197929ea7SJunchao Zhang static char help[]= " Test VecScatterRemap() on various vecscatter. \n\ 297929ea7SJunchao Zhang We may do optimization based on index patterns. After index remapping by VecScatterRemap(), we need to \n\ 397929ea7SJunchao Zhang make sure the vecscatter works as expected with the optimizaiton. \n\ 497929ea7SJunchao Zhang VecScatterRemap() does not support all kinds of vecscatters. In addition, it only supports remapping \n\ 597929ea7SJunchao Zhang entries where we read the data (i.e., todata in paralle scatter, fromdata in sequential scatter). This test \n\ 697929ea7SJunchao Zhang tests VecScatterRemap on parallel to paralle (PtoP) vecscatter, sequential general to sequential \n\ 797929ea7SJunchao Zhang general (SGToSG) vecscatter and sequential general to sequential stride 1 (SGToSS_Stride1) vecscatter.\n\n"; 897929ea7SJunchao Zhang 997929ea7SJunchao Zhang #include <petscvec.h> 1097929ea7SJunchao Zhang 1197929ea7SJunchao Zhang int main(int argc,char **argv) 1297929ea7SJunchao Zhang { 1397929ea7SJunchao Zhang PetscErrorCode ierr; 1497929ea7SJunchao Zhang PetscInt i,n,*ix,*iy,*tomap,start; 1597929ea7SJunchao Zhang Vec x,y; 1697929ea7SJunchao Zhang PetscMPIInt nproc,rank; 1797929ea7SJunchao Zhang IS isx,isy; 1897929ea7SJunchao Zhang const PetscInt *ranges; 1997929ea7SJunchao Zhang VecScatter vscat; 2097929ea7SJunchao Zhang 2197929ea7SJunchao Zhang PetscFunctionBegin; 2297929ea7SJunchao Zhang ierr = PetscInitialize(&argc,&argv,(char*)0,help);if (ierr) return ierr; 23ffc4695bSBarry Smith ierr = MPI_Comm_size(PETSC_COMM_WORLD,&nproc);CHKERRMPI(ierr); 24ffc4695bSBarry Smith ierr = MPI_Comm_rank(PETSC_COMM_WORLD,&rank);CHKERRMPI(ierr); 2597929ea7SJunchao Zhang 26*2c71b3e2SJacob Faibussowitsch PetscCheckFalse(nproc != 2,PETSC_COMM_WORLD,PETSC_ERR_WRONG_MPI_SIZE,"This test must run with exactly two MPI ranks"); 2797929ea7SJunchao Zhang 2897929ea7SJunchao Zhang /* ==================================================================== 2997929ea7SJunchao Zhang (1) test VecScatterRemap on a parallel to parallel (PtoP) vecscatter 3097929ea7SJunchao Zhang ==================================================================== 3197929ea7SJunchao Zhang */ 3297929ea7SJunchao Zhang 3397929ea7SJunchao Zhang n = 64; /* long enough to trigger memcpy optimizations both in local scatter and remote scatter */ 3497929ea7SJunchao Zhang 3597929ea7SJunchao Zhang /* create two MPI vectors x, y of length n=64, N=128 */ 3697929ea7SJunchao Zhang ierr = VecCreateMPI(PETSC_COMM_WORLD,n,PETSC_DECIDE,&x);CHKERRQ(ierr); 3797929ea7SJunchao Zhang ierr = VecDuplicate(x,&y);CHKERRQ(ierr); 3897929ea7SJunchao Zhang 3997929ea7SJunchao Zhang /* Initialize x as {0~127} */ 4097929ea7SJunchao Zhang ierr = VecGetOwnershipRanges(x,&ranges);CHKERRQ(ierr); 4197929ea7SJunchao Zhang for (i=ranges[rank]; i<ranges[rank+1]; i++) { ierr = VecSetValue(x,i,(PetscScalar)i,INSERT_VALUES);CHKERRQ(ierr); } 4297929ea7SJunchao Zhang ierr = VecAssemblyBegin(x);CHKERRQ(ierr); 4397929ea7SJunchao Zhang ierr = VecAssemblyEnd(x);CHKERRQ(ierr); 4497929ea7SJunchao Zhang 4597929ea7SJunchao Zhang /* create two general index sets isx = {0~127} and isy = {32~63,64~95,96~127,0~31}. isx is sequential, but we use 4697929ea7SJunchao Zhang it as general and let PETSc detect the pattern and optimize it. indices in isy are set to make the vecscatter 4797929ea7SJunchao Zhang have both local scatter and remote scatter (i.e., MPI communication) 4897929ea7SJunchao Zhang */ 4997929ea7SJunchao Zhang ierr = PetscMalloc2(n,&ix,n,&iy);CHKERRQ(ierr); 5097929ea7SJunchao Zhang start = ranges[rank]; 5197929ea7SJunchao Zhang for (i=ranges[rank]; i<ranges[rank+1]; i++) ix[i-start] = i; 5297929ea7SJunchao Zhang ierr = ISCreateGeneral(PETSC_COMM_WORLD,n,ix,PETSC_COPY_VALUES,&isx);CHKERRQ(ierr); 5397929ea7SJunchao Zhang 54dd400576SPatrick Sanan if (rank == 0) { for (i=0; i<n; i++) iy[i] = i+32; } 5597929ea7SJunchao Zhang else for (i=0; i<n/2; i++) { iy[i] = i+96; iy[i+n/2] = i; } 5697929ea7SJunchao Zhang 5797929ea7SJunchao Zhang ierr = ISCreateGeneral(PETSC_COMM_WORLD,n,iy,PETSC_COPY_VALUES,&isy);CHKERRQ(ierr); 5897929ea7SJunchao Zhang 5997929ea7SJunchao Zhang /* create a vecscatter that shifts x to the tail by quater periodically and puts the results in y */ 6097929ea7SJunchao Zhang ierr = VecScatterCreate(x,isx,y,isy,&vscat);CHKERRQ(ierr); 6197929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 6297929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 6397929ea7SJunchao Zhang 6497929ea7SJunchao Zhang /* view y to check the result. y should be {Q3,Q0,Q1,Q2} of x, that is {96~127,0~31,32~63,64~95} */ 6597929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"Before VecScatterRemap on PtoP, MPI vector y is:\n");CHKERRQ(ierr); 6697929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 6797929ea7SJunchao Zhang 6897929ea7SJunchao Zhang /* now call the weird subroutine VecScatterRemap to slightly change the vecscatter. It changes where we read vector 6997929ea7SJunchao Zhang x entries to send out, but does not change the communication pattern (i.e., send/recv pairs and msg lengths). 7097929ea7SJunchao Zhang 7197929ea7SJunchao Zhang We create tomap as {32~63,0~31}. Originaly, we read from indices {0~64} of the local x to send out. The remap 7297929ea7SJunchao Zhang does indices[i] = tomap[indices[i]]. Therefore, after the remap, we read from indices {32~63,0~31} of the local x. 7397929ea7SJunchao Zhang isy is unchanged. So, we will shift x to {Q2,Q1,Q0,Q3}, that is {64~95,32~63,0~31,96~127} 7497929ea7SJunchao Zhang */ 7597929ea7SJunchao Zhang ierr = PetscMalloc1(n,&tomap);CHKERRQ(ierr); 7697929ea7SJunchao Zhang for (i=0; i<n/2; i++) { tomap[i] = i+n/2; tomap[i+n/2] = i; }; 7797929ea7SJunchao Zhang ierr = VecScatterRemap(vscat,tomap,NULL);CHKERRQ(ierr); 7897929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 7997929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 8097929ea7SJunchao Zhang 8197929ea7SJunchao Zhang /* view y to check the result. y should be {64~95,32~63,0~31,96~127} */ 8297929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"After VecScatterRemap on PtoP, MPI vector y is:\n");CHKERRQ(ierr); 8397929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 8497929ea7SJunchao Zhang 8597929ea7SJunchao Zhang /* destroy everything before we recreate them in different types */ 8697929ea7SJunchao Zhang ierr = PetscFree2(ix,iy);CHKERRQ(ierr); 8797929ea7SJunchao Zhang ierr = VecDestroy(&x);CHKERRQ(ierr); 8897929ea7SJunchao Zhang ierr = VecDestroy(&y);CHKERRQ(ierr); 8997929ea7SJunchao Zhang ierr = ISDestroy(&isx);CHKERRQ(ierr); 9097929ea7SJunchao Zhang ierr = ISDestroy(&isy);CHKERRQ(ierr); 9197929ea7SJunchao Zhang ierr = PetscFree(tomap);CHKERRQ(ierr); 9297929ea7SJunchao Zhang ierr = VecScatterDestroy(&vscat);CHKERRQ(ierr); 9397929ea7SJunchao Zhang 9497929ea7SJunchao Zhang /* ========================================================================================== 9597929ea7SJunchao Zhang (2) test VecScatterRemap on a sequential general to sequential general (SGToSG) vecscatter 9697929ea7SJunchao Zhang ========================================================================================== 9797929ea7SJunchao Zhang */ 9897929ea7SJunchao Zhang n = 64; /* long enough to trigger memcpy optimizations in local scatter */ 9997929ea7SJunchao Zhang 10097929ea7SJunchao Zhang /* create two seq vectors x, y of length n */ 10197929ea7SJunchao Zhang ierr = VecCreateSeq(PETSC_COMM_SELF,n,&x);CHKERRQ(ierr); 10297929ea7SJunchao Zhang ierr = VecDuplicate(x,&y);CHKERRQ(ierr); 10397929ea7SJunchao Zhang 10497929ea7SJunchao Zhang /* Initialize x as {0~63} */ 10597929ea7SJunchao Zhang for (i=0; i<n; i++) { ierr = VecSetValue(x,i,(PetscScalar)i,INSERT_VALUES);CHKERRQ(ierr); } 10697929ea7SJunchao Zhang ierr = VecAssemblyBegin(x);CHKERRQ(ierr); 10797929ea7SJunchao Zhang ierr = VecAssemblyEnd(x);CHKERRQ(ierr); 10897929ea7SJunchao Zhang 10997929ea7SJunchao Zhang /* create two general index sets isx = isy = {0~63}, which are sequential, but we use them as 11097929ea7SJunchao Zhang general and let PETSc detect the pattern and optimize it */ 11197929ea7SJunchao Zhang ierr = PetscMalloc2(n,&ix,n,&iy);CHKERRQ(ierr); 11297929ea7SJunchao Zhang for (i=0; i<n; i++) ix[i] = i; 11397929ea7SJunchao Zhang ierr = ISCreateGeneral(PETSC_COMM_SELF,n,ix,PETSC_COPY_VALUES,&isx);CHKERRQ(ierr); 11497929ea7SJunchao Zhang ierr = ISDuplicate(isx,&isy);CHKERRQ(ierr); 11597929ea7SJunchao Zhang 11697929ea7SJunchao Zhang /* create a vecscatter that just copies x to y */ 11797929ea7SJunchao Zhang ierr = VecScatterCreate(x,isx,y,isy,&vscat);CHKERRQ(ierr); 11897929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 11997929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 12097929ea7SJunchao Zhang 12197929ea7SJunchao Zhang /* view y to check the result. y should be {0~63} */ 12297929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"\nBefore VecScatterRemap on SGToSG, SEQ vector y is:\n");CHKERRQ(ierr); 12397929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 12497929ea7SJunchao Zhang 12597929ea7SJunchao Zhang /* now call the weird subroutine VecScatterRemap to slightly change the vecscatter. 12697929ea7SJunchao Zhang 12797929ea7SJunchao Zhang Create tomap as {32~63,0~31}. Originaly, we read from indices {0~64} of seq x to write to y. The remap 12897929ea7SJunchao Zhang does indices[i] = tomap[indices[i]]. Therefore, after the remap, we read from indices{32~63,0~31} of seq x. 12997929ea7SJunchao Zhang */ 13097929ea7SJunchao Zhang ierr = PetscMalloc1(n,&tomap);CHKERRQ(ierr); 13197929ea7SJunchao Zhang for (i=0; i<n/2; i++) { tomap[i] = i+n/2; tomap[i+n/2] = i; }; 13297929ea7SJunchao Zhang ierr = VecScatterRemap(vscat,tomap,NULL);CHKERRQ(ierr); 13397929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 13497929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 13597929ea7SJunchao Zhang 13697929ea7SJunchao Zhang /* view y to check the result. y should be {32~63,0~31} */ 13797929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"After VecScatterRemap on SGToSG, SEQ vector y is:\n");CHKERRQ(ierr); 13897929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 13997929ea7SJunchao Zhang 14097929ea7SJunchao Zhang /* destroy everything before we recreate them in different types */ 14197929ea7SJunchao Zhang ierr = PetscFree2(ix,iy);CHKERRQ(ierr); 14297929ea7SJunchao Zhang ierr = VecDestroy(&x);CHKERRQ(ierr); 14397929ea7SJunchao Zhang ierr = VecDestroy(&y);CHKERRQ(ierr); 14497929ea7SJunchao Zhang ierr = ISDestroy(&isx);CHKERRQ(ierr); 14597929ea7SJunchao Zhang ierr = ISDestroy(&isy);CHKERRQ(ierr); 14697929ea7SJunchao Zhang ierr = PetscFree(tomap);CHKERRQ(ierr); 14797929ea7SJunchao Zhang ierr = VecScatterDestroy(&vscat);CHKERRQ(ierr); 14897929ea7SJunchao Zhang 14997929ea7SJunchao Zhang /* =================================================================================================== 15097929ea7SJunchao Zhang (3) test VecScatterRemap on a sequential general to sequential stride 1 (SGToSS_Stride1) vecscatter 15197929ea7SJunchao Zhang =================================================================================================== 15297929ea7SJunchao Zhang */ 15397929ea7SJunchao Zhang n = 64; /* long enough to trigger memcpy optimizations in local scatter */ 15497929ea7SJunchao Zhang 15597929ea7SJunchao Zhang /* create two seq vectors x of length n, and y of length n/2 */ 15697929ea7SJunchao Zhang ierr = VecCreateSeq(PETSC_COMM_SELF,n,&x);CHKERRQ(ierr); 15797929ea7SJunchao Zhang ierr = VecCreateSeq(PETSC_COMM_SELF,n/2,&y);CHKERRQ(ierr); 15897929ea7SJunchao Zhang 15997929ea7SJunchao Zhang /* Initialize x as {0~63} */ 16097929ea7SJunchao Zhang for (i=0; i<n; i++) { ierr = VecSetValue(x,i,(PetscScalar)i,INSERT_VALUES);CHKERRQ(ierr); } 16197929ea7SJunchao Zhang ierr = VecAssemblyBegin(x);CHKERRQ(ierr); 16297929ea7SJunchao Zhang ierr = VecAssemblyEnd(x);CHKERRQ(ierr); 16397929ea7SJunchao Zhang 16497929ea7SJunchao Zhang /* create a general index set isx = {0:63:2}, which actually is a stride IS with first=0, n=32, step=2, 16597929ea7SJunchao Zhang but we use it as general and let PETSc detect the pattern and optimize it. */ 16697929ea7SJunchao Zhang ierr = PetscMalloc2(n/2,&ix,n/2,&iy);CHKERRQ(ierr); 16797929ea7SJunchao Zhang for (i=0; i<n/2; i++) ix[i] = i*2; 16897929ea7SJunchao Zhang ierr = ISCreateGeneral(PETSC_COMM_SELF,n/2,ix,PETSC_COPY_VALUES,&isx);CHKERRQ(ierr); 16997929ea7SJunchao Zhang 17097929ea7SJunchao Zhang /* create a stride1 index set isy = {0~31}. We intentionally set the step to 1 to trigger optimizations */ 17197929ea7SJunchao Zhang ierr = ISCreateStride(PETSC_COMM_SELF,32,0,1,&isy);CHKERRQ(ierr); 17297929ea7SJunchao Zhang 17397929ea7SJunchao Zhang /* create a vecscatter that just copies even entries of x to y */ 17497929ea7SJunchao Zhang ierr = VecScatterCreate(x,isx,y,isy,&vscat);CHKERRQ(ierr); 17597929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 17697929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 17797929ea7SJunchao Zhang 17897929ea7SJunchao Zhang /* view y to check the result. y should be {0:63:2} */ 17997929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"\nBefore VecScatterRemap on SGToSS_Stride1, SEQ vector y is:\n");CHKERRQ(ierr); 18097929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 18197929ea7SJunchao Zhang 18297929ea7SJunchao Zhang /* now call the weird subroutine VecScatterRemap to slightly change the vecscatter. 18397929ea7SJunchao Zhang 18497929ea7SJunchao Zhang Create tomap as {32~63,0~31}. Originaly, we read from indices{0:63:2} of seq x to write to y. The remap 18597929ea7SJunchao Zhang does indices[i] = tomap[indices[i]]. Therefore, after the remap, we read from indices{32:63:2,0:31:2} of seq x. 18697929ea7SJunchao Zhang */ 18797929ea7SJunchao Zhang ierr = PetscMalloc1(n,&tomap);CHKERRQ(ierr); 18897929ea7SJunchao Zhang for (i=0; i<n/2; i++) { tomap[i] = i+n/2; tomap[i+n/2] = i; }; 18997929ea7SJunchao Zhang ierr = VecScatterRemap(vscat,tomap,NULL);CHKERRQ(ierr); 19097929ea7SJunchao Zhang ierr = VecScatterBegin(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 19197929ea7SJunchao Zhang ierr = VecScatterEnd(vscat,x,y,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 19297929ea7SJunchao Zhang 19397929ea7SJunchao Zhang /* view y to check the result. y should be {32:63:2,0:31:2} */ 19497929ea7SJunchao Zhang ierr = PetscPrintf(PETSC_COMM_WORLD,"After VecScatterRemap on SGToSS_Stride1, SEQ vector y is:\n");CHKERRQ(ierr); 19597929ea7SJunchao Zhang ierr = VecView(y,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 19697929ea7SJunchao Zhang 19797929ea7SJunchao Zhang /* destroy everything before PetscFinalize */ 19897929ea7SJunchao Zhang ierr = PetscFree2(ix,iy);CHKERRQ(ierr); 19997929ea7SJunchao Zhang ierr = VecDestroy(&x);CHKERRQ(ierr); 20097929ea7SJunchao Zhang ierr = VecDestroy(&y);CHKERRQ(ierr); 20197929ea7SJunchao Zhang ierr = ISDestroy(&isx);CHKERRQ(ierr); 20297929ea7SJunchao Zhang ierr = ISDestroy(&isy);CHKERRQ(ierr); 20397929ea7SJunchao Zhang ierr = PetscFree(tomap);CHKERRQ(ierr); 20497929ea7SJunchao Zhang ierr = VecScatterDestroy(&vscat);CHKERRQ(ierr); 20597929ea7SJunchao Zhang 20697929ea7SJunchao Zhang ierr = PetscFinalize(); 20797929ea7SJunchao Zhang return ierr; 20897929ea7SJunchao Zhang } 20997929ea7SJunchao Zhang 21097929ea7SJunchao Zhang /*TEST 21197929ea7SJunchao Zhang 21297929ea7SJunchao Zhang test: 21397929ea7SJunchao Zhang suffix: 1 21497929ea7SJunchao Zhang nsize: 2 21597929ea7SJunchao Zhang diff_args: -j 21697929ea7SJunchao Zhang requires: double 21797929ea7SJunchao Zhang TEST*/ 21897929ea7SJunchao Zhang 219