libgpuarray/src/gpuarray_buffer_blas.c at master · jrao1/libgpuarray · GitHub

History

248 lines (223 loc) · 9.34 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

#include "private.h"

#include <gpuarray/error.h>

int gpublas_setup(gpucontext *ctx) {

if (ctx->blas_ops == NULL)

return GA_UNSUPPORTED_ERROR;

return ctx->blas_ops->setup(ctx);

}

void gpublas_teardown(gpucontext *ctx) {

if (ctx->blas_ops != NULL)

return ctx->blas_ops->teardown(ctx);

}

const char *gpublas_error(gpucontext *ctx) {

if (ctx->blas_ops != NULL)

return ctx->blas_ops->error(ctx);

return "No blas ops available, API error.";

}

int gpublas_hdot(

size_t N,

gpudata *X, size_t offX, size_t incX,

gpudata *Y, size_t offY, size_t incY,

gpudata *Z, size_t offZ) {

return gpudata_context(X)->blas_ops->hdot(

N, X, offX, incX, Y, offY, incY, Z, offZ);

}

int gpublas_sdot(

size_t N,

gpudata *X, size_t offX, size_t incX,

gpudata *Y, size_t offY, size_t incY,

gpudata *Z, size_t offZ) {

return gpudata_context(X)->blas_ops->sdot(

N, X, offX, incX, Y, offY, incY, Z, offZ);

}

int gpublas_ddot(

size_t N,

gpudata *X, size_t offX, size_t incX,

gpudata *Y, size_t offY, size_t incY,

gpudata *Z, size_t offZ) {

return gpudata_context(X)->blas_ops->ddot(

N, X, offX, incX, Y, offY, incY, Z, offZ);

}

int gpublas_hgemv(cb_order order, cb_transpose transA,

size_t M, size_t N, float alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *X, size_t offX, int incX,

float beta,

gpudata *Y, size_t offY, int incY) {

return gpudata_context(A)->blas_ops->hgemv(

order, transA, M, N, alpha, A, offA, lda,

X, offX, incX, beta, Y, offY, incY);

}

int gpublas_sgemv(cb_order order, cb_transpose transA,

size_t M, size_t N, float alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *X, size_t offX, int incX,

float beta,

gpudata *Y, size_t offY, int incY) {

return gpudata_context(A)->blas_ops->sgemv(

order, transA, M, N, alpha, A, offA, lda,

X, offX, incX, beta, Y, offY, incY);

}

int gpublas_dgemv(cb_order order, cb_transpose transA,

size_t M, size_t N, double alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *X, size_t offX, int incX,

double beta,

gpudata *Y, size_t offY, int incY) {

return gpudata_context(A)->blas_ops->dgemv(

order, transA, M, N, alpha, A, offA, lda,

X, offX, incX, beta, Y, offY, incY);

}

int gpublas_hgemm(cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, float alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *B, size_t offB, size_t ldb,

float beta, gpudata *C, size_t offC, size_t ldc) {

return gpudata_context(A)->blas_ops->hgemm(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc);

}

int gpublas_sgemm(cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, float alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *B, size_t offB, size_t ldb,

float beta, gpudata *C, size_t offC, size_t ldc) {

return gpudata_context(A)->blas_ops->sgemm(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc);

}

int gpublas_dgemm(cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, double alpha,

gpudata *A, size_t offA, size_t lda,

gpudata *B, size_t offB, size_t ldb,

double beta, gpudata *C, size_t offC, size_t ldc) {

return gpudata_context(A)->blas_ops->dgemm(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc);

}

int gpublas_hger(cb_order order, size_t M, size_t N, float alpha,

gpudata *X, size_t offX, int incX,

gpudata *Y, size_t offY, int incY,

gpudata *A, size_t offA, size_t lda) {

return gpudata_context(X)->blas_ops->hger(

order, M, N, alpha, X, offX, incX, Y, offY, incY, A, offA, lda);

}

int gpublas_sger(cb_order order, size_t M, size_t N, float alpha,

gpudata *X, size_t offX, int incX,

gpudata *Y, size_t offY, int incY,

gpudata *A, size_t offA, size_t lda) {

return gpudata_context(X)->blas_ops->sger(

order, M, N, alpha, X, offX, incX, Y, offY, incY, A, offA, lda);

}

int gpublas_dger(cb_order order, size_t M, size_t N, double alpha,

gpudata *X, size_t offX, int incX,

gpudata *Y, size_t offY, int incY,

gpudata *A, size_t offA, size_t lda) {

return gpudata_context(X)->blas_ops->dger(

order, M, N, alpha, X, offX, incX, Y, offY, incY, A, offA, lda);

}

int gpublas_hgemmBatch(

cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, float alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **B, size_t *offB, size_t ldb,

float beta, gpudata **C, size_t *offC, size_t ldc,

size_t batchCount, int flags) {

if (flags != 0) return GA_INVALID_ERROR;

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->hgemmBatch(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc, batchCount);

}

int gpublas_sgemmBatch(

cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, float alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **B, size_t *offB, size_t ldb,

float beta, gpudata **C, size_t *offC, size_t ldc,

size_t batchCount, int flags) {

if (flags != 0) return GA_INVALID_ERROR;

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->sgemmBatch(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc, batchCount);

}

int gpublas_dgemmBatch(

cb_order order, cb_transpose transA, cb_transpose transB,

size_t M, size_t N, size_t K, double alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **B, size_t *offB, size_t ldb,

double beta, gpudata **C, size_t *offC, size_t ldc,

size_t batchCount, int flags) {

if (flags != 0) return GA_INVALID_ERROR;

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->dgemmBatch(

order, transA, transB, M, N, K, alpha, A, offA, lda,

B, offB, ldb, beta, C, offC, ldc, batchCount);

}

int gpublas_hgemvBatch(

cb_order order, cb_transpose transA,

size_t M, size_t N, float alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **x, size_t *offX, size_t incX,

float beta, gpudata **y, size_t *offY, size_t incY,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->hgemvBatch(

order, transA, M, N, alpha, A, offA, lda, x, offX, incX,

beta, y, offY, incY, batchCount, flags);

}

int gpublas_sgemvBatch(

cb_order order, cb_transpose transA,

size_t M, size_t N, float alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **x, size_t *offX, size_t incX,

float beta, gpudata **y, size_t *offY, size_t incY,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->sgemvBatch(

order, transA, M, N, alpha, A, offA, lda, x, offX, incX,

beta, y, offY, incY, batchCount, flags);

}

int gpublas_dgemvBatch(

cb_order order, cb_transpose transA,

size_t M, size_t N, double alpha,

gpudata **A, size_t *offA, size_t lda,

gpudata **x, size_t *offX, size_t incX,

double beta, gpudata **y, size_t *offY, size_t incY,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(A[0])->blas_ops->dgemvBatch(

order, transA, M, N, alpha, A, offA, lda, x, offX, incX,

beta, y, offY, incY, batchCount, flags);

}

int gpublas_hgerBatch(cb_order order, size_t M, size_t N, float alpha,

gpudata **x, size_t *offX, size_t incX,

gpudata **y, size_t *offY, size_t incY,

gpudata **A, size_t *offA, size_t lda,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(x[0])->blas_ops->hgerBatch(

order, M, N, alpha, x, offX, incX, y, offY, incY,

A, offA, lda, batchCount, flags);

}

int gpublas_sgerBatch(cb_order order, size_t M, size_t N, float alpha,

gpudata **x, size_t *offX, size_t incX,

gpudata **y, size_t *offY, size_t incY,

gpudata **A, size_t *offA, size_t lda,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(x[0])->blas_ops->sgerBatch(

order, M, N, alpha, x, offX, incX, y, offY, incY,

A, offA, lda, batchCount, flags);

}

int gpublas_dgerBatch(cb_order order, size_t M, size_t N, double alpha,

gpudata **x, size_t *offX, size_t incX,

gpudata **y, size_t *offY, size_t incY,

gpudata **A, size_t *offA, size_t lda,

size_t batchCount, int flags) {

if (batchCount == 0) return GA_NO_ERROR;

return gpudata_context(x[0])->blas_ops->dgerBatch(

order, M, N, alpha, x, offX, incX, y, offY, incY,

A, offA, lda, batchCount, flags);

}

libgpuarray/src/gpuarray_buffer_blas.c at master · jrao1/libgpuarray · GitHub

Expand file tree

Breadcrumbs

gpuarray_buffer_blas.c

Latest commit

History

Breadcrumbs

gpuarray_buffer_blas.c

File metadata and controls