|
| 1 | +# ********************************************************************************** |
| 2 | +# Copyright (c) 2022, The OpenBLAS Project |
| 3 | +# All rights reserved. |
| 4 | +# Redistribution and use in source and binary forms, with or without |
| 5 | +# modification, are permitted provided that the following conditions are |
| 6 | +# met: |
| 7 | +# 1. Redistributions of source code must retain the above copyright |
| 8 | +# notice, this list of conditions and the following disclaimer. |
| 9 | +# 2. Redistributions in binary form must reproduce the above copyright |
| 10 | +# notice, this list of conditions and the following disclaimer in |
| 11 | +# the documentation and/or other materials provided with the |
| 12 | +# distribution. |
| 13 | +# 3. Neither the name of the OpenBLAS project nor the names of |
| 14 | +# its contributors may be used to endorse or promote products |
| 15 | +# derived from this software without specific prior written permission. |
| 16 | +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 17 | +# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 18 | +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 19 | +# ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE |
| 20 | +# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 21 | +# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 22 | +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| 23 | +# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 24 | +# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE |
| 25 | +# USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 26 | +# ********************************************************************************** |
| 27 | + |
| 28 | +SAMAXKERNEL = amax_rvv.c |
| 29 | +DAMAXKERNEL = amax_rvv.c |
| 30 | +CAMAXKERNEL = zamax_rvv.c |
| 31 | +ZAMAXKERNEL = zamax_rvv.c |
| 32 | + |
| 33 | +SAMINKERNEL = amin_rvv.c |
| 34 | +DAMINKERNEL = amin_rvv.c |
| 35 | +CAMINKERNEL = zamin_rvv.c |
| 36 | +ZAMINKERNEL = zamin_rvv.c |
| 37 | + |
| 38 | +SMAXKERNEL = max_rvv.c |
| 39 | +DMAXKERNEL = max_rvv.c |
| 40 | + |
| 41 | +SMINKERNEL = min_rvv.c |
| 42 | +DMINKERNEL = min_rvv.c |
| 43 | + |
| 44 | +ISAMAXKERNEL = iamax_rvv.c |
| 45 | +IDAMAXKERNEL = iamax_rvv.c |
| 46 | +ICAMAXKERNEL = izamax_rvv.c |
| 47 | +IZAMAXKERNEL = izamax_rvv.c |
| 48 | + |
| 49 | +ISAMINKERNEL = iamin_rvv.c |
| 50 | +IDAMINKERNEL = iamin_rvv.c |
| 51 | +ICAMINKERNEL = izamin_rvv.c |
| 52 | +IZAMINKERNEL = izamin_rvv.c |
| 53 | + |
| 54 | +ISMAXKERNEL = imax_rvv.c |
| 55 | +IDMAXKERNEL = imax_rvv.c |
| 56 | + |
| 57 | +ISMINKERNEL = imin_rvv.c |
| 58 | +IDMINKERNEL = imin_rvv.c |
| 59 | + |
| 60 | +SASUMKERNEL = asum_rvv.c |
| 61 | +DASUMKERNEL = asum_rvv.c |
| 62 | +CASUMKERNEL = zasum_rvv.c |
| 63 | +ZASUMKERNEL = zasum_rvv.c |
| 64 | + |
| 65 | +SSUMKERNEL = sum_rvv.c |
| 66 | +DSUMKERNEL = sum_rvv.c |
| 67 | +CSUMKERNEL = zsum_rvv.c |
| 68 | +ZSUMKERNEL = zsum_rvv.c |
| 69 | + |
| 70 | +SAXPYKERNEL = axpy_rvv.c |
| 71 | +DAXPYKERNEL = axpy_rvv.c |
| 72 | +CAXPYKERNEL = zaxpy_rvv.c |
| 73 | +ZAXPYKERNEL = zaxpy_rvv.c |
| 74 | + |
| 75 | +SAXPBYKERNEL = axpby_rvv.c |
| 76 | +DAXPBYKERNEL = axpby_rvv.c |
| 77 | +CAXPBYKERNEL = zaxpby_rvv.c |
| 78 | +ZAXPBYKERNEL = zaxpby_rvv.c |
| 79 | + |
| 80 | +SCOPYKERNEL = copy_rvv.c |
| 81 | +DCOPYKERNEL = copy_rvv.c |
| 82 | +CCOPYKERNEL = zcopy_rvv.c |
| 83 | +ZCOPYKERNEL = zcopy_rvv.c |
| 84 | + |
| 85 | +SDOTKERNEL = dot_rvv.c |
| 86 | +DDOTKERNEL = dot_rvv.c |
| 87 | +CDOTKERNEL = zdot_rvv.c |
| 88 | +ZDOTKERNEL = zdot_rvv.c |
| 89 | +DSDOTKERNEL = dot_rvv.c |
| 90 | + |
| 91 | +SNRM2KERNEL = nrm2_rvv.c |
| 92 | +DNRM2KERNEL = nrm2_rvv.c |
| 93 | +CNRM2KERNEL = znrm2_rvv.c |
| 94 | +ZNRM2KERNEL = znrm2_rvv.c |
| 95 | + |
| 96 | +SROTKERNEL = rot_rvv.c |
| 97 | +DROTKERNEL = rot_rvv.c |
| 98 | +CROTKERNEL = zrot_rvv.c |
| 99 | +ZROTKERNEL = zrot_rvv.c |
| 100 | + |
| 101 | +SSCALKERNEL = scal_rvv.c |
| 102 | +DSCALKERNEL = scal_rvv.c |
| 103 | +CSCALKERNEL = zscal_rvv.c |
| 104 | +ZSCALKERNEL = zscal_rvv.c |
| 105 | + |
| 106 | +SSWAPKERNEL = swap_rvv.c |
| 107 | +DSWAPKERNEL = swap_rvv.c |
| 108 | +CSWAPKERNEL = zswap_rvv.c |
| 109 | +ZSWAPKERNEL = zswap_rvv.c |
| 110 | + |
| 111 | +SGEMVNKERNEL = gemv_n_rvv.c |
| 112 | +DGEMVNKERNEL = gemv_n_rvv.c |
| 113 | +CGEMVNKERNEL = zgemv_n_rvv.c |
| 114 | +ZGEMVNKERNEL = zgemv_n_rvv.c |
| 115 | + |
| 116 | +SGEMVTKERNEL = gemv_t_rvv.c |
| 117 | +DGEMVTKERNEL = gemv_t_rvv.c |
| 118 | +CGEMVTKERNEL = zgemv_t_rvv.c |
| 119 | +ZGEMVTKERNEL = zgemv_t_rvv.c |
| 120 | + |
| 121 | +CTRMMKERNEL = ztrmmkernel_2x2_rvv.c |
| 122 | +ZTRMMKERNEL = ztrmmkernel_2x2_rvv.c |
| 123 | + |
| 124 | +# SGEMM_UNROLL_N set in params.h |
| 125 | +ifeq ($(SGEMM_UNROLL_N), 2) |
| 126 | +SGEMMKERNEL = gemmkernel_2x2_rvv.c |
| 127 | +SGEMMONCOPY = gemm_ncopy_2_rvv.c |
| 128 | +SGEMMOTCOPY = gemm_tcopy_2_rvv.c |
| 129 | +SGEMMONCOPYOBJ = sgemm_oncopy.o |
| 130 | +SGEMMOTCOPYOBJ = sgemm_otcopy.o |
| 131 | + |
| 132 | +STRMMKERNEL = trmmkernel_2x2_rvv.c |
| 133 | +else ifeq ($(SGEMM_UNROLL_N), 4) |
| 134 | +SGEMMKERNEL = gemmkernel_4x4_rvv.c |
| 135 | +SGEMMONCOPY = gemm_ncopy_4_rvv.c |
| 136 | +SGEMMOTCOPY = ../generic/gemm_tcopy_4.c |
| 137 | +SGEMMONCOPYOBJ = sgemm_oncopy.o |
| 138 | +SGEMMOTCOPYOBJ = sgemm_otcopy.o |
| 139 | + |
| 140 | +STRMMKERNEL = trmmkernel_4x4_rvv.c |
| 141 | +else ifeq ($(SGEMM_UNROLL_N), 8) |
| 142 | +# UNROLL_M is VLMAX |
| 143 | +SGEMMKERNEL = gemmkernel_rvv_v1x8.c |
| 144 | +SGEMMINCOPY = gemm_ncopy_rvv_v1.c |
| 145 | +SGEMMITCOPY = gemm_tcopy_rvv_v1.c |
| 146 | +SGEMMONCOPY = gemm_ncopy_$(SGEMM_UNROLL_N)_rvv.c |
| 147 | +SGEMMOTCOPY = gemm_tcopy_$(SGEMM_UNROLL_N)_rvv.c |
| 148 | +SGEMMINCOPYOBJ = sgemm_incopy$(TSUFFIX).$(SUFFIX) |
| 149 | +SGEMMITCOPYOBJ = sgemm_itcopy$(TSUFFIX).$(SUFFIX) |
| 150 | +SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX) |
| 151 | +SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX) |
| 152 | + |
| 153 | +STRMMKERNEL = trmmkernel_rvv_v1x8.c |
| 154 | + |
| 155 | +STRMMUNCOPY_M = trmm_uncopy_rvv_v1.c |
| 156 | +STRMMLNCOPY_M = trmm_lncopy_rvv_v1.c |
| 157 | +STRMMUTCOPY_M = trmm_utcopy_rvv_v1.c |
| 158 | +STRMMLTCOPY_M = trmm_ltcopy_rvv_v1.c |
| 159 | + |
| 160 | +SSYMMUCOPY_M = symm_ucopy_rvv_v1.c |
| 161 | +SSYMMLCOPY_M = symm_lcopy_rvv_v1.c |
| 162 | +endif |
| 163 | + |
| 164 | +# SGEMM_UNROLL_N set in params.h |
| 165 | +ifeq ($(DGEMM_UNROLL_N), 2) |
| 166 | +DGEMMKERNEL = gemmkernel_2x2_rvv.c |
| 167 | +DGEMMONCOPY = gemm_ncopy_2_rvv.c |
| 168 | +DGEMMOTCOPY = gemm_tcopy_2_rvv.c |
| 169 | +DGEMMONCOPYOBJ = dgemm_oncopy.o |
| 170 | +DGEMMOTCOPYOBJ = dgemm_otcopy.o |
| 171 | + |
| 172 | +DTRMMKERNEL = trmmkernel_2x2_rvv.c |
| 173 | +else ifeq ($(DGEMM_UNROLL_N), 4) |
| 174 | +DGEMMKERNEL = gemmkernel_4x4_rvv.c |
| 175 | +DGEMMONCOPY = gemm_ncopy_4_rvv.c |
| 176 | +DGEMMOTCOPY = ../generic/gemm_tcopy_4.c |
| 177 | +DGEMMONCOPYOBJ = dgemm_oncopy.o |
| 178 | +DGEMMOTCOPYOBJ = dgemm_otcopy.o |
| 179 | + |
| 180 | +DTRMMKERNEL = trmmkernel_4x4_rvv.c |
| 181 | +else ifeq ($(DGEMM_UNROLL_N), 8) |
| 182 | +# UNROLL_M is VLMAX |
| 183 | +DGEMMKERNEL = gemmkernel_rvv_v1x8.c |
| 184 | +DGEMMINCOPY = gemm_ncopy_rvv_v1.c |
| 185 | +DGEMMITCOPY = gemm_tcopy_rvv_v1.c |
| 186 | +DGEMMONCOPY = gemm_ncopy_$(DGEMM_UNROLL_N)_rvv.c |
| 187 | +DGEMMOTCOPY = gemm_tcopy_$(DGEMM_UNROLL_N)_rvv.c |
| 188 | +DGEMMINCOPYOBJ = dgemm_incopy$(TSUFFIX).$(SUFFIX) |
| 189 | +DGEMMITCOPYOBJ = dgemm_itcopy$(TSUFFIX).$(SUFFIX) |
| 190 | +DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX) |
| 191 | +DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX) |
| 192 | + |
| 193 | +DTRMMKERNEL = trmmkernel_rvv_v1x8.c |
| 194 | +DTRMMUNCOPY_M = trmm_uncopy_rvv_v1.c |
| 195 | +DTRMMLNCOPY_M = trmm_lncopy_rvv_v1.c |
| 196 | +DTRMMUTCOPY_M = trmm_utcopy_rvv_v1.c |
| 197 | +DTRMMLTCOPY_M = trmm_ltcopy_rvv_v1.c |
| 198 | + |
| 199 | +DSYMMUCOPY_M = symm_ucopy_rvv_v1.c |
| 200 | +DSYMMLCOPY_M = symm_lcopy_rvv_v1.c |
| 201 | +endif |
| 202 | + |
| 203 | +CGEMMKERNEL = ../generic/zgemmkernel_2x2.c |
| 204 | +CGEMMONCOPY = ../generic/zgemm_ncopy_2.c |
| 205 | +CGEMMOTCOPY = ../generic/zgemm_tcopy_2.c |
| 206 | +CGEMMONCOPYOBJ = cgemm_oncopy.o |
| 207 | +CGEMMOTCOPYOBJ = cgemm_otcopy.o |
| 208 | + |
| 209 | +ZGEMMKERNEL = ../generic/zgemmkernel_2x2.c |
| 210 | +ZGEMMONCOPY = ../generic/zgemm_ncopy_2.c |
| 211 | +ZGEMMOTCOPY = ../generic/zgemm_tcopy_2.c |
| 212 | +ZGEMMONCOPYOBJ = zgemm_oncopy.o |
| 213 | +ZGEMMOTCOPYOBJ = zgemm_otcopy.o |
| 214 | + |
| 215 | +STRSMKERNEL_LN = trsm_kernel_LN_rvv_v1.c |
| 216 | +STRSMKERNEL_LT = trsm_kernel_LT_rvv_v1.c |
| 217 | +STRSMKERNEL_RN = trsm_kernel_RN_rvv_v1.c |
| 218 | +STRSMKERNEL_RT = trsm_kernel_RT_rvv_v1.c |
| 219 | + |
| 220 | +DTRSMKERNEL_LN = trsm_kernel_LN_rvv_v1.c |
| 221 | +DTRSMKERNEL_LT = trsm_kernel_LT_rvv_v1.c |
| 222 | +DTRSMKERNEL_RN = trsm_kernel_RN_rvv_v1.c |
| 223 | +DTRSMKERNEL_RT = trsm_kernel_RT_rvv_v1.c |
| 224 | + |
| 225 | +CTRSMKERNEL_LN = ../generic/trsm_kernel_LN.c |
| 226 | +CTRSMKERNEL_LT = ../generic/trsm_kernel_LT.c |
| 227 | +CTRSMKERNEL_RN = ../generic/trsm_kernel_RN.c |
| 228 | +CTRSMKERNEL_RT = ../generic/trsm_kernel_RT.c |
| 229 | + |
| 230 | +ZTRSMKERNEL_LN = ../generic/trsm_kernel_LN.c |
| 231 | +ZTRSMKERNEL_LT = ../generic/trsm_kernel_LT.c |
| 232 | +ZTRSMKERNEL_RN = ../generic/trsm_kernel_RN.c |
| 233 | +ZTRSMKERNEL_RT = ../generic/trsm_kernel_RT.c |
| 234 | + |
| 235 | +TRSMCOPYLN_M = trsm_lncopy_rvv_v1.c |
| 236 | +TRSMCOPYLT_M = trsm_ltcopy_rvv_v1.c |
| 237 | +TRSMCOPYUN_M = trsm_uncopy_rvv_v1.c |
| 238 | +TRSMCOPYUT_M = trsm_utcopy_rvv_v1.c |
| 239 | + |
| 240 | +SSYMV_U_KERNEL = symv_U_rvv.c |
| 241 | +SSYMV_L_KERNEL = symv_L_rvv.c |
| 242 | +DSYMV_U_KERNEL = symv_U_rvv.c |
| 243 | +DSYMV_L_KERNEL = symv_L_rvv.c |
| 244 | +CSYMV_U_KERNEL = ../generic/zsymv_k.c |
| 245 | +CSYMV_L_KERNEL = ../generic/zsymv_k.c |
| 246 | +ZSYMV_U_KERNEL = ../generic/zsymv_k.c |
| 247 | +ZSYMV_L_KERNEL = ../generic/zsymv_k.c |
| 248 | + |
| 249 | + |
| 250 | +LSAME_KERNEL = ../generic/lsame.c |
| 251 | + |
| 252 | +SCABS_KERNEL = ../generic/cabs.c |
| 253 | +DCABS_KERNEL = ../generic/cabs.c |
| 254 | +QCABS_KERNEL = ../generic/cabs.c |
| 255 | + |
| 256 | +ifndef SGEMM_BETA |
| 257 | +SGEMM_BETA = gemm_beta_rvv.c |
| 258 | +endif |
| 259 | +ifndef DGEMM_BETA |
| 260 | +DGEMM_BETA = gemm_beta_rvv.c |
| 261 | +endif |
| 262 | +ifndef CGEMM_BETA |
| 263 | +CGEMM_BETA = zgemm_beta_rvv.c |
| 264 | +endif |
| 265 | +ifndef ZGEMM_BETA |
| 266 | +ZGEMM_BETA = zgemm_beta_rvv.c |
| 267 | +endif |
0 commit comments