Code

Velocity SIMD CPU Runtime#891

Code

Closed

m4rs-mt wants to merge 66 commits intomasterm4rs-mt/ILGPU:masterfrom velocitym4rs-mt/ILGPU:velocityCopy head branch name to clipboard

+10,759-143

Commits

Commits on Mar 2, 2023

• Added .vscode directory to .gitignore file.

  m4rs-mt

  committed
  43e2831

• Enhanced comment in Device class.

  m4rs-mt

  committed
  eebd13d

• Changed visibility of Interop.WriteImplementation.

  m4rs-mt

  committed
  eec6210

• Added helper to get the field names of managed structure types.

  m4rs-mt

  committed
  249ccd2

• Added helper to get a managed type corresponding to ArithmeticBasicValueType.

  m4rs-mt

  committed
  5eed500

• Adapted PhiBindings to support gathering of all phi nodes in a function.

  m4rs-mt

  committed
  a04f4cb

• Fixed typo in EntryPoint class.

  m4rs-mt

  committed
  cc62d17

• Extended TypeNode functionality to support custom managed type generation.

  m4rs-mt

  committed
  bfbfb71

• Updated Atomic functions to use readonly structures.

  m4rs-mt

  committed
  31d1d9f

• Exposed Method of MethodEmitter in RuntimeSystem.

  m4rs-mt

  committed
  f555f7e

• Added more code-generation functions to IBackendCodeGenerator.

  m4rs-mt

  committed
  381de60

• Adapted Backends to updated IBackendCodeGenerator.

  m4rs-mt

  committed
  41196ed

• Adapted ILAcceleratorSpecializer to make it reusable.

  m4rs-mt

  committed
  aec3541

• Added ILEmitter extensions to simplify MSIL code generation.

  m4rs-mt

  committed
  0d9bafb

• Updated ArgumentMapper to use common field names.

  m4rs-mt

  committed
  77ca9b1

• Unsealed PTXArgumentMapper to make it reusable.

  m4rs-mt

  committed
  924c7bf

• Extended static TypeInformation classes to support additional Velocity-specific operations.

  m4rs-mt

  committed
  63d78f1

• Extended static TypeInformation.ttinclude to support valid mapping of types to BasicValueTypes.

  m4rs-mt

  committed
  94c8d34

• Extended static TypeInformation to support enhanced queries of Float16 types.

  m4rs-mt

  committed
  dce9ac9

• Fixed typo in CPURuntimeContext.

  m4rs-mt

  committed
  d819ce1

• Added simplification of IfBranch terminators.

  m4rs-mt

  committed
  8109545

• Extended ILEmitter by adding a new WriteLine instruction for debugging.

  m4rs-mt

  committed
  cd1a12a

• Extended declarative unary math functions with implementation details for Velocity backend.

  m4rs-mt

  committed
  57f1b74

• Extended declarative binary math functions with implementation details for Velocity backend.

  m4rs-mt

  committed
  0a86c90

• Extended declarative ternary math functions with implementation details for Velocity backend.

  m4rs-mt

  committed
  a144448

• Added VelocityWarpVerifier to verify specific warp-wide operations at runtime.

  m4rs-mt

  committed
  a4c48ce

• Added specialized 32bit warp implementation (VelocityWarp32).

  m4rs-mt

  committed
  63375a7

• Added specialized 32bit warp operations (VelocityWarp32.Operations).

  m4rs-mt

  committed
  a909c23

• Added specialized 64bit warp implementation (VelocityWarp64).

  m4rs-mt

  committed
  e8d2bf7

• Added specialized 64bit warp operations (VelocityWarp64.Operations).

  m4rs-mt

  committed
  d889743

• Added automatically generated and specialized VelocityWarp operations.

  m4rs-mt

  committed
  9fc7e26

• Added VelocityWarps.cs to .gitignore file.

  m4rs-mt

  committed
  0cf3014

• Added new VelocityLaneMask structure to track active lanes inside a warp.

  m4rs-mt

  committed
  0d370f8

• Added new VelocityCodeGenerator to build mask-based SIMD control flow to Velocity backend.

  m4rs-mt

  committed
  e88f06d

• Added new VelocityCodeGenerator.Views to generate code for View-based values to Velocity backend.

  m4rs-mt

  committed
  9a8d6dc

• Added new VelocityCodeGenerator.Values to generate code for general IR values to Velocity backend.

  m4rs-mt

  committed
  05ed2b3

• Added new VelocityCodeGenerator.Threads to generate code for Thread-based values to Velocity backend.

  m4rs-mt

  committed
  ca9edc0

• Added new VelocityCodeGenerator.Terminators to generate code for SIMD-based terminators to Velocity backend.

  m4rs-mt

  committed
  e965440

• Added new VelocityCodeGenerator.IO to generate code for SIMD-based IO operations to Velocity backend.

  m4rs-mt

  committed
  6d20a92

• Added additional Velocity-specific instructions to Velocity backend.

  m4rs-mt

  committed
  7889db1

• Added new Velocity type generator to created vectorized managed type instances.

  m4rs-mt

  committed
  cc7a87e

• Registered new Velocity backend type.

  m4rs-mt

  committed
  f4de003

• Added new VelocityArgumentMapper that constructs padded argument structures based on the PTXArgumentMapper.

  m4rs-mt

  committed
  cf09ebf

• Added VelocityKernelFunction generator to Velocity backend.

  m4rs-mt

  committed
  d53b856

• Added VelocityFunction generator to Velocity backend.

  m4rs-mt

  committed
  d252feb

• Added VelocityGenerationModule to map methods to the managed world and transfer scalar arguments into the vector world.

  m4rs-mt

  committed
  4048fab

• Added new Velocity argument mapper class to convert input .Net arguments.

  m4rs-mt

  committed
  7bbc079

• Added new VelocityCompiledKernel class to represent compiled Velocity kernels.

  m4rs-mt

  committed
  ec78daa

• Added new general VelocityBackend class to convert compile Velocity kernels.

  m4rs-mt

  committed
  6ad0bc5

• Added Velocity accelerator AcceleratorType enumeration and extension providers.

  m4rs-mt

  committed
  9972042

• Update runtime error messages.

  m4rs-mt

  committed
  e766187

• Added new VelocityAccelerator class to work with SIMD-based kernels.

  m4rs-mt

  committed
  178b3b9

• Refined peer-access implementation of CPUAccelerator.

  m4rs-mt

  committed
  af00ab3

• Added new VelocityDevice class to represent and configure SIMD-based velocity accelerators.

  m4rs-mt

  committed
  e372cbc

• Added new VelocityAccelerator builder extension methods.

  m4rs-mt

  committed
  9106dd1

• Added new VelocityStream to represent streams in Velocity land.

  m4rs-mt

  committed
  a426bdc

• Added new VelocityMultiprocessor to represent individual SIMD-cores targeted by Velocity kernels.

  m4rs-mt

  committed
  3e6cd50

• Added new VelocityMemoryBuffer to represent memory buffers in Velocity land.

  m4rs-mt

  committed
  8e96a23

• Adapted CPUMemoryBuffer to support new Velocity buffers.

  m4rs-mt

  committed
  235a537

• Added new VelocityKernel to represent optimized and compiled Velocity kernels.

  m4rs-mt

  committed
  544ffe6

• Added abstract internal VelocityParameters class to pass automatically generated parameter instances to Velocity kernels.

  m4rs-mt

  committed
  1047b0f

• Added new type extension helpers.

  m4rs-mt

  committed
  6de9171

• Adapted GroupOperation tests to support different group configurations.

  m4rs-mt

  committed
  67854be

• Adapted KernelEntryPoint tests to support different group configurations.

  m4rs-mt

  committed
  11c96ea

• Adapted MemoryBufferOperation tests to support different group configurations.

  m4rs-mt

  committed
  da556d4

• Added core ILGPU Velocity tests.

  m4rs-mt

  committed
  74c1871