add stopping criterion norm2 call

ginkgo-project · Feb 5, 2021 · 19f9867 · 19f9867
1 parent faa1a62
commit 19f9867
Show file tree

Hide file tree

Showing 5 changed files with 13 additions and 8 deletions.
diff --git a/core/solver/bicg.cpp b/core/solver/bicg.cpp
@@ -186,14 +186,15 @@ void Bicg<ValueType>::apply_impl(const LinOp *b, LinOp *x) const
 
     /* Memory movement summary:
      * Per iteration:
-     * 13.5n * values + matrix/preconditioner storage
+     * 14n * values + matrix/preconditioner storage
      * Two iterations:
-     * 27n * values + matrix/preconditioner storage + conj storage
+     * 28n * values + matrix/preconditioner storage + conj storage
      * 2x SpMV:                4n * values + storage + conj storage
      * 2x Preconditioner:      4n * values + storage + conj storage
      * 2x dot                  4n
      * 1x step 1 (axpys)       6n
      * 1x step 2 (axpys)       9n
+     * 1x norm2 residual        n
      */
     while (true) {
         get_preconditioner()->apply(r.get(), z.get());

diff --git a/core/solver/bicgstab.cpp b/core/solver/bicgstab.cpp
@@ -139,16 +139,17 @@ void Bicgstab<ValueType>::apply_impl(const LinOp *b, LinOp *x) const
 
     /* Memory movement summary:
      * Per iteration:
-     * 14.5n * values + matrix/preconditioner storage
+     * 15.5n * values + matrix/preconditioner storage
      * Two iterations:
-     * 29n * values + 2 * matrix/preconditioner storage
+     * 31n * values + 2 * matrix/preconditioner storage
      * 2x SpMV:                4n * values + 2 * storage
      * 2x Preconditioner:      4n * values + 2 * storage
      * 3x dot                  6n
      * 1x norm2                 n
      * 1x step 1 (fused axpys) 4n
      * 1x step 2 (axpy)        3n
      * 1x step 3 (fused axpys) 7n
+     * 2x norm2 residual       2n
      */
     while (true) {
         ++iter;

diff --git a/core/solver/cg.cpp b/core/solver/cg.cpp
@@ -129,12 +129,13 @@ void Cg<ValueType>::apply_impl(const LinOp *b, LinOp *x) const
 
     int iter = -1;
     /* Memory movement summary:
-     * 17n * values + matrix/preconditioner storage
+     * 18n * values + matrix/preconditioner storage
      * 1x SpMV:           2n * values + storage
      * 1x Preconditioner: 2n * values + storage
      * 2x dot             4n
      * 1x step 1 (axpy)   3n
      * 1x step 2 (axpys)  6n
+     * 1x norm2 residual   n
      */
     while (true) {
         get_preconditioner()->apply(r.get(), z.get());

diff --git a/core/solver/cgs.cpp b/core/solver/cgs.cpp
@@ -140,15 +140,16 @@ void Cgs<ValueType>::apply_impl(const LinOp *b, LinOp *x) const
     int iter = 0;
     /* Memory movement summary:
      * Per iteration:
-     * 13.5n * values + matrix/preconditioner storage
+     * 14n * values + matrix/preconditioner storage
      * Two iterations:
-     * 27n * values + 2 * matrix/preconditioner storage
+     * 28n * values + 2 * matrix/preconditioner storage
      * 2x SpMV:                4n * values + 2 * storage
      * 2x Preconditioner:      4n * values + 2 * storage
      * 2x dot                  4n
      * 1x step 1 (fused axpys) 5n
      * 1x step 2 (fused axpys) 4n
      * 1x step 3 (axpys)       6n
+     * 1x norm2 residual        n
      */
     while (true) {
         r->compute_dot(r_tld.get(), rho.get());

diff --git a/core/solver/fcg.cpp b/core/solver/fcg.cpp
@@ -131,12 +131,13 @@ void Fcg<ValueType>::apply_impl(const LinOp *b, LinOp *x) const
 
     int iter = -1;
     /* Memory movement summary:
-     * 20n * values + matrix/preconditioner storage
+     * 21n * values + matrix/preconditioner storage
      * 1x SpMV:                2n * values + storage
      * 1x Preconditioner:      2n * values + storage
      * 3x dot                  6n
      * 1x step 1 (axpy)        3n
      * 1x step 2 (fused axpys) 7n
+     * 1x norm2 residual        n
      */
     while (true) {
         get_preconditioner()->apply(r.get(), z.get());