kern_obj * spawn_process(const uchar * cd, ulong csz, ulong dsz, ulong bsz, kern_obj * par, uchar priv, uchar priority)
{
ulong i, pbase, ssz, absz, pgs;
kern_obj * proc = alloc_kobj();
ssz = csz + dsz;
absz = bsz;
if (csz % 0x1000 != 0) csz = csz - (csz % 0x1000) + 0x1000;
if (dsz % 0x1000 != 0) dsz = dsz - (dsz % 0x1000) + 0x1000;
if (bsz % 0x1000 != 0) bsz = bsz - (bsz % 0x1000) + 0x1000;
proc->type = KOBJ_PROCESS;
proc->u.proc.parent = par;
proc->u.proc.priv = priv;
proc->u.proc.priority = priority;
proc->u.proc.code_data_ws = alloc_ws();
pgs = csz / 0x1000 + dsz / 0x1000 + bsz / 0x1000;
pbase = alloc_pages(pgs + 0x200, PHYS_PAGES);
if (pbase == 0) {
free_kobj(proc);
return NULL;
}
add_pgs_to_ws(proc->u.proc.code_data_ws, USPACE_BOT, pbase, csz + dsz + bsz + 0x200000, 7);
for (i = 0; i < ssz + absz; i++) {
if (i % 0x1000 == 0) {
ATQ(KPT0_0_LOC) = (pbase + i) | 3;
INVLPG(KSPACE_LOC);
}
if (i < ssz) ATB(KSPACE_LOC + i % 0x1000) = cd[i];
else ATB(KSPACE_LOC + i % 0x1000) = 0;
}
if (spawn_thread(proc, (int (*)())USPACE_BOT) == NULL) {
free_kobj(proc);
free_pages(pbase, pgs, PHYS_PAGES);
return NULL;
}
getlock(&procthrd_lock, 0);
if (head_process == NULL) {
proc->u.proc.next = proc;
head_process = proc;
} else {
proc->u.proc.next = head_process->u.proc.next;
head_process->u.proc.next = proc;
}
unlock(&procthrd_lock, 0);
return proc;
}
void add_pgs_to_ws(kern_obj * o, ulong vbase, ulong pbase, uint size, ushort fl)
{
while (o->u.ws.next != NULL) o = o->u.ws.next;
if (o->u.ws.size != 0) {
o->u.ws.next = alloc_ws();
o = o->u.ws.next;
}
o->u.ws.flags = fl;
o->u.ws.vbase = vbase;
o->u.ws.pbase = pbase;
o->u.ws.size = size;
}
kern_obj * spawn_thread(kern_obj * proc, int (*func)())
{
ulong fl, pg;
kern_obj * thrd = alloc_kobj();
thrd->type = KOBJ_THREAD;
thrd->u.thrd.proc = proc;
thrd->u.thrd.sub = alloc_kobj();
thrd->u.thrd.sub->type = KOBJ_THREAD_SUB;
thrd->u.thrd.sub->u.thrd2.stack_ws = alloc_ws();
pg = alloc_pages(2, PHYS_PAGES);
if (pg == 0) {
free_kobj(thrd->u.thrd.sub);
free_kobj(thrd);
return 0;
}
add_pgs_to_ws(thrd->u.thrd.sub->u.thrd2.stack_ws, USPACE_TOP - 0x2000, pg, 0x2000, 7);
thrd->u.thrd.sub->u.thrd2.rsp = USPACE_TOP - 0x50;
ATQ(KPT0_0_LOC) = (pg + 0x1000) | 3;
INVLPG(KSPACE_LOC);
ATQ(KSPACE_LOC + 0xFB0) = (ulong)func;
ATQ(KSPACE_LOC + 0xFB8) = USER_CS | 3;
GET_FLAGS(fl);
ATQ(KSPACE_LOC + 0xFC0) = fl;
ATQ(KSPACE_LOC + 0xFC8) = USPACE_TOP - 0x1000;
ATQ(KSPACE_LOC + 0xFD0) = USER_DS | 3;
getlock(&procthrd_lock, 1);
if (head_thread == NULL) {
thrd->u.thrd.next = thrd;
head_thread = thrd;
} else {
thrd->u.thrd.next = head_thread->u.thrd.next;
head_thread->u.thrd.next = thrd;
}
unlock(&procthrd_lock, 1);
return thrd;
}
int main(int argc,char *argv[])
{
int irw,isp,ispp[2],status[6],mnkv;
int bs[4],Ns,nmx,nkv,nmr,ncy,ninv;
double kappa,m0,dm,mu0,mu,res,mres;
double sqne,sqnp[2];
complex_dble lnw1[2],lnr,dr,drmx;
solver_parms_t sp;
mrw_masses_t ms;
MPI_Init(&argc,&argv);
MPI_Comm_rank(MPI_COMM_WORLD,&my_rank);
if (my_rank==0)
{
flog=freopen("check2.log","w",stdout);
fin=freopen("check2.in","r",stdin);
printf("\n");
printf("Direct check of mrw2\n");
printf("----------------------\n\n");
printf("%dx%dx%dx%d lattice, ",NPROC0*L0,NPROC1*L1,NPROC2*L2,NPROC3*L3);
printf("%dx%dx%dx%d process grid, ",NPROC0,NPROC1,NPROC2,NPROC3);
printf("%dx%dx%dx%d local lattice\n\n",L0,L1,L2,L3);
}
mnkv=0;
mres=0.0;
for (isp=0;isp<3;isp++)
{
read_solver_parms(isp);
sp=solver_parms(isp);
if (sp.res>mres)
mres=sp.res;
if (sp.nkv>mnkv)
mnkv=sp.nkv;
}
read_bc_parms();
if (my_rank==0)
{
find_section("SAP");
read_line("bs","%d %d %d %d",bs,bs+1,bs+2,bs+3);
}
MPI_Bcast(bs,4,MPI_INT,0,MPI_COMM_WORLD);
set_sap_parms(bs,0,1,1);
if (my_rank==0)
{
find_section("Deflation subspace");
read_line("bs","%d %d %d %d",bs,bs+1,bs+2,bs+3);
read_line("Ns","%d",&Ns);
}
MPI_Bcast(bs,4,MPI_INT,0,MPI_COMM_WORLD);
MPI_Bcast(&Ns,1,MPI_INT,0,MPI_COMM_WORLD);
set_dfl_parms(bs,Ns);
if (my_rank==0)
{
find_section("Deflation subspace generation");
read_line("kappa","%lf",&kappa);
read_line("mu","%lf",&mu);
read_line("ninv","%d",&ninv);
read_line("nmr","%d",&nmr);
read_line("ncy","%d",&ncy);
}
MPI_Bcast(&kappa,1,MPI_DOUBLE,0,MPI_COMM_WORLD);
MPI_Bcast(&mu,1,MPI_DOUBLE,0,MPI_COMM_WORLD);
MPI_Bcast(&ninv,1,MPI_INT,0,MPI_COMM_WORLD);
MPI_Bcast(&nmr,1,MPI_INT,0,MPI_COMM_WORLD);
MPI_Bcast(&ncy,1,MPI_INT,0,MPI_COMM_WORLD);
set_dfl_gen_parms(kappa,mu,ninv,nmr,ncy);
if (my_rank==0)
{
find_section("Deflation projection");
read_line("nkv","%d",&nkv);
read_line("nmx","%d",&nmx);
read_line("res","%lf",&res);
fclose(fin);
}
MPI_Bcast(&nkv,1,MPI_INT,0,MPI_COMM_WORLD);
MPI_Bcast(&nmx,1,MPI_INT,0,MPI_COMM_WORLD);
MPI_Bcast(&res,1,MPI_DOUBLE,0,MPI_COMM_WORLD);
set_dfl_pro_parms(nkv,nmx,res);
set_lat_parms(6.0,1.0,0,NULL,1.234);
print_solver_parms(status,status+1);
print_sap_parms(0);
print_dfl_parms(0);
start_ranlux(0,1245);
geometry();
mnkv=2*mnkv+2;
if (mnkv<(Ns+2))
mnkv=Ns+2;
if (mnkv<5)
mnkv=5;
alloc_ws(mnkv);
alloc_wsd(7);
alloc_wv(2*nkv+2);
alloc_wvd(4);
drmx.re=0.0;
drmx.im=0.0;
for (irw=0;irw<3;irw++)
{
dm=1.0e-2;
for (isp=0;isp<3;isp++)
{
ispp[0]=isp;
ispp[1]=isp;
if (isp==0)
{
m0=1.0877;
mu0=0.1;
}
else if (isp==1)
{
m0=0.0877;
mu0=0.01;
}
else
{
m0=-0.0123;
mu0=0.001;
}
random_ud();
if (isp==2)
{
dfl_modes(status);
error_root(status[0]<0,1,"main [check2.c]",
"dfl_modes failed");
}
if (irw==0)
{
ms.m1=m0;
ms.d1=dm;
ms.mu1=mu0;
ms.m2=m0;
ms.d2=dm;
ms.mu2=mu0;
lnr=mrw2(ms,0,ispp,lnw1,sqnp,&sqne,status);
dr.re=fabs(lnw1[0].re-lnw1[1].re);
dr.im=fabs(lnw1[0].im-lnw1[1].im);
lnr=mrw2(ms,1,ispp,lnw1,sqnp,&sqne,status);
dr.re+=fabs(lnr.re-(2.0*mu0*dm+dm*dm)*sqnp[0]);
dr.re+=fabs(lnw1[0].re-lnw1[1].re);
dr.re+=fabs(sqnp[0]-sqnp[1]);
dr.im+=fabs(lnr.im);
dr.im+=fabs(lnw1[0].im-lnw1[1].im);
}
else if (irw==1)
{
ms.m1=m0;
ms.d1=dm;
ms.mu1=mu0;
ms.m2=m0;
ms.d2=-dm;
ms.mu2=mu0;
lnr=mrw2(ms,0,ispp,lnw1,sqnp,&sqne,status);
dr.re=fabs(lnw1[0].re+lnw1[1].re);
dr.im=fabs(lnw1[0].im+lnw1[1].im);
lnr=mrw2(ms,1,ispp,lnw1,sqnp,&sqne,status);
dr.re+=fabs(lnr.re+dm*dm*sqnp[0]);
dr.re+=fabs(lnw1[0].re+lnw1[1].re);
dr.re+=fabs(sqnp[0]-sqnp[1]);
dr.im+=fabs(lnr.im-2.0*lnw1[0].im);
dr.im+=fabs(lnw1[0].im+lnw1[1].im);
}
else
{
ms.m1=m0;
ms.d1=dm;
ms.mu1=mu0;
ms.m2=m0+dm;
ms.d2=-dm;
ms.mu2=mu0;
lnr=mrw2(ms,0,ispp,lnw1,sqnp,&sqne,status);
dr.re=fabs(lnr.re);
dr.im=fabs(lnr.im);
}
if (dr.re>drmx.re)
drmx.re=dr.re;
if (dr.im>drmx.im)
drmx.im=dr.im;
if (my_rank==0)
{
if (irw==0)
printf("mrw2(d2=d1): ");
else if (irw==1)
printf("mrw2(d2=-d1): ");
else
printf("mrw2(m2=m1+d1,d2=-d1): ");
if ((isp==0)||(isp==1))
printf("status = %d\n",status[0]);
else if (isp==2)
printf("status = (%d,%d,%d)\n",
status[0],status[1],status[2]);
printf("diff = %.1e + i%.1e\n\n",dr.re,dr.im);
}
error_chk();
}
}
if (my_rank==0)
{
printf("\n");
printf("max diff = %.1e + i%.1e\n",drmx.re,drmx.im);
printf("(should be smaller than %.1e)\n\n",mres*sqrt((double)(VOLUME*NPROC*24)));
fclose(flog);
}
MPI_Finalize();
exit(0);
}