58 KiB
Sherman-Morrison-Woodbury
- Headers
- Naïve Sherman-Morrison
- Woodbury 2x2
- Woodbury 3x3
- Sherman-Morrison with update splitting
- Woodbury 3x3 and 2x2 with Sherman-Morrison and update splitting
- Helper Functions
- End of files
Low- and high-level functions that use the Sherman-Morrison and Woodbury matrix inversion formulas to update the inverse of a non-singular matrix
Headers
#include "qmckl.h"
#include "assert.h"
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <math.h>
int main() {
qmckl_context context;
context = qmckl_context_create();
qmckl_exit_code rc;
Naïve Sherman-Morrison
qmckl_sherman_morrison_naive
This is the simplest of the available Sherman-Morrison-Woodbury kernels. It applies rank-1 updates one by one in the order that is given. It only checks if the denominator in the Sherman-Morrison formula is not too close to zero when an update is evaluated. It will exit with an error code of the denominator is too close to zero.
The formula for any update $u_j$ (index $j$ is suppresed for clarity) that is applied is \[ (S + uv^T)^{-1} = S^{-1} - \frac{S^{-1} uv^T S^{-1}}{1 + v^T S^{-1} u} \]
where $S$ is the Slater-matrix, $u$ and $v^T$ are the column and row vectors containing the updates, $S^{-1}$ is the inverse of the Slater-matrix.
Even though the Slater-matrix $S$ with all updates applied at once is invertable, during the course of applying updates to the inverse Slater-matrix $S^{-1}$ one-by-one it can happen that one of the intermediate inverse matrices $S^{-1}$ becomes singular. Therefore a global threshold value $\epsilon$ is defined that is used to evaluate each individual update $u_j$ when it is applied.
This value sets the lower bound for which the denominator $1+v_j^TS^{-1}u_j$ is considered to be too small and will most probably result in a singular matrix $S$, or at least in an inverse of $S$ of very poor numerical quality. Therefore, when $1+v_j^TS^{-1}u_j \geq \epsilon$, the update is applied as usual and the kernel exits with return code \texttt{QMCKL_SUCCESS}. If $1+v_j^TS^{-1}u_j \leq \epsilon$ the update is rejected and the kernel exits with return code \texttt{QMCKL_FAILURE}.
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
qmckl_context | context | in | Global state |
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
uint64_t | N_updates | in | Number of rank-1 updates to be applied to Slater_inv |
double | Updates[N_updates*Dim] | in | Array containing the updates |
uint64_t | Updates_index[N_updates] | in | Array containing the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse of a Slater-matrix |
double* | determinant | inout | Determinant of the Slater-matrix |
Requirements
context
is notQMCKL_NULL_CONTEXT
LDS >= 2
Dim >= 2
N_updates >= 1
Updates
is allocated with $N_updates \times Dim$ elementsUpdates_index
is allocated with $N_updates$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elements
C header
qmckl_exit_code qmckl_sherman_morrison_naive(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
#include <stdbool.h>
#include <math.h>
#include "qmckl.h"
#include "config.h"
// Order important because
// __GNUC__ also set in ICC, ICX and CLANG
// __clang__ also set in ICX
#if defined(__INTEL_COMPILER)
#define IVDEP _Pragma("ivdep")
#define ALIGNED _Pragma("vector aligned")
#elif defined(__INTEL_LLVM_COMPILER)
#define IVDEP _Pragma("ivdep")
#define ALIGNED _Pragma("vector aligned")
#elif defined(__clang__)
#define IVDEP _Pragma("clang loop vectorize(enable)")
#define ALIGNED
#elif defined(__GNUC__)
#define IVDEP _Pragma("GCC ivdep")
#define ALIGNED
#endif
qmckl_exit_code qmckl_sherman_morrison_naive_hpc(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* __restrict Updates,
const uint64_t* __restrict Updates_index,
const double breakdown,
double* __restrict Slater_inv,
double* __restrict determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith( context,
QMCKL_NULL_CONTEXT,
"qmckl_sherman_morrison_naive_hpc",
NULL);
}
double __attribute__((aligned(8))) C[Dim];
double __attribute__((aligned(8))) D[LDS];
uint64_t l = 0;
// For each update
while (l < N_updates) {
// C = S^{-1} x u_l
for (uint64_t i = 0; i < Dim; i++) {
C[i] = 0.0f;
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
C[i] += Slater_inv[i * LDS + j] * Updates[l * LDS + j];
}
}
// Denominator: v_l^T * C
const int cui = Updates_index[l] - 1;
double den = 1.0f + C[cui];
if (fabs(den) < breakdown) {
return QMCKL_FAILURE;
}
double iden = 1.0f / den;
// Update det(A)
if (determinant)
*determinant *= den;
// selecting column: v_l^T * S_inv
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
D[j] = Slater_inv[cui * LDS + j];
}
// A^{-1} = A^{-1} - C x D / den
for (uint64_t i = 0; i < Dim; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
const double update = C[i] * D[j] * iden;
Slater_inv[i * LDS + j] -= update;
}
}
l += 1;
}
return QMCKL_SUCCESS;
}
#+NAME:naive_template_code
static inline qmckl_exit_code qmckl_sherman_morrison_naive_{Dim}(
const qmckl_context context,
const uint64_t N_updates,
const double* __restrict Updates,
const uint64_t* __restrict Updates_index,
const double breakdown,
double* __restrict Slater_inv,
double* __restrict determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith( context,
QMCKL_NULL_CONTEXT,
"qmckl_sherman_morrison_naive_{Dim}",
NULL);
}
#define D{Dim}_P ((1+({Dim}-1)/SIMD_LENGTH)*SIMD_LENGTH)
double __attribute__((aligned(8))) C[{Dim}];
double __attribute__((aligned(8))) D[D{Dim}_P];
uint64_t l = 0;
// For each update
while (l < N_updates) {
// C = A^{-1} x U_l
for (uint64_t i = 0; i < {Dim}; i++) {
C[i] = 0;
IVDEP
ALIGNED
for (uint64_t j = 0; j < D{Dim}_P; j++) {
C[i] += Slater_inv[i * D{Dim}_P + j] * Updates[l * D{Dim}_P + j];
}
}
// Denominator
const int cui = Updates_index[l] - 1;
double den = 1.0f + C[cui];
if (fabs(den) < breakdown) {
return QMCKL_FAILURE;
}
double iden = 1.0f / den;
// Update det(A)
if (determinant)
*determinant *= den;
// selecting column: D = v_l^T * S_inv
IVDEP
ALIGNED
for (uint64_t j = 0; j < D{Dim}_P; j++) {
D[j] = Slater_inv[cui * D{Dim}_P + j];
}
// A^{-1} = A^{-1} - C x D / den
for (uint64_t i = 0; i < {Dim}; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < D{Dim}_P; j++) {
double update = C[i] * D[j] * iden;
Slater_inv[i * D{Dim}_P + j] -= update;
}
}
l += 1;
}
return QMCKL_SUCCESS;
}
#+NAME:naive_template_generator
#+NAME:naive_switch_generator
<<naive_template_generator()>>
qmckl_exit_code qmckl_sherman_morrison_naive(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_sherman_morrison_naive",
NULL);
}
if (LDS == (1+(Dim-1)/SIMD_LENGTH)*SIMD_LENGTH) { // Most cases
switch (Dim) {
<<naive_switch_generator()>>
}
}
else { // When SIMD_LENGTH > 1, called with LDS == Dim AND Dim != (1+(Dim-1)/SIMD_LENGTH)*SIMD_LENGTH)
return qmckl_sherman_morrison_naive_hpc(context,
LDS,
Dim,
N_updates,
Updates,
Updates_index,
breakdown,
Slater_inv,
determinant);
}
return QMCKL_FAILURE;
}
Performance
This function performs best when there is only 1 rank-1 update in the update cycle. It is not useful to use Sherman-Morrison with update splitting for these cycles since splitting can never resolve a situation where applying the update causes singular behaviour.
Woodbury 2x2
qmckl_woodbury_2x2
The Woodbury 2x2 kernel. It is used to apply two rank-1 updates at once. The formula used in this algorithm is called the Woodbury Matrix Identity \[ (S + U V)^{-1} = S^{-1} - C B^{-1} D \] where $S$ is the Slater-matrix $U$ and $V$ are the matrices containing the updates and the canonical basis matrix $S^{-1}$ is the inverse of the Slater-matrix $C:= S^{-1}U$, a Dim $\times 2$ matrix $B := 1 + VC$, the $2 \times 2$ matrix that is going to be inverted $D := VS^{-1}$, a $2 \times Dim$ matrix
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
qmckl_context | context | in | Global state |
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
double | Updates[2*Dim] | in | Array containing the updates |
uint64_t | Updates_index[2] | in | Array containing the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse of a Slater-matrix |
double* | determinant | inout | Determinant of Slater-matrix |
Requirements
context
is notqmckl_null_context
LDS >= 2
Dim >= 2
Updates
is allocated with $2 \times Dim$ elementsUpdates_index
is allocated with $2$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elements
C header
qmckl_exit_code qmckl_woodbury_2x2(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
C source
qmckl_exit_code qmckl_woodbury_2x2_hpc(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const double* __restrict Updates,
const uint64_t* __restrict Updates_index,
const double breakdown,
double* __restrict Slater_inv,
double* __restrict determinant) {
/*
C := S^{-1} * U, dim x 2
B := 1 + V * C, 2 x 2
D := V * S^{-1}, 2 x dim
*/
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_woodbury_2x2_hpc",
NULL);
}
const uint64_t row1 = (Updates_index[0] - 1);
const uint64_t row2 = (Updates_index[1] - 1);
// Compute C = (S^T)^{-1}U : Dim x 2
double __attribute__((aligned(8))) C[2 * Dim];
for (uint64_t i = 0; i < Dim; i++) {
C[i * 2] = 0;
C[i * 2 + 1] = 0;
IVDEP
ALIGNED
for (uint64_t k = 0; k < LDS; k++) {
C[i * 2] += Slater_inv[i * LDS + k] * Updates[k];
C[i * 2 + 1] += Slater_inv[i * LDS + k] * Updates[LDS + k];
}
}
// Compute B = 1 + VC : 2 x 2
const double B0 = C[row1 * 2] + 1;
const double B1 = C[row1 * 2 + 1];
const double B2 = C[row2 * 2];
const double B3 = C[row2 * 2 + 1] + 1;
// Check if determinant of inverted matrix is not zero
double det = B0 * B3 - B1 * B2;
if (fabs(det) < breakdown) {
return QMCKL_FAILURE;
}
// Update det(S) when passed
if (determinant)
*determinant *= det;
// Compute B^{-1} with explicit formula for 2 x 2 inversion
double __attribute__((aligned(8))) Binv[4], idet = 1.0 / det;
Binv[0] = idet * B3;
Binv[1] = -1.0 * idet * B1;
Binv[2] = -1.0 * idet * B2;
Binv[3] = idet * B0;
// tmp = B^{-1}D : 2 x LDS
double __attribute__((aligned(8))) tmp[2 * LDS];
double* r1dim = &(Slater_inv[row1 * LDS]);
double* r2dim = &(Slater_inv[row2 * LDS]);
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
tmp[j] = Binv[0] * r1dim[j] + Binv[1] * r2dim[j];
tmp[LDS + j] = Binv[2] * r1dim[j] + Binv[3] * r2dim[j];
}
// Compute (S^T)^{-1} - C * tmp : Dim x LDS
for (uint64_t i = 0; i < Dim; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
Slater_inv[i * LDS + j] -= C[i * 2] * tmp[j];
Slater_inv[i * LDS + j] -= C[i * 2 + 1] * tmp[LDS + j];
}
}
return QMCKL_SUCCESS;
}
#+NAME:woodbury_2x2_template_code
static inline qmckl_exit_code qmckl_woodbury_2x2_{Dim}(
const qmckl_context context,
const double* __restrict Updates,
const uint64_t* __restrict Updates_index,
const double breakdown,
double* __restrict Slater_inv,
double* __restrict determinant) {
/*
C := S^{-1} * U, dim x 2
B := 1 + V * C, 2 x 2
D := V * S^{-1}, 2 x dim
*/
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_woodbury_2x2_{Dim}",
NULL);
}
const uint64_t row1 = (Updates_index[0] - 1);
const uint64_t row2 = (Updates_index[1] - 1);
// Compute C = (S^T)^{-1}U : {Dim} x 2
double __attribute__((aligned(8))) C[2 * {Dim}];
for (uint64_t i = 0; i < {Dim}; i++) {
C[i * 2] = 0;
C[i * 2 + 1] = 0;
IVDEP
ALIGNED
for (uint64_t k = 0; k < D{Dim}_P; k++) {
C[i * 2] += Slater_inv[i * D{Dim}_P + k] * Updates[k];
C[i * 2 + 1] += Slater_inv[i * D{Dim}_P + k] * Updates[D{Dim}_P + k];
}
}
// Compute B = 1 + VC : 2 x 2
const double B0 = C[row1 * 2] + 1;
const double B1 = C[row1 * 2 + 1];
const double B2 = C[row2 * 2];
const double B3 = C[row2 * 2 + 1] + 1;
// Check if determinant of inverted matrix is not zero
double det = B0 * B3 - B1 * B2;
if (fabs(det) < breakdown) {
return QMCKL_FAILURE;
}
// Update det(S) when passed
if (determinant)
*determinant *= det;
// Compute B^{-1} with explicit formula for 2 x 2 inversion
double __attribute__((aligned(8))) Binv[4], idet = 1.0 / det;
Binv[0] = idet * B3;
Binv[1] = -1.0 * idet * B1;
Binv[2] = -1.0 * idet * B2;
Binv[3] = idet * B0;
// tmp = B^{-1}D : 2 x D{Dim}_P
double __attribute__((aligned(8))) tmp[2 * D{Dim}_P];
double* r1dim = &(Slater_inv[row1 * D{Dim}_P]);
double* r2dim = &(Slater_inv[row2 * D{Dim}_P]);
IVDEP
ALIGNED
for (uint64_t j = 0; j < D{Dim}_P; j++) {
tmp[j] = Binv[0] * r1dim[j] + Binv[1] * r2dim[j];
tmp[D{Dim}_P + j] = Binv[2] * r1dim[j] + Binv[3] * r2dim[j];
}
// Compute (S^T)^{-1} - C * tmp : {Dim} x D{Dim}_P
for (uint64_t i = 0; i < {Dim}; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < D{Dim}_P; j++) {
Slater_inv[i * D{Dim}_P + j] -= C[i * 2] * tmp[j];
Slater_inv[i * D{Dim}_P + j] -= C[i * 2 + 1] * tmp[D{Dim}_P + j];
}
}
return QMCKL_SUCCESS;
}
#+NAME:woodbury_2x2_template_generator
#+NAME:woodbury_2x2_switch_generator
<<woodbury_2x2_template_generator()>>
qmckl_exit_code qmckl_woodbury_2x2(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_woodbury_2x2",
NULL);
}
if (LDS == (1+(Dim-1)/SIMD_LENGTH)*SIMD_LENGTH) { // Most cases
switch (Dim) {
<<woodbury_2x2_switch_generator()>>
}
}
else { // When SIMD_LENGTH > 1, called with LDS == Dim AND Dim != (1+(Dim-1)/SIMD_LENGTH)*SIMD_LENGTH)
return qmckl_woodbury_2x2_hpc(context,
LDS,
Dim,
Updates,
Updates_index,
breakdown,
Slater_inv,
determinant);
}
return QMCKL_FAILURE;
}
Performance
This function is most efficient when used in cases where there are only 2 rank-1 updates and it is sure they will not result in a singular matrix.
Woodbury 3x3
qmckl_woodbury_3x3
The 3x3 version of the Woodbury 2x2 kernel. It is used to apply three rank-1 updates at once. The formula used in this kernel is the same as for Woodbury 2x2, except for the sizes of the following matrices:
$C:= S^{-1}U$, a Dim $\times 3$ matrix $B := 1 + VC$, the $3 \times 3$ matrix that is going to be inverted $D := VS^{-1}$, a $3 \times Dim$ matrix
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix. #pragma ivdep #pragma vector aligned
qmckl_context | context | in | Global state |
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
double | Updates[3*Dim] | in | Array containing the updates |
uint64_t | Updates_index[3] | in | Array containing the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse of a Slater-matrix |
double* | determinant | inout | Determinant of Slater-matrix |
Requirements
context
is notqmckl_null_context
LDS >= 2
Dim >= 2
Updates
is allocated with $3 \times Dim$ elementsUpdates_index
is allocated with $3$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elements
C header
qmckl_exit_code qmckl_woodbury_3x3(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
C source
qmckl_exit_code qmckl_woodbury_3x3(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant) {
/*
C := S^{-1} * U, dim x 3
B := 1 + V * C, 3 x 3
D := V * S^{-1}, 3 x dim
,*/
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_woodbury_3x3",
NULL);
}
const uint64_t row1 = (Updates_index[0] - 1);
const uint64_t row2 = (Updates_index[1] - 1);
const uint64_t row3 = (Updates_index[2] - 1);
// Compute C = (S^T)^{-1}U : Dim x 3
double __attribute__((aligned(8))) C[3 * Dim];
for (uint64_t i = 0; i < Dim; i++) {
C[i * 3] = 0;
C[i * 3 + 1] = 0;
C[i * 3 + 2] = 0;
IVDEP
ALIGNED
for (uint64_t k = 0; k < LDS; k++) {
C[i * 3] += Slater_inv[i * LDS + k] * Updates[k];
C[i * 3 + 1] += Slater_inv[i * LDS + k] * Updates[LDS + k];
C[i * 3 + 2] += Slater_inv[i * LDS + k] * Updates[2 * LDS + k];
}
}
// Compute B = 1 + VC : 3 x 3
const double B0 = C[row1 * 3] + 1;
const double B1 = C[row1 * 3 + 1];
const double B2 = C[row1 * 3 + 2];
const double B3 = C[row2 * 3];
const double B4 = C[row2 * 3 + 1] + 1;
const double B5 = C[row2 * 3 + 2];
const double B6 = C[row3 * 3];
const double B7 = C[row3 * 3 + 1];
const double B8 = C[row3 * 3 + 2] + 1;
// Check if determinant of B is not too close to zero
double det;
det = B0 * (B4 * B8 - B5 * B7) - B1 * (B3 * B8 - B5 * B6) +
B2 * (B3 * B7 - B4 * B6);
if (fabs(det) < breakdown) {
return QMCKL_FAILURE;
}
// Update det(Slater) if passed
if (determinant)
*determinant *= det;
// Compute B^{-1} with explicit formula for 3 x 3 inversion
double __attribute__((aligned(8))) Binv[9], idet = 1.0 / det;
Binv[0] = (B4 * B8 - B7 * B5) * idet;
Binv[1] = -(B1 * B8 - B7 * B2) * idet;
Binv[2] = (B1 * B5 - B4 * B2) * idet;
Binv[3] = -(B3 * B8 - B6 * B5) * idet;
Binv[4] = (B0 * B8 - B6 * B2) * idet;
Binv[5] = -(B0 * B5 - B3 * B2) * idet;
Binv[6] = (B3 * B7 - B6 * B4) * idet;
Binv[7] = -(B0 * B7 - B6 * B1) * idet;
Binv[8] = (B0 * B4 - B3 * B1) * idet;
// tmp = B^{-1}D : 3 x LDS
double __attribute__((aligned(8))) tmp[3 * LDS];
double* r1dim = &(Slater_inv[row1 * LDS]);
double* r2dim = &(Slater_inv[row2 * LDS]);
double* r3dim = &(Slater_inv[row3 * LDS]);
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
tmp[j] = Binv[0] * r1dim[j] + Binv[1] * r2dim[j] + Binv[2] * r3dim[j];
tmp[LDS + j] =
Binv[3] * r1dim[j] + Binv[4] * r2dim[j] + Binv[5] * r3dim[j];
tmp[2 * LDS + j] =
Binv[6] * r1dim[j] + Binv[7] * r2dim[j] + Binv[8] * r3dim[j];
}
// Compute (S^T)^{-1} - C * tmp : Dim x LDS
for (uint64_t i = 0; i < Dim; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
Slater_inv[i * LDS + j] -= C[i * 3] * tmp[j];
Slater_inv[i * LDS + j] -= C[i * 3 + 1] * tmp[LDS + j];
Slater_inv[i * LDS + j] -= C[i * 3 + 2] * tmp[2 * LDS + j];
}
}
return QMCKL_SUCCESS;
}
Performance…
This function is most efficient when used in cases where there are only 3 rank-1 updates and it is sure they will not result in a singular matrix.
Sherman-Morrison with update splitting
qmckl_sherman_morrison_splitting
This is a variation on the 'Naive' Sherman-Morrison kernel. Whenever the denominator $1+v_j^T S^{-1} u_j$ in the Sherman-Morrison formula is deemed to be too close to zero, the update $u_j$ is split in half: $u_j \rightarrow \frac{1}{2} u_j$. One half is applied immediately –necessarily increasing the value of the denominator because of the split– while the other halve is put in a queue that will be applied when all the remaining updates have been treated.
The kernel is executed recursively until the queue is eiter empty and all updates are applied successfully, or the size of the queue equals the number of initial updates. In the last case the Slater-matrix that would have resulted from applying the updates is singular and therefore the kernel exits with an exit code.
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
qmckl_context | context | in | Global state |
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
uint64_t | N_updates | in | Number of rank-1 updates to be applied to Slater_inv |
double | Updates[N_updates*Dim] | in | Array containing the updates |
uint64_t | Updates_index[N_updates] | in | Array containing the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse of a Slater-matrix |
double* | determinant | inout | Determinant of the Slater-matrix |
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
Requirements
context
is notQMCKL_NULL_CONTEXT
LDS >= 2
Dim >= 2
N_updates >= 1
Updates
is allocated with $N_updates \times Dim$ elementsUpdates_index
is allocated with $N_updates$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elements
C header
qmckl_exit_code qmckl_sherman_morrison_splitting(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
C source
qmckl_exit_code qmckl_sherman_morrison_splitting(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_sherman_morrison_splitting",
NULL);
}
double __attribute__((aligned(8))) later_updates[LDS * N_updates];
uint64_t later_index[N_updates];
uint64_t later = 0;
qmckl_exit_code rc = qmckl_slagel_splitting(
LDS, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv,
later_updates, later_index, &later, determinant);
if (rc != QMCKL_SUCCESS) return QMCKL_FAILURE;
if (later > 0) {
qmckl_exit_code rc = qmckl_sherman_morrison_splitting(
context, LDS, Dim, later, later_updates, later_index, breakdown,
Slater_inv, determinant);
if (rc != QMCKL_SUCCESS) return QMCKL_FAILURE;
}
return QMCKL_SUCCESS;
}
Performance…
This kernel performs best when there are 2 or more rank-1 update cycles and fail-rate is high.
Woodbury 3x3 and 2x2 with Sherman-Morrison and update splitting
qmckl_sherman_morrison_smw32s
The Woodbury 3x3 and 2x2 kernel with Sherman-Morrison and update splitting combines the low-level Woodbury 3x3 kernel, the Woobury 2x2 kernel and Sherman-Morrison with update splitting. It works the almost the same as Woodbury 3x3 with Sherman-Morrison and update splitting, except that when there is a remainder of two rank-1 updates, it is first tried with Woodbury 2x2 instead of sending them all to Sherman-Morrison with update splitting. For example, in the case of 5 updates the updates are applied in 1 block of 3 updates end 1 block of 2 updates.
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
qmckl_context | context | in | Global state |
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
uint64_t | N_updates | in | Number of rank-1 updates to be applied to Slater_inv |
double | Updates[N_updates*Dim] | in | Array containing the updates |
uint64_t | Updates_index[N_updates] | in | Array containing the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse of a Slater-matrix |
double* | determinant | inout | Determinant of the Slater-matrix |
Requirements
context
is notQMCKL_NULL_CONTEXT
LDS >= 2
Dim >= 2
N_updates >= 1
Updates
is allocated with $N_updates \times Dim$ elementsUpdates_index
is allocated with $N_updates$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elements
C header
qmckl_exit_code qmckl_sherman_morrison_smw32s(
const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
C source
qmckl_exit_code qmckl_sherman_morrison_smw32s(const qmckl_context context,
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant) {
if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) {
return qmckl_failwith(context,
QMCKL_NULL_CONTEXT,
"qmckl_sherman_morrison_smw32s",
NULL);
}
double __attribute__((aligned(8))) later_updates[LDS * N_updates];
uint64_t later_index[N_updates];
uint64_t later = 0;
// Special case for 4 rank-1 updates: 2+2
if (N_updates == 4) {
qmckl_exit_code rc =
qmckl_woodbury_2x2(context, LDS, Dim, Updates, Updates_index,
breakdown, Slater_inv, determinant);
if (rc != QMCKL_SUCCESS) { // Send the entire block to slagel_splitting
uint64_t l = 0;
rc = qmckl_slagel_splitting(LDS, Dim, 2, Updates, Updates_index,
breakdown, Slater_inv,
later_updates + (LDS * later),
later_index + later, &l, determinant);
later += l;
}
rc = qmckl_woodbury_2x2(context, LDS, Dim, &Updates[2 * LDS],
&Updates_index[2], breakdown, Slater_inv,
determinant);
if (rc != QMCKL_SUCCESS) { // Send the entire block to slagel_splitting
uint64_t l = 0;
rc = qmckl_slagel_splitting(
LDS, Dim, 2, &Updates[2 * LDS], &Updates_index[2], breakdown,
Slater_inv, later_updates + (LDS * later), later_index + later,
&l, determinant);
later += l;
}
if (later > 0) {
rc = qmckl_sherman_morrison_splitting(
context, LDS, Dim, later, later_updates, later_index, breakdown,
Slater_inv, determinant);
}
return QMCKL_SUCCESS;
}
// And for the other cases != 4
// Apply first 3*n_of_3blocks updates in n_of_3blocks blocks of 3 updates
// with Woodbury 3x3 kernel
uint64_t n_of_3blocks = N_updates / 3;
uint64_t remainder = N_updates % 3;
uint64_t length_3block = 3 * LDS;
if (n_of_3blocks > 0) {
for (uint64_t i = 0; i < n_of_3blocks; i++) {
const double* Updates_3block = &Updates[i * length_3block];
const uint64_t* Updates_index_3block = &Updates_index[i * 3];
qmckl_exit_code rc = qmckl_woodbury_3x3(
context, LDS, Dim, Updates_3block, Updates_index_3block,
breakdown, Slater_inv, determinant);
if (rc != QMCKL_SUCCESS) { // Send the entire block to slagel_splitting
uint64_t l = 0;
rc = qmckl_slagel_splitting(
LDS, Dim, 3, Updates_3block, Updates_index_3block,
breakdown, Slater_inv, later_updates + (LDS * later),
later_index + later, &l, determinant);
later += l;
}
}
}
// Apply last remaining block of 2 updates with Woodbury 2x2 kernel
if (remainder == 2) {
const double* Updates_2block = &Updates[n_of_3blocks * length_3block];
const uint64_t* Updates_index_2block = &Updates_index[3 * n_of_3blocks];
qmckl_exit_code rc = qmckl_woodbury_2x2(
context, LDS, Dim, Updates_2block, Updates_index_2block,
breakdown, Slater_inv, determinant);
if (rc != QMCKL_SUCCESS) { // Send the entire block to slagel_splitting
uint64_t l = 0;
rc = qmckl_slagel_splitting(
LDS, Dim, 2, Updates_2block, Updates_index_2block, breakdown,
Slater_inv, later_updates + (LDS * later), later_index + later,
&l, determinant);
later += l;
}
}
// Apply last remaining update with slagel_splitting
if (remainder == 1) {
const double* Updates_1block = &Updates[n_of_3blocks * length_3block];
const uint64_t* Updates_index_1block = &Updates_index[3 * n_of_3blocks];
uint64_t l = 0;
qmckl_exit_code rc = qmckl_slagel_splitting(
LDS, Dim, 1, Updates_1block, Updates_index_1block, breakdown,
Slater_inv, later_updates + (LDS * later), later_index + later, &l,
determinant);
if (rc != QMCKL_SUCCESS) return QMCKL_FAILURE;
later += l;
}
if (later > 0) {
qmckl_exit_code rc = qmckl_sherman_morrison_splitting(
context, LDS, Dim, later, later_updates, later_index, breakdown,
Slater_inv, determinant);
if (rc != QMCKL_SUCCESS) return QMCKL_FAILURE;
}
return QMCKL_SUCCESS;
}
Performance…
This kernel performs best for update cycles with 2 or more rank-1 updates and the fail-rate is low.
Helper Functions
Private helper-functions that are used by the Sherman-Morrison-Woodbury kernels. These functions can only be used internally by the kernels in this module.
qmckl_slagel_splitting
qmckl_slagel_splitting
is the non-recursive, inner part of the 'Sherman-Morrison with update splitting'-kernel.
It is used internally to apply a collection of $N$ rank-1 updates to the inverse Slater-matrix $S^{-1}$ and
splitting an update in two equal pieces if necessary. In case of a split, it applies the first half of the update,
while putting the second half in a waiting queue to be applied at the end.
Therefore, when $1+v_j^TS^{-1}u_j \geq \epsilon$, the update is applied as usual. Otherwise, $u_j$ will be redefined as $\frac{1}{2}u_j$. One half is applied immediately, the other half will be applied at the end of the algorithm, using vectors $u_{j'}=\frac{1}{2}u_j$ and $v_{j'}^T=v_{j}^T$, which are stored in the array \texttt{later_updates}.
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting from applying the updates to the original matrix.
uint64_t | LDS | in | Leading dimension of Slater_inv |
uint64_t | Dim | in | Dimension of Slater_inv |
uint64_t | N_updates | in | Number of rank-1 updates to be applied to Slater_inv |
double | Updates[N_updates*Dim] | in | Array containing the rank-1 updates |
uint64_t | Updates_index[N_updates] | in | Array containing positions of the rank-1 updates |
double | breakdown | in | Break-down parameter on which to fail or not |
double | Slater_inv[LDS*Dim] | inout | Array containing the inverse Slater-matrix |
double | later_updates[Dim * N_updates] | inout | Array containing the split updates for later |
uint64_t | later_index[N_updates] | inout | Array containing the positions of the split updates for later |
uint64_t | later | inout | Number of split updates for later |
double* | determinant | inout | Determinant of the Slater-matrix |
Requirements
LDS >= 2
Dim >= 2
N_updates >= 1
Updates
is allocated with $N_updates \times Dim$ elementsUpdates_index
is allocated with $N_updates$ elementsbreakdown
is a small number such that $0 < breakdown << 1$Slater_inv
is allocated with $Dim \times Dim$ elementslater_updates
is allocated with $later \times Dim$ elementslater_index
is allocated with $N_updates$ elementslater >= 0
C header
qmckl_exit_code qmckl_slagel_splitting (
const uint64_t LDS,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* later_updates,
uint64_t* later_index,
uint64_t* later,
double* determinant);
C source
qmckl_exit_code qmckl_slagel_splitting(uint64_t LDS,
uint64_t Dim,
uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* later_updates,
uint64_t* later_index,
uint64_t* later,
double* determinant) {
double __attribute__((aligned(8))) C[LDS];
double __attribute__((aligned(8))) D[LDS];
uint64_t l = 0;
// For each update
while (l < N_updates) {
// C = S^{-1} x U_l
for (uint64_t i = 0; i < Dim; i++) {
C[i] = 0.0f;
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
C[i] += Slater_inv[i * LDS + j] * Updates[l * LDS + j];
}
}
// Denominator
const int cui = Updates_index[l] - 1;
double den = 1.0f + C[cui];
if (fabs(den) < breakdown) {
// U_l = U_l / 2: split the update in 2 equal halves and save the
// second halve in later_updates
IVDEP
ALIGNED
for (uint64_t i = 0; i < LDS; i++) {
later_updates[*later * LDS + i] = Updates[l * LDS + i] * 0.5f;
C[i] *= 0.5f;
}
later_index[*later] = Updates_index[l];
(*later)++;
den = 1.0f + C[cui];
} // From here onwards we continue with applying the first halve of the
// update to Slater_inv
double iden = 1.0f / den;
if (determinant)
*determinant *= den;
// D = v^T x S^{-1} : 1 x LDS
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
D[j] = Slater_inv[cui * LDS + j];
}
// S^{-1} = S^{-1} - C x D / den
for (uint64_t i = 0; i < Dim; i++) {
IVDEP
ALIGNED
for (uint64_t j = 0; j < LDS; j++) {
const double update = C[i] * D[j] * iden;
Slater_inv[i * LDS + j] -= update;
}
}
l += 1;
}
return QMCKL_SUCCESS;
}
Performance
This function cannot be used by itself and is used in Sherman-Morrison with update splitting and Woodbury 3x3 and 2x2 with Sherman-Morrison and update splitting. Please look at the performance reccomendations for those two kernels.
End of files
assert (qmckl_context_destroy(context) == QMCKL_SUCCESS);
return 0;
}