Actual source code: cusparsematimpl.h


  4: #include <petscpkg_version.h>
  5: #include <petsc/private/cudavecimpl.h>

  7: #include <cusparse_v2.h>

  9: #include <algorithm>
 10: #include <vector>

 12: #include <thrust/device_vector.h>
 13: #include <thrust/device_ptr.h>
 14: #include <thrust/device_malloc_allocator.h>
 15: #include <thrust/transform.h>
 16: #include <thrust/functional.h>
 17: #include <thrust/sequence.h>
 18: #include <thrust/system/system_error.h>

 20: #if (CUSPARSE_VER_MAJOR > 10 || CUSPARSE_VER_MAJOR == 10 && CUSPARSE_VER_MINOR >= 2) /* According to cuda/10.1.168 on OLCF Summit */
 21: #define CHKERRCUSPARSE(stat)\
 22: do {\
 23:   if (PetscUnlikely(stat)) {\
 24:     const char *name  = cusparseGetErrorName(stat);\
 25:     const char *descr = cusparseGetErrorString(stat);\
 26:     if ((stat == CUSPARSE_STATUS_NOT_INITIALIZED) || (stat == CUSPARSE_STATUS_ALLOC_FAILED)) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_GPU_RESOURCE,"cuSPARSE error %d (%s) : %s. Reports not initialized or alloc failed; this indicates the GPU has run out resources",(int)stat,name,descr); \
 27:     else SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_GPU,"cuSPARSE error %d (%s) : %s",(int)stat,name,descr);\
 28:   }\
 29: } while (0)
 30: #else
 31: #define CHKERRCUSPARSE(stat) do {if (PetscUnlikely(stat)) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_GPU,"cusparse error %d",(int)stat);} while (0)
 32: #endif

 34: #define PetscStackCallThrust(body) do {                                     \
 35:     try {                                                                   \
 36:       body;                                                                 \
 37:     } catch(thrust::system_error& e) {                                      \
 38:       SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_LIB,"Error in Thrust %s",e.what());\
 39:     }                                                                       \
 40:   } while (0)

 42: #if defined(PETSC_USE_COMPLEX)
 43:   #if defined(PETSC_USE_REAL_SINGLE)
 44:     const cuComplex PETSC_CUSPARSE_ONE        = {1.0f, 0.0f};
 45:     const cuComplex PETSC_CUSPARSE_ZERO       = {0.0f, 0.0f};
 46:   #elif defined(PETSC_USE_REAL_DOUBLE)
 47:     const cuDoubleComplex PETSC_CUSPARSE_ONE  = {1.0, 0.0};
 48:     const cuDoubleComplex PETSC_CUSPARSE_ZERO = {0.0, 0.0};
 49:   #endif
 50: #else
 51:   const PetscScalar PETSC_CUSPARSE_ONE        = 1.0;
 52:   const PetscScalar PETSC_CUSPARSE_ZERO       = 0.0;
 53: #endif

 55: #if PETSC_PKG_CUDA_VERSION_GE(9,0,0)
 56:   #define cusparse_create_analysis_info  cusparseCreateCsrsv2Info
 57:   #define cusparse_destroy_analysis_info cusparseDestroyCsrsv2Info
 58:   #if defined(PETSC_USE_COMPLEX)
 59:     #if defined(PETSC_USE_REAL_SINGLE)
 60:       #define cusparse_get_svbuffsize(a,b,c,d,e,f,g,h,i,j) cusparseCcsrsv2_bufferSize(a,b,c,d,e,(cuComplex*)(f),g,h,i,j)
 61:       #define cusparse_analysis(a,b,c,d,e,f,g,h,i,j,k)     cusparseCcsrsv2_analysis(a,b,c,d,e,(const cuComplex*)(f),g,h,i,j,k)
 62:       #define cusparse_solve(a,b,c,d,e,f,g,h,i,j,k,l,m,n)  cusparseCcsrsv2_solve(a,b,c,d,(const cuComplex*)(e),f,(const cuComplex*)(g),h,i,j,(const cuComplex*)(k),(cuComplex*)(l),m,n)
 63:     #elif defined(PETSC_USE_REAL_DOUBLE)
 64:       #define cusparse_get_svbuffsize(a,b,c,d,e,f,g,h,i,j) cusparseZcsrsv2_bufferSize(a,b,c,d,e,(cuDoubleComplex*)(f),g,h,i,j)
 65:       #define cusparse_analysis(a,b,c,d,e,f,g,h,i,j,k)     cusparseZcsrsv2_analysis(a,b,c,d,e,(const cuDoubleComplex*)(f),g,h,i,j,k)
 66:       #define cusparse_solve(a,b,c,d,e,f,g,h,i,j,k,l,m,n)  cusparseZcsrsv2_solve(a,b,c,d,(const cuDoubleComplex*)(e),f,(const cuDoubleComplex*)(g),h,i,j,(const cuDoubleComplex*)(k),(cuDoubleComplex*)(l),m,n)
 67:     #endif
 68:   #else /* not complex */
 69:     #if defined(PETSC_USE_REAL_SINGLE)
 70:       #define cusparse_get_svbuffsize cusparseScsrsv2_bufferSize
 71:       #define cusparse_analysis       cusparseScsrsv2_analysis
 72:       #define cusparse_solve          cusparseScsrsv2_solve
 73:     #elif defined(PETSC_USE_REAL_DOUBLE)
 74:       #define cusparse_get_svbuffsize cusparseDcsrsv2_bufferSize
 75:       #define cusparse_analysis       cusparseDcsrsv2_analysis
 76:       #define cusparse_solve          cusparseDcsrsv2_solve
 77:     #endif
 78:   #endif
 79: #else
 80:   #define cusparse_create_analysis_info  cusparseCreateSolveAnalysisInfo
 81:   #define cusparse_destroy_analysis_info cusparseDestroySolveAnalysisInfo
 82:   #if defined(PETSC_USE_COMPLEX)
 83:     #if defined(PETSC_USE_REAL_SINGLE)
 84:       #define cusparse_solve(a,b,c,d,e,f,g,h,i,j,k) cusparseCcsrsv_solve((a),(b),(c),(cuComplex*)(d),(e),(cuComplex*)(f),(g),(h),(i),(cuComplex*)(j),(cuComplex*)(k))
 85:       #define cusparse_analysis(a,b,c,d,e,f,g,h,i)  cusparseCcsrsv_analysis((a),(b),(c),(d),(e),(cuComplex*)(f),(g),(h),(i))
 86:     #elif defined(PETSC_USE_REAL_DOUBLE)
 87:       #define cusparse_solve(a,b,c,d,e,f,g,h,i,j,k) cusparseZcsrsv_solve((a),(b),(c),(cuDoubleComplex*)(d),(e),(cuDoubleComplex*)(f),(g),(h),(i),(cuDoubleComplex*)(j),(cuDoubleComplex*)(k))
 88:       #define cusparse_analysis(a,b,c,d,e,f,g,h,i)  cusparseZcsrsv_analysis((a),(b),(c),(d),(e),(cuDoubleComplex*)(f),(g),(h),(i))
 89:     #endif
 90:   #else /* not complex */
 91:     #if defined(PETSC_USE_REAL_SINGLE)
 92:       #define cusparse_solve    cusparseScsrsv_solve
 93:       #define cusparse_analysis cusparseScsrsv_analysis
 94:     #elif defined(PETSC_USE_REAL_DOUBLE)
 95:       #define cusparse_solve    cusparseDcsrsv_solve
 96:       #define cusparse_analysis cusparseDcsrsv_analysis
 97:     #endif
 98:   #endif
 99: #endif

101: #if PETSC_PKG_CUDA_VERSION_GE(11,0,0)
102:   #define cusparse_csr2csc cusparseCsr2cscEx2
103:   #if defined(PETSC_USE_COMPLEX)
104:     #if defined(PETSC_USE_REAL_SINGLE)
105:       #define cusparse_scalartype CUDA_C_32F
106:       #define cusparse_csr_spgeam(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t)            cusparseCcsrgeam2(a,b,c,(cuComplex*)d,e,f,(cuComplex*)g,h,i,(cuComplex*)j,k,l,(cuComplex*)m,n,o,p,(cuComplex*)q,r,s,t)
107:       #define cusparse_csr_spgeam_bufferSize(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t) cusparseCcsrgeam2_bufferSizeExt(a,b,c,(cuComplex*)d,e,f,(cuComplex*)g,h,i,(cuComplex*)j,k,l,(cuComplex*)m,n,o,p,(cuComplex*)q,r,s,t)
108:     #elif defined(PETSC_USE_REAL_DOUBLE)
109:       #define cusparse_scalartype CUDA_C_64F
110:       #define cusparse_csr_spgeam(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t)            cusparseZcsrgeam2(a,b,c,(cuDoubleComplex*)d,e,f,(cuDoubleComplex*)g,h,i,(cuDoubleComplex*)j,k,l,(cuDoubleComplex*)m,n,o,p,(cuDoubleComplex*)q,r,s,t)
111:       #define cusparse_csr_spgeam_bufferSize(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t) cusparseZcsrgeam2_bufferSizeExt(a,b,c,(cuDoubleComplex*)d,e,f,(cuDoubleComplex*)g,h,i,(cuDoubleComplex*)j,k,l,(cuDoubleComplex*)m,n,o,p,(cuDoubleComplex*)q,r,s,t)
112:     #endif
113:   #else /* not complex */
114:     #if defined(PETSC_USE_REAL_SINGLE)
115:       #define cusparse_scalartype CUDA_R_32F
116:       #define cusparse_csr_spgeam            cusparseScsrgeam2
117:       #define cusparse_csr_spgeam_bufferSize cusparseScsrgeam2_bufferSizeExt
118:     #elif defined(PETSC_USE_REAL_DOUBLE)
119:       #define cusparse_scalartype CUDA_R_64F
120:       #define cusparse_csr_spgeam            cusparseDcsrgeam2
121:       #define cusparse_csr_spgeam_bufferSize cusparseDcsrgeam2_bufferSizeExt
122:     #endif
123:   #endif
124: #else
125:   #if defined(PETSC_USE_COMPLEX)
126:     #if defined(PETSC_USE_REAL_SINGLE)
127:       #define cusparse_csr_spmv(a,b,c,d,e,f,g,h,i,j,k,l,m)       cusparseCcsrmv((a),(b),(c),(d),(e),(cuComplex*)(f),(g),(cuComplex*)(h),(i),(j),(cuComplex*)(k),(cuComplex*)(l),(cuComplex*)(m))
128:       #define cusparse_csr_spmm(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) cusparseCcsrmm((a),(b),(c),(d),(e),(f),(cuComplex*)(g),(h),(cuComplex*)(i),(j),(k),(cuComplex*)(l),(m),(cuComplex*)(n),(cuComplex*)(o),(p))
129:       #define cusparse_csr2csc(a,b,c,d,e,f,g,h,i,j,k,l)          cusparseCcsr2csc((a),(b),(c),(d),(cuComplex*)(e),(f),(g),(cuComplex*)(h),(i),(j),(k),(l))
130:       #define cusparse_hyb_spmv(a,b,c,d,e,f,g,h)                 cusparseChybmv((a),(b),(cuComplex*)(c),(d),(e),(cuComplex*)(f),(cuComplex*)(g),(cuComplex*)(h))
131:       #define cusparse_csr2hyb(a,b,c,d,e,f,g,h,i,j)              cusparseCcsr2hyb((a),(b),(c),(d),(cuComplex*)(e),(f),(g),(h),(i),(j))
132:       #define cusparse_hyb2csr(a,b,c,d,e,f)                      cusparseChyb2csr((a),(b),(c),(cuComplex*)(d),(e),(f))
133:       #define cusparse_csr_spgemm(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t) cusparseCcsrgemm(a,b,c,d,e,f,g,h,(cuComplex*)i,j,k,l,m,(cuComplex*)n,o,p,q,(cuComplex*)r,s,t)
134:       #define cusparse_csr_spgeam(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s)   cusparseCcsrgeam(a,b,c,(cuComplex*)d,e,f,(cuComplex*)g,h,i,(cuComplex*)j,k,l,(cuComplex*)m,n,o,p,(cuComplex*)q,r,s)
135:     #elif defined(PETSC_USE_REAL_DOUBLE)
136:       #define cusparse_csr_spmv(a,b,c,d,e,f,g,h,i,j,k,l,m)       cusparseZcsrmv((a),(b),(c),(d),(e),(cuDoubleComplex*)(f),(g),(cuDoubleComplex*)(h),(i),(j),(cuDoubleComplex*)(k),(cuDoubleComplex*)(l),(cuDoubleComplex*)(m))
137:       #define cusparse_csr_spmm(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) cusparseZcsrmm((a),(b),(c),(d),(e),(f),(cuDoubleComplex*)(g),(h),(cuDoubleComplex*)(i),(j),(k),(cuDoubleComplex*)(l),(m),(cuDoubleComplex*)(n),(cuDoubleComplex*)(o),(p))
138:       #define cusparse_csr2csc(a,b,c,d,e,f,g,h,i,j,k,l)          cusparseZcsr2csc((a),(b),(c),(d),(cuDoubleComplex*)(e),(f),(g),(cuDoubleComplex*)(h),(i),(j),(k),(l))
139:       #define cusparse_hyb_spmv(a,b,c,d,e,f,g,h)                 cusparseZhybmv((a),(b),(cuDoubleComplex*)(c),(d),(e),(cuDoubleComplex*)(f),(cuDoubleComplex*)(g),(cuDoubleComplex*)(h))
140:       #define cusparse_csr2hyb(a,b,c,d,e,f,g,h,i,j)              cusparseZcsr2hyb((a),(b),(c),(d),(cuDoubleComplex*)(e),(f),(g),(h),(i),(j))
141:       #define cusparse_hyb2csr(a,b,c,d,e,f)                      cusparseZhyb2csr((a),(b),(c),(cuDoubleComplex*)(d),(e),(f))
142:       #define cusparse_csr_spgemm(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t) cusparseZcsrgemm(a,b,c,d,e,f,g,h,(cuDoubleComplex*)i,j,k,l,m,(cuDoubleComplex*)n,o,p,q,(cuDoubleComplex*)r,s,t)
143:       #define cusparse_csr_spgeam(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s)   cusparseZcsrgeam(a,b,c,(cuDoubleComplex*)d,e,f,(cuDoubleComplex*)g,h,i,(cuDoubleComplex*)j,k,l,(cuDoubleComplex*)m,n,o,p,(cuDoubleComplex*)q,r,s)
144:     #endif
145:   #else
146:     #if defined(PETSC_USE_REAL_SINGLE)
147:       #define cusparse_csr_spmv cusparseScsrmv
148:       #define cusparse_csr_spmm cusparseScsrmm
149:       #define cusparse_csr2csc  cusparseScsr2csc
150:       #define cusparse_hyb_spmv cusparseShybmv
151:       #define cusparse_csr2hyb  cusparseScsr2hyb
152:       #define cusparse_hyb2csr  cusparseShyb2csr
153:       #define cusparse_csr_spgemm cusparseScsrgemm
154:       #define cusparse_csr_spgeam cusparseScsrgeam
155:     #elif defined(PETSC_USE_REAL_DOUBLE)
156:       #define cusparse_csr_spmv cusparseDcsrmv
157:       #define cusparse_csr_spmm cusparseDcsrmm
158:       #define cusparse_csr2csc  cusparseDcsr2csc
159:       #define cusparse_hyb_spmv cusparseDhybmv
160:       #define cusparse_csr2hyb  cusparseDcsr2hyb
161:       #define cusparse_hyb2csr  cusparseDhyb2csr
162:       #define cusparse_csr_spgemm cusparseDcsrgemm
163:       #define cusparse_csr_spgeam cusparseDcsrgeam
164:     #endif
165:   #endif
166: #endif

168: #define THRUSTINTARRAY32 thrust::device_vector<int>
169: #define THRUSTINTARRAY thrust::device_vector<PetscInt>
170: #define THRUSTARRAY thrust::device_vector<PetscScalar>

172: /* A CSR matrix structure */
173: struct CsrMatrix {
174:   PetscInt         num_rows;
175:   PetscInt         num_cols;
176:   PetscInt         num_entries;
177:   THRUSTINTARRAY32 *row_offsets;
178:   THRUSTINTARRAY32 *column_indices;
179:   THRUSTARRAY      *values;
180: };

182: /* This is struct holding the relevant data needed to a MatSolve */
183: struct Mat_SeqAIJCUSPARSETriFactorStruct {
184:   /* Data needed for triangular solve */
185:   cusparseMatDescr_t          descr;
186:   cusparseOperation_t         solveOp;
187:   CsrMatrix                   *csrMat;
188:  #if PETSC_PKG_CUDA_VERSION_GE(9,0,0)
189:   csrsv2Info_t                solveInfo;
190:  #else
191:   cusparseSolveAnalysisInfo_t solveInfo;
192:  #endif
193:   cusparseSolvePolicy_t       solvePolicy;     /* whether level information is generated and used */
194:   int                         solveBufferSize;
195:   void                        *solveBuffer;
196:   size_t                      csr2cscBufferSize; /* to transpose the triangular factor (only used for CUDA >= 11.0) */
197:   void                        *csr2cscBuffer;
198:   PetscScalar                 *AA_h; /* managed host buffer for moving values to the GPU */
199: };

201: /* This is a larger struct holding all the triangular factors for a solve, transpose solve, and any indices used in a reordering */
202: struct Mat_SeqAIJCUSPARSETriFactors {
203:   Mat_SeqAIJCUSPARSETriFactorStruct *loTriFactorPtr; /* pointer for lower triangular (factored matrix) on GPU */
204:   Mat_SeqAIJCUSPARSETriFactorStruct *upTriFactorPtr; /* pointer for upper triangular (factored matrix) on GPU */
205:   Mat_SeqAIJCUSPARSETriFactorStruct *loTriFactorPtrTranspose; /* pointer for lower triangular (factored matrix) on GPU for the transpose (useful for BiCG) */
206:   Mat_SeqAIJCUSPARSETriFactorStruct *upTriFactorPtrTranspose; /* pointer for upper triangular (factored matrix) on GPU for the transpose (useful for BiCG)*/
207:   THRUSTINTARRAY                    *rpermIndices;  /* indices used for any reordering */
208:   THRUSTINTARRAY                    *cpermIndices;  /* indices used for any reordering */
209:   THRUSTARRAY                       *workVector;
210:   cusparseHandle_t                  handle;   /* a handle to the cusparse library */
211:   PetscInt                          nnz;      /* number of nonzeros ... need this for accurate logging between ICC and ILU */
212:   PetscScalar                       *a_band_d; /* GPU data for banded CSR LU factorization matrix diag(L)=1 */
213:   int                               *i_band_d; /* this could be optimized away */
214: };

216: struct Mat_CusparseSpMV {
217:   PetscBool             initialized;    /* Don't rely on spmvBuffer != NULL to test if the struct is initialized, */
218:   size_t                spmvBufferSize; /* since I'm not sure if smvBuffer can be NULL even after cusparseSpMV_bufferSize() */
219:   void                  *spmvBuffer;
220:  #if PETSC_PKG_CUDA_VERSION_GE(11,0,0)  /* these are present from CUDA 10.1, but PETSc code makes use of them from CUDA 11 on */
221:   cusparseDnVecDescr_t  vecXDescr,vecYDescr; /* descriptor for the dense vectors in y=op(A)x */
222:  #endif
223: };

225: /* This is struct holding the relevant data needed to a MatMult */
226: struct Mat_SeqAIJCUSPARSEMultStruct {
227:   void               *mat;  /* opaque pointer to a matrix. This could be either a cusparseHybMat_t or a CsrMatrix */
228:   cusparseMatDescr_t descr; /* Data needed to describe the matrix for a multiply */
229:   THRUSTINTARRAY     *cprowIndices;   /* compressed row indices used in the parallel SpMV */
230:   PetscScalar        *alpha_one; /* pointer to a device "scalar" storing the alpha parameter in the SpMV */
231:   PetscScalar        *beta_zero; /* pointer to a device "scalar" storing the beta parameter in the SpMV as zero*/
232:   PetscScalar        *beta_one; /* pointer to a device "scalar" storing the beta parameter in the SpMV as one */
233:  #if PETSC_PKG_CUDA_VERSION_GE(11,0,0)
234:   cusparseSpMatDescr_t  matDescr;  /* descriptor for the matrix, used by SpMV and SpMM */
235:   Mat_CusparseSpMV      cuSpMV[3]; /* different Mat_CusparseSpMV structs for non-transpose, transpose, conj-transpose */
236:   Mat_SeqAIJCUSPARSEMultStruct() : matDescr(NULL) {
237:     for (int i=0; i<3; i++) cuSpMV[i].initialized = PETSC_FALSE;
238:   }
239:  #endif
240: };

242: /* This is a larger struct holding all the matrices for a SpMV, and SpMV Tranpose */
243: struct Mat_SeqAIJCUSPARSE {
244:   Mat_SeqAIJCUSPARSEMultStruct *mat;            /* pointer to the matrix on the GPU */
245:   Mat_SeqAIJCUSPARSEMultStruct *matTranspose;   /* pointer to the matrix on the GPU (for the transpose ... useful for BiCG) */
246:   THRUSTARRAY                  *workVector;     /* pointer to a workvector to which we can copy the relevant indices of a vector we want to multiply */
247:   THRUSTINTARRAY32             *rowoffsets_gpu; /* rowoffsets on GPU in non-compressed-row format. It is used to convert CSR to CSC */
248:   PetscInt                     nrows;           /* number of rows of the matrix seen by GPU */
249:   MatCUSPARSEStorageFormat     format;          /* the storage format for the matrix on the device */
250:   cudaStream_t                 stream;          /* a stream for the parallel SpMV ... this is not owned and should not be deleted */
251:   cusparseHandle_t             handle;          /* a handle to the cusparse library ... this may not be owned (if we're working in parallel i.e. multiGPUs) */
252:   PetscObjectState             nonzerostate;    /* track nonzero state to possibly recreate the GPU matrix */
253:  #if PETSC_PKG_CUDA_VERSION_GE(11,0,0)
254:   size_t                       csr2cscBufferSize; /* stuff used to compute the matTranspose above */
255:   void                         *csr2cscBuffer;    /* This is used as a C struct and is calloc'ed by PetscNewLog() */
256:   cusparseCsr2CscAlg_t         csr2cscAlg;        /* algorithms can be selected from command line options */
257:   cusparseSpMVAlg_t            spmvAlg;
258:   cusparseSpMMAlg_t            spmmAlg;
259:  #endif
260:   THRUSTINTARRAY               *csr2csc_i;
261:   PetscSplitCSRDataStructure   *deviceMat;       /* Matrix on device for, eg, assembly */
262:   THRUSTINTARRAY               *cooPerm;
263:   THRUSTINTARRAY               *cooPerm_a;
264: };

266: PETSC_INTERN PetscErrorCode MatCUSPARSECopyToGPU(Mat);
267: PETSC_INTERN PetscErrorCode MatCUSPARSESetStream(Mat, const cudaStream_t stream);
268: PETSC_INTERN PetscErrorCode MatCUSPARSESetHandle(Mat, const cusparseHandle_t handle);
269: PETSC_INTERN PetscErrorCode MatCUSPARSEClearHandle(Mat);
270: PETSC_INTERN PetscErrorCode MatSetPreallocationCOO_SeqAIJCUSPARSE(Mat,PetscInt,const PetscInt[],const PetscInt[]);
271: PETSC_INTERN PetscErrorCode MatSetValuesCOO_SeqAIJCUSPARSE(Mat,const PetscScalar[],InsertMode);
272: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSEGetArrayRead(Mat,const PetscScalar**);
273: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSERestoreArrayRead(Mat,const PetscScalar**);
274: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSEGetArrayWrite(Mat,PetscScalar**);
275: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSERestoreArrayWrite(Mat,PetscScalar**);
276: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSEGetArray(Mat,PetscScalar**);
277: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSERestoreArray(Mat,PetscScalar**);
278: PETSC_INTERN PetscErrorCode MatSeqAIJCUSPARSEMergeMats(Mat,Mat,MatReuse,Mat*);

280: PETSC_STATIC_INLINE bool isCudaMem(const void *data)
281: {
282:   cudaError_t                  cerr;
283:   struct cudaPointerAttributes attr;
284:   enum cudaMemoryType          mtype;
285:   cerr = cudaPointerGetAttributes(&attr,data); /* Do not check error since before CUDA 11.0, passing a host pointer returns cudaErrorInvalidValue */
286:   cudaGetLastError(); /* Reset the last error */
287:   #if (CUDART_VERSION < 10000)
288:     mtype = attr.memoryType;
289:   #else
290:     mtype = attr.type;
291:   #endif
292:   if (cerr == cudaSuccess && mtype == cudaMemoryTypeDevice) return true;
293:   else return false;
294: }

296: #endif