/*
* Check handling of conversions between exceptional entries and full bitmaps.
*/
static void ida_check_conv(struct ida *ida)
{
unsigned long i;
for (i = 0; i < IDA_BITMAP_BITS * 2; i += IDA_BITMAP_BITS) {
IDA_BUG_ON(ida, ida_alloc_min(ida, i + 1, GFP_KERNEL) != i + 1);
IDA_BUG_ON(ida, ida_alloc_min(ida, i + BITS_PER_LONG,
GFP_KERNEL) != i + BITS_PER_LONG);
ida_free(ida, i + 1);
ida_free(ida, i + BITS_PER_LONG);
IDA_BUG_ON(ida, !ida_is_empty(ida));
}
for (i = 0; i < IDA_BITMAP_BITS * 2; i++)
IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
for (i = IDA_BITMAP_BITS * 2; i > 0; i--)
ida_free(ida, i - 1);
IDA_BUG_ON(ida, !ida_is_empty(ida));
for (i = 0; i < IDA_BITMAP_BITS + BITS_PER_LONG - 4; i++)
IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
for (i = IDA_BITMAP_BITS + BITS_PER_LONG - 4; i > 0; i--)
ida_free(ida, i - 1);
IDA_BUG_ON(ida, !ida_is_empty(ida));
}
static int hash__init_new_context(struct mm_struct *mm)
{
int index;
index = hash__alloc_context_id();
if (index < 0)
return index;
mm->context.hash_context = kmalloc(sizeof(struct hash_mm_context),
GFP_KERNEL);
if (!mm->context.hash_context) {
ida_free(&mmu_context_ida, index);
return -ENOMEM;
}
/*
* The old code would re-promote on fork, we don't do that when using
* slices as it could cause problem promoting slices that have been
* forced down to 4K.
*
* For book3s we have MMU_NO_CONTEXT set to be ~0. Hence check
* explicitly against context.id == 0. This ensures that we properly
* initialize context slice details for newly allocated mm's (which will
* have id == 0) and don't alter context slice inherited via fork (which
* will have id != 0).
*
* We should not be calling init_new_context() on init_mm. Hence a
* check against 0 is OK.
*/
if (mm->context.id == 0) {
memset(mm->context.hash_context, 0, sizeof(struct hash_mm_context));
slice_init_new_context_exec(mm);
} else {
/* This is fork. Copy hash_context details from current->mm */
memcpy(mm->context.hash_context, current->mm->context.hash_context, sizeof(struct hash_mm_context));
#ifdef CONFIG_PPC_SUBPAGE_PROT
/* inherit subpage prot detalis if we have one. */
if (current->mm->context.hash_context->spt) {
mm->context.hash_context->spt = kmalloc(sizeof(struct subpage_prot_table),
GFP_KERNEL);
if (!mm->context.hash_context->spt) {
ida_free(&mmu_context_ida, index);
kfree(mm->context.hash_context);
return -ENOMEM;
}
}
#endif
}
pkey_mm_init(mm);
return index;
}
int
ida_detach(device_t dev)
{
struct ida_softc *ida;
int error;
ida = (struct ida_softc *)device_get_softc(dev);
error = bus_generic_detach(dev);
if (error)
return (error);
error = device_delete_children(dev);
if (error)
return (error);
/*
* XXX
* before detaching, we must make sure that the system is
* quiescent; nothing mounted, no pending activity.
*/
/*
* XXX
* now, how are we supposed to maintain a list of our drives?
* iterate over our "child devices"?
*/
destroy_dev(ida->ida_dev_t);
ida_free(ida);
return (error);
}
static void destroy_contexts(mm_context_t *ctx)
{
int index, context_id;
for (index = 0; index < ARRAY_SIZE(ctx->extended_id); index++) {
context_id = ctx->extended_id[index];
if (context_id)
ida_free(&mmu_context_ida, context_id);
}
}
static void chan_dev_release(struct device *dev)
{
struct dma_chan_dev *chan_dev;
chan_dev = container_of(dev, typeof(*chan_dev), device);
if (atomic_dec_and_test(chan_dev->idr_ref)) {
ida_free(&dma_ida, chan_dev->dev_id);
kfree(chan_dev->idr_ref);
}
kfree(chan_dev);
}
/*
* Straightforward checks that allocating and freeing IDs work.
*/
static void ida_check_alloc(struct ida *ida)
{
int i, id;
for (i = 0; i < 10000; i++)
IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
ida_free(ida, 20);
ida_free(ida, 21);
for (i = 0; i < 3; i++) {
id = ida_alloc(ida, GFP_KERNEL);
IDA_BUG_ON(ida, id < 0);
if (i == 2)
IDA_BUG_ON(ida, id != 10000);
}
for (i = 0; i < 5000; i++)
ida_free(ida, i);
IDA_BUG_ON(ida, ida_alloc_min(ida, 5000, GFP_KERNEL) != 10001);
ida_destroy(ida);
IDA_BUG_ON(ida, !ida_is_empty(ida));
}
/*
* Check what happens when we fill a leaf and then delete it. This may
* discover mishandling of IDR_FREE.
*/
static void ida_check_leaf(struct ida *ida, unsigned int base)
{
unsigned long i;
for (i = 0; i < IDA_BITMAP_BITS; i++) {
IDA_BUG_ON(ida, ida_alloc_min(ida, base, GFP_KERNEL) !=
base + i);
}
ida_destroy(ida);
IDA_BUG_ON(ida, !ida_is_empty(ida));
IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != 0);
IDA_BUG_ON(ida, ida_is_empty(ida));
ida_free(ida, 0);
IDA_BUG_ON(ida, !ida_is_empty(ida));
}
/***** FINALIZE ********************************************************/
char* fmmv_finalize(FmmvHandle *FMMV, struct FmmvStatistics *statistics)
{
ida_free(FMMV);
if (FMMV->targets) {
FMMV_FREE(FMMV, FMMV->perm, FMMV->NParticles*sizeof(int));
FMMV_FREE(FMMV, FMMV->permTargets, FMMV->NTargets*sizeof(int));
freeTree_ST(FMMV, FMMV->firstSourceBoxOfLevel[0]);
}
else {
FMMV_FREE(FMMV, FMMV->perm, FMMV->NParticles*sizeof(int));
freeTree(FMMV, FMMV->firstSourceBoxOfLevel[0]);
}
finish_all(FMMV);
if (FMMV->allocatedMemory!=0) {
printf("*** ALLOCATED MEMORY NOT FREED: %u ***\n", FMMV->allocatedMemory);
}
stat_stop(FMMV, STAT_TOTAL);
if (statistics) {
*statistics = FMMV->statistics;
statistics->maxNoOfStoredXin = FMMV->maxNoOfStoredXin;
statistics->maxAllocatedMemory = FMMV->maxAllocatedMemory;
statistics->noOfDirectInteractions = FMMV->noOfDirectInteractions;
statistics->PAPIeventSet = FMMV->statistics.PAPIeventSet;
}
free(FMMV);
return 0;
/* //TODO: error handling
_err:
return errbuf;
*/
}
int main(int argc, char**argv)
{
int err;
int i;
double t0, t1;
FmmvHandle _FMMV;
FmmvHandle *FMMV = &_FMMV;
void (*GEN_M)(FmmvHandle *FMMV, Box *box);
void (*EVAL_M)(FmmvHandle *FMMV, Box *target, Box *source);
void (*EVAL_L)(FmmvHandle *FMMV, Box *box);
void (*GEN_L)(FmmvHandle *FMMV, Box *target, Box *source);
void (*GEN_L_EVAL_M)(FmmvHandle *FMMV, Box *target, Box *source);
void (*EVAL_DIRECT)(FmmvHandle *FMMV, Box *target, Box *source) = FMMV->eval_direct;
Box _box_A;
Box _box_B;
Box *box_A = &_box_A;
Box *box_B = &_box_B;
Box _child_A[4];
Box _child_B[4];
int NParticles = 100;
int NTargets = 100;
_FLOAT_ (*particles)[2];
_FLOAT_ (*targets)[2] = NULL;
_FLOAT_ *charges;
_FLOAT_ (*dipoleMoments)[2] = NULL;
_FLOAT_ x, y;
_FLOAT_ *potM;
_FLOAT_ *potL;
_FLOAT_ (*gradM)[2] = NULL;
_FLOAT_ (*gradL)[2] = NULL;
_FLOAT_ *exPot;
_FLOAT_ (*exGrad)[2] = NULL;
_FLOAT_ errL2, errL2grad;
double beta = 0.0;
int pM = 8;
int pL = 8;
int s = 8;
int *perm;
int *permTargets;
int with_dipoleSources=0;
int with_gradients=0;
int LM_combined=0;
int with_M2M=0;
int with_L2L=0;
int printResult=0;
#ifdef USE_SINGLE_PRECISION
int exAcc=1;
#else
int exAcc=2;
#endif
//int exAcc=0;
_FLOAT_ x_A = .125;
_FLOAT_ y_A = .125;
double size_A = .0625;
int level_A = 2;
_FLOAT_ x_B = .875;
_FLOAT_ y_B = .875;
double size_B = .0625;
int level_B = 2;
feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW );
set_from_command_line_double(argc, argv, "beta", &beta);
set_from_command_line_bool(argc, argv, "dipoleSources", &with_dipoleSources);
set_from_command_line_bool(argc, argv, "gradients", &with_gradients);
set_from_command_line_bool(argc, argv, "printResult", &printResult);
set_from_command_line_bool(argc, argv, "LM_combined", &LM_combined);
set_from_command_line_bool(argc, argv, "M2M", &with_M2M);
set_from_command_line_bool(argc, argv, "L2L", &with_L2L);
set_from_command_line_int(argc, argv, "directEvalAccuracy", &exAcc);
if (with_dipoleSources) {
if (with_gradients) {
FMMV->dataKind = FMM_ST_DIPOLE_GRAD;
GEN_M = gen_M_ST_dipole_grad;
EVAL_M = eval_M_ST_dipole_grad;
GEN_L = gen_L_ST_dipole_grad;
EVAL_L = eval_L_ST_dipole_grad;
GEN_L_EVAL_M = gen_L_eval_M_dipole_grad;
switch (exAcc) {
case 0: EVAL_DIRECT = eval_direct_ST_dipole_grad_acc0; break;
case 1: EVAL_DIRECT = eval_direct_ST_dipole_grad_acc1; break;
#if FMM_PRECISION==1
case 2: EVAL_DIRECT = eval_direct_ST_dipole_grad_acc2; break;
#endif
}
}
else {
FMMV->dataKind = FMM_ST_DIPOLE;
GEN_M = gen_M_ST_dipole;
EVAL_M = eval_M_ST_dipole;
GEN_L = gen_L_ST_dipole;
EVAL_L = eval_L_ST_dipole;
GEN_L_EVAL_M = gen_L_eval_M_dipole;
switch (exAcc) {
case 0: EVAL_DIRECT = eval_direct_ST_dipole_acc0; break;
case 1: EVAL_DIRECT = eval_direct_ST_dipole_acc1; break;
#if FMM_PRECISION==1
case 2: EVAL_DIRECT = eval_direct_ST_dipole_acc2; break;
#endif
}
}
}
else {
if (with_gradients) {
FMMV->dataKind = FMM_ST_GRAD;
GEN_M = gen_M_ST_grad;
EVAL_M = eval_M_ST_grad;
GEN_L = gen_L_ST_grad;
EVAL_L = eval_L_ST_grad;
GEN_L_EVAL_M = gen_L_eval_M_grad;
switch (exAcc) {
case 0: EVAL_DIRECT = eval_direct_ST_grad_acc0; break;
case 1: EVAL_DIRECT = eval_direct_ST_grad_acc1; break;
#if FMM_PRECISION==1
case 2: EVAL_DIRECT = eval_direct_ST_grad_acc2; break;
#endif
}
}
else {
FMMV->dataKind = FMM_ST_STANDARD;
GEN_M = gen_M_ST_standard;
EVAL_M = eval_M_ST_standard;
GEN_L = gen_L_ST_standard;
EVAL_L = eval_L_ST_standard;
GEN_L_EVAL_M = gen_L_eval_M_standard;
switch (exAcc) {
case 0: EVAL_DIRECT = eval_direct_ST_standard_acc0; break;
case 1: EVAL_DIRECT = eval_direct_ST_standard_acc1; break;
#if FMM_PRECISION==1
case 2: EVAL_DIRECT = eval_direct_ST_standard_acc2; break;
#endif
}
}
}
set_from_command_line_int(argc, argv, "NParticles", &NParticles);
set_from_command_line_int(argc, argv, "NTargets", &NTargets);
set_from_command_line_int(argc, argv, "pM", &pM);
set_from_command_line_int(argc, argv, "pL", &pL);
set_from_command_line_int(argc, argv, "s", &s);
set_from_command_line_int(argc, argv, "level_A", &level_A);
set_from_command_line_int(argc, argv, "level_B", &level_B);
// set_from_command_line_double(argc, argv, "size_A", &size_A);
// set_from_command_line_double(argc, argv, "size_B", &size_B);
size_A = ldexp(1.0, -level_A);
size_B = ldexp(1.0, -level_B);
particles = (_FLOAT_ (*)[2]) calloc(NParticles, 2*sizeof(_FLOAT_));
charges = (_FLOAT_ *) calloc(NParticles, sizeof(_FLOAT_));
perm = (int *) calloc(NParticles, sizeof(int));
permTargets = (int *) calloc(NTargets, sizeof(int));
targets = (_FLOAT_ (*)[2]) calloc(NTargets, 2*sizeof(_FLOAT_));
potM = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
potL = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
exPot = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
if (with_gradients) {
gradM = (_FLOAT_ (*)[2]) calloc(NTargets, 2*sizeof(_FLOAT_));
gradL = (_FLOAT_ (*)[2]) calloc(NTargets, 2*sizeof(_FLOAT_));
exGrad = (_FLOAT_ (*)[2]) calloc(NTargets, 2*sizeof(_FLOAT_));
}
if (with_dipoleSources) {
dipoleMoments = (_FLOAT_ (*)[2]) calloc(NParticles, 2*sizeof(_FLOAT_));
}
FMMV->beta = beta;
FMMV->pM = pM;
FMMV->pL = pL;
FMMV->s_eps = s;
FMMV->particles = particles;
FMMV->charges = charges;
FMMV->dipoleMoments = dipoleMoments;
FMMV->perm = perm;
FMMV->NParticles = NParticles;
FMMV->targets = targets;
FMMV->permTargets = permTargets;
FMMV->NTargets = NTargets;
FMMV->scale = 1.0;
FMMV->lambda = 0;
my_srand(1481765933);
box_A->firstParticle = 0;
box_A->noOfParticles = NParticles;
box_A->x = x_A;
box_A->y = y_A;
box_A->level = level_A;
box_A->M = 0;
/* if (with_M2M) */ {
int m = NParticles/4;
double d = 0.25*size_A;
int c;
for (c=0; c<4; c++) {
box_A->child[c] = &_child_A[c];
box_A->child[c]->level = level_A + 1;
box_A->child[c]->M = 0;
box_A->child[c]->firstParticle = c*m;
box_A->child[c]->noOfParticles = m;
}
box_A->child[3]->noOfParticles = NParticles - box_A->child[3]->firstParticle;
box_A->child[SW]->x = x_A - d;
box_A->child[SW]->y = y_A - d;
box_A->child[NW]->x = x_A - d;
box_A->child[NW]->y = y_A + d;
box_A->child[SE]->x = x_A + d;
box_A->child[SE]->y = y_A - d;
box_A->child[NE]->x = x_A + d;
box_A->child[NE]->y = y_A + d;
for (c=0; c<4; c++) {
for (i = box_A->child[c]->firstParticle; i < box_A->child[c]->firstParticle+box_A->child[c]->noOfParticles; i++) {
x = box_A->child[c]->x + 2.0*d*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
y = box_A->child[c]->y + 2.0*d*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
particles[i][0] = x;
particles[i][1] = y;
}
}
}
/* else {
double d = size_A/2.0;
for (i=0;i<NParticles;i++) {
x = x_A + size_A*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
y = y_A + size_A*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
particles[i][0] = x;
particles[i][1] = y;
}
}*/
for (i=0;i<NParticles;i++) {
perm[i] = i;
charges[i] = 1.0;
}
if (with_dipoleSources) {
for (i=0;i<NParticles;i++) {
x = (my_rand() / (_FLOAT_) MY_RAND_MAX) - 0.5;
y = (my_rand() / (_FLOAT_) MY_RAND_MAX) - 0.5;
dipoleMoments[i][0] = x;
dipoleMoments[i][1] = y;
}
}
box_B->firstTarget = 0;
box_B->noOfTargets = NTargets;
box_B->x = x_B;
box_B->y = y_B;
box_B->level = level_B;
box_B->L = 0;
/* if (with_L2L) */ {
int m = NTargets/4;
double d = 0.25*size_B;
int c;
for (c=0; c<4; c++) {
box_B->child[c] = &_child_B[c];
box_B->child[c]->level = level_B + 1;
box_B->child[c]->L = 0;
box_B->child[c]->firstTarget = c*m;
box_B->child[c]->noOfTargets = m;
}
box_B->child[3]->noOfTargets = NTargets - box_B->child[3]->firstTarget;
box_B->child[SW]->x = x_B - d;
box_B->child[SW]->y = y_B - d;
box_B->child[NW]->x = x_B - d;
box_B->child[NW]->y = y_B + d;
box_B->child[SE]->x = x_B + d;
box_B->child[SE]->y = y_B - d;
box_B->child[NE]->x = x_B + d;
box_B->child[NE]->y = y_B + d;
for (c=0; c<4; c++) {
for (i = box_B->child[c]->firstTarget; i < box_B->child[c]->firstTarget+box_B->child[c]->noOfTargets; i++) {
x = box_B->child[c]->x + 2.0*d*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
y = box_B->child[c]->y + 2.0*d*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
targets[i][0] = x;
targets[i][1] = y;
}
}
}
/* else {
double d = size_B/2.0;
for (i=0;i<NTargets;i++) {
x = x_B + size_B*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
y = y_B + size_B*(my_rand() / (_FLOAT_) MY_RAND_MAX) - d;
targets[i][0] = x;
targets[i][1] = y;
}
} */
for (i=0;i<NTargets;i++) {
permTargets[i] = i;
}
init_all(FMMV);
err = ida_allocate(FMMV);
copy_particles(FMMV);
copy_charges(FMMV);
zero_pot(FMMV);
t0 = (double) clock()/CLOCKS_PER_SEC;
EVAL_DIRECT(FMMV, box_B, box_A);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (eval_direct) : %7.4f\n", t1);
FMMV->potentials = exPot;
FMMV->gradients = exGrad;
backcopy_pot(FMMV);
if (with_M2M||with_L2L) { // in init_M2M is initialisation als necessary for L2L
init_M2M(FMMV, -1);
}
if (with_M2M) {
int c;
t0 = (double) clock()/CLOCKS_PER_SEC;
for (c=0; c<4; c++) {
GEN_M(FMMV, box_A->child[c]);
}
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (GEN_M) : %7.4f\n", t1);
//init_M2M(FMMV, -1);
init_M2M(FMMV, level_A);
t0 = (double) clock()/CLOCKS_PER_SEC;
M2M(FMMV, box_A);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (M2M) : %7.4f\n", t1);
}
else {
t0 = (double) clock()/CLOCKS_PER_SEC;
GEN_M(FMMV, box_A);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (GEN_M) : %7.4f\n", t1);
}
/*
printf("M:\n");
for (i=0;i<=pM;i++) {
printf("%3i %24.16e %24.16e\n", i, box_A->M[2*i], box_A->M[2*i+1]);
}
*/
zero_pot(FMMV);
t0 = (double) clock()/CLOCKS_PER_SEC;
EVAL_M(FMMV, box_B, box_A);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (EVAL_M) : %7.4f\n", t1);
FMMV->potentials = potM;
FMMV->gradients = gradM;
backcopy_pot(FMMV);
t0 = (double) clock()/CLOCKS_PER_SEC;
GEN_L(FMMV, box_B, box_A);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (GEN_L) : %7.4f\n", t1);
zero_pot(FMMV);
if (with_L2L) {
int c;
init_L2L(FMMV, -1);
init_L2L(FMMV, level_B);
t0 = (double) clock()/CLOCKS_PER_SEC;
L2L(FMMV, box_B);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (L2L) : %7.4f\n", t1);
t0 = (double) clock()/CLOCKS_PER_SEC;
for (c=0; c<4; c++) {
EVAL_L(FMMV, box_B->child[c]);
}
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (EVAL_L) : %7.4f\n", t1);
printf("L:\n");
for (i=0;i<=pL;i++) {
printf("%3i %24.16e %24.16e\n", i, box_B->child[3]->L[2*i], box_B->child[3]->L[2*i+1]);
}
}
else {
t0 = (double) clock()/CLOCKS_PER_SEC;
EVAL_L(FMMV, box_B);
t1 = (double) clock()/CLOCKS_PER_SEC - t0;
printf("time (EVAL_L) : %7.4f\n", t1);
printf("L:\n");
for (i=0;i<=pL;i++) {
printf("%3i %24.16e %24.16e\n", i, box_B->L[2*i], box_B->L[2*i+1]);
}
}
FMMV->potentials = potL;
FMMV->gradients = gradL;
backcopy_pot(FMMV);
ida_free(FMMV);
if (with_gradients) {
relL2Err(0, 0, 0); /* init */
relL2Err2(0, NULL, NULL); /* init */
if (printResult) {
printf("\n Pot(M) Pot(exact) rel.err. rel.err.(grad)\n");
printf("============================================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potM[i], exPot[i]); /* accumulate */
relL2Err2(1, gradM[i], exGrad[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e %12.4e\n", i, (double) potM[i], (double) exPot[i], (double) relErr(potM[i], exPot[i]),(_FLOAT_) relErr2(gradM[i], exGrad[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
errL2grad = relL2Err2(2, NULL, NULL); /*finalize*/
printf ("\nerr_L2(potM) = %.4e err_L2(gradM) = %.4e\n", errL2, errL2grad);
printf("%24.16e %24.16e\n", exGrad[0][0], exGrad[0][1]);
printf("%24.16e %24.16e\n", gradM[0][0], gradM[0][1]);
relL2Err(0, 0, 0); /* init */
relL2Err2(0, NULL, NULL); /* init */
if (printResult) {
printf("\n Pot(L) Pot(exact) rel.err. rel.err.(grad)\n");
printf("============================================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potL[i], exPot[i]); /* accumulate */
relL2Err2(1, gradL[i], exGrad[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e %12.4e\n", i, (double) potL[i], (double) exPot[i], (double) relErr(potL[i], exPot[i]),(_FLOAT_) relErr2(gradL[i], exGrad[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
errL2grad = relL2Err2(2, NULL, NULL); /*finalize*/
printf ("\nerr_L2(potL) = %.4e err_L2(gradL) = %.4e\n", errL2, errL2grad);
printf("%24.16e %24.16e\n", exGrad[0][0], exGrad[0][1]);
printf("%24.16e %24.16e\n", gradL[0][0], gradL[0][1]);
}
else {
relL2Err(0, 0, 0);
if (printResult) {
printf("\n Pot(M) Pot(exact) rel.err.\n");
printf("============================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potM[i], exPot[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e\n", i, (double) potM[i], (double) exPot[i], (double) relErr(potM[i], exPot[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
printf ("\nerr_L2(potM) = %.4e\n", errL2);
relL2Err(0, 0, 0);
if (printResult) {
printf("\n Pot(L) Pot(exact) rel.err.\n");
printf("============================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potL[i], exPot[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e\n", i, (double) potL[i], (double) exPot[i], (double) relErr(potL[i], exPot[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
printf ("\nerr_L2(potL) = %.4e\n", errL2);
}
// TODO: deallocate...
finish_all(FMMV);
return 0;
}
int main(int argc, char**argv)
{
int err;
int i;
FmmvHandle _FMMV;
FmmvHandle *FMMV = &_FMMV;
void (*GEN_M)(FmmvHandle *FMMV, Box *box);
void (*GEN_L)(FmmvHandle *FMMV, Box *target, Box *source);
void (*EVAL_L)(FmmvHandle *FMMV, Box *box);
void (*EVAL_DIRECT)(FmmvHandle *FMMV, Box *target, Box *source) = FMMV->eval_direct;
_FLOAT_ X1[8*FMM_S_EXP_MAX];
_FLOAT_ X2[8*FMM_S_EXP_MAX];
_FLOAT_ X[FMM_S_EXP_MAX];
_FLOAT_ D[FMM_S_EXP_MAX];
Box _box_A;
Box _box_B;
Box _parent_A;
Box _parent_B;
int which_child_A = SWD;
int which_child_B = SWD;
Box *box_A = &_box_A;
Box *box_B = &_box_B;
Box *parent_A = &_parent_A;
Box *parent_B = &_parent_B;
int NParticles = 100;
int NTargets = 100;
_FLOAT_ (*particles)[3];
_FLOAT_ (*targets)[3] = NULL;
_FLOAT_ *charges;
_FLOAT_ x, y, z;
_FLOAT_ x_B, y_B, z_B;
_FLOAT_ x_A, y_A, z_A;
_FLOAT_ *potX;
_FLOAT_ *potL;
_FLOAT_ *exPot;
_FLOAT_ errL2;
double beta = 0.0;
int pM = 8;
int pL = 8;
int s = 8;
int *perm;
int *permTargets;
int printResult=0;
int level = 2;
_FLOAT_ size;
_FLOAT_ dx = 1;
_FLOAT_ dy = 1;
_FLOAT_ dz = 2;
feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW );
FMMV->dataKind = FMM_ST_STANDARD;
GEN_M = gen_M_ST_standard;
GEN_L = gen_L_ST_standard;
EVAL_L = eval_L_ST_standard;
#if (FMM_PRECISION==0)
EVAL_DIRECT = eval_direct_ST_standard_acc1;
#else
EVAL_DIRECT = eval_direct_ST_standard_acc2;
#endif
set_from_command_line_double(argc, argv, "beta", &beta);
set_from_command_line_bool(argc, argv, "printResult", &printResult);
set_from_command_line_int(argc, argv, "NParticles", &NParticles);
set_from_command_line_int(argc, argv, "NTargets", &NTargets);
set_from_command_line_int(argc, argv, "pM", &pM);
set_from_command_line_int(argc, argv, "pL", &pL);
set_from_command_line_int(argc, argv, "s", &s);
set_from_command_line_int(argc, argv, "level", &level);
size = ldexp(1.0, -level);
particles = (_FLOAT_ (*)[3]) calloc(NParticles, 3*sizeof(_FLOAT_));
charges = (_FLOAT_ *) calloc(NParticles, sizeof(_FLOAT_));
perm = (int *) calloc(NParticles, sizeof(int));
permTargets = (int *) calloc(NTargets, sizeof(int));
targets = (_FLOAT_ (*)[3]) calloc(NTargets, 3*sizeof(_FLOAT_));
potX = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
potL = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
exPot = (_FLOAT_ *) calloc(NTargets, sizeof(_FLOAT_));
FMMV->beta = beta;
FMMV->pM = pM;
FMMV->pL = pL;
FMMV->s_eps = s;
FMMV->particles = particles;
FMMV->charges = charges;
FMMV->perm = perm;
FMMV->NParticles = NParticles;
FMMV->targets = targets;
FMMV->permTargets = permTargets;
FMMV->NTargets = NTargets;
FMMV->scale = 1.0;
FMMV->lambda = 0;
my_srand(1481765933);
x_A = 0.5*size;
y_A = 0.5*size;
z_A = 0.5*size;
box_A->firstParticle = 0;
box_A->noOfParticles = NParticles;
parent_A->firstParticle = 0;
parent_A->noOfParticles = NParticles;
box_A->x = x_A;
box_A->y = y_A;
box_A->z = z_A;
box_A->level = level;
parent_A->level = level-1;
box_A->M = 0;
x_B = x_A + size*dx;
y_B = y_A + size*dy;
z_B = z_A + size*dz;
box_B->firstTarget = 0;
box_B->noOfTargets = NTargets;
parent_B->firstTarget = 0;
parent_B->noOfTargets = NTargets;
box_B->x = x_B;
box_B->y = y_B;
box_B->z = z_B;
box_B->level = level;
parent_B->level = level-1;
box_B->L = 0;
for (i=0; i<8; i++) {
parent_A->child[i] = 0;
parent_B->child[i] = 0;
}
parent_A->child[which_child_A] = box_A;
parent_B->child[which_child_B] = box_B;
for (i=0;i<NParticles;i++) {
x = x_A + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
y = y_A + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
z = z_A + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
particles[i][0] = x;
particles[i][1] = y;
particles[i][2] = z;
}
for (i=0;i<NParticles;i++) {
perm[i] = i;
charges[i] = 1.0;
}
for (i=0;i<NTargets;i++) {
x = x_B + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
y = y_B + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
z = z_B + size*(my_rand() / (_FLOAT_) MY_RAND_MAX - 0.5);
targets[i][0] = x;
targets[i][1] = y;
targets[i][2] = z;
}
for (i=0;i<NTargets;i++) {
permTargets[i] = i;
}
init_all(FMMV);
err = ida_allocate(FMMV);
copy_particles(FMMV);
copy_charges(FMMV);
zero_pot(FMMV);
EVAL_DIRECT(FMMV, box_B, box_A);
FMMV->potentials = exPot;
backcopy_pot(FMMV);
GEN_M(FMMV, box_A);
GEN_L(FMMV, box_B, box_A);
zero_pot(FMMV);
EVAL_L(FMMV, box_B);
FMMV->potentials = potL;
backcopy_pot(FMMV);
/**********************************************/
for (i=0;i<6;i++) {
box_B->X[i] = 0;
}
box_B->X[XU] = X;
for (i=0;i<pL*(pL+1);i++) {
box_B->L[i] = 0.0;
}
init_M2L(FMMV, -1);
init_M2L(FMMV, level);
M2X(FMMV, 0, parent_A, X1, X2);
gen_diag_X2X(ldexp(beta,-level), FMMV->lambda, FMMV->M, FMMV->s_eps, FMMV->s_exp, dx, dy, dz, D);
VEC_MUL_C(FMMV->s_exp/2, D, X1+FMMV->s_exp*which_child_A, box_B->X[XU]);
X2L(FMMV, 0, parent_B, 0);
finish_M2L(FMMV);
zero_pot(FMMV);
EVAL_L(FMMV, box_B);
FMMV->potentials = potX;
backcopy_pot(FMMV);
/**********************************************/
ida_free(FMMV);
relL2Err(0, 0, 0);
if (printResult) {
printf("\n Pot(L) Pot(exact) rel.err.\n");
printf("============================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potL[i], exPot[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e\n", i, (double) potL[i], (double) exPot[i], (double) relErr(potL[i], exPot[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
printf ("\nerr_L2(potL) = %.4e\n", errL2);
relL2Err(0, 0, 0);
if (printResult) {
printf("\n Pot(X) Pot(exact) rel.err.\n");
printf("============================================\n");
}
for (i=0; i<NTargets; i++) {
relL2Err(1, potX[i], exPot[i]); /* accumulate */
if (printResult) {
printf("%3i %12.4e %12.4e %12.4e\n", i, (double) potX[i], (double) exPot[i], (double) relErr(potX[i], exPot[i]));
}
}
errL2 = relL2Err(2, 0, 0); /*finalize*/
printf ("\nerr_L2(potX) = %.4e\n", errL2);
// TODO: deallocate...
finish_all(FMMV);
return 0;
}
void __destroy_context(int context_id)
{
ida_free(&mmu_context_ida, context_id);
}
static int
ida_eisa_attach(device_t dev)
{
struct ida_softc *ida;
struct ida_board *board;
int error;
int rid;
ida = device_get_softc(dev);
ida->dev = dev;
board = ida_eisa_match(eisa_get_id(dev));
ida->cmd = *board->accessor;
ida->flags = board->flags;
mtx_init(&ida->lock, "ida", NULL, MTX_DEF);
callout_init_mtx(&ida->ch, &ida->lock, 0);
ida->regs_res_type = SYS_RES_IOPORT;
ida->regs_res_id = 0;
ida->regs = bus_alloc_resource_any(dev, ida->regs_res_type,
&ida->regs_res_id, RF_ACTIVE);
if (ida->regs == NULL) {
device_printf(dev, "can't allocate register resources\n");
return (ENOMEM);
}
error = bus_dma_tag_create(
/* parent */ bus_get_dma_tag(dev),
/* alignment */ 0,
/* boundary */ 0,
/* lowaddr */ BUS_SPACE_MAXADDR_32BIT,
/* highaddr */ BUS_SPACE_MAXADDR,
/* filter */ NULL,
/* filterarg */ NULL,
/* maxsize */ MAXBSIZE,
/* nsegments */ IDA_NSEG,
/* maxsegsize */ BUS_SPACE_MAXSIZE_32BIT,
/* flags */ BUS_DMA_ALLOCNOW,
/* lockfunc */ NULL,
/* lockarg */ NULL,
&ida->parent_dmat);
if (error != 0) {
device_printf(dev, "can't allocate DMA tag\n");
ida_free(ida);
return (ENOMEM);
}
rid = 0;
ida->irq_res_type = SYS_RES_IRQ;
ida->irq = bus_alloc_resource_any(dev, ida->irq_res_type, &rid,
RF_ACTIVE | RF_SHAREABLE);
if (ida->irq == NULL) {
ida_free(ida);
return (ENOMEM);
}
error = bus_setup_intr(dev, ida->irq, INTR_TYPE_BIO | INTR_ENTROPY | INTR_MPSAFE,
NULL, ida_intr, ida, &ida->ih);
if (error) {
device_printf(dev, "can't setup interrupt\n");
ida_free(ida);
return (ENOMEM);
}
error = ida_init(ida);
if (error) {
ida_free(ida);
return (error);
}
return (0);
}
static int
ida_pci_attach(device_t dev)
{
struct ida_board *board = ida_pci_match(dev);
u_int32_t id = pci_get_devid(dev);
struct ida_softc *ida;
u_int command;
int error, rid;
command = pci_read_config(dev, PCIR_COMMAND, 1);
/*
* it appears that this board only does MEMIO access.
*/
if ((command & PCIM_CMD_MEMEN) == 0) {
device_printf(dev, "Only memory mapped I/O is supported\n");
return (ENXIO);
}
ida = (struct ida_softc *)device_get_softc(dev);
ida->dev = dev;
ida->cmd = *board->accessor;
ida->flags = board->flags;
ida->regs_res_type = SYS_RES_MEMORY;
ida->regs_res_id = IDA_PCI_MEMADDR;
if (id == IDA_DEVICEID_DEC_SMART)
ida->regs_res_id = PCIR_BAR(0);
ida->regs = bus_alloc_resource_any(dev, ida->regs_res_type,
&ida->regs_res_id, RF_ACTIVE);
if (ida->regs == NULL) {
device_printf(dev, "can't allocate memory resources\n");
return (ENOMEM);
}
error = bus_dma_tag_create(
/* parent */ NULL,
/* alignment */ 1,
/* boundary */ 0,
/* lowaddr */ BUS_SPACE_MAXADDR_32BIT,
/* highaddr */ BUS_SPACE_MAXADDR,
/* filter */ NULL,
/* filterarg */ NULL,
/* maxsize */ MAXBSIZE,
/* nsegments */ IDA_NSEG,
/* maxsegsize */ BUS_SPACE_MAXSIZE_32BIT,
/* flags */ BUS_DMA_ALLOCNOW,
/* lockfunc */ NULL,
/* lockarg */ NULL,
&ida->parent_dmat);
if (error != 0) {
device_printf(dev, "can't allocate DMA tag\n");
ida_free(ida);
return (ENOMEM);
}
rid = 0;
ida->irq_res_type = SYS_RES_IRQ;
ida->irq = bus_alloc_resource_any(dev, ida->irq_res_type, &rid,
RF_ACTIVE | RF_SHAREABLE);
if (ida->irq == NULL) {
ida_free(ida);
return (ENOMEM);
}
error = bus_setup_intr(dev, ida->irq, INTR_TYPE_BIO | INTR_ENTROPY,
NULL, ida_intr, ida, &ida->ih);
if (error) {
device_printf(dev, "can't setup interrupt\n");
ida_free(ida);
return (ENOMEM);
}
error = ida_init(ida);
if (error) {
ida_free(ida);
return (error);
}
ida_attach(ida);
ida->flags |= IDA_ATTACHED;
return (0);
}
