ROCm/Tensile

Issues

• [Bug]: Rocblas errors out on fp16 gemm operations

  #2044 opened a month ago by IMbackK
  17

• [Question] How does tensile avoid bank conflict for matrix core instructions?

  #2043 opened a month ago by LeiWang1999
  2

• [Feature]: More sample benchmark config.yaml needed to support bf16 and I8II on RDNA arches

  #1955 opened a month ago by likelovewant
  1

• [Issue]: Should Tensile/Tests/disabled be removed?

  #2038 opened 2 months ago by littlewu2508
  2

• [Issue]: Calling min(size_t, int) in ContractionSolution.cpp causes ambiguity

  #1977 opened 2 months ago by yxsamliu
  1

• how to enable t_debug?

  #1722 opened 2 months ago by jdgh000
  2

• [Issue]: How is a decision tree type YAML trained?

  #1975 opened 5 months ago by lty-qd
  2

• [Feature]: Generalize linux distro handling

  #1969 opened 2 months ago by trixirt
  6

• [Question] | How to generate config YAMLs for GridBased tensile tuning?

  #2033 opened 2 months ago by rtmadduri
  2

• hipErrorSharedObjectInitFailed when testing default example

  #1478 opened 2 months ago by flint-stone
  9

• [Feature]: Offload compression

  #2031 opened 2 months ago by LunNova
  1

• [Feature]: Optional use of joblib

  #1989 opened 2 months ago by trixirt
  5

• undefined symbol: Tensile::TypedContractionInputs<Tensile::Half>::TypedContractionInputs()

  #1833 opened 2 months ago by xiakubaobaore
  4

• [Bug]: rocBLAS error: Cannot read TensileLibrary.dat: No such file or directory

  #1936 opened 2 months ago by slipperyslipped
  28

• Help enabling wmma instructions

  #1943 opened 2 months ago by spayne
  7

• hipblasdgemm not getting close to peak

  #1705 opened 3 months ago by JorgeG94
  6

• Is Tensile adapted to RDNA2 ?

  #1579 opened 4 months ago by v01dXYZ
  2

• [Feature]: Restructure the code to build a wheel and use importlib to embed non-python files

  #1874 opened a year ago by bioinfornatics
  1

• TypeError: sequence item 0: expected str instance, NoneType found for compilerArgs

  #1982 opened 5 months ago by waheedi
  8

• kernel.cpp without assembly kernel implement

  #1832 opened 5 months ago by DoubleClark
  2

• [Feature]: support for gfx1103

  #1922 opened 7 months ago by NeoChen1024
  4

• Kernels source code are not generated

  #1769 opened 5 months ago by mabdallah89
  1

• [Feature]: about 7900xtx benchmark.

  #1935 opened 5 months ago by Axl-zhang
  7

• [Feature]: Further FP32 GEMM optimization for gfx11

  #1715 opened 5 months ago by littlewu2508
  2

• Out of date tuning documentation

  #1310 opened 4 years ago by kahmed10
  3

• [Feature]: Support for gfx1036

  #1907 opened 8 months ago by bitozoid
  0

• Tensile won't produce backend libraries for archs without optimized logic files when using --separate-architectures

  #1757 opened 10 months ago by ulyssesrr
  33

• The solution ends with an error.

  #1413 opened 2 years ago by vasslavich
  1

• Please enable two factor authentication in your github account

  #1502 opened 2 years ago by DeeptimayeeSethi
  0

• How to get both LdcEqualsLdd as true and false in rocblas logic yaml?

  #1252 opened 2 years ago by hanjiemicro
  1

• invalid output with gfx803 and BUILD_WITH_TENSILE_HOST ON

  #1287 opened 2 years ago by xuhuisheng
  0

• build fails with rocm4.1

  #1323 opened 2 years ago by gggh000
  5

• How to point an ISA in yaml (example)?

  #1412 opened 2 years ago by vasslavich
  1

• GlobalReadCoalesceGroupA[B] leads to wrong results

  #1426 opened 2 years ago by vasslavich
  1

• Set constStrideC0 to 0 instead of 1

  #1515 opened 2 years ago by NevesLucas
  1

• How to determine the input size to be tested？

  #1558 opened 2 years ago by zeroMrCc
  2

• Rouding error with Gfx90aFp16altSupport

  #1583 opened 2 years ago by sumin-hong
  1

• Asking for advices about tuning small GEMM with large batch count

  #1565 opened 2 years ago by littlewu2508
  4

• Enabling UnrollLoopEfficiencyEnable leads to crash during kernel generation

  #1500 opened 3 years ago by NevesLucas
  3

• `getKernel` should return `hipErrorNotFound` if no module

  #1494 opened 3 years ago by v01dXYZ
  2

• Trying RX 6700XT

  #1410 opened 3 years ago by littlewu2508
  11

• Why MT not equal to WG*TT?

  #1482 opened 3 years ago by lingjiew93
  2

• mixed precision casues failure to benchmark

  #1490 opened 3 years ago by NevesLucas
  2

• ‘JoinParameters’ is no longer supported

  #1479 opened 3 years ago by flint-stone
  5

• Trying to compile with upstream llvm-13

  #1455 opened 3 years ago by Maxzor
  12

• rocblas installation fails during Tensile run with os.fork(): OSError: [Errno 12] Cannot allocate memory

  #1415 opened 3 years ago by UweSauter
  2

• No such file or directory: '/root/workspace/ROCm/Tensile/repo/Tensile/Source/SolutionMapper.h'

  #1414 opened 3 years ago by vasslavich
  1

• tensile_client throwing std::out_of_range

  #1341 opened 3 years ago by elliottbinder
  3

• TypeError: string indices must be integers in BenchmarkStructs.py:62

  #1282 opened 3 years ago by powderluv
  6

• How to specify serveral architectures to be genrated

  #1395 opened 3 years ago by littlewu2508
  5