56         rtype vv[JJ][CC / 4];  // Output tile to be held in registers  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()
78         rtype vv[JJ][CC / 4];  // Output tile to be held in registers  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()
121         rtype vv[JJ];  // Output tile to be held in registers  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()
132             fmadd(vv[jj], tqu, set1(t[(j + jj) * t_stride_0 + b * t_stride_1]));  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()
166       rtype vv[AA][JJ / 4];  // Output tile to be held in registers  in CeedTensorContract_Avx_Single()  local
169         for (CeedInt jj = 0; jj < JJ / 4; jj++) vv[aa][jj] = loadu(&v[(a + aa) * J + j + jj * 4]);  in CeedTensorContract_Avx_Single()
177             fmadd(vv[aa][jj], tqv, set1(u[(a + aa) * B + b]));  in CeedTensorContract_Avx_Single()
182         for (CeedInt jj = 0; jj < JJ / 4; jj++) storeu(&v[(a + aa) * J + j + jj * 4], vv[aa][jj]);  in CeedTensorContract_Avx_Single()
190     rtype vv[AA][JJ / 4];  // Output tile to be held in registers  in CeedTensorContract_Avx_Single()  local
193       for (CeedInt jj = 0; jj < JJ / 4; jj++) vv[aa][jj] = loadu(&v[(a + aa) * J + j + jj * 4]);  in CeedTensorContract_Avx_Single()
201           fmadd(vv[aa][jj], tqv, set1(u[(a + aa) * B + b]));  in CeedTensorContract_Avx_Single()
206       for (CeedInt jj = 0; jj < JJ / 4; jj++) storeu(&v[(a + aa) * J + j + jj * 4], vv[aa][jj]);  in CeedTensorContract_Avx_Single()
216       rtype vv[AA];  // Output tile to be held in registers  in CeedTensorContract_Avx_Single()  local
218       for (CeedInt aa = 0; aa < AA; aa++) vv[aa] = loadu(&v[(a + aa) * J + j]);  in CeedTensorContract_Avx_Single()
234           fmadd(vv[aa], tqv, set1(u[(a + aa) * B + b]));  in CeedTensorContract_Avx_Single()
237       for (CeedInt aa = 0; aa < AA; aa++) storeu(&v[(a + aa) * J + j], vv[aa]);  in CeedTensorContract_Avx_Single()