This is an automated email from the ASF dual-hosted git repository.
markd pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/systemds.git.
from 846d1e2 [MINOR] Fix null value check in EncoderMVImpute
new 0084f24 [SYSTEMDS-2826] Sparse input support for CUDA codegen * Code
template handling refactor * A few code snippets from the row template that
the diff didn't cleanly separate (so things might not compile/run without
the other commit (7bc6379d59a0c19d881fdac8229be64d880d30cc)). Intent was to
split it in smaller chunks with moderate effort.
new c291bdb [SYSTEMDS-2827] Rowwise template for CUDA codegen
new a7efd67 [SYSTEMDS-2852] Improve SPOOF CUDA compilation
new 889d575 [SYSTEMDS-2853] Refactor spoof cuda runtime operations
new 510838a [MINOR] fix jitify submodule config
The 5 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
.gitmodules | 5 +-
pom.xml | 11 +
.../cpp/lib/libsystemds_spoof_cuda-Linux-x86_64.so | Bin 266240 -> 265584
bytes
.../lib/libsystemds_spoof_cuda-Windows-AMD64.dll | Bin 219648 -> 222208
bytes
src/main/cuda/CMakeLists.txt | 44 +-
src/main/cuda/ext/jitify | 1 +
src/main/cuda/headers/Matrix.h | 309 ++
.../headers/TempStorage.cuh} | 50 +-
src/main/cuda/headers/agg_ops.cuh | 196 +-
src/main/cuda/headers/operators.cuh | 120 +
src/main/cuda/headers/reduction.cuh | 185 +-
src/main/cuda/headers/spoof_utils.cuh | 674 ++-
.../headers/vector_add.cuh} | 36 +-
src/main/cuda/headers/vector_write.cuh | 172 +
src/main/cuda/kernels/reduction.cu | 311 +-
src/main/cuda/kernels/reduction.ptx | 4626 ++++++--------------
src/main/cuda/spoof-launcher/SpoofCUDAContext.cpp | 196 +-
src/main/cuda/spoof-launcher/SpoofCUDAContext.h | 312 +-
src/main/cuda/spoof-launcher/SpoofCellwise.h | 203 +
src/main/cuda/spoof-launcher/SpoofOperator.h | 104 +
src/main/cuda/spoof-launcher/SpoofRowwise.h | 77 +
src/main/cuda/spoof-launcher/host_utils.h | 13 +-
src/main/cuda/spoof-launcher/jni_bridge.cpp | 217 +-
src/main/cuda/spoof-launcher/jni_bridge.h | 62 +-
src/main/cuda/spoof/cellwise.cu | 65 +-
src/main/cuda/spoof/rowwise.cu | 80 +
src/main/java/org/apache/sysds/api/DMLScript.java | 12 +-
src/main/java/org/apache/sysds/common/Types.java | 37 +-
.../apache/sysds/hops/codegen/SpoofCompiler.java | 57 +-
.../org/apache/sysds/hops/codegen/cplan/CNode.java | 87 +-
.../sysds/hops/codegen/cplan/CNodeBinary.java | 28 +-
.../apache/sysds/hops/codegen/cplan/CNodeCell.java | 47 +-
.../apache/sysds/hops/codegen/cplan/CNodeData.java | 3 +-
.../sysds/hops/codegen/cplan/CNodeMultiAgg.java | 10 +
.../apache/sysds/hops/codegen/cplan/CNodeNary.java | 17 +-
.../hops/codegen/cplan/CNodeOuterProduct.java | 10 +
.../apache/sysds/hops/codegen/cplan/CNodeRow.java | 116 +-
.../apache/sysds/hops/codegen/cplan/CNodeTpl.java | 2 +
.../sysds/hops/codegen/cplan/CNodeUnary.java | 15 +-
.../sysds/hops/codegen/cplan/CodeTemplate.java | 58 +-
.../sysds/hops/codegen/cplan/cuda/Binary.java | 67 +-
.../sysds/hops/codegen/cplan/cuda/CellWise.java | 77 -
.../sysds/hops/codegen/cplan/cuda/Ternary.java | 40 +-
.../sysds/hops/codegen/cplan/cuda/Unary.java | 42 +-
.../sysds/hops/codegen/cplan/java/Binary.java | 26 +-
.../sysds/hops/codegen/cplan/java/CellWise.java | 79 -
.../Cellwise.java.template} | 31 +-
.../hops/codegen/cplan/java/Rowwise.java.template | 41 +
.../sysds/hops/codegen/cplan/java/Ternary.java | 27 +-
.../sysds/hops/codegen/cplan/java/Unary.java | 25 +-
.../apache/sysds/runtime/codegen/CodegenUtils.java | 18 +-
.../apache/sysds/runtime/codegen/SpoofCUDA.java | 122 -
.../sysds/runtime/codegen/SpoofCUDACellwise.java | 156 +
.../sysds/runtime/codegen/SpoofCUDAOperator.java | 173 +
.../sysds/runtime/codegen/SpoofCUDARowwise.java | 123 +
.../sysds/runtime/codegen/SpoofCellwise.java | 37 +-
.../sysds/runtime/codegen/SpoofMultiAggregate.java | 5 +
.../sysds/runtime/codegen/SpoofOperator.java | 11 +-
.../sysds/runtime/codegen/SpoofOuterProduct.java | 5 +
.../apache/sysds/runtime/codegen/SpoofRowwise.java | 105 +-
.../controlprogram/context/ExecutionContext.java | 21 +-
.../instructions/gpu/SpoofCUDAInstruction.java | 111 +-
.../instructions/gpu/context/GPUObject.java | 15 +-
.../test/functions/codegen/CellwiseTmplTest.java | 15 +-
.../test/functions/codegen/RowAggTmplTest.java | 3 +-
65 files changed, 5181 insertions(+), 4762 deletions(-)
create mode 160000 src/main/cuda/ext/jitify
create mode 100644 src/main/cuda/headers/Matrix.h
copy src/main/{java/org/apache/sysds/runtime/functionobjects/BitwShiftR.java
=> cuda/headers/TempStorage.cuh} (51%)
create mode 100644 src/main/cuda/headers/operators.cuh
copy src/main/{java/org/apache/sysds/runtime/lineage/LineageCodegenItem.java
=> cuda/headers/vector_add.cuh} (55%)
create mode 100644 src/main/cuda/headers/vector_write.cuh
create mode 100644 src/main/cuda/spoof-launcher/SpoofCellwise.h
create mode 100644 src/main/cuda/spoof-launcher/SpoofOperator.h
create mode 100644 src/main/cuda/spoof-launcher/SpoofRowwise.h
create mode 100644 src/main/cuda/spoof/rowwise.cu
delete mode 100644
src/main/java/org/apache/sysds/hops/codegen/cplan/cuda/CellWise.java
delete mode 100644
src/main/java/org/apache/sysds/hops/codegen/cplan/java/CellWise.java
copy src/main/java/org/apache/sysds/hops/codegen/cplan/{CodeTemplate.java =>
java/Cellwise.java.template} (57%)
create mode 100644
src/main/java/org/apache/sysds/hops/codegen/cplan/java/Rowwise.java.template
delete mode 100644
src/main/java/org/apache/sysds/runtime/codegen/SpoofCUDA.java
create mode 100644
src/main/java/org/apache/sysds/runtime/codegen/SpoofCUDACellwise.java
create mode 100644
src/main/java/org/apache/sysds/runtime/codegen/SpoofCUDAOperator.java
create mode 100644
src/main/java/org/apache/sysds/runtime/codegen/SpoofCUDARowwise.java