int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLOAT *sb, BLASLONG dummy) {
BLASLONG m, n, lda, ldb;
FLOAT *beta, *a, *b;
BLASLONG ls, is, js;
BLASLONG min_l, min_i, min_j;
BLASLONG jjs, min_jj;
m = args -> m;
n = args -> n;
a = (FLOAT *)args -> a;
b = (FLOAT *)args -> b;
lda = args -> lda;
ldb = args -> ldb;
beta = (FLOAT *)args -> beta;
if (range_m) {
BLASLONG m_from = *(((BLASLONG *)range_m) + 0);
BLASLONG m_to = *(((BLASLONG *)range_m) + 1);
m = m_to - m_from;
b += m_from * COMPSIZE;
}
if (beta) {
#ifndef COMPLEX
if (beta[0] != ONE)
GEMM_BETA(m, n, 0, beta[0], NULL, 0, NULL, 0, b, ldb);
if (beta[0] == ZERO) return 0;
#else
if ((beta[0] != ONE) || (beta[1] != ZERO))
GEMM_BETA(m, n, 0, beta[0], beta[1], NULL, 0, NULL, 0, b, ldb);
if ((beta[0] == ZERO) && (beta[1] == ZERO)) return 0;
#endif
}
#if (defined(UPPER) && !defined(TRANSA)) || (!defined(UPPER) && defined(TRANSA))
for(js = 0; js < n; js += GEMM_R){
min_j = n - js;
if (min_j > GEMM_R) min_j = GEMM_R;
for(ls = 0; ls < js; ls += GEMM_Q){
min_l = js - ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
min_i = m;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
for(jjs = js; jjs < js + min_j; jjs += min_jj){
min_jj = min_j + js - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#ifndef TRANSA
GEMM_ONCOPY(min_l, min_jj, a + (ls + jjs * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
#else
GEMM_OTCOPY(min_l, min_jj, a + (jjs + ls * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
#endif
GEMM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb + min_l * (jjs - js) * COMPSIZE,
b + (jjs * ldb) * COMPSIZE, ldb);
}
for(is = min_i; is < m; is += GEMM_P){
min_i = m - is;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
GEMM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb, b + (is + js * ldb) * COMPSIZE, ldb);
}
}
for(ls = js; ls < js + min_j; ls += GEMM_Q){
min_l = js + min_j - ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
min_i = m;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
#ifndef TRANSA
TRSM_OUNCOPY(min_l, min_l, a + (ls + ls * lda) * COMPSIZE, lda, 0, sb);
#else
TRSM_OLTCOPY(min_l, min_l, a + (ls + ls * lda) * COMPSIZE, lda, 0, sb);
#endif
TRSM_KERNEL(min_i, min_l, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb,
b + (ls * ldb) * COMPSIZE, ldb, 0);
for(jjs = 0; jjs < min_j - min_l - ls + js; jjs += min_jj){
min_jj = min_j - min_l - ls + js - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#ifndef TRANSA
GEMM_ONCOPY (min_l, min_jj, a + (ls + (ls + min_l + jjs) * lda) * COMPSIZE, lda,
sb + min_l * (min_l + jjs) * COMPSIZE);
#else
GEMM_OTCOPY (min_l, min_jj, a + ((ls + min_l + jjs) + ls * lda) * COMPSIZE, lda,
sb + min_l * (min_l + jjs) * COMPSIZE);
#endif
GEMM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb + min_l * (min_l + jjs) * COMPSIZE,
b + (min_l + ls + jjs) * ldb * COMPSIZE, ldb);
}
for(is = min_i; is < m; is += GEMM_P){
min_i = m - is;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
TRSM_KERNEL(min_i, min_l, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb,
b + (is + ls * ldb) * COMPSIZE, ldb, 0);
GEMM_KERNEL(min_i, min_j - min_l + js - ls, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb + min_l * min_l * COMPSIZE,
b + (is + ( min_l + ls) * ldb) * COMPSIZE, ldb);
}
}
}
#else
BLASLONG start_ls;
for(js = n; js > 0; js -= GEMM_R){
min_j = js;
if (min_j > GEMM_R) min_j = GEMM_R;
for (ls = js; ls < n; ls += GEMM_Q) {
min_l = n - ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
min_i = m;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
for(jjs = js; jjs < js + min_j; jjs += min_jj){
min_jj = min_j + js - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#ifndef TRANSA
GEMM_ONCOPY(min_l, min_jj, a + (ls + (jjs - min_j) * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
#else
GEMM_OTCOPY(min_l, min_jj, a + ((jjs - min_j) + ls * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
#endif
GEMM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb + min_l * (jjs - js) * COMPSIZE,
b + (jjs - min_j) * ldb * COMPSIZE, ldb);
}
for(is = min_i; is < m; is += GEMM_P){
min_i = m - is;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
GEMM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb, b + (is + (js - min_j) * ldb) * COMPSIZE, ldb);
}
}
start_ls = js - min_j;
while (start_ls + GEMM_Q < js) start_ls += GEMM_Q;
for(ls = start_ls; ls >= js - min_j; ls -= GEMM_Q){
min_l = js - ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
min_i = m;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
#ifndef TRANSA
TRSM_OLNCOPY(min_l, min_l, a + (ls + ls * lda) * COMPSIZE, lda,
0, sb + min_l * (min_j - js + ls) * COMPSIZE);
#else
TRSM_OUTCOPY(min_l, min_l, a + (ls + ls * lda) * COMPSIZE, lda,
0, sb + min_l * (min_j - js + ls) * COMPSIZE);
#endif
TRSM_KERNEL(min_i, min_l, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb + min_l * (min_j - js + ls) * COMPSIZE,
b + (ls * ldb) * COMPSIZE, ldb, 0);
for(jjs = 0; jjs < min_j - js + ls; jjs += min_jj){
min_jj = min_j - js + ls - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
#ifndef TRANSA
GEMM_ONCOPY (min_l, min_jj, a + (ls + (js - min_j + jjs) * lda) * COMPSIZE, lda,
sb + min_l * jjs * COMPSIZE);
#else
GEMM_OTCOPY (min_l, min_jj, a + ((js - min_j + jjs) + ls * lda) * COMPSIZE, lda,
sb + min_l * jjs * COMPSIZE);
#endif
GEMM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb + min_l * jjs * COMPSIZE,
b + (js - min_j + jjs) * ldb * COMPSIZE, ldb);
}
for(is = min_i; is < m; is += GEMM_P){
min_i = m - is;
if (min_i > GEMM_P) min_i = GEMM_P;
GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
TRSM_KERNEL(min_i, min_l, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb + min_l * (min_j - js + ls) * COMPSIZE,
b + (is + ls * ldb) * COMPSIZE, ldb, 0);
GEMM_KERNEL(min_i, min_j - js + ls, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa,
sb,
b + (is + (js - min_j) * ldb) * COMPSIZE, ldb);
}
}
}
#endif
return 0;
}
int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLOAT *sb, BLASLONG dummy) {
BLASLONG m, n, lda, ldb;
FLOAT *beta, *a, *b;
BLASLONG ls, is, js;
BLASLONG min_l, min_i, min_j;
BLASLONG jjs, min_jj;
m = args -> m;
n = args -> n;
a = (FLOAT *)args -> a;
b = (FLOAT *)args -> b;
lda = args -> lda;
ldb = args -> ldb;
beta = (FLOAT *)args -> beta;
if (range_n) {
BLASLONG n_from = *(((BLASLONG *)range_n) + 0);
BLASLONG n_to = *(((BLASLONG *)range_n) + 1);
n = n_to - n_from;
b += n_from * ldb * COMPSIZE;
}
if (beta) {
#ifndef COMPLEX
if (beta[0] != ONE)
GEMM_BETA(m, n, 0, beta[0], NULL, 0, NULL, 0, b, ldb);
if (beta[0] == ZERO) return 0;
#else
if ((beta[0] != ONE) || (beta[1] != ZERO))
GEMM_BETA(m, n, 0, beta[0], beta[1], NULL, 0, NULL, 0, b, ldb);
if ((beta[0] == ZERO) && (beta[1] == ZERO)) return 0;
#endif
}
for(js = 0; js < n; js += GEMM_R){
min_j = n - js;
if (min_j > GEMM_R) min_j = GEMM_R;
#if (!defined(UPPER) && !defined(TRANSA)) || (defined(UPPER) && defined(TRANSA))
for(ls = 0; ls < m; ls += GEMM_Q){
min_l = m - ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
min_i = min_l;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
TRSM_ILTCOPY(min_l, min_i, a + (ls + ls * lda) * COMPSIZE, lda, 0, sa);
#else
TRSM_IUNCOPY(min_l, min_i, a + (ls + ls * lda) * COMPSIZE, lda, 0, sa);
#endif
for(jjs = js; jjs < js + min_j; jjs += min_jj){
min_jj = min_j + js - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
GEMM_ONCOPY(min_l, min_jj, b + (ls + jjs * ldb) * COMPSIZE, ldb, sb + min_l * (jjs - js) * COMPSIZE);
TRSM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb + min_l * (jjs - js) * COMPSIZE,
b + (ls + jjs * ldb) * COMPSIZE, ldb, 0);
}
for(is = ls + min_i; is < ls + min_l; is += GEMM_P){
min_i = ls + min_l - is;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
TRSM_ILTCOPY(min_l, min_i, a + (is + ls * lda) * COMPSIZE, lda, is - ls, sa);
#else
TRSM_IUNCOPY(min_l, min_i, a + (ls + is * lda) * COMPSIZE, lda, is - ls, sa);
#endif
TRSM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb, b + (is + js * ldb) * COMPSIZE, ldb, is - ls);
}
for(is = ls + min_l; is < m; is += GEMM_P){
min_i = m - is;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
GEMM_ITCOPY(min_l, min_i, a + (is + ls * lda) * COMPSIZE, lda, sa);
#else
GEMM_INCOPY(min_l, min_i, a + (ls + is * lda) * COMPSIZE, lda, sa);
#endif
GEMM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb, b + (is + js * ldb) * COMPSIZE, ldb);
}
}
#else
BLASLONG start_is;
for(ls = m; ls > 0; ls -= GEMM_Q){
min_l = ls;
if (min_l > GEMM_Q) min_l = GEMM_Q;
start_is = ls - min_l;
while (start_is + GEMM_P < ls) start_is += GEMM_P;
min_i = ls - start_is;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
TRSM_IUTCOPY(min_l, min_i, a + (start_is + (ls - min_l) * lda) * COMPSIZE, lda, start_is - (ls - min_l), sa);
#else
TRSM_ILNCOPY(min_l, min_i, a + ((ls - min_l) + start_is * lda) * COMPSIZE, lda, start_is - (ls - min_l), sa);
#endif
for(jjs = js; jjs < js + min_j; jjs += min_jj){
min_jj = min_j + js - jjs;
if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
GEMM_ONCOPY(min_l, min_jj, b + (ls - min_l + jjs * ldb) * COMPSIZE, ldb, sb + min_l * (jjs - js) * COMPSIZE);
TRSM_KERNEL(min_i, min_jj, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb + min_l * (jjs - js) * COMPSIZE,
b + (start_is + jjs * ldb) * COMPSIZE, ldb, start_is - ls + min_l);
}
for(is = start_is - GEMM_P; is >= ls - min_l; is -= GEMM_P){
min_i = ls - is;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
TRSM_IUTCOPY(min_l, min_i, a + (is + (ls - min_l) * lda) * COMPSIZE, lda, is - (ls - min_l), sa);
#else
TRSM_ILNCOPY(min_l, min_i, a + ((ls - min_l) + is * lda) * COMPSIZE, lda, is - (ls - min_l), sa);
#endif
TRSM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb,
b + (is + js * ldb) * COMPSIZE, ldb, + is - (ls - min_l) );
}
for(is = 0; is < ls - min_l; is += GEMM_P){
min_i = ls - min_l - is;
if (min_i > GEMM_P) min_i = GEMM_P;
#ifndef TRANSA
GEMM_ITCOPY(min_l, min_i, a + (is + (ls - min_l) * lda) * COMPSIZE, lda, sa);
#else
GEMM_INCOPY(min_l, min_i, a + ((ls - min_l) + is * lda) * COMPSIZE, lda, sa);
#endif
GEMM_KERNEL(min_i, min_j, min_l, dm1,
#ifdef COMPLEX
ZERO,
#endif
sa, sb, b + (is + js * ldb) * COMPSIZE, ldb);
}
}
#endif
}
return 0;
}
int CNAME(BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG offset, int flag){
BLASLONG i, j;
BLASLONG loop;
FLOAT subbuffer[GEMM_UNROLL_MN * GEMM_UNROLL_MN * COMPSIZE];
if (m + offset < 0) {
#ifndef LOWER
GEMM_KERNEL(m, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
return 0;
}
if (n < offset) {
#ifdef LOWER
GEMM_KERNEL(m, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
return 0;
}
if (offset > 0) {
#ifdef LOWER
GEMM_KERNEL(m, offset, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
b += offset * k * COMPSIZE;
c += offset * ldc * COMPSIZE;
n -= offset;
offset = 0;
if (n <= 0) return 0;
}
if (n > m + offset) {
#ifndef LOWER
GEMM_KERNEL(m, n - m - offset, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a,
b + (m + offset) * k * COMPSIZE,
c + (m + offset) * ldc * COMPSIZE, ldc);
#endif
n = m + offset;
if (n <= 0) return 0;
}
if (offset < 0) {
#ifndef LOWER
GEMM_KERNEL(-offset, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
a -= offset * k * COMPSIZE;
c -= offset * COMPSIZE;
m += offset;
offset = 0;
if (m <= 0) return 0;
}
if (m > n - offset) {
#ifdef LOWER
GEMM_KERNEL(m - n + offset, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + (n - offset) * k * COMPSIZE,
b,
c + (n - offset) * COMPSIZE, ldc);
#endif
m = n + offset;
if (m <= 0) return 0;
}
for (loop = 0; loop < n; loop += GEMM_UNROLL_MN) {
int mm, nn;
mm = (loop & ~(GEMM_UNROLL_MN - 1));
nn = MIN(GEMM_UNROLL_MN, n - loop);
#ifndef LOWER
GEMM_KERNEL(mm, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b + loop * k * COMPSIZE, c + loop * ldc * COMPSIZE, ldc);
#endif
if (flag) {
GEMM_BETA(nn, nn, 0, ZERO,
#ifdef COMPLEX
ZERO,
#endif
NULL, 0, NULL, 0, subbuffer, nn);
GEMM_KERNEL(nn, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + loop * k * COMPSIZE, b + loop * k * COMPSIZE, subbuffer, nn);
#ifndef LOWER
for (j = 0; j < nn; j ++) {
for (i = 0; i <= j; i ++) {
c[(i + loop + (j + loop) * ldc) * 2 + 0] +=
subbuffer[(i + j * nn) * 2 + 0] + subbuffer[(j + i * nn) * 2 + 0];
if (i != j) {
c[(i + loop + (j + loop) * ldc) * 2 + 1] +=
subbuffer[(i + j * nn) * 2 + 1] - subbuffer[(j + i * nn) * 2 + 1];
} else {
c[(i + loop + (j + loop) * ldc) * 2 + 1] = ZERO;
}
}
}
#else
for (j = 0; j < nn; j ++) {
for (i = j; i < nn; i ++) {
c[(i + loop + (j + loop) * ldc) * 2 + 0] +=
subbuffer[(i + j * nn) * 2 + 0] + subbuffer[(j + i * nn) * 2 + 0];
if (i != j) {
c[(i + loop + (j + loop) * ldc) * 2 + 1] +=
subbuffer[(i + j * nn) * 2 + 1] - subbuffer[(j + i * nn) * 2 + 1];
} else {
c[(i + loop + (j + loop) * ldc) * 2 + 1] = ZERO;
}
}
}
#endif
}
#ifdef LOWER
GEMM_KERNEL(m - mm - nn, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + (mm + nn) * k * COMPSIZE, b + loop * k * COMPSIZE,
c + (mm + nn + loop * ldc) * COMPSIZE, ldc);
#endif
}
return 0;
}
int CNAME(BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha_r,
#ifdef COMPLEX
FLOAT alpha_i,
#endif
FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG offset){
BLASLONG i, j;
BLASLONG loop;
FLOAT *cc, *ss;
FLOAT subbuffer[GEMM_UNROLL_MN * (GEMM_UNROLL_MN + 1) * COMPSIZE];
if (m + offset < 0) {
#ifndef LOWER
GEMM_KERNEL(m, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
return 0;
}
if (n < offset) {
#ifdef LOWER
GEMM_KERNEL(m, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
return 0;
}
if (offset > 0) {
#ifdef LOWER
GEMM_KERNEL(m, offset, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
b += offset * k * COMPSIZE;
c += offset * ldc * COMPSIZE;
n -= offset;
offset = 0;
if (n <= 0) return 0;
}
if (n > m + offset) {
#ifndef LOWER
GEMM_KERNEL(m, n - m - offset, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a,
b + (m + offset) * k * COMPSIZE,
c + (m + offset) * ldc * COMPSIZE, ldc);
#endif
n = m + offset;
if (n <= 0) return 0;
}
if (offset < 0) {
#ifndef LOWER
GEMM_KERNEL(-offset, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b, c, ldc);
#endif
a -= offset * k * COMPSIZE;
c -= offset * COMPSIZE;
m += offset;
offset = 0;
if (m <= 0) return 0;
}
if (m > n - offset) {
#ifdef LOWER
GEMM_KERNEL(m - n + offset, n, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + (n - offset) * k * COMPSIZE,
b,
c + (n - offset) * COMPSIZE, ldc);
#endif
m = n + offset;
if (m <= 0) return 0;
}
for (loop = 0; loop < n; loop += GEMM_UNROLL_MN) {
int mm, nn;
mm = (loop/GEMM_UNROLL_MN) * GEMM_UNROLL_MN;
nn = MIN(GEMM_UNROLL_MN, n - loop);
#ifndef LOWER
GEMM_KERNEL(mm, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a, b + loop * k * COMPSIZE, c + loop * ldc * COMPSIZE, ldc);
#endif
GEMM_BETA(nn, nn, 0, ZERO,
#ifdef COMPLEX
ZERO,
#endif
NULL, 0, NULL, 0, subbuffer, nn);
GEMM_KERNEL(nn, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + loop * k * COMPSIZE, b + loop * k * COMPSIZE, subbuffer, nn);
cc = c + (loop + loop * ldc) * COMPSIZE;
ss = subbuffer;
#ifndef LOWER
for (j = 0; j < nn; j ++) {
for (i = 0; i <= j; i ++) {
#ifndef COMPLEX
cc[i] += ss[i];
#else
cc[i * 2 + 0] += ss[i * 2 + 0];
cc[i * 2 + 1] += ss[i * 2 + 1];
#endif
}
ss += nn * COMPSIZE;
cc += ldc * COMPSIZE;
}
#else
for (j = 0; j < nn; j ++) {
for (i = j; i < nn; i ++) {
#ifndef COMPLEX
cc[i] += ss[i];
#else
cc[i * 2 + 0] += ss[i * 2 + 0];
cc[i * 2 + 1] += ss[i * 2 + 1];
#endif
}
ss += nn * COMPSIZE;
cc += ldc * COMPSIZE;
}
#endif
#ifdef LOWER
GEMM_KERNEL(m - mm - nn, nn, k,
alpha_r,
#ifdef COMPLEX
alpha_i,
#endif
a + (mm + nn) * k * COMPSIZE, b + loop * k * COMPSIZE,
c + (mm + nn + loop * ldc) * COMPSIZE, ldc);
#endif
}
return 0;
}
