Lines Matching refs:jj
57 for (CeedInt jj = 0; jj < JJ; jj++) { 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()
61 for (CeedInt jj = 0; jj < JJ; jj++) { // unroll in CeedTensorContract_Avx_Blocked() local
62 rtype tqv = set1(t[(j + jj) * t_stride_0 + b * t_stride_1]); in CeedTensorContract_Avx_Blocked()
64 fmadd(vv[jj][cc], tqv, loadu(&u[(a * B + b) * C + c + cc * 4])); in CeedTensorContract_Avx_Blocked()
68 for (CeedInt jj = 0; jj < JJ; jj++) { in CeedTensorContract_Avx_Blocked() local
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()
80 for (CeedInt jj = 0; jj < J - j; jj++) { 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()
84 for (CeedInt jj = 0; jj < J - j; jj++) { // doesn't unroll in CeedTensorContract_Avx_Blocked() local
85 rtype tqv = set1(t[(j + jj) * t_stride_0 + b * t_stride_1]); in CeedTensorContract_Avx_Blocked()
88 fmadd(vv[jj][cc], tqv, loadu(&u[(a * B + b) * C + c + cc * 4])); in CeedTensorContract_Avx_Blocked()
92 for (CeedInt jj = 0; jj < J - j; jj++) { in CeedTensorContract_Avx_Blocked() local
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()
123 for (CeedInt jj = 0; jj < JJ; jj++) vv[jj] = loadu(&v[(a * J + j + jj) * C + c]); in CeedTensorContract_Avx_Remainder() local
131 for (CeedInt jj = 0; jj < JJ; jj++) { // unroll in CeedTensorContract_Avx_Remainder() local
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() 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() local
172 for (CeedInt jj = 0; jj < JJ / 4; jj++) { // unroll in CeedTensorContract_Avx_Single() local
173 …rtype tqv = set(t[(j + jj * 4 + 3) * t_stride_0 + b * t_stride_1], t[(j + jj * 4 + 2) * t_stride_0… in CeedTensorContract_Avx_Single()
174 …t[(j + jj * 4 + 1) * t_stride_0 + b * t_stride_1], t[(j + jj * 4 + 0) * t_stride_0 + b * t_stride_… 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() 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() local
196 for (CeedInt jj = 0; jj < JJ / 4; jj++) { // unroll in CeedTensorContract_Avx_Single() local
197 …rtype tqv = set(t[(j + jj * 4 + 3) * t_stride_0 + b * t_stride_1], t[(j + jj * 4 + 2) * t_stride_0… in CeedTensorContract_Avx_Single()
198 …t[(j + jj * 4 + 1) * t_stride_0 + b * t_stride_1], t[(j + jj * 4 + 0) * t_stride_0 + b * t_stride_… 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() local