Contents

• 1. Introduction
  • 1.1. cuSolverDN: Dense LAPACK
  • 1.2. cuSolverSP: Sparse LAPACK
  • 1.3. cuSolverRF: Refactorization
  • 1.4. Naming Conventions
  • 1.5. Asynchronous Execution
  • 1.6. Library Property
  • 1.7. High Precision Package
• 2. Using the CUSOLVER API
  • 2.1. General Description
    • 2.1.1. Thread Safety
    • 2.1.2. Scalar Parameters
    • 2.1.3. Parallelism with Streams
    • 2.1.4. How to Link cusolver Library
    • 2.1.5. Link Third-party LAPACK Library
    • 2.1.6. Convention of info
    • 2.1.7. Usage of _bufferSize
    • 2.1.8. Deterministic Results
  • 2.2. cuSolver Types Reference
    • 2.2.1. cuSolverDN Types
      • 2.2.1.1. cusolverDnHandle_t
      • 2.2.1.2. cublasFillMode_t
      • 2.2.1.3. cublasOperation_t
      • 2.2.1.4. cusolverEigType_t
      • 2.2.1.5. cusolverEigMode_t
      • 2.2.1.6. cusolverIRSRefinement_t
      • 2.2.1.7. cusolverDnIRSParams_t
      • 2.2.1.8. cusolverDnIRSInfos_t
      • 2.2.1.9. cusolverDnFunction_t
      • 2.2.1.10. cusolverAlgMode_t
      • 2.2.1.11. cusolverStatus_t
      • 2.2.1.12. cusolverDeterministicMode_t
    • 2.2.2. cuSolverSP Types
      • 2.2.2.1. cusolverSpHandle_t
      • 2.2.2.2. cusparseMatDescr_t
      • 2.2.2.3. cusolverStatus_t
    • 2.2.3. cuSolverRF Types
      • 2.2.3.1. cusolverRfHandle_t
      • 2.2.3.2. cusolverRfMatrixFormat_t
      • 2.2.3.3. cusolverRfNumericBoostReport_t
      • 2.2.3.4. cusolverRfResetValuesFastMode_t
      • 2.2.3.5. cusolverRfFactorization_t
      • 2.2.3.6. cusolverRfTriangularSolve_t
      • 2.2.3.7. cusolverRfUnitDiagonal_t
      • 2.2.3.8. cusolverStatus_t
  • 2.3. cuSolver Formats Reference
    • 2.3.1. Index Base Format
    • 2.3.2. Vector (Dense) Format
    • 2.3.3. Matrix (Dense) Format
    • 2.3.4. Matrix (CSR) Format
    • 2.3.5. Matrix (CSC) Format
  • 2.4. cuSolverDN: dense LAPACK Function Reference
    • 2.4.1. cuSolverDN Helper Function Reference
      • 2.4.1.1. cusolverDnCreate()
      • 2.4.1.2. cusolverDnDestroy()
      • 2.4.1.3. cusolverDnSetStream()
      • 2.4.1.4. cusolverDnGetStream()
      • 2.4.1.5. cusolverDnSetDeterministicMode()
      • 2.4.1.6. cusolverDnGetDeterministicMode()
      • 2.4.1.7. cusolverDnCreateSyevjInfo()
      • 2.4.1.8. cusolverDnDestroySyevjInfo()
      • 2.4.1.9. cusolverDnXsyevjSetTolerance()
      • 2.4.1.10. cusolverDnXsyevjSetMaxSweeps()
      • 2.4.1.11. cusolverDnXsyevjSetSortEig()
      • 2.4.1.12. cusolverDnXsyevjGetResidual()
      • 2.4.1.13. cusolverDnXsyevjGetSweeps()
      • 2.4.1.14. cusolverDnCreateGesvdjInfo()
      • 2.4.1.15. cusolverDnDestroyGesvdjInfo()
      • 2.4.1.16. cusolverDnXgesvdjSetTolerance()
      • 2.4.1.17. cusolverDnXgesvdjSetMaxSweeps()
      • 2.4.1.18. cusolverDnXgesvdjSetSortEig()
      • 2.4.1.19. cusolverDnXgesvdjGetResidual()
      • 2.4.1.20. cusolverDnXgesvdjGetSweeps()
      • 2.4.1.21. cusolverDnIRSParamsCreate()
      • 2.4.1.22. cusolverDnIRSParamsDestroy()
      • 2.4.1.23. cusolverDnIRSParamsSetSolverPrecisions()
      • 2.4.1.24. cusolverDnIRSParamsSetSolverMainPrecision()
      • 2.4.1.25. cusolverDnIRSParamsSetSolverLowestPrecision()
      • 2.4.1.26. cusolverDnIRSParamsSetRefinementSolver()
      • 2.4.1.27. cusolverDnIRSParamsSetTol()
      • 2.4.1.28. cusolverDnIRSParamsSetTolInner()
      • 2.4.1.29. cusolverDnIRSParamsSetMaxIters()
      • 2.4.1.30. cusolverDnIRSParamsSetMaxItersInner()
      • 2.4.1.31. cusolverDnIRSParamsEnableFallback()
      • 2.4.1.32. cusolverDnIRSParamsDisableFallback()
      • 2.4.1.33. cusolverDnIRSParamsGetMaxIters()
      • 2.4.1.34. cusolverDnIRSInfosCreate()
      • 2.4.1.35. cusolverDnIRSInfosDestroy()
      • 2.4.1.36. cusolverDnIRSInfosGetMaxIters()
      • 2.4.1.37. cusolverDnIRSInfosGetNiters()
      • 2.4.1.38. cusolverDnIRSInfosGetOuterNiters()
      • 2.4.1.39. cusolverDnIRSInfosRequestResidual()
      • 2.4.1.40. cusolverDnIRSInfosGetResidualHistory()
      • 2.4.1.41. cusolverDnCreateParams()
      • 2.4.1.42. cusolverDnDestroyParams()
      • 2.4.1.43. cusolverDnSetAdvOptions()
    • 2.4.2. Dense Linear Solver Reference (legacy)
      • 2.4.2.1. cusolverDn<t>potrf()
      • 2.4.2.2. cusolverDnPotrf()[DEPRECATED]
      • 2.4.2.3. cusolverDn<t>potrs()
      • 2.4.2.4. cusolverDnPotrs()[DEPRECATED]
      • 2.4.2.5. cusolverDn<t>potri()
      • 2.4.2.6. cusolverDn<t>getrf()
      • 2.4.2.7. cusolverDnGetrf()[DEPRECATED]
      • 2.4.2.8. cusolverDn<t>getrs()
      • 2.4.2.9. cusolverDnGetrs()[DEPRECATED]
      • 2.4.2.10. cusolverDn<t1><t2>gesv()
      • 2.4.2.11. cusolverDnIRSXgesv()
      • 2.4.2.12. cusolverDn<t>geqrf()
      • 2.4.2.13. cusolverDnGeqrf()[DEPRECATED]
      • 2.4.2.14. cusolverDn<t1><t2>gels()
      • 2.4.2.15. cusolverDnIRSXgels()
      • 2.4.2.16. cusolverDn<t>ormqr()
      • 2.4.2.17. cusolverDn<t>orgqr()
      • 2.4.2.18. cusolverDn<t>sytrf()
      • 2.4.2.19. cusolverDn<t>potrfBatched()
      • 2.4.2.20. cusolverDn<t>potrsBatched()
    • 2.4.3. Dense Eigenvalue Solver Reference (legacy)
      • 2.4.3.1. cusolverDn<t>gebrd()
      • 2.4.3.2. cusolverDn<t>orgbr()
      • 2.4.3.3. cusolverDn<t>sytrd()
      • 2.4.3.4. cusolverDn<t>ormtr()
      • 2.4.3.5. cusolverDn<t>orgtr()
      • 2.4.3.6. cusolverDn<t>gesvd()
      • 2.4.3.7. cusolverDnGesvd()[DEPRECATED]
      • 2.4.3.8. cusolverDn<t>gesvdj()
      • 2.4.3.9. cusolverDn<t>gesvdjBatched()
      • 2.4.3.10. cusolverDn<t>gesvdaStridedBatched()
      • 2.4.3.11. cusolverDn<t>syevd()
      • 2.4.3.12. cusolverDnSyevd()[DEPRECATED]
      • 2.4.3.13. cusolverDn<t>syevdx()
      • 2.4.3.14. cusolverDnSyevdx()[DEPRECATED]
      • 2.4.3.15. cusolverDn<t>sygvd()
      • 2.4.3.16. cusolverDn<t>sygvdx()
      • 2.4.3.17. cusolverDn<t>syevj()
      • 2.4.3.18. cusolverDn<t>sygvj()
      • 2.4.3.19. cusolverDn<t>syevjBatched()
    • 2.4.4. Dense Linear Solver Reference (64-bit API)
      • 2.4.4.1. cusolverDnXpotrf()
      • 2.4.4.2. cusolverDnXpotrs()
      • 2.4.4.3. cusolverDnXgetrf()
      • 2.4.4.4. cusolverDnXgetrs()
      • 2.4.4.5. cusolverDnXgeqrf()
      • 2.4.4.6. cusolverDnXsytrs()
      • 2.4.4.7. cusolverDnXtrtri()
    • 2.4.5. Dense Eigenvalue Solver Reference (64-bit API)
      • 2.4.5.1. cusolverDnXgesvd()
      • 2.4.5.2. cusolverDnXgesvdp()
      • 2.4.5.3. cusolverDnXgesvdr()
      • 2.4.5.4. cusolverDnXsyevd()
      • 2.4.5.5. cusolverDnXsyevdx()
  • 2.5. cuSolverSP: sparse LAPACK Function Reference
    • 2.5.1. Helper Function Reference
      • 2.5.1.1. cusolverSpCreate()
      • 2.5.1.2. cusolverSpDestroy()
      • 2.5.1.3. cusolverSpSetStream()
      • 2.5.1.4. cusolverSpXcsrissym()
    • 2.5.2. High Level Function Reference
      • 2.5.2.1. cusolverSp<t>csrlsvlu()
      • 2.5.2.2. cusolverSp<t>csrlsvqr()
      • 2.5.2.3. cusolverSp<t>csrlsvchol()
      • 2.5.2.4. cusolverSp<t>csrlsqvqr()
      • 2.5.2.5. cusolverSp<t>csreigvsi()
      • 2.5.2.6. cusolverSp<t>csreigs()
    • 2.5.3. Low Level Function Reference
      • 2.5.3.1. cusolverSpXcsrsymrcm()
      • 2.5.3.2. cusolverSpXcsrsymmdq()
      • 2.5.3.3. cusolverSpXcsrsymamd()
      • 2.5.3.4. cusolverSpXcsrmetisnd()
      • 2.5.3.5. cusolverSpXcsrzfd()
      • 2.5.3.6. cusolverSpXcsrperm()
      • 2.5.3.7. cusolverSpXcsrqrBatched()
  • 2.6. cuSolverRF: Refactorization Reference
    • 2.6.1. cusolverRfAccessBundledFactors()
    • 2.6.2. cusolverRfAnalyze()
    • 2.6.3. cusolverRfSetupDevice()
    • 2.6.4. cusolverRfSetupHost()
    • 2.6.5. cusolverRfCreate()
    • 2.6.6. cusolverRfExtractBundledFactorsHost()
    • 2.6.7. cusolverRfExtractSplitFactorsHost()
    • 2.6.8. cusolverRfDestroy()
    • 2.6.9. cusolverRfGetMatrixFormat()
    • 2.6.10. cusolverRfGetNumericProperties()
    • 2.6.11. cusolverRfGetNumericBoostReport()
    • 2.6.12. cusolverRfGetResetValuesFastMode()
    • 2.6.13. cusolverRfGet_Algs()
    • 2.6.14. cusolverRfRefactor()
    • 2.6.15. cusolverRfResetValues()
    • 2.6.16. cusolverRfSetMatrixFormat()
    • 2.6.17. cusolverRfSetNumericProperties()
    • 2.6.18. cusolverRfSetResetValuesFastMode()
    • 2.6.19. cusolverRfSetAlgs()
    • 2.6.20. cusolverRfSolve()
    • 2.6.21. cusolverRfBatchSetupHost()
    • 2.6.22. cusolverRfBatchAnalyze()
    • 2.6.23. cusolverRfBatchResetValues()
    • 2.6.24. cusolverRfBatchRefactor()
    • 2.6.25. cusolverRfBatchSolve()
    • 2.6.26. cusolverRfBatchZeroPivot()
• 3. Using the CUSOLVERMG API
  • 3.1. General Description
    • 3.1.1. Thread Safety
    • 3.1.2. Determinism
    • 3.1.3. Tile Strategy
    • 3.1.4. Global Matrix Versus Local Matrix
    • 3.1.5. Usage of _bufferSize
    • 3.1.6. Synchronization
    • 3.1.7. Context Switch
    • 3.1.8. NVLINK
  • 3.2. cuSolverMG Types Reference
    • 3.2.1. cuSolverMG Types
    • 3.2.2. cusolverMgHandle_t
    • 3.2.3. cusolverMgGridMapping_t
    • 3.2.4. cudaLibMgGrid_t
    • 3.2.5. cudaLibMgMatrixDesc_t
  • 3.3. Helper Function Reference
    • 3.3.1. cusolverMgCreate()
    • 3.3.2. cusolverMgDestroy()
    • 3.3.3. cusolverMgDeviceSelect()
    • 3.3.4. cusolverMgCreateDeviceGrid()
    • 3.3.5. cusolverMgDestroyGrid()
    • 3.3.6. cusolverMgCreateMatDescr()
    • 3.3.7. cusolverMgDestroyMatrixDesc()
  • 3.4. Dense Linear Solver Reference
    • 3.4.1. cusolverMgPotrf()
    • 3.4.2. cusolverMgPotrs()
    • 3.4.3. cusolverMgPotri()
    • 3.4.4. cusolverMgGetrf()
    • 3.4.5. cusolverMgGetrs()
  • 3.5. Dense Eigenvalue Solver Reference
    • 3.5.1. cusolverMgSyevd()
• 4. Acknowledgements
• 5. Bibliography
• 6. Notices
  • 6.1. Notice
  • 6.2. OpenCL
  • 6.3. Trademarks