 'LIBBLAS = -fopenmp {0}'.format(blas_flags), parallel=False)
+        else:
+            make('check',
+                 'LIBBLAS = {0}'.format(blas_flags), parallel=False)
+
     def install(self, spec, prefix):
         """make install function"""
         # make install in parallel fails with message 'File already exists'
diff --git a/var/spack/repos/builtin/packages/amdlibm/package.py b/var/spack/repos/builtin/packages/amdlibm/package.py
index c38c98fd1e..fa88f3bf56 100644
--- a/var/spack/repos/builtin/packages/amdlibm/package.py
+++ b/var/spack/repos/builtin/packages/amdlibm/package.py
@@ -25,6 +25,7 @@ class Amdlibm(SConsPackage):
     # of master branch.
     # To install amdlibm from latest master branch:
     # spack install amdlibm ^amdlibm@master
+    version("3.1", branch="aocl-3.1")
     version("3.0", branch="aocl-3.0")
     version("2.2", commit="4033e022da428125747e118ccd6fdd9cee21c470")
 
@@ -32,6 +33,7 @@ class Amdlibm(SConsPackage):
             description="Building with verbosity")
 
     # Mandatory dependencies
+    depends_on("python@3.6.2", when="%aocc@3.2.0:", type=("build", "run"))
     depends_on("python@3.6.1:", type=("build", "run"))
     depends_on("scons@3.1.2:", type=("build"))
     depends_on("mpfr", type=("link"))
@@ -39,25 +41,35 @@ class Amdlibm(SConsPackage):
     patch("0001-libm-ose-Scripts-cleanup-pyc-files.patch", when="@2.2")
     patch("0002-libm-ose-prevent-log-v3.c-from-building.patch", when="@2.2")
 
-    conflicts("%gcc@:9.1", msg="Minimum required GCC version is 9.2.0")
+    conflicts(
+        '%gcc@:9.1.0',
+        msg='Minimum required GCC version is 9.2.0')
+    conflicts(
+        '%gcc@11.2.0:',
+        msg='Maximum required GCC version is 11.1.0')
+    conflicts(
+        '%aocc@3.2.0',
+        when='@2.2:3.0',
+        msg='amdlibm 2.2 and 3.0 versions are not supported with AOCC 3.2.0')
 
     def build_args(self, spec, prefix):
         """Setting build arguments for amdlibm """
         args = ["--prefix={0}".format(prefix)]
 
-        if "%aocc" in spec:
-            args.append("--compiler=aocc")
-
         # we are circumventing the use of
         # Spacks compiler wrappers because
         # SCons wipes out all environment variables.
-        args.append("CC={0}".format(self.compiler.cc))
-        args.append("CXX={0}".format(self.compiler.cxx))
+        if spec.satisfies("@:3.0 %aocc"):
+            args.append("--compiler=aocc")
+
+        var_prefix = '' if spec.satisfies("@:3.0") else 'ALM_'
+        args.append("{0}CC={1}".format(var_prefix, self.compiler.cc))
+        args.append("{0}CXX={1}".format(var_prefix, self.compiler.cxx))
 
         if "+verbose" in spec:
-            args.append("verbose=1")
+            args.append("--verbose=1")
         else:
-            args.append("verbose=0")
+            args.append("--verbose=0")
 
         return args
 
diff --git a/var/spack/repos/builtin/packages/amdscalapack/package.py b/var/spack/repos/builtin/packages/amdscalapack/package.py
index 23ccc2ccb3..d427760dcb 100644
--- a/var/spack/repos/builtin/packages/amdscalapack/package.py
+++ b/var/spack/repos/builtin/packages/amdscalapack/package.py
@@ -20,11 +20,11 @@ class Amdscalapack(ScalapackBase):
 
     _name = 'amdscalapack'
     homepage = "https://developer.amd.com/amd-aocl/scalapack/"
-    url = "https://github.com/amd/scalapack/archive/3.0.tar.gz"
     git = "https://github.com/amd/scalapack.git"
 
     maintainers = ['amd-toolchain-support']
 
+    version('3.1', sha256='4c2ee2c44644a0feec0c6fc1b1a413fa9028f14d7035d43a398f5afcfdbacb98')
     version('3.0', sha256='6e6f3578f44a8e64518d276e7580530599ecfa8729f568303ed2590688e7096f')
     version('2.2', sha256='2d64926864fc6d12157b86e3f88eb1a5205e7fc157bf67e7577d0f18b9a7484c')
 
@@ -33,6 +33,21 @@ class Amdscalapack(ScalapackBase):
         default='Release',
         description='CMake build type',
         values=('Release', 'RelWithDebInfo'))
+    variant(
+        'ilp64',
+        default=False,
+        description='Build with ILP64 support')
+
+    conflicts('+ilp64', when="@:3.0",
+              msg="ILP64 is supported from 3.1 onwards")
+
+    def url_for_version(self, version):
+        if version == Version('3.1'):
+            return "https://github.com/amd/aocl-scalapack/archive/3.1.tar.gz"
+        elif version == Version('3.0'):
+            return "https://github.com/amd/scalapack/archive/3.0.tar.gz"
+        elif version == Version('2.2'):
+            return "https://github.com/amd/scalapack/archive/2.2.tar.gz"
 
     def cmake_args(self):
         """ cmake_args function"""
@@ -47,8 +62,13 @@ class Amdscalapack(ScalapackBase):
             args.extend(['-DUSE_DOTC_WRAPPER:BOOL=%s' % (
                         'ON' if spec.satisfies('%aocc ^amdblis') else 'OFF')])
 
+        # -DENABLE_ILP64:BOOL=ON
+        args.extend([self.define_from_variant('ENABLE_ILP64', 'ilp64')])
+
+        # -DUSE_F2C:BOOL=ON
+        args.extend([self.define('USE_F2C', spec.satisfies('@:3.0'))])
+
         args.extend([
-            '-DUSE_F2C=ON',
             '-DLAPACK_FOUND=true',
             '-DCMAKE_C_COMPILER=%s' % spec['mpi'].mpicc,
             '-DCMAKE_Fortran_COMPILER=%s' % spec['mpi'].mpifc,
diff --git a/var/spack/repos/builtin/packages/aocl-sparse/package.py b/var/spack/repos/builtin/packages/aocl-sparse/package.py
index f23ef06027..426e17faa3 100644
--- a/var/spack/repos/builtin/packages/aocl-sparse/package.py
+++ b/var/spack/repos/builtin/packages/aocl-sparse/package.py
@@ -20,6 +20,7 @@ class AoclSparse(CMakePackage):
 
     maintainers = ['amd-toolchain-support']
 
+    version('3.1',  sha256='8536f06095c95074d4297a3d2910654085dd91bce82e116c10368a9f87e9c7b9')
     version('3.0',  sha256='1d04ba16e04c065051af916b1ed9afce50296edfa9b1513211a7378e1d6b952e')
     version('2.2',  sha256='33c2ed6622cda61d2613ee63ff12c116a6cd209c62e54307b8fde986cd65f664')
 
@@ -34,6 +35,7 @@ class AoclSparse(CMakePackage):
             description='Build with ILP64 support')
 
     depends_on('boost', when='@2.2')
+    depends_on('cmake@3.5:', type='build')
 
     @property
     def build_directory(self):
diff --git a/var/spack/repos/builtin/packages/mumps/package.py b/var/spack/repos/builtin/packages/mumps/package.py
index 9f5d6194cd..957a4ba9dc 100644
--- a/var/spack/repos/builtin/packages/mumps/package.py
+++ b/var/spack/repos/builtin/packages/mumps/package.py
@@ -182,11 +182,14 @@ class Mumps(Package):
                 optf.append('-qfixed')
 
         # As of version 5.2.0, MUMPS is able to take advantage
-        # of the GEMMT BLAS extension. MKL is currently the only
+        # of the GEMMT BLAS extension. MKL and amdblis are the only
         # known BLAS implementation supported.
         if '@5.2.0: ^mkl' in self.spec:
             optf.append('-DGEMMT_AVAILABLE')
 
+        if '@5.2.0: ^amdblis@3.0:' in self.spec:
+            optf.append('-DGEMMT_AVAILABLE')
+
         if '+openmp' in self.spec:
             optc.append(self.compiler.openmp_flag)
             optf.append(self.compiler.openmp_flag)