void ATL_crefgbmvT
(
const int M,
const int N,
const int KL,
const int KU,
const float * ALPHA,
const float * A,
const int LDA,
const float * X,
const int INCX,
const float * BETA,
float * Y,
const int INCY
)
{
/*
* Purpose
* =======
*
* ATL_crefgbmvT( ... ) <=> ATL_crefgbmv( AtlasTrans, ... )
*
* See ATL_crefgbmv for details.
*
* ---------------------------------------------------------------------
*/
/*
* .. Local Variables ..
*/
register float t0_i, t0_r;
int i, i0, i1, iaij, ix, j, jaj, jy, k, kx=0, ky=0;
int incx2 = 2 * INCX, incy2 = 2 * INCY,
lda2 = ( LDA << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, jaj = 0, jy = ky; j < M; j++, jaj += lda2, jy += incy2 )
{
Mset( ATL_sZERO, ATL_sZERO, t0_r, t0_i );
k = KU - j; i0 = ( j - KU > 0 ? j - KU : 0 );
i1 = ( N - 1 > j + KL ? j + KL : N - 1 );
for( i = i0, iaij = ((k+i0) << 1)+jaj, ix = kx;
i <= i1; i++, iaij += 2, ix += incx2 )
{ Mmla( A[iaij], A[iaij+1], X[ix], X[ix+1], t0_r, t0_i ); }
Mcelscal( BETA[0], BETA[1], Y[jy], Y[jy+1] );
Mmla( ALPHA[0], ALPHA[1], t0_r, t0_i, Y[jy], Y[jy+1] );
if( j >= KU ) kx += incx2;
}
/*
* End of ATL_crefgbmvT
*/
}
void ATL_crefsyrkLT
(
const int N,
const int K,
const float * ALPHA,
const float * A,
const int LDA,
const float * BETA,
float * C,
const int LDC
)
{
/*
* Purpose
* =======
*
* ATL_crefsyrkLT( ... )
*
* <=>
*
* ATL_crefsyrk( AtlasLower, AtlasTrans, ... )
*
* See ATL_crefsyrk for details.
*
* ---------------------------------------------------------------------
*/
/*
* .. Local Variables ..
*/
register float t0_i, t0_r;
int i, iali, ialj, icij, j, jai, jaj, jcj, l,
lda2 = ( LDA << 1 ), ldc2 = ( LDC << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, jaj = 0, jcj = 0; j < N; j++, jaj += lda2, jcj += ldc2 )
{
for( i = j, jai = j*lda2, icij = (j << 1)+jcj;
i < N; i++, jai += lda2, icij += 2 )
{
Mset( ATL_sZERO, ATL_sZERO, t0_r, t0_i );
for( l = 0, iali = jai, ialj = jaj; l < K; l++, iali += 2, ialj += 2 )
{ Mmla( A[iali], A[iali+1], A[ialj], A[ialj+1], t0_r, t0_i ); }
Mcelscal( BETA[0], BETA[1], C[icij], C[icij+1] );
Mmla( ALPHA[0], ALPHA[1], t0_r, t0_i, C[icij], C[icij+1] );
}
}
/*
* End of ATL_crefsyrkLT
*/
}
void ATL_crefgpmvLH
(
const int M,
const int N,
const float * ALPHA,
const float * A,
const int LDA,
const float * X,
const int INCX,
const float * BETA,
float * Y,
const int INCY
)
{
/*
* Purpose
* =======
*
* ATL_crefgpmvLH( ... )
*
* <=>
*
* ATL_crefgpmv( AtlasLower, AtlasConjTrans, ... )
*
* See ATL_crefgpmv for details.
*
* ---------------------------------------------------------------------
*/
/*
* .. Local Variables ..
*/
register float t0_i, t0_r;
int i, iaij, incx2 = 2 * INCX, incy2 = 2 * INCY,
ix, j, jaj, jy, lda2 = ( LDA << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, jaj = 0, jy = 0; j < M; j++, jy += incy2 )
{
Mset( ATL_sZERO, ATL_sZERO, t0_r, t0_i );
for( i = 0, iaij = jaj, ix = 0; i < N; i++, iaij += 2, ix += incx2 )
{ Mmla( A[iaij], -A[iaij+1], X[ix], X[ix+1], t0_r, t0_i ); }
Mcelscal( BETA[0], BETA[1], Y[jy], Y[jy+1] );
Mmla( ALPHA[0], ALPHA[1], t0_r, t0_i, Y[jy], Y[jy+1] );
lda2 -= 2; jaj += lda2;
}
/*
* End of ATL_crefgpmvLH
*/
}
void ATL_crefgemmCT
(
const int M,
const int N,
const int K,
const float * ALPHA,
const float * A,
const int LDA,
const float * B,
const int LDB,
const float * BETA,
float * C,
const int LDC
)
{
/*
* .. Local Variables ..
*/
register float t0_i, t0_r;
int i, iali, ibj, ibjl, icij, j, jai, jcj, l,
lda2 = ( LDA << 1 ), ldb2 = ( LDB << 1 ),
ldc2 = ( LDC << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, ibj = 0, jcj = 0; j < N; j++, ibj += 2, jcj += ldc2 )
{
for( i = 0, icij = jcj, jai = 0; i < M; i++, icij += 2, jai += lda2 )
{
Mset( ATL_sZERO, ATL_sZERO, t0_r, t0_i );
for( l = 0, iali = jai, ibjl = ibj;
l < K; l++, iali += 2, ibjl += ldb2 )
{ Mmla( A[iali], -A[iali+1], B[ibjl], B[ibjl+1], t0_r, t0_i ); }
Mcelscal( BETA[0], BETA[1], C[icij], C[icij+1] );
Mmla( ALPHA[0], ALPHA[1], t0_r, t0_i, C[icij], C[icij+1] );
}
}
/*
* End of ATL_crefgemmCT
*/
}
void ATL_crefhemmRL
(
const int M,
const int N,
const float * ALPHA,
const float * A,
const int LDA,
const float * B,
const int LDB,
const float * BETA,
float * C,
const int LDC
)
{
/*
* Purpose
* =======
*
* ATL_crefhemmRL( ... )
*
* <=>
*
* ATL_crefhemm( AtlasRight, AtlasLower, ... )
*
* See ATL_crefhemm for details.
*
* ---------------------------------------------------------------------
*/
/*
* .. Local Variables ..
*/
register float t0_i, t0_r;
int i, iaj, iajk, iakj, ibij, ibik, icij, j,
jaj, jbj, jbk, jcj, k, lda2 = ( LDA << 1 ),
ldb2 = ( LDB << 1 ), ldc2 = ( LDC << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, iaj = 0, jaj = 0, jbj = 0, jcj = 0;
j < N; j++, iaj += 2, jaj += lda2, jbj += ldb2, jcj += ldc2 )
{
i = ( j << 1 ) + jaj;
Mset( ALPHA[0] * A[i], ALPHA[1] * A[i], t0_r, t0_i );
for( i = 0, ibij = jbj, icij = jcj; i < M; i++, ibij += 2, icij += 2 )
{
Mcelscal( BETA[0], BETA[1], C[icij], C[icij+1] );
Mmla( t0_r, t0_i, B[ibij], B[ibij+1], C[icij], C[icij+1] );
}
for( k = 0, iajk = iaj, jbk = 0; k < j; k++, iajk += lda2, jbk += ldb2 )
{
Mmul( ALPHA[0], ALPHA[1], A[iajk], -A[iajk+1], t0_r, t0_i );
for( i = 0, ibik = jbk, icij = jcj; i < M; i++, ibik += 2, icij += 2 )
{ Mmla( t0_r, t0_i, B[ibik], B[ibik+1], C[icij], C[icij+1] ); }
}
for( k = j+1, iakj = ((j+1) << 1)+jaj, jbk = (j+1)*ldb2;
k < N; k++, iakj += 2, jbk += ldb2 )
{
Mmul( ALPHA[0], ALPHA[1], A[iakj], A[iakj+1], t0_r, t0_i );
for( i = 0, ibik = jbk, icij = jcj; i < M; i++, ibik += 2, icij += 2 )
{ Mmla( t0_r, t0_i, B[ibik], B[ibik+1], C[icij], C[icij+1] ); }
}
}
/*
* End of ATL_crefhemmRL
*/
}
void ATL_crefsymmLL
(
const int M,
const int N,
const float * ALPHA,
const float * A,
const int LDA,
const float * B,
const int LDB,
const float * BETA,
float * C,
const int LDC
)
{
/*
* Purpose
* =======
*
* ATL_crefsymmLL( ... )
*
* <=>
*
* ATL_crefsymm( AtlasLeft, AtlasLower, ... )
*
* See ATL_crefsymm for details.
*
* ---------------------------------------------------------------------
*/
/*
* .. Local Variables ..
*/
register float t0_i, t0_r, t1_i, t1_r;
int i, iaki, ibij, ibkj, icij, ickj, j, jai,
jbj, jcj, k, lda2 = ( LDA << 1 ),
ldb2 = ( LDB << 1 ), ldc2 = ( LDC << 1 );
/* ..
* .. Executable Statements ..
*
*/
for( j = 0, jbj = 0, jcj = 0; j < N; j++, jbj += ldb2, jcj += ldc2 )
{
for( i = M-1, jai = (M-1)*lda2, ibij = ( (M-1) << 1 )+jbj,
icij = ( (M-1) << 1 )+jcj; i >= 0;
i--, jai -= lda2, ibij -= 2, icij -= 2 )
{
Mmul( ALPHA[0], ALPHA[1], B[ibij], B[ibij+1], t0_r, t0_i );
Mset( ATL_sZERO, ATL_sZERO, t1_r, t1_i );
for( k = i+1, iaki = ( (i+1) << 1 )+jai, ibkj = ( (i+1) << 1 )+jbj,
ickj = ( (i+1) << 1 )+jcj; k < M;
k++, iaki += 2, ibkj += 2, ickj += 2 )
{
Mmla( A[iaki], A[iaki+1], t0_r, t0_i, C[ickj], C[ickj+1] );
Mmla( B[ibkj], B[ibkj+1], A[iaki], A[iaki+1], t1_r, t1_i );
}
Mcelscal( BETA[0], BETA[1], C[icij], C[icij+1] );
k = ( i << 1 ) + jai;
Mmla( t0_r, t0_i, A[k], A[k+1], C[icij], C[icij+1] );
Mmla( ALPHA[0], ALPHA[1], t1_r, t1_i, C[icij], C[icij+1] );
}
}
/*
* End of ATL_crefsymmLL
*/
}