Merge pull request #110 from ecmwf-ifs/naml-remove-claw

CLOUDSC: Purge Loki-CLAW and GPU-CLAW variants from dwarf

.github/scripts/verify-targets.sh

@@ -18,10 +18,6 @@ if [[ "$build_flags" == *"--with-gpu"* ]]
 then
   targets+=(dwarf-cloudsc-gpu-scc dwarf-cloudsc-gpu-scc-hoist dwarf-cloudsc-gpu-scc-k-caching)
   targets+=(dwarf-cloudsc-gpu-omp-scc-hoist)
-  if [[ "$build_flags" == *"--with-claw"* ]]
-  then
-    targets+=(dwarf-cloudsc-gpu-claw)
-  fi
   if [[ "$build_flags" == *"--with-field"* ]]
   then
     targets+=(dwarf-cloudsc-gpu-scc-field)
@@ -35,17 +31,12 @@ fi
 
 if [[ "$build_flags" == *"--with-loki"* ]]
 then
-  targets+=(dwarf-cloudsc-loki-idem dwarf-cloudsc-loki-sca)
   targets+=(dwarf-cloudsc-loki-scc dwarf-cloudsc-loki-scc-hoist)
   targets+=(dwarf-cloudsc-loki-idem-stack dwarf-cloudsc-loki-scc-stack)
   if [[ "$build_flags" != *"--single-precision"* ]]
   then
     targets+=(dwarf-cloudsc-loki-c)
   fi
-  if [[ "$build_flags" == *"--with-claw"* ]]
-  then
-    targets+=(dwarf-cloudsc-loki-claw-cpu dwarf-cloudsc-loki-claw-gpu)
-  fi
   if [[ "$build_flags" == *"--with-cuda"* ]]
   then
     targets+=(dwarf-cloudsc-loki-scc-cuf-hoist dwarf-cloudsc-loki-scc-cuf-parametrise)

README.md

@@ -37,12 +37,6 @@ Balthasar Reuter ([email protected])
 - **dwarf-cloudsc-gpu-kernels**: GPU-enabled version of the CLOUDSC dwarf
   that uses OpenACC and relies on the `!$acc kernels` directive to offload
   the computational kernel.
-- **dwarf-cloudsc-gpu-claw** (deprecated!): GPU-enabled and optimized version of
-  CLOUDSC that is based on an auto-generated version of CLOUDSC based on the CLAW
-  tool. The kernel in this demonstrator has been further optimized with gang-level
-  loop blocking to demonstrate potential performance gains. This variant is defunct
-  on current Nvidia GPUs and therefore deactivated by default, requiring explicit
-  `--with-claw` flag to build.
 - **dwarf-cloudsc-gpu-scc**: GPU-enabled and optimized version of
   CLOUDSC that utilises the native blocked IFS memory layout via a
   "single-column coalesced" (SCC) loop layout. Here the outer NPROMA
@@ -191,7 +185,7 @@ device. This can be achieved using the `CUDA_VISIBLE_DEVICES` environment
 variable:
 
 ```sh
-mpirun -np 2 bash -c "CUDA_VISIBLE_DEVICES=\${OMPI_COMM_WORLD_RANK} bin/dwarf-cloudsc-gpu-claw 1 163840 8192"
+mpirun -np 2 bash -c "CUDA_VISIBLE_DEVICES=\${OMPI_COMM_WORLD_RANK} bin/dwarf-cloudsc-gpu-scc-stack 1 163840 128"
 ```
 
 ### Choosing between HDF5 and Serialbox input file format
@@ -349,8 +343,7 @@ source-to-source translation tool that allows us to create bespoke
 transformations for the IFS to target and experiment with emerging HPC
 architectures and programming models. We use the CLOUDSC dwarf as a demonstrator
 for targeted transformation capabilities of physics and grid point computations
-kernels, including conversion to C and GPU, directly or via downstream tools
-like CLAW.
+kernels, including conversion to C and GPU.
 
 The following build flags enable the demonstrator build targets on the
 ECMWF Atos HPC facility's GPU partition:
@@ -368,21 +361,10 @@ The following Loki modes are included in the dwarf, each with a bespoke demonstr
 - **cloudsc-loki-sca**: Pure single-column mode that strips all horizontal
   vector loops from the kernel and introduces an outer "column-loop"
   at the driver level.
-- **cloudsc-loki-claw-cpu** (deprecated): Same as SCA, but also adds the
-  necessary CLAW annotations. The resulting cloudsc.claw.F90 file is then
-  processed by CLAW to re-insert vector loops for optimal CPU execution.
-- **cloudsc-loki-claw-gpu** (deprecated): Creates the same CLAW-ready kernel
-  file, but triggers the GPU-specific optimizations in the CLAW compiler to insert
-  OpenACC-offload instructions in the driver and an OpenACC parallel loop inside
-  the kernel for each block. This needs to be run with large block sizes (eg.
-  NPROMA=1024-8192).
 - **cloudsc-loki-c**: A prototype C transpilation pipeline that converts
   the kernel to C and calls it via iso_c_bindings interfaces from the
   driver.
 
-To enable the deprecated and, on GPU, defunct CLAW variants, the build-flag
-`--with-claw` needs to be specified explicitly.
-
 ## Python-driven CLOUDSC variants
 The following partly or fully Python-based CLOUDSC are available:
 - **cloudsc-python**: GT4PY based Python-only implementation. Refer to `src/cloudsc_python`
@@ -402,10 +384,6 @@ Loki currently supports three frontends to parse the Fortran source code:
 - [FParser](https://github.com/stfc/fparser) (`loki-frontend=fp`):
   The preferred default; developed by STFC for PsyClone.
 - [OMNI](https://github.com/omni-compiler/omni-compiler) frontend (`loki-frontend=omni`):
-  Generates the same AST as used by CLAW.
-- [OFP](https://github.com/OpenFortranProject/open-fortran-parser),
-  a Python wrapper around the ROSE frontend (`loki-frontend=ofp`):
-  Supported, but bugged in some places and slow; use with care.
 
 For completeness, all three frontends are tested in our CI, which
 means we require the `.xmod` module description files for utility

bundle.yml

@@ -156,18 +156,10 @@ options :
             ENABLE_CLOUDSC_LOKI=ON
             LOKI_ENABLE_NO_INSTALL=OFF
 
-    - with-claw :
-        help  : Enable deprecated (and defunct) CLAW-generated variants
-        cmake : >
-            ENABLE_CLOUDSC_GPU_CLAW=ON
-            ENABLE_CLOUDSC_LOKI_CLAW=ON
-            LOKI_ENABLE_CLAW=ON
-
     - without-loki-install :
         help  : Skip installation of Loki (Requires Loki to be on the PATH)
         cmake : >
             LOKI_ENABLE_NO_INSTALL=ON
-            LOKI_ENABLE_CLAW=OFF
 
     - loki-frontend :
         help  : Frontend parser to use for Loki transformations
@@ -213,18 +205,10 @@ options :
         help  : Build the C version of CLOUDSC [ON|OFF]
         cmake : ENABLE_CLOUDSC_C={{value}}
 
-    - cloudsc-gpu-claw :
-        help  : Build the deprecated CLAW-based GPU version CLOUDSC  [ON|OFF]
-        cmake : ENABLE_CLOUDSC_GPU_CLAW={{value}}
-
     - cloudsc-loki :
         help  : Build the optimized Loki-based GPU version CLOUDSC  [ON|OFF]
         cmake : ENABLE_CLOUDSC_LOKI={{value}}
 
-    - cloudsc-loki-claw :
-        help  : Build the deprecated Loki+CLAW-based GPU version CLOUDSC  [ON|OFF]
-        cmake : ENABLE_CLOUDSC_LOKI_CLAW={{value}}
-
     - cloudsc-python-f2py :
         help  : Enable dedicated pure Python variant of CLOUDSC [ON|OFF]
         cmake : ENABLE_CLOUDSC_PYTHON_F2PY={{value}}

src/cloudsc_gpu/CMakeLists.txt

@@ -12,11 +12,6 @@ ecbuild_add_option( FEATURE CLOUDSC_GPU_KERNELS
     CONDITION Serialbox_FOUND OR HDF5_FOUND
 )
 
-# Define the CLAW-based GPU dwarf variant as an ECBuild feature
-ecbuild_add_option( FEATURE CLOUDSC_GPU_CLAW
-    DESCRIPTION "Build optimized GPU version of CLOUDSC derived from CLAW using OpenACC" DEFAULT OFF
-    CONDITION Serialbox_FOUND OR HDF5_FOUND
-)
 
 ecbuild_add_option( FEATURE CLOUDSC_GPU_SCC
     DESCRIPTION "Build optimized GPU version of CLOUDSC using SCC layout and OpenACC" DEFAULT OFF
@@ -74,29 +69,6 @@ ecbuild_add_option( FEATURE CLOUDSC_GPU_SCC_FIELD
 )
 
 
-if( HAVE_CLOUDSC_GPU_CLAW )
-    ecbuild_add_executable(
-        TARGET dwarf-cloudsc-gpu-claw
-        SOURCES
-            dwarf_cloudsc_gpu.F90
-            cloudsc_driver_gpu_claw_mod.F90
-            cloudsc.claw.gpu.F90
-        LIBS
-            cloudsc-common-lib
-        DEFINITIONS ${CLOUDSC_DEFINITIONS} CLOUDSC_GPU_CLAW
-    )
-
-    ecbuild_add_test(
-        TARGET dwarf-cloudsc-gpu-claw-serial
-        COMMAND bin/dwarf-cloudsc-gpu-claw
-        ARGS 1 1280 128
-        WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/../../..
-        OMP 1
-        ENABLED OFF  # CLAW variant is currently broken
-    )
-endif()
-
-
 if( HAVE_CLOUDSC_GPU_SCC )
     ecbuild_add_executable(
         TARGET dwarf-cloudsc-gpu-scc