xref: /petsc/src/mat/impls/aij/seq/seqcusparse/cusparsematimpl.h (revision 7e4e2a2d5b77d05327773006136c5762506227f3)
1 #if !defined(__CUSPARSEMATIMPL)
2 #define __CUSPARSEMATIMPL
3 
4 #include <../src/vec/vec/impls/seq/seqcuda/cudavecimpl.h>
5 
6 #include <cusparse_v2.h>
7 
8 #include <algorithm>
9 #include <vector>
10 
11 #include <thrust/device_vector.h>
12 #include <thrust/device_ptr.h>
13 #include <thrust/device_malloc_allocator.h>
14 #include <thrust/transform.h>
15 #include <thrust/functional.h>
16 #include <thrust/sequence.h>
17 
18 #if defined(PETSC_USE_COMPLEX)
19 #if defined(PETSC_USE_REAL_SINGLE)
20 #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))
21 #define cusparse_analysis(a,b,c,d,e,f,g,h,i)         cusparseCcsrsv_analysis((a),(b),(c),(d),(e),(cuComplex*)(f),(g),(h),(i))
22 #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))
23 #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))
24 #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))
25 #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))
26 #define cusparse_hyb2csr(a,b,c,d,e,f)                cusparseChyb2csr((a),(b),(c),(cuComplex*)(d),(e),(f))
27 cuFloatComplex PETSC_CUSPARSE_ONE  = {1.0f, 0.0f};
28 cuFloatComplex PETSC_CUSPARSE_ZERO = {0.0f, 0.0f};
29 #elif defined(PETSC_USE_REAL_DOUBLE)
30 #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))
31 #define cusparse_analysis(a,b,c,d,e,f,g,h,i)         cusparseZcsrsv_analysis((a),(b),(c),(d),(e),(cuDoubleComplex*)(f),(g),(h),(i))
32 #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))
33 #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))
34 #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))
35 #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))
36 #define cusparse_hyb2csr(a,b,c,d,e,f)                cusparseZhyb2csr((a),(b),(c),(cuDoubleComplex*)(d),(e),(f))
37 cuDoubleComplex PETSC_CUSPARSE_ONE  = {1.0, 0.0};
38 cuDoubleComplex PETSC_CUSPARSE_ZERO = {0.0, 0.0};
39 #endif
40 #else
41 PetscScalar PETSC_CUSPARSE_ONE  = 1.0;
42 PetscScalar PETSC_CUSPARSE_ZERO = 0.0;
43 #if defined(PETSC_USE_REAL_SINGLE)
44 #define cusparse_solve    cusparseScsrsv_solve
45 #define cusparse_analysis cusparseScsrsv_analysis
46 #define cusparse_csr_spmv cusparseScsrmv
47 #define cusparse_csr2csc  cusparseScsr2csc
48 #define cusparse_hyb_spmv cusparseShybmv
49 #define cusparse_csr2hyb  cusparseScsr2hyb
50 #define cusparse_hyb2csr  cusparseShyb2csr
51 #elif defined(PETSC_USE_REAL_DOUBLE)
52 #define cusparse_solve    cusparseDcsrsv_solve
53 #define cusparse_analysis cusparseDcsrsv_analysis
54 #define cusparse_csr_spmv cusparseDcsrmv
55 #define cusparse_csr2csc  cusparseDcsr2csc
56 #define cusparse_hyb_spmv cusparseDhybmv
57 #define cusparse_csr2hyb  cusparseDcsr2hyb
58 #define cusparse_hyb2csr  cusparseDhyb2csr
59 #endif
60 #endif
61 
62 #define THRUSTINTARRAY32 thrust::device_vector<int>
63 #define THRUSTINTARRAY thrust::device_vector<PetscInt>
64 #define THRUSTARRAY thrust::device_vector<PetscScalar>
65 
66 /* A CSR matrix structure */
67 struct CsrMatrix {
68   PetscInt         num_rows;
69   PetscInt         num_cols;
70   PetscInt         num_entries;
71   THRUSTINTARRAY32 *row_offsets;
72   THRUSTINTARRAY32 *column_indices;
73   THRUSTARRAY      *values;
74 };
75 
76 //#define CUSPMATRIXCSR32 cusp::csr_matrix<int,PetscScalar,cusp::device_memory>
77 
78 /* This is struct holding the relevant data needed to a MatSolve */
79 struct Mat_SeqAIJCUSPARSETriFactorStruct {
80   /* Data needed for triangular solve */
81   cusparseMatDescr_t          descr;
82   cusparseSolveAnalysisInfo_t solveInfo;
83   cusparseOperation_t         solveOp;
84   CsrMatrix                   *csrMat;
85 };
86 
87 /* This is struct holding the relevant data needed to a MatMult */
88 struct Mat_SeqAIJCUSPARSEMultStruct {
89   void               *mat;  /* opaque pointer to a matrix. This could be either a cusparseHybMat_t or a CsrMatrix */
90   cusparseMatDescr_t descr; /* Data needed to describe the matrix for a multiply */
91   THRUSTINTARRAY     *cprowIndices;   /* compressed row indices used in the parallel SpMV */
92   PetscScalar        *alpha; /* pointer to a device "scalar" storing the alpha parameter in the SpMV */
93   PetscScalar        *beta_zero; /* pointer to a device "scalar" storing the beta parameter in the SpMV as zero*/
94   PetscScalar        *beta_one; /* pointer to a device "scalar" storing the beta parameter in the SpMV as one */
95 };
96 
97 /* This is a larger struct holding all the triangular factors for a solve, transpose solve, and
98  any indices used in a reordering */
99 struct Mat_SeqAIJCUSPARSETriFactors {
100   Mat_SeqAIJCUSPARSETriFactorStruct *loTriFactorPtr; /* pointer for lower triangular (factored matrix) on GPU */
101   Mat_SeqAIJCUSPARSETriFactorStruct *upTriFactorPtr; /* pointer for upper triangular (factored matrix) on GPU */
102   Mat_SeqAIJCUSPARSETriFactorStruct *loTriFactorPtrTranspose; /* pointer for lower triangular (factored matrix) on GPU for the transpose (useful for BiCG) */
103   Mat_SeqAIJCUSPARSETriFactorStruct *upTriFactorPtrTranspose; /* pointer for upper triangular (factored matrix) on GPU for the transpose (useful for BiCG)*/
104   THRUSTINTARRAY                    *rpermIndices;  /* indices used for any reordering */
105   THRUSTINTARRAY                    *cpermIndices;  /* indices used for any reordering */
106   THRUSTARRAY                       *workVector;
107   cusparseHandle_t                  handle;   /* a handle to the cusparse library */
108   PetscInt                          nnz;      /* number of nonzeros ... need this for accurate logging between ICC and ILU */
109 };
110 
111 /* This is a larger struct holding all the matrices for a SpMV, and SpMV Tranpose */
112 struct Mat_SeqAIJCUSPARSE {
113   Mat_SeqAIJCUSPARSEMultStruct *mat; /* pointer to the matrix on the GPU */
114   Mat_SeqAIJCUSPARSEMultStruct *matTranspose; /* pointer to the matrix on the GPU (for the transpose ... useful for BiCG) */
115   THRUSTARRAY                  *workVector; /*pointer to a workvector to which we can copy the relevant indices of a vector we want to multiply */
116   PetscInt                     nonzerorow; /* number of nonzero rows ... used in the flop calculations */
117   MatCUSPARSEStorageFormat     format;   /* the storage format for the matrix on the device */
118   cudaStream_t                 stream;   /* a stream for the parallel SpMV ... this is not owned and should not be deleted */
119   cusparseHandle_t             handle;   /* a handle to the cusparse library ... this may not be owned (if we're working in parallel i.e. multiGPUs) */
120   PetscObjectState             nonzerostate;
121 };
122 
123 PETSC_INTERN PetscErrorCode MatCUSPARSECopyToGPU(Mat);
124 PETSC_INTERN PetscErrorCode MatCUSPARSESetStream(Mat, const cudaStream_t stream);
125 PETSC_INTERN PetscErrorCode MatCUSPARSESetHandle(Mat, const cusparseHandle_t handle);
126 PETSC_INTERN PetscErrorCode MatCUSPARSEClearHandle(Mat);
127 #endif
128