/*************************************************************************** Copyright (c) 2013, The OpenBLAS Project All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the OpenBLAS project nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. *****************************************************************************/ #include "common.h" #if !defined(DOUBLE) #define VSETVL(n) RISCV_RVV(vsetvl_e32m4)(n) #define VSETVL_MAX RISCV_RVV(vsetvlmax_e32m1)() #define FLOAT_V_T vfloat32m4_t #define FLOAT_V_T_M1 vfloat32m1_t #define VFMVFS_FLOAT RISCV_RVV(vfmv_f_s_f32m1_f32) #define VLSEV_FLOAT RISCV_RVV(vlse32_v_f32m4) #define VSSEV_FLOAT RISCV_RVV(vsse32_v_f32m4) #ifdef RISCV_0p10_INTRINSICS #define VFREDSUM_FLOAT(va, vb, gvl) RISCV_RVV(vfredusum_vs_f32m4_f32m1)(v_res, va, vb, gvl) #else #define VFREDSUM_FLOAT RISCV_RVV(vfredusum_vs_f32m4_f32m1) #endif #define VFMACCVV_FLOAT RISCV_RVV(vfmacc_vv_f32m4) #define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f32m4) #define VFMVVF_FLOAT RISCV_RVV(vfmv_v_f_f32m4) #define VFMVVF_FLOAT_M1 RISCV_RVV(vfmv_v_f_f32m1) #define VFMULVV_FLOAT RISCV_RVV(vfmul_vv_f32m4) #define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f32m4) #define VFNMSACVV_FLOAT RISCV_RVV(vfnmsac_vv_f32m4) #else #define VSETVL(n) RISCV_RVV(vsetvl_e64m4)(n) #define VSETVL_MAX RISCV_RVV(vsetvlmax_e64m1)() #define FLOAT_V_T vfloat64m4_t #define FLOAT_V_T_M1 vfloat64m1_t #define VFMVFS_FLOAT RISCV_RVV(vfmv_f_s_f64m1_f64) #define VLSEV_FLOAT RISCV_RVV(vlse64_v_f64m4) #define VSSEV_FLOAT RISCV_RVV(vsse64_v_f64m4) #ifdef RISCV_0p10_INTRINSICS #define VFREDSUM_FLOAT(va, vb, gvl) RISCV_RVV(vfredusum_vs_f64m4_f64m1)(v_res, va, vb, gvl) #else #define VFREDSUM_FLOAT RISCV_RVV(vfredusum_vs_f64m4_f64m1) #endif #define VFMACCVV_FLOAT RISCV_RVV(vfmacc_vv_f64m4) #define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f64m4) #define VFMVVF_FLOAT RISCV_RVV(vfmv_v_f_f64m4) #define VFMVVF_FLOAT_M1 RISCV_RVV(vfmv_v_f_f64m1) #define VFMULVV_FLOAT RISCV_RVV(vfmul_vv_f64m4) #define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f64m4) #define VFNMSACVV_FLOAT RISCV_RVV(vfnmsac_vv_f64m4) #endif int CNAME(BLASLONG m, BLASLONG offset, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG incx, FLOAT *y, BLASLONG incy, FLOAT *buffer){ BLASLONG i, j, k; BLASLONG ix, iy, ia; BLASLONG jx, jy, ja; FLOAT temp_r1, temp_i1; FLOAT temp_r2, temp_i2; FLOAT *a_ptr = a; unsigned int gvl = 0; FLOAT_V_T_M1 v_res, v_z0; gvl = VSETVL_MAX; v_res = VFMVVF_FLOAT_M1(0, gvl); v_z0 = VFMVVF_FLOAT_M1(0, gvl); FLOAT_V_T va0, va1, vx0, vx1, vy0, vy1, vr0, vr1; BLASLONG stride_x, stride_y, stride_a, inc_xv, inc_yv, inc_av, len, lda2; BLASLONG inc_x2 = incx * 2; BLASLONG inc_y2 = incy * 2; stride_x = inc_x2 * sizeof(FLOAT); stride_y = inc_y2 * sizeof(FLOAT); stride_a = 2 * sizeof(FLOAT); lda2 = lda * 2; jx = 0; jy = 0; ja = 0; for(j = 0; j < offset; j++){ temp_r1 = alpha_r * x[jx] - alpha_i * x[jx+1];; temp_i1 = alpha_r * x[jx+1] + alpha_i * x[jx]; temp_r2 = 0; temp_i2 = 0; y[jy] += temp_r1 * a_ptr[ja]; y[jy+1] += temp_i1 * a_ptr[ja]; ix = jx + inc_x2; iy = jy + inc_y2; ia = ja + 2; i = j + 1; len = m - i; if(len > 0){ gvl = VSETVL(len); inc_xv = incx * gvl * 2; inc_yv = incy * gvl * 2; inc_av = gvl * 2; vr0 = VFMVVF_FLOAT(0, gvl); vr1 = VFMVVF_FLOAT(0, gvl); for(k = 0; k < len / gvl; k++){ va0 = VLSEV_FLOAT(&a_ptr[ia], stride_a, gvl); va1 = VLSEV_FLOAT(&a_ptr[ia+1], stride_a, gvl); vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl); vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl); #ifndef HEMVREV vy0 = VFMACCVF_FLOAT(vy0, temp_r1, va0, gvl); vy0 = VFNMSACVF_FLOAT(vy0, temp_i1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_r1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_i1, va0, gvl); #else vy0 = VFMACCVF_FLOAT(vy0, temp_r1, va0, gvl); vy0 = VFMACCVF_FLOAT(vy0, temp_i1, va1, gvl); vy1 = VFNMSACVF_FLOAT(vy1, temp_r1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_i1, va0, gvl); #endif VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl); VSSEV_FLOAT(&y[iy+1], stride_y, vy1, gvl); vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl); vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl); #ifndef HEMVREV vr0 = VFMACCVV_FLOAT(vr0, vx0, va0, gvl); vr0 = VFMACCVV_FLOAT(vr0, vx1, va1, gvl); vr1 = VFMACCVV_FLOAT(vr1, vx1, va0, gvl); vr1 = VFNMSACVV_FLOAT(vr1, vx0, va1, gvl); #else vr0 = VFMACCVV_FLOAT(vr0, vx0, va0, gvl); vr0 = VFNMSACVV_FLOAT(vr0, vx1, va1, gvl); vr1 = VFMACCVV_FLOAT(vr1, vx1, va0, gvl); vr1 = VFMACCVV_FLOAT(vr1, vx0, va1, gvl); #endif i += gvl; ix += inc_xv; iy += inc_yv; ia += inc_av; } v_res = VFREDSUM_FLOAT(vr0, v_z0, gvl); temp_r2 = VFMVFS_FLOAT(v_res); v_res = VFREDSUM_FLOAT(vr1, v_z0, gvl); temp_i2 = VFMVFS_FLOAT(v_res); if(i < m){ gvl = VSETVL(m-i); va0 = VLSEV_FLOAT(&a_ptr[ia], stride_a, gvl); va1 = VLSEV_FLOAT(&a_ptr[ia+1], stride_a, gvl); vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl); vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl); #ifndef HEMVREV vy0 = VFMACCVF_FLOAT(vy0, temp_r1, va0, gvl); vy0 = VFNMSACVF_FLOAT(vy0, temp_i1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_r1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_i1, va0, gvl); #else vy0 = VFMACCVF_FLOAT(vy0, temp_r1, va0, gvl); vy0 = VFMACCVF_FLOAT(vy0, temp_i1, va1, gvl); vy1 = VFNMSACVF_FLOAT(vy1, temp_r1, va1, gvl); vy1 = VFMACCVF_FLOAT(vy1, temp_i1, va0, gvl); #endif VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl); VSSEV_FLOAT(&y[iy+1], stride_y, vy1, gvl); vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl); vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl); #ifndef HEMVREV vr0 = VFMULVV_FLOAT(vx0, va0, gvl); vr0 = VFMACCVV_FLOAT(vr0, vx1, va1, gvl); vr1 = VFMULVV_FLOAT(vx1, va0, gvl); vr1 = VFNMSACVV_FLOAT(vr1, vx0, va1, gvl); #else vr0 = VFMULVV_FLOAT(vx0, va0, gvl); vr0 = VFNMSACVV_FLOAT(vr0, vx1, va1, gvl); vr1 = VFMULVV_FLOAT(vx1, va0, gvl); vr1 = VFMACCVV_FLOAT(vr1, vx0, va1, gvl); #endif v_res = VFREDSUM_FLOAT(vr0, v_z0, gvl); temp_r2 += VFMVFS_FLOAT(v_res); v_res = VFREDSUM_FLOAT(vr1, v_z0, gvl); temp_i2 += VFMVFS_FLOAT(v_res); } } y[jy] += alpha_r * temp_r2 - alpha_i * temp_i2; y[jy+1] += alpha_r * temp_i2 + alpha_i * temp_r2; jx += inc_x2; jy += inc_y2; ja += 2; a_ptr += lda2; } return(0); }