:orphan:
# KSPCGUseSingleReduction
Merge the two inner products needed in `KSPCG` into a single `MPI_Allreduce()` call. 
## Synopsis
```
#include "petscksp.h" 
PetscErrorCode KSPCGUseSingleReduction(KSP ksp, PetscBool flg)
```
Logically Collective


## Input Parameters

- ***ksp -*** the iterative context
- ***flg -*** turn on or off the single reduction



## Options Database Key

- ***-ksp_cg_single_reduction <bool> -*** Merge inner products into single `MPI_Allreduce()`





## Notes
The algorithm used in this case is described as Method 1 in [1]. V. Eijkhout credits the algorithm initially to Chronopoulos and Gear.

It requires two extra work vectors than the conventional implementation in PETSc.

See also `KSPPIPECG`, `KSPPIPECR`, and `KSPGROPPCG` that use non-blocking reductions. [](sec_pipelineksp),


## References

- ***[1] -*** Lapack Working Note 56, "Conjugate Gradient Algorithms with Reduced Synchronization Overhead
Distributed Memory Multiprocessors", by E. F. D'Azevedo, V. L. Eijkhout, and C. H. Romine, December 3, 1999.



## See Also
 [](ch_ksp), [](sec_pipelineksp), `KSP`, `KSPCG`, `KSPGMRES`, `KSPPIPECG`, `KSPPIPECR`, and `KSPGROPPCG`

## Level
intermediate

## Location
<A HREF="PETSC_DOC_OUT_ROOT_PLACEHOLDER/src/ksp/ksp/impls/cg/cgtype.c.html#KSPCGUseSingleReduction">src/ksp/ksp/impls/cg/cgtype.c</A>

## Implementations

<A HREF="PETSC_DOC_OUT_ROOT_PLACEHOLDER/src/ksp/ksp/impls/cg/cg.c.html#KSPCGUseSingleReduction_CG">KSPCGUseSingleReduction_CG in src/ksp/ksp/impls/cg/cg.c</A><BR>


---
[Edit on GitLab](https://gitlab.com/petsc/petsc/-/edit/release/src/ksp/ksp/impls/cg/cgtype.c)


[Index of all KSP routines](index.md)  
[Table of Contents for all manual pages](/manualpages/index.md)  
[Index of all manual pages](/manualpages/singleindex.md)  
