Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

refactor: update implementations for blas/base Level 1 routines #2517

Merged
merged 2 commits into from
Jul 6, 2024
Merged

refactor: update implementations for blas/base Level 1 routines #2517

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4e3169a
refactor: update implementations for BLAS Level 1 routines
aman-095 Jul 6, 2024
cc274dc
refactor: update implementations for Level 1double-precision
aman-095 Jul 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 6 additions & 41 deletions lib/node_modules/@stdlib/blas/base/daxpy/lib/daxpy.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,10 @@

'use strict';

// VARIABLES //
// MODULES //

var M = 4;
var stride2offset = require( '@stdlib/strided/base/stride2offset' );
var ndarray = require('./ndarray.js');


// MAIN //
Expand Down Expand Up @@ -49,48 +50,12 @@ var M = 4;
function daxpy( N, alpha, x, strideX, y, strideY ) {
var ix;
var iy;
var m;
var i;
if ( N <= 0 || alpha === 0.0 ) {
return y;
}
// Use unrolled loops if both strides are equal to `1`...
if ( strideX === 1 && strideY === 1 ) {
m = N % M;

// If we have a remainder, run a clean-up loop...
if ( m > 0 ) {
for ( i = 0; i < m; i++ ) {
y[ i ] += alpha * x[ i ];
}
}
if ( N < M ) {
return y;
}
for ( i = m; i < N; i += M ) {
y[ i ] += alpha * x[ i ];
y[ i+1 ] += alpha * x[ i+1 ];
y[ i+2 ] += alpha * x[ i+2 ];
y[ i+3 ] += alpha * x[ i+3 ];
}
return y;
}
if ( strideX < 0 ) {
ix = (1-N) * strideX;
} else {
ix = 0;
}
if ( strideY < 0 ) {
iy = (1-N) * strideY;
} else {
iy = 0;
}
for ( i = 0; i < N; i++ ) {
y[ iy ] += alpha * x[ ix ];
ix += strideX;
iy += strideY;
}
return y;
ix = stride2offset( N, strideX );
iy = stride2offset( N, strideY );
return ndarray( N, alpha, x, strideX, ix, y, strideY, iy );
}


Expand Down
49 changes: 7 additions & 42 deletions lib/node_modules/@stdlib/blas/base/ddot/lib/ddot.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,10 @@

'use strict';

// VARIABLES //
// MODULES //

var M = 5;
var stride2offset = require( '@stdlib/strided/base/stride2offset' );
var ndarray = require( './ndarray.js' );


// MAIN //
Expand All @@ -45,50 +46,14 @@ var M = 5;
* // returns -5.0
*/
function ddot( N, x, strideX, y, strideY ) {
var dot;
var ix;
var iy;
var m;
var i;

dot = 0.0;
if ( N <= 0 ) {
return dot;
}
// Use unrolled loops if both strides are equal to `1`...
if ( strideX === 1 && strideY === 1 ) {
m = N % M;

// If we have a remainder, run a clean-up loop...
if ( m > 0 ) {
for ( i = 0; i < m; i++ ) {
dot += x[ i ] * y[ i ];
}
}
if ( N < M ) {
return dot;
}
for ( i = m; i < N; i += M ) {
dot += ( x[ i ] * y[ i ] ) + ( x[ i+1 ] * y[ i+1 ] ) + ( x[ i+2 ] * y[ i+2 ] ) + ( x[ i+3 ] * y[ i+3 ] ) + ( x[ i+4 ] * y[ i+4 ] ); // eslint-disable-line max-len
}
return dot;
}
if ( strideX < 0 ) {
ix = ( 1-N ) * strideX;
} else {
ix = 0;
}
if ( strideY < 0 ) {
iy = ( 1-N ) * strideY;
} else {
iy = 0;
}
for ( i = 0; i < N; i++ ) {
dot += ( x[ ix ] * y[ iy ] );
ix += strideX;
iy += strideY;
return 0.0;
}
return dot;
ix = stride2offset( N, strideX );
iy = stride2offset( N, strideY );
return ndarray( N, x, strideX, ix, y, strideY, iy );
}


Expand Down
39 changes: 9 additions & 30 deletions lib/node_modules/@stdlib/blas/base/drot/lib/drot.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,12 @@

'use strict';

// MODULES //

var stride2offset = require( '@stdlib/strided/base/stride2offset' );
var ndarray = require( './ndarray.js' );


// MAIN //

/**
Expand All @@ -43,42 +49,15 @@
* // y => <Float64Array>[ ~4.2, 4.4, 4.6, 4.8, 5.0 ]
*/
function drot( N, x, strideX, y, strideY, c, s ) {
var tmp;
var ix;
var iy;
var i;

if ( N <= 0 ) {
return y;
}
// If both strides are equal to `1`...
if ( strideX === 1 && strideY === 1 ) {
for ( i = 0; i < N; i++ ) {
tmp = ( c * x[ i ] ) + ( s * y[ i ] );
y[ i ] = ( c * y[ i ] ) - ( s * x[ i ] );
x[ i ] = tmp;
}
return y;
}
// If both strides are not equal to `1`...
if ( strideX < 0 ) {
ix = ( 1 - N ) * strideX;
} else {
ix = 0;
}
if ( strideY < 0 ) {
iy = ( 1 - N ) * strideY;
} else {
iy = 0;
}
for ( i = 0; i < N; i++ ) {
tmp = ( c * x[ ix ] ) + ( s * y[ iy ] );
y[ iy ] = ( c * y[ iy ] ) - ( s * x[ ix ] );
x[ ix ] = tmp;
ix += strideX;
iy += strideY;
}
return y;
ix = stride2offset( N, strideX );
iy = stride2offset( N, strideY );
return ndarray( N, x, strideX, ix, y, strideY, iy, c, s );
}


Expand Down
104 changes: 9 additions & 95 deletions lib/node_modules/@stdlib/blas/base/drotm/lib/drotm.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,12 @@

'use strict';

// MODULES //

var stride2offset = require( '@stdlib/strided/base/stride2offset' );
var ndarray = require( './ndarray.js' );


// MAIN //

/**
Expand All @@ -44,108 +50,16 @@
*/
function drotm( N, x, strideX, y, strideY, param ) {
var dflag;
var dh11;
var dh12;
var dh21;
var dh22;
var ix;
var iy;
var i;
var w;
var z;

dflag = param[ 0 ];
if ( N <= 0 || dflag === -2.0 ) {
return y;
}
if ( strideX === strideY && strideX > 0 ) {
ix = 0;
if ( dflag < 0.0 ) {
dh11 = param[ 1 ];
dh12 = param[ 3 ];
dh21 = param[ 2 ];
dh22 = param[ 4 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ ix ];
x[ ix ] = ( w * dh11 ) + ( z * dh12 );
y[ ix ] = ( w * dh21 ) + ( z * dh22 );
ix += strideX;
}
return y;
}
if ( dflag === 0.0 ) {
dh12 = param[ 3 ];
dh21 = param[ 2 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ ix ];
x[ ix ] = w + ( z * dh12 );
y[ ix ] = ( w * dh21 ) + z;
ix += strideX;
}
return y;
}
dh11 = param[ 1 ];
dh22 = param[ 4 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ ix ];
x[ ix ] = ( w * dh11 ) + z;
y[ ix ] = -w + ( z * dh22 );
ix += strideX;
}
return y;
}
if ( strideX < 0 ) {
ix = ( 1 - N ) * strideX;
} else {
ix = 0;
}
if ( strideY < 0 ) {
iy = ( 1 - N ) * strideY;
} else {
iy = 0;
}
if ( dflag < 0.0 ) {
dh11 = param[ 1 ];
dh12 = param[ 3 ];
dh21 = param[ 2 ];
dh22 = param[ 4 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ iy ];
x[ ix ] = ( w * dh11 ) + ( z * dh12 );
y[ iy ] = ( w * dh21 ) + ( z * dh22 );
ix += strideX;
iy += strideY;
}
return y;
}
if ( dflag === 0.0 ) {
dh12 = param[ 3 ];
dh21 = param[ 2 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ iy ];
x[ ix ] = w + ( z * dh12 );
y[ iy ] = ( w * dh21 ) + z;
ix += strideX;
iy += strideY;
}
return y;
}
dh11 = param[ 1 ];
dh22 = param[ 4 ];
for ( i = 0; i < N; i++ ) {
w = x[ ix ];
z = y[ iy ];
x[ ix ] = ( w * dh11 ) + z;
y[ iy ] = -w + ( z * dh22 );
ix += strideX;
iy += strideY;
}
return y;
ix = stride2offset( N, strideX );
iy = stride2offset( N, strideY );
return ndarray( N, x, strideX, ix, y, strideY, iy, param );
}


Expand Down
49 changes: 7 additions & 42 deletions lib/node_modules/@stdlib/blas/base/dsdot/lib/dsdot.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,9 +18,10 @@

'use strict';

// VARIABLES //
// MODULES //

var M = 5;
var stride2offset = require( '@stdlib/strided/base/stride2offset' );
var ndarray = require( './ndarray.js' );


// MAIN //
Expand All @@ -45,50 +46,14 @@ var M = 5;
* // returns -5.0
*/
function dsdot( N, x, strideX, y, strideY ) {
var dot;
var ix;
var iy;
var m;
var i;

dot = 0.0;
if ( N <= 0 ) {
return dot;
}
// Use unrolled loops if both strides are equal to `1`...
if ( strideX === 1 && strideY === 1 ) {
m = N % M;

// If we have a remainder, run a clean-up loop...
if ( m > 0 ) {
for ( i = 0; i < m; i++ ) {
dot += x[ i ] * y[ i ];
}
}
if ( N < M ) {
return dot;
}
for ( i = m; i < N; i += M ) {
dot += ( x[i]*y[i] ) + ( x[i+1]*y[i+1] ) + ( x[i+2]*y[i+2] ) + ( x[i+3]*y[i+3] ) + ( x[i+4]*y[i+4] ); // eslint-disable-line max-len
}
return dot;
}
if ( strideX < 0 ) {
ix = ( 1-N ) * strideX;
} else {
ix = 0;
}
if ( strideY < 0 ) {
iy = ( 1-N ) * strideY;
} else {
iy = 0;
}
for ( i = 0; i < N; i++ ) {
dot += x[ ix ] * y[ iy ];
ix += strideX;
iy += strideY;
return 0.0;
}
return dot;
ix = stride2offset( N, strideX );
iy = stride2offset( N, strideY );
return ndarray( N, x, strideX, ix, y, strideY, iy );
}


Expand Down
Loading
Loading