eyalroz/gpu-kernel-runner

Issues

• Support printing generated scalar arguments to the log

  #92 opened 10 months ago by eyalroz
  0

• kernel_adapter defined_value methods incorrectly flip the semantics of user_specified

  #91 opened 10 months ago by eyalroz
  0

• Don't enable the CUDA language - we're not actually compiling any CUDA

  #90 opened a year ago by eyalroz
  0

• Avoid warning about CL_TARGET_OPENCL_VERSION when compiling with CUDA 12

  #85 opened a year ago by eyalroz
  1

• Avoid OpenCL headers warning about non-specification of target OpenCL version

  #89 opened a year ago by eyalroz
  0

• Getting warnings with CUDA 12 about undefined OpenCL target version

  #88 opened a year ago by eyalroz
  0

• Implement CUDA <-> OpenCL compatibility support for half-precision types

  #87 opened a year ago by eyalroz
  0

• Support control of synchronization resolution

  #86 opened 2 years ago by eyalroz
  0

• Support flushing the L2 cache before each invocation

  #84 opened 2 years ago by eyalroz
  0

• Custom compilation options may clash with options the runner adds itself

  #72 opened 2 years ago by eyalroz
  1

• Error using an OpenCL platform other than the first one

  #83 opened 2 years ago by eyalroz
  0

• Support listing OpenCL platforms

  #82 opened 2 years ago by eyalroz
  0

• Ensure generated preprocessor option values don't override user-specified ones

  #81 opened 2 years ago by eyalroz
  1

• When repeat registration is attempted, try printing the key

  #80 opened 2 years ago by eyalroz
  0

• When receiving multiple preprocessor definitions, need to keep the last

  #78 opened 2 years ago by eyalroz
  0

• Collect the valued and valueless preprocessor definitions into a single struct

  #79 opened 2 years ago by eyalroz
  0

• Adapter does not get chance to generate arguments

  #77 opened 2 years ago by eyalroz
  0

• Try creating missing output directories on specified output path

  #76 opened 2 years ago by eyalroz
  0

• Allow kernel adapters to set a default language standard

  #75 opened 2 years ago by eyalroz
  0

• Support --ecosystem="foo" rather than --foo

  #73 opened 2 years ago by eyalroz
  0

• Try all aliases of a buffer's name for fallback names

  #71 opened 2 years ago by eyalroz
  0

• Support --ecosystem="foo" rather than --foo

  #74 opened 2 years ago by eyalroz
  1

• Output buffer zero'ing code assumes CUDA stream exists

  #70 opened 2 years ago by eyalroz
  0

• Unuseful error message when a scalar argument is missing

  #69 opened 2 years ago by eyalroz
  0

• Prefer user-specified definitions over generated ones

  #68 opened 2 years ago by eyalroz
  0

• Use a straight-up vector for preprocessor definition details

  #67 opened 2 years ago by eyalroz
  0

• Support parameter name aliases when specifying arguments

  #63 opened 2 years ago by eyalroz
  0

• Default to writing output inout buffers files to the current directory

  #66 opened 2 years ago by eyalroz
  0

• Split off an IDE-parser header for CUDA

  #65 opened 2 years ago by eyalroz
  0

• Expand CDT-parser-specific OpenCL header to also cover JetBrains CLion

  #64 opened 2 years ago by eyalroz
  0

• Trouble generating sources for multiple adapters with the same filename

  #59 opened 2 years ago by eyalroz
  1

• Support adapter generation of additional preprocessor definitions

  #62 opened 2 years ago by eyalroz
  0

• Trouble generating sources for multiple adapters with the same filename

  #60 opened 2 years ago by eyalroz
  0

• Validate launch configurations

  #61 opened 2 years ago by eyalroz
  0

• Re-generate adapter source if the generation script has changed

  #56 opened 2 years ago by eyalroz
  0

• Support kernel adapters both inside and outside of the kernel_adapters namespace

  #58 opened 2 years ago by eyalroz
  0

• `make clean` doesn't remove generated sources

  #57 opened 2 years ago by eyalroz
  1

• Support specifying output buffer sizes via the command-line

  #55 opened 2 years ago by eyalroz
  0

• Support building for OpenCL only with no CUDA support

  #54 opened 2 years ago by eyalroz
  0

• Drop the parameter and preprocessor definition description fields

  #53 opened 2 years ago by eyalroz
  0

• Make it optional for a kernel adapter to privde a map of preprocessor defines

  #52 opened 2 years ago by eyalroz
  0

• Set intersection and difference impl bug

  #51 opened 2 years ago by eyalroz
  0

• Use fmt formatters rather than printing via stringstreams

  #50 opened 2 years ago by eyalroz
  0

• Useless check for whether we've "gotten" all buffers which need reading

  #49 opened 2 years ago by eyalroz
  0

• Add more alias for command-line options

  #48 opened 2 years ago by eyalroz
  0

• Support dynamic shared memory in OpenCL

  #47 opened 2 years ago by eyalroz
  0

• Drop two-phase command-line parsing

  #46 opened 2 years ago by eyalroz
  0

• kernel adapter instantiation-source generator python script fails to ensure directory existence

  #45 opened 2 years ago by eyalroz
  0

• Support single-precision CUDA math primitives syntax when porting to OpenCL

  #44 opened 2 years ago by eyalroz
  0

• Support AMD hardware

  #42 opened 2 years ago by eyalroz
  0