Refactor the code and update the imports

SeisSol · Jun 5, 2024 · 17def2e · 17def2e
1 parent 800142e
commit 17def2e
Show file tree

Hide file tree

Showing 227 changed files with 2,775 additions and 1,805 deletions.
diff --git a/.gitignore b/.gitignore
@@ -31,3 +31,5 @@ yateto.egg-info
 
 *.swo
 *.swp
+
+examples/created_code/
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
@@ -164,7 +164,7 @@ pep8:
     - helper
   allow_failure: true
   script:
-    - pep8 kernelforge
+    - pep8 tensorforge
 
 
 pylint:
@@ -174,7 +174,7 @@ pylint:
     - helper
   allow_failure: true
   script:
-    - pylint kernelforge
+    - pylint tensorforge
 
 
 install:
@@ -186,9 +186,9 @@ install:
     matrix:
       - BACKEND: [cuda, hipsycl, hip]
   before_script:
-  - pip3 install --user git+https://github.com/seissol/kernelforge.git@$CI_COMMIT_SHA
+  - pip3 install --user git+https://github.com/seissol/tensorforge.git@$CI_COMMIT_SHA
   script:
-    - export isntall_path=$(python3 -c 'import kernelforge, os; print(os.path.dirname(kernelforge.__file__))')
+    - export isntall_path=$(python3 -c 'import tensorforge, os; print(os.path.dirname(tensorforge.__file__))')
     - tree $isntall_path
     - export root_dir=$PWD
     - cd ./tests/cmake_integration

diff --git a/benchmarks/double-gemm/CMakeLists.txt b/benchmarks/double-gemm/CMakeLists.txt
@@ -27,19 +27,19 @@ set(GPU_TARGET_INCLUDE_DIRS ../../submodules/Device
 if(${DEVICE_BACKEND} STREQUAL "cuda")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cu
-        include/kernelforge_aux.cu)
+        include/tensorforge_aux.cu)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cu
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif(${DEVICE_BACKEND} STREQUAL "hip")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cpp
-        include/kernelforge_aux.cpp)
+        include/tensorforge_aux.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif((${DEVICE_BACKEND} STREQUAL "oneapi") OR (${DEVICE_BACKEND} STREQUAL "hipsycl"))
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cpp
-        include/kernelforge_aux_sycl.cpp)
+        include/tensorforge_aux_sycl.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 else()

diff --git a/benchmarks/double-gemm/generate.py b/benchmarks/double-gemm/generate.py
@@ -1,6 +1,6 @@
-from kernelforge import GenerationError, GemmGenerator
-from kernelforge.common.vm.vm import vm_factory
-from kernelforge.common.matrix.tensor import Tensor
+from tensorforge import GenerationError, GemmGenerator
+from tensorforge.common.vm.vm import vm_factory
+from tensorforge.common.matrix.tensor import Tensor
 import os
 import yaml
 import argparse

diff --git a/benchmarks/double-gemm/main.cpp b/benchmarks/double-gemm/main.cpp
@@ -4,21 +4,22 @@
 #include "kernels.h"
 #include "stop_watch.h"
 #include "gemm.h"
-#include "kernelforge_aux.h"
+#include "tensorforge_aux.h"
 #include "yaml-cpp/yaml.h"
 #include <device.h>
 #include <iostream>
 #include <tuple>
 #include <vector>
 #include <string>
 
-using namespace kernelforge;
+using namespace tensorforge;
 using namespace reference;
 using namespace device;
 
 int estimateNumElements(int SizeA, int SizeB, int SizeC, int SizeD, int SizeTmp, double AllowedSpaceInGB);
 
-int main(int Argc, char* Arcv[]) {
+int main(int Argc, char *Arcv[])
+{
 
   YAML::Node Params = YAML::LoadFile("./params.yaml");
   YAML::Node MatrixASpec = Params["MatA"];
@@ -41,7 +42,7 @@ int main(int Argc, char* Arcv[]) {
   int N = BboxB[3] - BboxB[1];
   int K = BboxA[3] - BboxA[1];
 
-  int SizeTemp = M * N;  // !< required only the exact size
+  int SizeTemp = M * N; // !< required only the exact size
 
   real Alpha = Params["alpha"].as<real>();
   real Beta = Params["beta"].as<real>();
@@ -83,7 +84,6 @@ int main(int Argc, char* Arcv[]) {
   real *DeviceD{};
   std::tie(DeviceC, std::ignore, DeviceD) = FirstDriver.getDeviceRawData();
 
-
   // Check correctness
   std::cout << "INFO: computing on CPU started" << std::endl;
   unsigned NextTmp = SizeTemp;
@@ -103,7 +103,8 @@ int main(int Argc, char* Arcv[]) {
   int Ldd = MatrixDSpec["num_rows"].as<int>();
   int LdTemp = M;
 
-  auto computeOffset = [](const int LidDim, const std::vector<int> &Bbox) {
+  auto computeOffset = [](const int LidDim, const std::vector<int> &Bbox)
+  {
     return LidDim * Bbox[1] + Bbox[0];
   };
 
@@ -113,23 +114,21 @@ int main(int Argc, char* Arcv[]) {
   int OffsetD = computeOffset(Ldd, BboxD);
   int OffsetTemp = 0;
 
-
-  kernelforge::reference::gemm(TransA, TransB,
-                             M, N, K,
-                             1.0, &HostA[OffsetA], Lda,
-                             &HostB[OffsetB], Ldb,
-                             0.0, HostTmp, LdTemp,
-                             NextA, NextB, NextTmp,
-                             NumElements);
-
-
-  kernelforge::reference::gemm(TransC, reference::LayoutType::NoTrans,
-                             L, N, M,
-                             Alpha, &HostC[OffsetC], Ldc,
-                             HostTmp, M,
-                             Beta, &HostD[OffsetD], Ldd,
-                             NextC, NextTmp, NextD,
-                             NumElements);
+  tensorforge::reference::gemm(TransA, TransB,
+                               M, N, K,
+                               1.0, &HostA[OffsetA], Lda,
+                               &HostB[OffsetB], Ldb,
+                               0.0, HostTmp, LdTemp,
+                               NextA, NextB, NextTmp,
+                               NumElements);
+
+  tensorforge::reference::gemm(TransC, reference::LayoutType::NoTrans,
+                               L, N, M,
+                               Alpha, &HostC[OffsetC], Ldc,
+                               HostTmp, M,
+                               Beta, &HostD[OffsetD], Ldd,
+                               NextC, NextTmp, NextD,
+                               NumElements);
 
   std::cout << "INFO: computing on GPU started" << std::endl;
   callFirstGemm(DeviceA, 0, DeviceB, 0, DeviceTmp, 0, NumElements, nullptr, FirstDriver.getTestStream());
@@ -139,20 +138,23 @@ int main(int Argc, char* Arcv[]) {
 
   std::cout << "INFO: comparsion started" << std::endl;
 
-  SecondDriver.packResults(L,  Ldd, N, OffsetD, SizeD, NumElements);
+  SecondDriver.packResults(L, Ldd, N, OffsetD, SizeD, NumElements);
   bool IsPassed = SecondDriver.isTestPassed<SimpleComparator>();
-  if (IsPassed) {
+  if (IsPassed)
+  {
     std::cout << "INFO: Results are correct" << std::endl;
   }
-  else {
+  else
+  {
     std::cout << "WARNING: Test failed" << std::endl;
   }
 
   // Measure performance
   utils::StopWatch<std::chrono::duration<double, std::chrono::nanoseconds::period>> Timer;
   int NumRepeats = Config["num_repeats"].as<int>();
   Timer.start();
-  for (int Repeat = 0; Repeat < NumRepeats; ++Repeat) {
+  for (int Repeat = 0; Repeat < NumRepeats; ++Repeat)
+  {
     callFirstGemm(DeviceA, 0, DeviceB, 0, DeviceTmp, 0, NumElements, nullptr, FirstDriver.getTestStream());
     callSecondGemm(DeviceC, 0, DeviceTmp, 0, DeviceD, 0, NumElements, nullptr, SecondDriver.getTestStream());
   }
@@ -171,11 +173,10 @@ int main(int Argc, char* Arcv[]) {
   SecondDriver.TearDown();
   device.api->finalize();
   return 0;
-
 }
 
-
-int estimateNumElements(int SizeA, int SizeB, int SizeC, int SizeD, int SizeTmp, double AllowedSpaceInGB) {
+int estimateNumElements(int SizeA, int SizeB, int SizeC, int SizeD, int SizeTmp, double AllowedSpaceInGB)
+{
   // Note: We are going to use only one matrix C. However, memory is going
   // to get allocated for all elements
   long long ElementSizeInBytes = (SizeD + SizeC + SizeTmp + SizeA + SizeB) * sizeof(real);

diff --git a/benchmarks/fused/CMakeLists.txt b/benchmarks/fused/CMakeLists.txt
@@ -25,7 +25,7 @@ if (${BACKEND} STREQUAL "cuda")
                                             ./tmp/kernel.cu
                                             ./common/aux.cpp
                                             ./common/gemm.cpp
-                                            ../kernelforge/include/kernelforge_aux.cu)
+                                            ../tensorforge/include/tensorforge_aux.cu)
   target_compile_definitions(${CMAKE_PROJECT_NAME} PRIVATE CUDA_UNDERHOOD)
 elseif(${BACKEND} STREQUAL "hip")
   if(NOT DEFINED HIP_PATH)
@@ -45,7 +45,7 @@ elseif(${BACKEND} STREQUAL "hip")
                  ./tmp/kernel.cpp
                  ./common/aux.cpp
                  ./common/gemm.cpp
-                 ../kernelforge/include/kernelforge_aux.hip.cpp)
+                 ../tensorforge/include/tensorforge_aux.hip.cpp)
 
   set_source_files_properties(${DEVICE_SRC} PROPERTIES HIP_SOURCE_PROPERTY_FORMAT 1)
   set(_HIPCC --offload-arch=${ARCH}; -std=c++14; -O3)
@@ -60,7 +60,7 @@ endif()
 
 
 target_include_directories(${CMAKE_PROJECT_NAME} PRIVATE ./common
-                                                         ../kernelforge/include)
+                                                         ../tensorforge/include)
 target_compile_definitions(${CMAKE_PROJECT_NAME} PRIVATE "CONCRETE_CPU_BACKEND=${REFERENCE_IMPL}"
                                                          "REAL_SIZE=${REAL_SIZE}")
 

diff --git a/benchmarks/fused/glang.py b/benchmarks/fused/glang.py
@@ -1,7 +1,7 @@
-from kernelforge.frontend import Parser, PostProcessor
-from kernelforge.common import FloatingPointType
-from kernelforge.generators.generator import Generator
-from kernelforge.common.context import Context
+from tensorforge.frontend import Parser, PostProcessor
+from tensorforge.common import FloatingPointType
+from tensorforge.generators.generator import Generator
+from tensorforge.common.context import Context
 from internals import BenchGenerator, EnryPointGenerator, Aux
 from os import path, makedirs
 import sys
@@ -98,7 +98,7 @@ def main():
 
   # write kernel, launcher and header to files
   with open(path.join(tmp_dir, f'kernel.{file_suffix}'), 'w') as file:
-    file.write('#include \"kernelforge_aux.h\"\n')
+    file.write('#include \"tensorforge_aux.h\"\n')
     for kernel, launcher in zip(kernels, launchers):
       file.write(kernel)
       file.write(launcher)

diff --git a/benchmarks/fused/internals/bench_generator.py b/benchmarks/fused/internals/bench_generator.py
@@ -1,6 +1,6 @@
 from .gpu_api import GpuAPI
 from .aux import Aux
-from kernelforge.common.basic_types import Addressing
+from tensorforge.common.basic_types import Addressing
 
 
 class BenchGenerator:

diff --git a/benchmarks/fused/internals/frontend/aux.py b/benchmarks/fused/internals/frontend/aux.py
@@ -1,5 +1,5 @@
-from kernelforge.common import Addressing
-from kernelforge.common.matrix.tensor import Tensor
+from tensorforge.common import Addressing
+from tensorforge.common.matrix.tensor import Tensor
 
 
 class VarFactory:

diff --git a/benchmarks/fused/internals/frontend/symbol_table.py b/benchmarks/fused/internals/frontend/symbol_table.py
@@ -1,7 +1,7 @@
 import enum
 from typing import Dict, Union
-from kernelforge.common import Addressing
-from kernelforge.common.matrix.tensor import Tensor
+from tensorforge.common import Addressing
+from tensorforge.common.matrix.tensor import Tensor
 
 
 class ObjType(enum.Enum):

diff --git a/benchmarks/fused/internals/frontend/traversals.py b/benchmarks/fused/internals/frontend/traversals.py
@@ -1,7 +1,7 @@
 import os
 from graphviz import Digraph
-from kernelforge.common import generate_tmp_matrix
-from kernelforge.generators.descriptions import GemmDescr
+from tensorforge.common import generate_tmp_matrix
+from tensorforge.generators.descriptions import GemmDescr
 from .nodes import VarNode, ScalarNode, MatrixNode, DeadNode
 from .nodes import BinarryOps, AssignNode, AddNode, MultNode
 from .nodes import StatementsNode, GemmListNode

diff --git a/benchmarks/fused/internals/templates/entry_point.tmpl b/benchmarks/fused/internals/templates/entry_point.tmpl
@@ -1,6 +1,6 @@
 #include "aux.h"
 #include "stop_watch.h"
-#include "kernelforge_aux.h"
+#include "tensorforge_aux.h"
 #include "gemm.h"
 #include "kernel.h"
 #include <iostream>

diff --git a/benchmarks/gemm-chain/CMakeLists.txt b/benchmarks/gemm-chain/CMakeLists.txt
@@ -24,21 +24,21 @@ if(${DEVICE_BACKEND} STREQUAL "cuda")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
                                 ${CMAKE_SOURCE_DIR}/gen_code/main.cu
                                 gen_code/kernels.cu
-                                include/kernelforge_aux.cu)
+                                include/tensorforge_aux.cu)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/main.cu
                           ${CMAKE_SOURCE_DIR}/gen_code/kernels.h
                           ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif(${DEVICE_BACKEND} STREQUAL "hip")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
                                 gen_code/main.cpp
                                 gen_code/kernels.cpp
-                                include/kernelforge_aux.cpp)
+                                include/tensorforge_aux.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif((${DEVICE_BACKEND} STREQUAL "oneapi") OR (${DEVICE_BACKEND} STREQUAL "hipsycl"))
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
                                 gen_code/kernels.cpp
                                 gen_code/main.cpp
-                                include/kernelforge_aux_sycl.cpp)
+                                include/tensorforge_aux_sycl.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 else()
     message(FATAL_ERROR "unknown backend, given: ${DEVICE_BACKEND}")

diff --git a/benchmarks/gemm-chain/bench.tmpl b/benchmarks/gemm-chain/bench.tmpl
@@ -1,5 +1,5 @@
 #include "kernels.h"
-#include "kernelforge_aux.h"
+#include "tensorforge_aux.h"
 #include <device.h>
 #include <chrono>
 #include <iostream>

diff --git a/benchmarks/gemm-chain/generate.py b/benchmarks/gemm-chain/generate.py
@@ -1,6 +1,6 @@
-from kernelforge import GenerationError, GemmGenerator
-from kernelforge.common.matrix.tensor import Tensor
-from kernelforge.common.vm.vm import vm_factory
+from tensorforge import GenerationError, GemmGenerator
+from tensorforge.common.matrix.tensor import Tensor
+from tensorforge.common.vm.vm import vm_factory
 from jinja2 import Environment, FileSystemLoader
 import os
 import yaml

diff --git a/benchmarks/simple-gemm/CMakeLists.txt b/benchmarks/simple-gemm/CMakeLists.txt
@@ -25,19 +25,19 @@ set(GPU_TARGET_INCLUDE_DIRS ../../submodules/Device
 if(${DEVICE_BACKEND} STREQUAL "cuda")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cu
-        include/kernelforge_aux.cu)
+        include/tensorforge_aux.cu)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cu
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif(${DEVICE_BACKEND} STREQUAL "hip")
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cpp
-        include/kernelforge_aux.cpp)
+        include/tensorforge_aux.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 elseif((${DEVICE_BACKEND} STREQUAL "oneapi") OR (${DEVICE_BACKEND} STREQUAL "hipsycl"))
     set(GPU_TARGET_SOURCE_FILES common/test_drivers/simple_driver.cpp
         gen_code/kernels.cpp
-        include/kernelforge_aux_sycl.cpp)
+        include/tensorforge_aux_sycl.cpp)
     set(GEN_COPY_PRODUCTS ${CMAKE_SOURCE_DIR}/gen_code/kernels.cpp
         ${CMAKE_SOURCE_DIR}/gen_code/kernels.h)
 else()

diff --git a/benchmarks/simple-gemm/generate.py b/benchmarks/simple-gemm/generate.py
@@ -1,7 +1,7 @@
-from kernelforge import GenerationError
-from kernelforge.common.matrix.tensor import Tensor
-from kernelforge import GemmGenerator, GemmKernelType
-from kernelforge.common.vm.vm import vm_factory
+from tensorforge import GenerationError
+from tensorforge.common.matrix.tensor import Tensor
+from tensorforge import GemmGenerator, GemmKernelType
+from tensorforge.common.vm.vm import vm_factory
 import os
 import yaml
 import argparse
Original file line number	Diff line number	Diff line change
Expand Up		@@ -31,3 +31,5 @@ yateto.egg-info

		*.swo
		*.swp

		examples/created_code/