21 #define fmadd(c, a, b) (c) = _mm256_fmadd_pd((a), (b), (c))  argument
23 #define fmadd(c, a, b) (c) += _mm256_mul_pd((a), (b))  argument
33 #define fmadd(c, a, b) (c) = _mm_fmadd_ps((a), (b), (c))  argument
35 #define fmadd(c, a, b) (c) += _mm_mul_ps((a), (b))  argument
55       for (CeedInt c = 0; c < (C / CC) * CC; c += CC) {  in CeedTensorContract_Avx_Blocked()  local
58 … for (CeedInt cc = 0; cc < CC / 4; cc++) vv[jj][cc] = loadu(&v[(a * J + j + jj) * C + c + cc * 4]);  in CeedTensorContract_Avx_Blocked()
64               fmadd(vv[jj][cc], tqv, loadu(&u[(a * B + b) * C + c + cc * 4]));  in CeedTensorContract_Avx_Blocked()
69 … for (CeedInt cc = 0; cc < CC / 4; cc++) storeu(&v[(a * J + j + jj) * C + c + cc * 4], vv[jj][cc]);  in CeedTensorContract_Avx_Blocked()
77       for (CeedInt c = 0; c < (C / CC) * CC; c += CC) {  in CeedTensorContract_Avx_Blocked()  local
81 … for (CeedInt cc = 0; cc < CC / 4; cc++) vv[jj][cc] = loadu(&v[(a * J + j + jj) * C + c + cc * 4]);  in CeedTensorContract_Avx_Blocked()
88               fmadd(vv[jj][cc], tqv, loadu(&u[(a * B + b) * C + c + cc * 4]));  in CeedTensorContract_Avx_Blocked()
93 … for (CeedInt cc = 0; cc < CC / 4; cc++) storeu(&v[(a * J + j + jj) * C + c + cc * 4], vv[jj][cc]);  in CeedTensorContract_Avx_Blocked()
118     for (CeedInt c = (C / CC) * CC; c < C; c += 4) {  in CeedTensorContract_Avx_Remainder()  local
123         for (CeedInt jj = 0; jj < JJ; jj++) vv[jj] = loadu(&v[(a * J + j + jj) * C + c]);  in CeedTensorContract_Avx_Remainder()
127           if (C - c == 1) tqu = set(0.0, 0.0, 0.0, u[(a * B + b) * C + c + 0]);  in CeedTensorContract_Avx_Remainder()
128 …  else if (C - c == 2) tqu = set(0.0, 0.0, u[(a * B + b) * C + c + 1], u[(a * B + b) * C + c + 0]);  in CeedTensorContract_Avx_Remainder()
129 …else if (C - c == 3) tqu = set(0.0, u[(a * B + b) * C + c + 2], u[(a * B + b) * C + c + 1], u[(a *…  in CeedTensorContract_Avx_Remainder()
130           else tqu = loadu(&u[(a * B + b) * C + c]);  in CeedTensorContract_Avx_Remainder()
135         for (CeedInt jj = 0; jj < JJ; jj++) storeu(&v[(a * J + j + jj) * C + c], vv[jj]);  in CeedTensorContract_Avx_Remainder()
143 … for (CeedInt c = (C / CC) * CC; c < C; c++) v[(a * J + j) * C + c] += tq * u[(a * B + b) * C + c];  in CeedTensorContract_Avx_Remainder()  local