Drop OCCA backends

.gitlab-ci.yml

@@ -21,21 +21,14 @@ noether-cpu:
 # -- LIBXSMM v1.17
     - cd .. && export XSMM_VERSION=libxsmm-1.17 && { [[ -d $XSMM_VERSION ]] || { git clone --depth 1 --branch 1.17 https://github.com/hfp/libxsmm.git $XSMM_VERSION && make -C $XSMM_VERSION -j$(nproc); }; } && export XSMM_DIR=$PWD/$XSMM_VERSION && cd libCEED
     - echo "-------------- LIBXSMM -------------" && git -C $XSMM_DIR describe --tags
-# -- OCCA v1.1.0
-    - cd .. && export OCCA_VERSION=occa-1.1.0 OCCA_OPENCL_ENABLED=0 && { [[ -d $OCCA_VERSION ]] || { git clone --depth 1 --branch v1.1.0 https://github.com/libocca/occa.git $OCCA_VERSION && make -C $OCCA_VERSION -j$(nproc); }; } && export OCCA_DIR=$PWD/$OCCA_VERSION && cd libCEED
-    - echo "-------------- OCCA ----------------" && make -C $OCCA_DIR info
   script:
     - rm -f .SUCCESS
 # libCEED
     - make configure OPT='-O -march=native -ffp-contract=fast'
-#    Note: OCCA backends currently disabled in CI
-    - BACKENDS_CPU=$(OCCA_DIR= make info-backends-all | grep -o '/cpu[^ ]*')
+    - BACKENDS_CPU=$(make info-backends-all | grep -o '/cpu[^ ]*')
     - echo "-------------- libCEED -------------" && make info
     - echo "-------------- BACKENDS_CPU --------" && echo $BACKENDS_CPU
     - make -j$NPROC_CPU
-# Remove OCCA after verifying it compiles
-# CodeCov of OCCA backend not useful since testing is intentionally disabled
-    - export OCCA_DIR= && make -j$NPROC_CPU -W build/interface/ceed-register.o
 # -- libCEED only tests
     - echo "-------------- core tests ----------"
     - echo '[{"subject":"/","metrics":[{"name":"Transfer Size (KB)","value":"19.5","desiredSize":"smaller"},{"name":"Speed Index","value":0,"desiredSize":"smaller"},{"name":"Total Score","value":92,"desiredSize":"larger"},{"name":"Requests","value":4,"desiredSize":"smaller"}]}]' > performance.json

Makefile

@@ -66,9 +66,6 @@ export CEED_DIR = $(abspath .)
 # XSMM_DIR env variable should point to XSMM master (github.com/hfp/libxsmm)
 XSMM_DIR ?= ../libxsmm
 
-# OCCA_DIR env variable should point to OCCA master (github.com/libocca/occa)
-OCCA_DIR ?= ../occa
-
 # env variable MAGMA_DIR can be used too
 MAGMA_DIR ?= ../magma
 
@@ -216,7 +213,7 @@ fluidsexamples  := $(fluidsexamples.c:examples/fluids/%.c=$(OBJDIR)/fluids-%)
 solidsexamples.c := $(sort $(wildcard examples/solids/*.c))
 solidsexamples   := $(solidsexamples.c:examples/solids/%.c=$(OBJDIR)/solids-%)
 
-# Backends/[ref, blocked, memcheck, opt, avx, occa, magma]
+# Backends/[ref, blocked, memcheck, opt, avx, magma]
 ref.c          := $(sort $(wildcard backends/ref/*.c))
 blocked.c      := $(sort $(wildcard backends/blocked/*.c))
 ceedmemcheck.c := $(sort $(wildcard backends/memcheck/*.c))
@@ -232,7 +229,6 @@ cuda-shared.cu := $(sort $(wildcard backends/cuda-shared/kernels/*.cu))
 cuda-gen.c     := $(sort $(wildcard backends/cuda-gen/*.c))
 cuda-gen.cpp   := $(sort $(wildcard backends/cuda-gen/*.cpp))
 cuda-gen.cu    := $(sort $(wildcard backends/cuda-gen/kernels/*.cu))
-occa.cpp       := $(sort $(shell find backends/occa -type f -name *.cpp))
 magma.c        := $(sort $(wildcard backends/magma/*.c))
 magma.cu       := $(sort $(wildcard backends/magma/kernels/cuda/*.cu))
 magma.hip      := $(sort $(wildcard backends/magma/kernels/hip/*.hip.cpp))
@@ -363,22 +359,6 @@ ifneq ($(wildcard $(XSMM_DIR)/lib/libxsmm.*),)
   BACKENDS_MAKE += $(XSMM_BACKENDS)
 endif
 
-# OCCA Backends
-OCCA_BACKENDS = /cpu/self/occa
-ifneq ($(wildcard $(OCCA_DIR)/lib/libocca.*),)
-  OCCA_MODES := $(shell $(OCCA_DIR)/bin/occa modes)
-  OCCA_BACKENDS += $(if $(filter OpenMP,$(OCCA_MODES)),/cpu/openmp/occa)
-# OCCA_BACKENDS += $(if $(filter OpenCL,$(OCCA_MODES)),/gpu/opencl/occa)
-  OCCA_BACKENDS += $(if $(filter HIP,$(OCCA_MODES)),/gpu/hip/occa)
-  OCCA_BACKENDS += $(if $(filter CUDA,$(OCCA_MODES)),/gpu/cuda/occa)
-
-  $(libceeds) : CPPFLAGS += -I$(OCCA_DIR)/include
-  PKG_LIBS += -L$(abspath $(OCCA_DIR))/lib -locca
-  LIBCEED_CONTAINS_CXX = 1
-  libceed.cpp += $(occa.cpp)
-  BACKENDS_MAKE += $(OCCA_BACKENDS)
-endif
-
 # CUDA Backends
 CUDA_LIB_DIR := $(wildcard $(foreach d,lib lib64,$(CUDA_DIR)/$d/libcudart.${SO_EXT}))
 CUDA_LIB_DIR := $(patsubst %/,%,$(dir $(firstword $(CUDA_LIB_DIR))))

README.md

@@ -153,13 +153,6 @@ There are multiple supported backends, which can be selected at runtime in the e
 | `/gpu/cuda/magma/det`      | CUDA MAGMA kernels                                | Yes                   |
 | `/gpu/hip/magma`           | HIP MAGMA kernels                                 | No                    |
 | `/gpu/hip/magma/det`       | HIP MAGMA kernels                                 | Yes                   |
-||
-| **OCCA**                   |
-| `/*/occa`                  | Selects backend based on available OCCA modes     | Yes                   |
-| `/cpu/self/occa`           | OCCA backend with serial CPU kernels              | Yes                   |
-| `/cpu/openmp/occa`         | OCCA backend with OpenMP kernels                  | Yes                   |
-| `/gpu/cuda/occa`           | OCCA backend with CUDA kernels                    | Yes                   |
-| `/gpu/hip/occa`~           | OCCA backend with HIP kernels                     | Yes                   |
 
 The `/cpu/self/*/serial` backends process one element at a time and are intended for meshes with a smaller number of high order elements.
 The `/cpu/self/*/blocked` backends process blocked batches of eight interlaced elements and are intended for meshes with higher numbers of elements.
@@ -196,14 +189,6 @@ For example:
 
 > - `/gpu/cuda/gen:device_id=1`
 
-The `/*/occa` backends rely upon the [OCCA](http://github.com/libocca/occa) package to provide cross platform performance.
-To enable the OCCA backend, the environment variable `OCCA_DIR` must point to the top-level OCCA directory, with the OCCA library located in the `${OCCA_DIR}/lib` (By default, `OCCA_DIR` is set to `../occa`).
-
-Additionally, users can pass specific OCCA device properties after setting the CEED resource.
-For example:
-
-> - `"/*/occa:mode='CUDA',device_id=0"`
-
 Bit-for-bit reproducibility is important in some applications.
 However, some libCEED backends use non-deterministic operations, such as `atomicAdd` for increased performance.
 The backends which are capable of generating reproducible results, with the proper compilation options, are highlighted in the list above.

backends/ceed-backend-list.h

@@ -24,7 +24,6 @@ MACRO(CeedRegister_Magma, 2, "/gpu/cuda/magma", "/gpu/hip/magma")
 MACRO(CeedRegister_Magma_Det, 2, "/gpu/cuda/magma/det", "/gpu/hip/magma/det")
 MACRO(CeedRegister_Memcheck_Blocked, 1, "/cpu/self/memcheck/blocked")
 MACRO(CeedRegister_Memcheck_Serial, 1, "/cpu/self/memcheck/serial")
-MACRO(CeedRegister_Occa, 4, "/cpu/self/occa", "/cpu/openmp/occa", "/gpu/hip/occa", "/gpu/cuda/occa")
 MACRO(CeedRegister_Opt_Blocked, 1, "/cpu/self/opt/blocked")
 MACRO(CeedRegister_Opt_Serial, 1, "/cpu/self/opt/serial")
 MACRO(CeedRegister_Ref, 1, "/cpu/self/ref/serial")