I am converting a 3-D Jacobi solver from pure MPI to Hybrid MPI+OpenMP. I have a 192x192x192 array which is divided among 24 processes in Pure MPI in 1-D decomposition i.e. each process has 192/24 x 192 x 192 = 8 x 192 x 192 slab of data. Now I do :
for(i=0 ; i <= 7; i++)
for(j=0; j<= 191; j++)
for(k=0; k<= 191; k++)
{
unew[i][j][k] = 1/6.0 * (u[i+1][j][k]+u[i-1][j][k]+
u[i][j+1][k]+u[i][j-1][k]+
u[i][j][k+1]+u[i][j][k-1]);
}
This update takes around 60 seconds for each process.
Now with Hybrid MPI, I run two processes (1 process per socket --bind-to socket --map-by socket and OMP_PROC_PLACES=coreswith OMP_PROC_BIND=close). I create 12 threads per MPI Process (i.e. 12 threads per socket or processor). Now each MPI process has an array of size : 192/2 x 192 x 192 = 96x192x192 elements. Each thread works on 96/12 x 192 x 192 = 8 x 192 x 192 portion of the array owned by each process. I do the same triple loop update using threads but the time is approximately 76 seconds for each thread. The load balance is perfect in both the problems. What could be the possible causes of performance degradation ? Is is False Sharing because threads could be invalidating the cache lines close to each other's chunk of data ? If yes, then how do I reduce this performance degradation ? (I have purposefully not mentioned ghost data but initially I am NOT overlapping communication with computation.)
In response to the comments below, am posting the code. Apologies for the long MWE but you can very safely ignore (1) Header files declaration (2) Variable Declaration (3) Memory allocation routine (4) Formation of Cartesian Topology (5) Setting boundary conditions in parallel using OpenMP parallel region (6) Declaration of MPI_Type_subarray datatype (7) MPI_Isend() and MPI_Irecv() calls and just concentrate on (a) INDEPENDENT UPDATE OpenMP parallel region (b) independent_update(...) routine being called from here.
/* IGNORE THIS PORTION */
#include<mpi.h>
#include<omp.h>
#include<stdio.h>
#include<stdlib.h>
#include<math.h>
#define MIN(a,b) (a < b ? a : b)
#define Tol 0.00001
/* IGNORE THIS ROUTINE */
void input(int *X, int *Y, int *Z)
{
int a=193, b=193, c=193;
*X = a;
*Y = b;
*Z = c;
}
/* IGNORE THIS ROUTINE */
float*** allocate_mem(int X, int Y, int Z)
{
int i,j;
float ***matrix;
float *arr;
arr = (float*)calloc(X*Y*Z, sizeof(float));
matrix = (float***)calloc(X, sizeof(float**));
for(i = 0 ; i<= X-1; i++)
matrix[i] = (float**)calloc(Y, sizeof(float*));
for(i = 0 ; i <= X-1; i++)
for(j=0; j<= Y-1; j++)
matrix[i][j] = &(arr[i*Y*Z + j*Z]);
return matrix ;
}
/* THIS ROUTINE IS IMPORTANT */
float independent_update(float ***old, float ***new, int NX, int NY, int NZ, int tID, int chunk)
{
int i,j,k, start, end;
float error = 0.0;
float diff;
start = tID * chunk + 1;
end = MIN( (tID+1)*chunk, NX-2 );
for(i = start; i <= end ; i++)
{
for(j = 1; j<= NY-2; j++)
{
#pragma omp simd
for(k = 1; k<= NZ-2; k++)
{
new[i][j][k] = (1/6.0) *(old[i-1][j][k] + old[i+1][j][k] + old[i][j-1][k] + old[i][j+1][k] + old[i][j][k-1] + old[i][j][k+1] );
diff = 1.0 - new[i][j][k];
diff = (diff > 0 ? diff : -1.0 * diff );
if(diff > error)
error = diff;
}
}
}
return error;
}
int main(int argc, char *argv[])
{
/* IGNORE VARIABLE DECLARATION */
int size, rank; //Size of old_comm and rank of process
int i, j, k,l; //General loop variables
MPI_Comm old_comm, new_comm; //MPI_COMM_WORLD handle and for MPI_Cart_create()
int N[3]; //For taking input of size of matrix from user
int P; //Represent number of processes i.e. same as size
int dims[3]; //For dimensions of Cartesian topology
int PX, PY, PZ; //X dim, Y dim, Z dim of each process
float ***old, ***new, ***temp; //Matrices for results dimensions is (Px+2)*(PY+2)*(PZ+2)
int period[3]; //Periodicity for each dimension
int reorder; //Whether processes should be reordered in new cartesian topology
int ndims; //Number of dimensions (which is 3)
int Z_TOWARDS_U, Z_AWAY_U; //Z neighbour towards you and away from you (Z const)
int X_DOWN, X_UP; //Below plane and above plane (X const)
int Y_LEFT, Y_RIGHT; //Left plane and right plane (Y const)
int coords[3]; //Finding coordinates of processes
int dimension; //Used in MPI_Cart_shift() , values = 0, 1,2
int displacement; //Used in MPI_Cart_shift(), values will be +1 to find immediate neighbours
float l_max_err; //Local maximum error on process
float l_max_err_new; //For dependent faces.
float G_max_err = 1.0; //Maximum error for stopping criterion
int iterations = 0 ; //Counting number of iterations
MPI_Request send[6], recv[6]; //For MPI_Isend and MPI_Irecv
int start[3]; //Start will be defined in MPI_Isend() and MPI_Irecv()
int gsize[3]; //Defining global size of subarray
MPI_Datatype x_subarray; //For sending X_UP and X_DOWN
int local_x[3]; //Defining local plane size for X_UP/X_DOWN
MPI_Datatype y_subarray; //For sending Y_LEFT and Y_RIGHT
int local_y[3]; //Defining local plane for Y_LEFT/Y_RIGHT
MPI_Datatype z_subarray; //For sending Z_TOWARDS_U and Z_AWAY_U
int local_z[3]; //Defining local plan size for XY plane i.e. where Z=0
double strt, end; //For measuring time
double strt1, end1, delta1; //For measuring trivial time 1
double strt2, end2, delta2; //For measuring trivial time 2
double t_i_strt, t_i_end, t_i_sum=0; //Time for independent computational kernel
double t_up_strt, t_up_end, t_up_sum=0; //Time for X_UP
double t_down_strt, t_down_end, t_down_sum=0; //Time for X_DOWN
double t_left_strt, t_left_end, t_left_sum=0; //Time for Y_LEFT
double t_right_strt, t_right_end, t_right_sum=0; //Time for Y_RIGHT
double t_towards_strt, t_towards_end, t_towards_sum=0; //For Z_TOWARDS_U
double t_away_strt, t_away_end, t_away_sum=0; //For Z_AWAY_U
double t_comm_strt, t_comm_end, t_comm_sum=0; //Time comm + independent update (need to subtract to get comm time)
double t_setup_strt,t_setup_end; //Set-up start and end time
double t_allred_strt,t_allred_end,t_allred_total=0.0; //Measuring Allreduce time separately.
int threadID; //ID of a thread
int nthreads; //Total threads in OpenMP region
int chunk; //chunk - used to calculate iterations of a thread
/* IGNORE MPI STARTUP ETC */
MPI_Init(&argc, &argv);
t_setup_strt = MPI_Wtime();
old_comm = MPI_COMM_WORLD;
MPI_Comm_size(old_comm, &size);
MPI_Comm_rank(old_comm, &rank);
P = size;
if(rank == 0)
{
input(&N[0], &N[1], &N[2]);
}
MPI_Bcast(N, 3, MPI_INT, 0, old_comm);
dims[0] = 0;
dims[1] = 0;
dims[2] = 0;
period[0] = period[1] = period[2] = 0; //All dimensions aperiodic
reorder = 0 ; //No reordering of ranks in new_comm
ndims = 3;
MPI_Dims_create(P,ndims,dims);
MPI_Cart_create(old_comm, ndims, dims, period, reorder, &new_comm);
if( (N[0]-1) % dims[0] == 0 && (N[1]-1) % dims[1] == 0 && (N[2]-1) % dims[2] == 0 )
{
PX = (N[0]-1)/dims[0]; //Rows of unknowns each process gets
PY = (N[1]-1)/dims[1]; //Columns of unknowns each process gets
PZ = (N[2]-1)/dims[2]; //Depth of unknowns each process gets
}
old = allocate_mem(PX+2, PY+2, PZ+2); //3D arrays with ghost points
new = allocate_mem(PX+2, PY+2, PZ+2); //3D arrays with ghost points
dimension = 0;
displacement = 1;
MPI_Cart_shift(new_comm, dimension, displacement, &X_UP, &X_DOWN); //Find UP and DOWN neighbours
dimension = 1;
MPI_Cart_shift(new_comm, dimension, displacement, &Y_LEFT, &Y_RIGHT); //Find UP and DOWN neighbours
dimension = 2;
MPI_Cart_shift(new_comm, dimension, displacement, &Z_TOWARDS_U, &Z_AWAY_U); //Find UP and DOWN neighbours
/* IGNORE BOUNDARY SETUPS FOR PDE */
#pragma omp parallel for default(none) shared(old,new,PX,PY,PZ) private(i,j,k) schedule(static)
for(i = 0; i <= PX+1; i++)
{
for(j = 0; j <= PY+1; j++)
{
for(k = 0; k <= PZ+1; k++)
{
old[i][j][k] = 0.0;
new[i][j][k] = 0.0;
}
}
}
#pragma omp parallel default(none) shared(X_DOWN,X_UP,Y_LEFT,Y_RIGHT,Z_TOWARDS_U,Z_AWAY_U,old,new,PX,PY,PZ) private(i,j,k,threadID,nthreads)
{
threadID = omp_get_thread_num();
nthreads = omp_get_num_threads();
if(threadID == 0)
{
if(X_DOWN == MPI_PROC_NULL) //X is constant here, this is YZ upper plane
{
for(j = 1 ; j<= PY ; j++)
for(k = 1 ; k<= PZ ; k++)
{
old[0][j][k] = 1;
new[0][j][k] = 1; //Set boundaries in new also
}
}
}
if(threadID == (nthreads-1))
{
if(X_UP == MPI_PROC_NULL) //YZ lower plane
{
for(j = 1 ; j<= PY ; j++)
for(k = 1; k<= PZ ; k++)
{
old[PX+1][j][k] = 1;
new[PX+1][j][k] = 1;
}
}
}
if(Y_LEFT == MPI_PROC_NULL) //Y is constant, this is left XZ plane, possibly can use collapse(2)
{
#pragma omp for schedule(static)
for(i = 1 ; i<= PX ; i++)
for(k = 1; k<= PZ; k++)
{
old[i][0][k] = 1;
new[i][0][k] = 1;
}
}
if(Y_RIGHT == MPI_PROC_NULL) //XZ right plane, again collapse(2) potential
{
#pragma omp for schedule(static)
for(i = 1 ; i<= PX; i++)
for(k = 1; k<= PZ ; k++)
{
old[i][PY+1][k] = 1;
new[i][PY+1][k] = 1;
}
}
if(Z_TOWARDS_U == MPI_PROC_NULL) //Z is constant here, towards you XY plane, collapse(2)
{
#pragma omp for schedule(static)
for(i = 1 ; i<= PX ; i++)
for(j = 1; j<= PY ; j++)
{
old[i][j][0] = 1;
new[i][j][0] = 1;
}
}
if(Z_AWAY_U == MPI_PROC_NULL) //Away from you XY plane, collapse(2)
{
#pragma omp for schedule(static)
for(i = 1 ; i<= PX; i++)
for(j = 1; j<= PY ; j++)
{
old[i][j][PZ+1] = 1;
new[i][j][PZ+1] = 1;
}
}
}
/* IGNORE SUBARRAY DECLARATION */
gsize[0] = PX+2; //Global sizes of 3-D cubes for each process
gsize[1] = PY+2;
gsize[2] = PZ+2;
start[0] = 0; //Will specify starting location while sending/receiving
start[1] = 0;
start[2] = 0;
local_x[0] = 1;
local_x[1] = PY;
local_x[2] = PZ;
MPI_Type_create_subarray(ndims, gsize, local_x, start, MPI_ORDER_C, MPI_FLOAT, &x_subarray);
MPI_Type_commit(&x_subarray);
local_y[0] = PX;
local_y[1] = 1;
local_y[2] = PZ;
MPI_Type_create_subarray(ndims, gsize, local_y, start, MPI_ORDER_C, MPI_FLOAT, &y_subarray);
MPI_Type_commit(&y_subarray);
local_z[0] = PX;
local_z[1] = PY;
local_z[2] = 1;
MPI_Type_create_subarray(ndims, gsize, local_z, start, MPI_ORDER_C, MPI_FLOAT, &z_subarray);
MPI_Type_commit(&z_subarray);
t_setup_end = MPI_Wtime();
strt = MPI_Wtime();
while(G_max_err > Tol) //iterations < ITERATIONS)
{
iterations++ ;
t_comm_strt = MPI_Wtime();
/* IGNORE MPI COMMUNICATION */
MPI_Irecv(&old[0][1][1], 1, x_subarray, X_DOWN, 10, new_comm, &recv[0]);
MPI_Irecv(&old[PX+1][1][1], 1, x_subarray, X_UP, 20, new_comm, &recv[1]);
MPI_Irecv(&old[1][PY+1][1], 1, y_subarray, Y_RIGHT, 30, new_comm, &recv[2]);
MPI_Irecv(&old[1][0][1], 1, y_subarray, Y_LEFT, 40, new_comm, &recv[3]);
MPI_Irecv(&old[1][1][PZ+1], 1, z_subarray, Z_AWAY_U, 50, new_comm, &recv[4]);
MPI_Irecv(&old[1][1][0], 1, z_subarray, Z_TOWARDS_U, 60, new_comm, &recv[5]);
MPI_Isend(&old[PX][1][1], 1, x_subarray, X_UP, 10, new_comm, &send[0]);
MPI_Isend(&old[1][1][1], 1, x_subarray, X_DOWN, 20, new_comm, &send[1]);
MPI_Isend(&old[1][1][1], 1, y_subarray, Y_LEFT, 30, new_comm, &send[2]);
MPI_Isend(&old[1][PY][1], 1, y_subarray, Y_RIGHT, 40, new_comm, &send[3]);
MPI_Isend(&old[1][1][1], 1, z_subarray, Z_TOWARDS_U, 50, new_comm, &send[4]);
MPI_Isend(&old[1][1][PZ], 1, z_subarray, Z_AWAY_U, 60, new_comm, &send[5]);
MPI_Waitall(6, send, MPI_STATUSES_IGNORE);
MPI_Waitall(6, recv, MPI_STATUSES_IGNORE);
t_comm_end = MPI_Wtime();
t_comm_sum = t_comm_sum + (t_comm_end - t_comm_strt);
/* Use threads in Independent update */
t_i_strt = MPI_Wtime();
l_max_err = 0.0; //Very important, Reduction result is combined with this !
/* THIS IS THE IMPORTANT REGION */
#pragma omp parallel default(none) shared(old,new,PX,PY,PZ,chunk) private(threadID,nthreads) reduction(max:l_max_err)
{
nthreads = omp_get_num_threads();
threadID = omp_get_thread_num();
chunk = (PX-1+1) / nthreads ;
l_max_err = independent_update(old, new, PX+2, PY+2, PZ+2, threadID, chunk);
}
t_i_end = MPI_Wtime();
t_i_sum = t_i_sum + (t_i_end - t_i_strt) ;
/* IGNORE THE REMAINING CODE */
t_allred_strt = MPI_Wtime();
MPI_Allreduce(&l_max_err, &G_max_err, 1, MPI_FLOAT, MPI_MAX, new_comm);
t_allred_end = MPI_Wtime();
t_allred_total = t_allred_total + (t_allred_end - t_allred_strt);
temp = new ;
new = old;
old = temp;
}
MPI_Barrier(new_comm);
end = MPI_Wtime();
if( rank == 0)
{
printf("\nIterations = %d, G_max_err = %f", iterations, G_max_err);
printf("\nThe total SET-UP time for MPI and boundary conditions is %lf", (t_setup_end-t_setup_strt));
printf("\nThe total time for SOLVING is %lf", (end-strt));
printf("\nThe total time for INDEPENDENT COMPUTE %lf", t_i_sum);
printf("\nThe total time for COMMUNICATION OVERHEAD is %lf", t_comm_sum);
printf("\nThe total time for MPI_ALLREDUCE() is %lf", t_allred_total);
}
MPI_Type_free(&x_subarray);
MPI_Type_free(&y_subarray);
MPI_Type_free(&z_subarray);
free(&old[0][0][0]);
free(&new[0][0][0]);
MPI_Finalize();
return 0;
}
P.S. : I am almost sure that the cost of spawning/waking the threads is not the reason for such a huge difference in the timing.
Please find attached Scalasca snapshot for INDEPENDENT COMPUTE of the Hybrid Program.
Using loop simd construct
#pragma omp parallel default(none) shared(old,new,PX,PY,PZ,l_max_err) private(i,j,k,diff)
{
#pragma omp for simd schedule(static) reduction(max:l_max_err)
for(i = 1; i <= PX ; i++)
{
for(j = 1; j<= PY; j++)
{
for(k = 1; k<= PZ; k++)
{
new[i][j][k] = (1/6.0) *(old[i-1][j][k] + old[i+1][j][k] + old[i][j-1][k] + old[i][j+1][k] + old[i][j][k-1] + old[i][j][k+1] );
diff = 1.0 - new[i][j][k];
diff = (diff > 0 ? diff : -1.0 * diff );
if(diff > l_max_err)
l_max_err = diff;
}
}
}
}
You frequently get memory access and cache issues when you just do one MPI process per socket on a CPU with multiple memory controllers. It can be on either the read or the write side, so you can't really say which. This is especially an issue when doing thread-parallel execution with lightweight compute tasks (e.g. math on arrays). One MPI process per socket in this case tends to fare significantly worse than pure MPI.
In your BIOS, set up whatever the maximal NUMA per socket option is
Use one MPI process per NUMA node.
Try some different parameter values in schedule(static). I've rarely found the default to be best.
Essentially what this will do is ensure each bundle of threads only works on a single pool of memory.
The Data:
A list of integers increasing in order (0,1,2,3,4,5.......)
A list of values that belong to those integers. As an example, 0 = 33, 1 = 45, 2 = 21, ....etc.
And an incrementing variable x which represent a minimum jump value.
x is the value of each jump. For example if x = 2, if 1 is chosen you cannot choose 2.
I need to determine the best way to choose integers, given some (x), that produce the highest total value from the value list.
EXAMPLE:
A = a set of 1 foot intervals (0,1,2,3,4,5,6,7,8,9)
B = the amount of money at each interval (9,5,7,3,2,7,8,10,21,12)
Distance = the minimum distance you can cover
- i.e. if the minimum distance is 3, you must skip 2 feet and leave the money, then you can
pick up the amount at the 3rd interval.
if you pick up at 0, the next one you can pick up is 3, if you choose 3 you can
next pick up 6 (after skipping 4 and 5). BUT, you dont have to pick up 6, you
could pick up 7 if it is worth more. You just can't pick up early.
So, how can I programmatically make the best jumps and end with the most money at the end?
So I am using the below equation for computing the opt value in the dynamic programming:
Here d is distance.
if (i -d) >= 0
opt(i) = max (opt(i-1), B[i] + OPT(i-d));
else
opt(i) = max (opt(i-1), B[i]);
Psuedo-code for computing the OPT value:
int A[] = {integers list}; // This is redundant if the integers are consecutive and are always from 0..n.
int B[] = {values list};
int i = 0;
int d = distance; // minimum distance between two picks.
int numIntegers = sizeof(A)/sizeof(int);
int opt[numIntegers];
opt[0] = B[0]; // For the first one Optimal value is picking itself.
for (i=1; i < numIntegers; i++) {
if ((i-d) < 0) {
opt[i] = max (opt[i-1], B[i]);
} else {
opt[i] = max (opt[i-1], B[i] + opt[i-d]);
}
}
EDIT based on OP's requirement about getting the selected integers from B:
for (i=numIntegres - 1; i >= 0;) {
if ((i == 0) && (opt[i] > 0)) {
printf ("%d ", i);
break;
}
if (opt[i] > opt[i-1]) {
printf ("%d ", i);
i = i -d;
} else {
i = i - 1;
}
}
If A[] does not have consecutive integers from 0 to n.
int A[] = {integers list}; // Here the integers may not be consecutive
int B[] = {values list};
int i = 0, j = 0;
int d = distance; // minimum distance between two picks.
int numAs = sizeof(A)/sizeof(int);
int numIntegers = A[numAs-1]
int opt[numIntegers];
opt[0] = 0;
if (A[0] == 0) {
opt[0] = B[0]; // For the first one Optimal value is picking itself.
j = 1;
}
for (i=1; i < numIntegers && j < numAs; i++, j++) {
if (i < A[j]) {
while (i < A[j]) {
opt[i] = opt[i -1];
i = i + 1:
}
}
if ((i-d) < 0) {
opt[i] = max (opt[i-1], B[j]);
} else {
opt[i] = max (opt[i-1], B[j] + opt[i-d]);
}
}
I'm posting this although much has already been posted about this question. I didn't want to post as an answer since it's not working. The answer to this post (Finding the rank of the Given string in list of all possible permutations with Duplicates) did not work for me.
So I tried this (which is a compilation of code I've plagiarized and my attempt to deal with repetitions). The non-repeating cases work fine. BOOKKEEPER generates 83863, not the desired 10743.
(The factorial function and letter counter array 'repeats' are working correctly. I didn't post to save space.)
while (pointer != length)
{
if (sortedWordChars[pointer] != wordArray[pointer])
{
// Swap the current character with the one after that
char temp = sortedWordChars[pointer];
sortedWordChars[pointer] = sortedWordChars[next];
sortedWordChars[next] = temp;
next++;
//For each position check how many characters left have duplicates,
//and use the logic that if you need to permute n things and if 'a' things
//are similar the number of permutations is n!/a!
int ct = repeats[(sortedWordChars[pointer]-64)];
// Increment the rank
if (ct>1) { //repeats?
System.out.println("repeating " + (sortedWordChars[pointer]-64));
//In case of repetition of any character use: (n-1)!/(times)!
//e.g. if there is 1 character which is repeating twice,
//x* (n-1)!/2!
int dividend = getFactorialIter(length - pointer - 1);
int divisor = getFactorialIter(ct);
int quo = dividend/divisor;
rank += quo;
} else {
rank += getFactorialIter(length - pointer - 1);
}
} else
{
pointer++;
next = pointer + 1;
}
}
Note: this answer is for 1-based rankings, as specified implicitly by example. Here's some Python that works at least for the two examples provided. The key fact is that suffixperms * ctr[y] // ctr[x] is the number of permutations whose first letter is y of the length-(i + 1) suffix of perm.
from collections import Counter
def rankperm(perm):
rank = 1
suffixperms = 1
ctr = Counter()
for i in range(len(perm)):
x = perm[((len(perm) - 1) - i)]
ctr[x] += 1
for y in ctr:
if (y < x):
rank += ((suffixperms * ctr[y]) // ctr[x])
suffixperms = ((suffixperms * (i + 1)) // ctr[x])
return rank
print(rankperm('QUESTION'))
print(rankperm('BOOKKEEPER'))
Java version:
public static long rankPerm(String perm) {
long rank = 1;
long suffixPermCount = 1;
java.util.Map<Character, Integer> charCounts =
new java.util.HashMap<Character, Integer>();
for (int i = perm.length() - 1; i > -1; i--) {
char x = perm.charAt(i);
int xCount = charCounts.containsKey(x) ? charCounts.get(x) + 1 : 1;
charCounts.put(x, xCount);
for (java.util.Map.Entry<Character, Integer> e : charCounts.entrySet()) {
if (e.getKey() < x) {
rank += suffixPermCount * e.getValue() / xCount;
}
}
suffixPermCount *= perm.length() - i;
suffixPermCount /= xCount;
}
return rank;
}
Unranking permutations:
from collections import Counter
def unrankperm(letters, rank):
ctr = Counter()
permcount = 1
for i in range(len(letters)):
x = letters[i]
ctr[x] += 1
permcount = (permcount * (i + 1)) // ctr[x]
# ctr is the histogram of letters
# permcount is the number of distinct perms of letters
perm = []
for i in range(len(letters)):
for x in sorted(ctr.keys()):
# suffixcount is the number of distinct perms that begin with x
suffixcount = permcount * ctr[x] // (len(letters) - i)
if rank <= suffixcount:
perm.append(x)
permcount = suffixcount
ctr[x] -= 1
if ctr[x] == 0:
del ctr[x]
break
rank -= suffixcount
return ''.join(perm)
If we use mathematics, the complexity will come down and will be able to find rank quicker. This will be particularly helpful for large strings.
(more details can be found here)
Suggest to programmatically define the approach shown here (screenshot attached below) given below)
I would say David post (the accepted answer) is super cool. However, I would like to improve it further for speed. The inner loop is trying to find inverse order pairs, and for each such inverse order, it tries to contribute to the increment of rank. If we use an ordered map structure (binary search tree or BST) in that place, we can simply do an inorder traversal from the first node (left-bottom) until it reaches the current character in the BST, rather than traversal for the whole map(BST). In C++, std::map is a perfect one for BST implementation. The following code reduces the necessary iterations in loop and removes the if check.
long long rankofword(string s)
{
long long rank = 1;
long long suffixPermCount = 1;
map<char, int> m;
int size = s.size();
for (int i = size - 1; i > -1; i--)
{
char x = s[i];
m[x]++;
for (auto it = m.begin(); it != m.find(x); it++)
rank += suffixPermCount * it->second / m[x];
suffixPermCount *= (size - i);
suffixPermCount /= m[x];
}
return rank;
}
#Dvaid Einstat, this was really helpful. It took me a WHILE to figure out what you were doing as I am still learning my first language(C#). I translated it into C# and figured that I'd give that solution as well since this listing helped me so much!
Thanks!
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Text.RegularExpressions;
namespace CsharpVersion
{
class Program
{
//Takes in the word and checks to make sure that the word
//is between 1 and 25 charaters inclusive and only
//letters are used
static string readWord(string prompt, int high)
{
Regex rgx = new Regex("^[a-zA-Z]+$");
string word;
string result;
do
{
Console.WriteLine(prompt);
word = Console.ReadLine();
} while (word == "" | word.Length > high | rgx.IsMatch(word) == false);
result = word.ToUpper();
return result;
}
//Creates a sorted dictionary containing distinct letters
//initialized with 0 frequency
static SortedDictionary<char,int> Counter(string word)
{
char[] wordArray = word.ToCharArray();
int len = word.Length;
SortedDictionary<char,int> count = new SortedDictionary<char,int>();
foreach(char c in word)
{
if(count.ContainsKey(c))
{
}
else
{
count.Add(c, 0);
}
}
return count;
}
//Creates a factorial function
static int Factorial(int n)
{
if (n <= 1)
{
return 1;
}
else
{
return n * Factorial(n - 1);
}
}
//Ranks the word input if there are no repeated charaters
//in the word
static Int64 rankWord(char[] wordArray)
{
int n = wordArray.Length;
Int64 rank = 1;
//loops through the array of letters
for (int i = 0; i < n-1; i++)
{
int x=0;
//loops all letters after i and compares them for factorial calculation
for (int j = i+1; j<n ; j++)
{
if (wordArray[i] > wordArray[j])
{
x++;
}
}
rank = rank + x * (Factorial(n - i - 1));
}
return rank;
}
//Ranks the word input if there are repeated charaters
//in the word
static Int64 rankPerm(String word)
{
Int64 rank = 1;
Int64 suffixPermCount = 1;
SortedDictionary<char, int> counter = Counter(word);
for (int i = word.Length - 1; i > -1; i--)
{
char x = Convert.ToChar(word.Substring(i,1));
int xCount;
if(counter[x] != 0)
{
xCount = counter[x] + 1;
}
else
{
xCount = 1;
}
counter[x] = xCount;
foreach (KeyValuePair<char,int> e in counter)
{
if (e.Key < x)
{
rank += suffixPermCount * e.Value / xCount;
}
}
suffixPermCount *= word.Length - i;
suffixPermCount /= xCount;
}
return rank;
}
static void Main(string[] args)
{
Console.WriteLine("Type Exit to end the program.");
string prompt = "Please enter a word using only letters:";
const int MAX_VALUE = 25;
Int64 rank = new Int64();
string theWord;
do
{
theWord = readWord(prompt, MAX_VALUE);
char[] wordLetters = theWord.ToCharArray();
Array.Sort(wordLetters);
bool duplicate = false;
for(int i = 0; i< theWord.Length - 1; i++)
{
if(wordLetters[i] < wordLetters[i+1])
{
duplicate = true;
}
}
if(duplicate)
{
SortedDictionary<char, int> counter = Counter(theWord);
rank = rankPerm(theWord);
Console.WriteLine("\n" + theWord + " = " + rank);
}
else
{
char[] letters = theWord.ToCharArray();
rank = rankWord(letters);
Console.WriteLine("\n" + theWord + " = " + rank);
}
} while (theWord != "EXIT");
Console.WriteLine("\nPress enter to escape..");
Console.Read();
}
}
}
If there are k distinct characters, the i^th character repeated n_i times, then the total number of permutations is given by
(n_1 + n_2 + ..+ n_k)!
------------------------------------------------
n_1! n_2! ... n_k!
which is the multinomial coefficient.
Now we can use this to compute the rank of a given permutation as follows:
Consider the first character(leftmost). say it was the r^th one in the sorted order of characters.
Now if you replace the first character by any of the 1,2,3,..,(r-1)^th character and consider all possible permutations, each of these permutations will precede the given permutation. The total number can be computed using the above formula.
Once you compute the number for the first character, fix the first character, and repeat the same with the second character and so on.
Here's the C++ implementation to your question
#include<iostream>
using namespace std;
int fact(int f) {
if (f == 0) return 1;
if (f <= 2) return f;
return (f * fact(f - 1));
}
int solve(string s,int n) {
int ans = 1;
int arr[26] = {0};
int len = n - 1;
for (int i = 0; i < n; i++) {
s[i] = toupper(s[i]);
arr[s[i] - 'A']++;
}
for(int i = 0; i < n; i++) {
int temp = 0;
int x = 1;
char c = s[i];
for(int j = 0; j < c - 'A'; j++) temp += arr[j];
for (int j = 0; j < 26; j++) x = (x * fact(arr[j]));
arr[c - 'A']--;
ans = ans + (temp * ((fact(len)) / x));
len--;
}
return ans;
}
int main() {
int i,n;
string s;
cin>>s;
n=s.size();
cout << solve(s,n);
return 0;
}
Java version of unrank for a String:
public static String unrankperm(String letters, int rank) {
Map<Character, Integer> charCounts = new java.util.HashMap<>();
int permcount = 1;
for(int i = 0; i < letters.length(); i++) {
char x = letters.charAt(i);
int xCount = charCounts.containsKey(x) ? charCounts.get(x) + 1 : 1;
charCounts.put(x, xCount);
permcount = (permcount * (i + 1)) / xCount;
}
// charCounts is the histogram of letters
// permcount is the number of distinct perms of letters
StringBuilder perm = new StringBuilder();
for(int i = 0; i < letters.length(); i++) {
List<Character> sorted = new ArrayList<>(charCounts.keySet());
Collections.sort(sorted);
for(Character x : sorted) {
// suffixcount is the number of distinct perms that begin with x
Integer frequency = charCounts.get(x);
int suffixcount = permcount * frequency / (letters.length() - i);
if (rank <= suffixcount) {
perm.append(x);
permcount = suffixcount;
if(frequency == 1) {
charCounts.remove(x);
} else {
charCounts.put(x, frequency - 1);
}
break;
}
rank -= suffixcount;
}
}
return perm.toString();
}
See also n-th-permutation-algorithm-for-use-in-brute-force-bin-packaging-parallelization.