src/rt/raypcalls.c

#ifndef lint
static const char       RCSid[] = "$Id: raypcalls.c,v 2.21 2009/12/12 00:03:42 greg Exp $";
#endif
/*
 *  raypcalls.c - interface for parallel rendering using Radiance
 *
 *  External symbols declared in ray.h
 */

#include "copyright.h"

/*
 *  These calls are designed similarly to the ones in raycalls.c,
 *  but allow for multiple rendering processes on the same host
 *  machine.  There is no sense in specifying more child processes
 *  than you have processor cores, but one child may help by allowing
 *  asynchronous ray computation in an interactive program, and
 *  will protect the caller from fatal rendering errors.
 *
 *  You should first read and understand the header in raycalls.c,
 *  as some things are explained there that are not repated here.
 *
 *  The first step is opening one or more rendering processes
 *  with a call to ray_pinit(oct, nproc).  Before calling fork(),
 *  ray_pinit() loads the octree and data structures into the
 *  caller's memory, and ray_popen() synchronizes the ambient
 *  file, if any.  Shared memory permits all sorts of queries
 *  that wouldn't be possible otherwise without causing any real
 *  memory overhead, since all the static data are shared
 *  between processes.  Rays are traced using a simple
 *  queuing mechanism, explained below.
 *
 *  The ray queue buffers RAYQLEN rays before sending to
 *  children, each of which may internally buffer RAYQLEN rays
 *  during evaluation.  Rays are not returned in the order
 *  they are sent when multiple processes are open.
 *
 *  Rays are queued and returned by a single
 *  ray_pqueue() call.  A ray_pqueue() return
 *  value of 0 indicates that no rays are ready
 *  and the queue is not yet full.  A return value of 1
 *  indicates that a ray was returned, though it is probably
 *  not the one you just requested.  Rays may be identified by
 *  the rno member of the RAY struct, which is incremented
 *  by the rayorigin() call, or may be set explicitly by
 *  the caller.  Below is an example call sequence:
 *
 *      myRay.rorg = ( ray origin point )
 *      myRay.rdir = ( normalized ray direction )
 *      myRay.rmax = ( maximum length, or zero for no limit )
 *      rayorigin(&myRay, PRIMARY, NULL, NULL);
 *      myRay.rno = ( my personal ray identifier )
 *      if (ray_pqueue(&myRay) == 1)
 *              { do something with results }
 *
 *  Note the differences between this and the simpler ray_trace()
 *  call.  In particular, the call may or may not return a value
 *  in the passed ray structure.  Also, you need to call rayorigin()
 *  yourself, which is normally called for you by ray_trace().  The
 *  benefit is that ray_pqueue() will trace rays faster in
 *  proportion to the number of CPUs you have available on your
 *  system.  If the ray queue is full before the call, ray_pqueue()
 *  will block until a result is ready so it can queue this one.
 *  The global int ray_pnidle indicates the number of currently idle
 *  children.  If you want to check for completed rays without blocking,
 *  or get the results from rays that have been queued without
 *  queuing any new ones, the ray_presult() call is for you:
 *
 *      if (ray_presult(&myRay, 1) == 1)
 *              { do something with results }
 *
 *  If the second argument is 1, the call won't block when
 *  results aren't ready, but will immediately return 0.
 *  If the second argument is 0, the call will block
 *  until a value is available, returning 0 only if the
 *  queue is completely empty.  A negative return value
 *  indicates that a rendering process died.  If this
 *  happens, ray_pclose(0) is automatically called to close
 *  all child processes, and ray_pnprocs is set to zero.
 *
 *  If you just want to fill the ray queue without checking for
 *  results, check ray_pnidle and call ray_psend():
 *
 *      while (ray_pnidle) {
 *              ( set up ray )
 *              ray_psend(&myRay);
 *      }
 *
 *  Note that it is a fatal error to call ra_psend() when
 *  ray_pnidle is zero.  The ray_presult() and/or ray_pqueue()
 *  functions may be called subsequently to read back the results.
 *
 *  When you are done, you may call ray_pdone(1) to close
 *  all child processes and clean up memory used by Radiance.
 *  Any queued ray calculations will be awaited and discarded.
 *  As with ray_done(), ray_pdone(0) hangs onto data files
 *  and fonts that are likely to be used in subsequent renderings.
 *  Whether you need to clean up memory or not, you should
 *  at least call ray_pclose(0) to await the child processes.
 *
 *  Warning:  You cannot affect any of the rendering processes
 *  by changing global parameter values onece ray_pinit() has
 *  been called.  Changing global parameters will have no effect
 *  until the next call to ray_pinit(), which restarts everything.
 *  If you just want to reap children so that you can alter the
 *  rendering parameters without reloading the scene, use the
 *  ray_pclose(0) and ray_popen(nproc) calls to close
 *  then restart the child processes after the changes are made.
 *
 *  Note:  These routines are written to coordinate with the
 *  definitions in raycalls.c, and in fact depend on them.
 *  If you want to trace a ray and get a result synchronously,
 *  use the ray_trace() call to compute it in the parent process.
 *  This will not interfere with any subprocess calculations,
 *  but beware that a fatal error may end with a call to quit().
 *
 *  Note:  One of the advantages of using separate processes
 *  is that it gives the calling program some immunity from
 *  fatal rendering errors.  As discussed in raycalls.c,
 *  Radiance tends to throw up its hands and exit at the
 *  first sign of trouble, calling quit() to return control
 *  to the top level.  Although you can avoid exit() with
 *  your own longjmp() in quit(), the cleanup afterwards
 *  is always suspect.  Through the use of subprocesses,
 *  we avoid this pitfall by closing the processes and
 *  returning a negative value from ray_pqueue() or
 *  ray_presult().  If you get a negative value from either
 *  of these calls, you can assume that the processes have
 *  been cleaned up with a call to ray_pclose(), though you
 *  will have to call ray_pdone() yourself if you want to
 *  free memory.  Obviously, you cannot continue rendering
 *  without risking further errors, but otherwise your
 *  process should not be compromised.
 */

#include  "rtprocess.h"
#include  "ray.h"
#include  "ambient.h"
#include  <sys/types.h>
#include  <sys/wait.h>
#include  "selcall.h"

#ifndef RAYQLEN
#define RAYQLEN         12              /* # rays to send at once */
#endif

#ifndef MAX_RPROCS
#if (FD_SETSIZE/2-4 < 64) 
#define MAX_NPROCS      (FD_SETSIZE/2-4)
#else
#define MAX_NPROCS      64              /* max. # rendering processes */
#endif
#endif

extern char     *shm_boundary;          /* boundary of shared memory */

int             ray_pnprocs = 0;        /* number of child processes */
int             ray_pnidle = 0;         /* number of idle children */

static struct child_proc {
        int     pid;                            /* child process id */
        int     fd_send;                        /* write to child here */
        int     fd_recv;                        /* read from child here */
        int     npending;                       /* # rays in process */
        RNUMBER rno[RAYQLEN];                   /* working on these rays */
} r_proc[MAX_NPROCS];                   /* our child processes */

static RAY      r_queue[2*RAYQLEN];     /* ray i/o buffer */
static int      r_send_next;            /* next send ray placement */
static int      r_recv_first;           /* position of first unreported ray */
static int      r_recv_next;            /* next received ray placement */

#define sendq_full()    (r_send_next >= RAYQLEN)

static int ray_pflush(void);
static void ray_pchild(int fd_in, int fd_out);


void
ray_pinit(              /* initialize ray-tracing processes */
        char    *otnm,
        int     nproc
)
{
        if (nobjects > 0)               /* close old calculation */
                ray_pdone(0);

        ray_init(otnm);                 /* load the shared scene */

        r_send_next = 0;                /* set up queue */
        r_recv_first = r_recv_next = RAYQLEN;

        ray_popen(nproc);               /* fork children */
}


static int
ray_pflush(void)                        /* send queued rays to idle children */
{
        int     nc, n, nw, i, sfirst;

        if ((ray_pnidle <= 0) | (r_send_next <= 0))
                return(0);              /* nothing we can send */
        
        sfirst = 0;                     /* divvy up labor */
        nc = ray_pnidle;
        for (i = ray_pnprocs; nc && i--; ) {
                if (r_proc[i].npending > 0)
                        continue;       /* child looks busy */
                n = (r_send_next - sfirst)/nc--;
                if (!n)
                        continue;
                                        /* smuggle set size in crtype */
                r_queue[sfirst].crtype = n;
                nw = writebuf(r_proc[i].fd_send, (char *)&r_queue[sfirst],
                                sizeof(RAY)*n);
                if (nw != sizeof(RAY)*n)
                        return(-1);     /* write error */
                r_proc[i].npending = n;
                while (n--)             /* record ray IDs */
                        r_proc[i].rno[n] = r_queue[sfirst+n].rno;
                sfirst += r_proc[i].npending;
                ray_pnidle--;           /* now she's busy */
        }
        if (sfirst != r_send_next)
                error(CONSISTENCY, "code screwup in ray_pflush");
        r_send_next = 0;
        return(sfirst);                 /* return total # sent */
}


void
ray_psend(                      /* add a ray to our send queue */
        RAY     *r
)
{
        if (r == NULL)
                return;
                                        /* flush output if necessary */
        if (sendq_full() && ray_pflush() <= 0)
                error(INTERNAL, "ray_pflush failed in ray_psend");

        r_queue[r_send_next++] = *r;
}


int
ray_pqueue(                     /* queue a ray for computation */
        RAY     *r
)
{
        if (r == NULL)
                return(0);
                                        /* check for full send queue */
        if (sendq_full()) {
                RAY     mySend = *r;
                                        /* wait for a result */
                if (ray_presult(r, 0) <= 0)
                        return(-1);
                                        /* put new ray in queue */
                r_queue[r_send_next++] = mySend;
                                /* XXX r_send_next may now be > RAYQLEN */
                return(1);
        }
                                        /* else add ray to send queue */
        r_queue[r_send_next++] = *r;
                                        /* check for returned ray... */
        if (r_recv_first >= r_recv_next)
                return(0);
                                        /* ...one is sitting in queue */
        *r = r_queue[r_recv_first++];
        return(1);
}


int
ray_presult(            /* check for a completed ray */
        RAY     *r,
        int     poll
)
{
        static struct timeval   tpoll;  /* zero timeval struct */
        static fd_set   readset, errset;
        int     n, ok;
        register int    pn;

        if (r == NULL)
                return(0);
                                        /* check queued results first */
        if (r_recv_first < r_recv_next) {
                *r = r_queue[r_recv_first++];
                return(1);
        }
        n = ray_pnprocs - ray_pnidle;   /* pending before flush? */

        if (ray_pflush() < 0)           /* send new rays to process */
                return(-1);
                                        /* reset receive queue */
        r_recv_first = r_recv_next = RAYQLEN;

        if (!poll)                      /* count newly sent unless polling */
                n = ray_pnprocs - ray_pnidle;
        if (n <= 0)                     /* return if nothing to await */
                return(0);
        if (!poll && ray_pnprocs == 1)  /* one process -> skip select() */
                FD_SET(r_proc[0].fd_recv, &readset);

getready:                               /* any children waiting for us? */
        for (pn = ray_pnprocs; pn--; )
                if (FD_ISSET(r_proc[pn].fd_recv, &readset) ||
                                FD_ISSET(r_proc[pn].fd_recv, &errset))
                        break;
                                        /* call select() if we must */
        if (pn < 0) {
                FD_ZERO(&readset); FD_ZERO(&errset); n = 0;
                for (pn = ray_pnprocs; pn--; ) {
                        if (r_proc[pn].npending > 0)
                                FD_SET(r_proc[pn].fd_recv, &readset);
                        FD_SET(r_proc[pn].fd_recv, &errset);
                        if (r_proc[pn].fd_recv >= n)
                                n = r_proc[pn].fd_recv + 1;
                }
                                        /* find out who is ready */
                while ((n = select(n, &readset, (fd_set *)NULL, &errset,
                                poll ? &tpoll : (struct timeval *)NULL)) < 0)
                        if (errno != EINTR) {
                                error(WARNING,
                                        "select call failed in ray_presult");
                                ray_pclose(0);
                                return(-1);
                        }
                if (n > 0)              /* go back and get it */
                        goto getready;
                return(0);              /* else poll came up empty */
        }
        if (r_recv_next + r_proc[pn].npending > sizeof(r_queue)/sizeof(RAY))
                error(CONSISTENCY, "buffer shortage in ray_presult()");

                                        /* read rendered ray data */
        n = readbuf(r_proc[pn].fd_recv, (char *)&r_queue[r_recv_next],
                        sizeof(RAY)*r_proc[pn].npending);
        if (n > 0) {
                r_recv_next += n/sizeof(RAY);
                ok = (n == sizeof(RAY)*r_proc[pn].npending);
        } else
                ok = 0;
                                        /* reset child's status */
        FD_CLR(r_proc[pn].fd_recv, &readset);
        if (n <= 0)
                FD_CLR(r_proc[pn].fd_recv, &errset);
        r_proc[pn].npending = 0;
        ray_pnidle++;
                                        /* check for rendering errors */
        if (!ok) {
                ray_pclose(0);          /* process died -- clean up */
                return(-1);
        }
                                        /* preen returned rays */
        for (n = r_recv_next - r_recv_first; n--; ) {
                register RAY    *rp = &r_queue[r_recv_first + n];
                rp->rno = r_proc[pn].rno[n];
                rp->parent = NULL;
                rp->newcset = rp->clipset = NULL;
                rp->rox = NULL;
                rp->slights = NULL;
        }
                                        /* return first ray received */
        *r = r_queue[r_recv_first++];
        return(1);
}


void
ray_pdone(              /* reap children and free data */
        int     freall
)
{
        ray_pclose(0);                  /* close child processes */

        if (shm_boundary != NULL) {     /* clear shared memory boundary */
                free((void *)shm_boundary);
                shm_boundary = NULL;
        }
        ray_done(freall);               /* free rendering data */
}


static void
ray_pchild(     /* process rays (never returns) */
        int     fd_in,
        int     fd_out
)
{
        int     n;
        register int    i;
                                        /* flag child process for quit() */
        ray_pnprocs = -1;
                                        /* read each ray request set */
        while ((n = read(fd_in, (char *)r_queue, sizeof(r_queue))) > 0) {
                int     n2;
                if (n < sizeof(RAY))
                        break;
                                        /* get smuggled set length */
                n2 = sizeof(RAY)*r_queue[0].crtype - n;
                if (n2 < 0)
                        error(INTERNAL, "buffer over-read in ray_pchild");
                if (n2 > 0) {           /* read the rest of the set */
                        i = readbuf(fd_in, (char *)r_queue + n, n2);
                        if (i != n2)
                                break;
                        n += n2;
                }
                n /= sizeof(RAY);
                                        /* evaluate rays */
                for (i = 0; i < n; i++) {
                        r_queue[i].crtype = r_queue[i].rtype;
                        r_queue[i].parent = NULL;
                        r_queue[i].clipset = NULL;
                        r_queue[i].slights = NULL;
                        r_queue[i].rlvl = 0;
                        samplendx++;
                        rayclear(&r_queue[i]);
                        rayvalue(&r_queue[i]);
                }
                                        /* write back our results */
                i = writebuf(fd_out, (char *)r_queue, sizeof(RAY)*n);
                if (i != sizeof(RAY)*n)
                        error(SYSTEM, "write error in ray_pchild");
        }
        if (n)
                error(SYSTEM, "read error in ray_pchild");
        ambsync();
        quit(0);                        /* normal exit */
}


void
ray_popen(                      /* open the specified # processes */
        int     nadd
)
{
                                        /* check if our table has room */
        if (ray_pnprocs + nadd > MAX_NPROCS)
                nadd = MAX_NPROCS - ray_pnprocs;
        if (nadd <= 0)
                return;
        ambsync();                      /* load any new ambient values */
        if (shm_boundary == NULL) {     /* first child process? */
                preload_objs();         /* preload auxiliary data */
                                        /* set shared memory boundary */
                shm_boundary = (char *)malloc(16);
                strcpy(shm_boundary, "SHM_BOUNDARY");
        }
        fflush(NULL);                   /* clear pending output */
        while (nadd--) {                /* fork each new process */
                int     p0[2], p1[2];
                if (pipe(p0) < 0 || pipe(p1) < 0)
                        error(SYSTEM, "cannot create pipe");
                if ((r_proc[ray_pnprocs].pid = fork()) == 0) {
                        int     pn;     /* close others' descriptors */
                        for (pn = ray_pnprocs; pn--; ) {
                                close(r_proc[pn].fd_send);
                                close(r_proc[pn].fd_recv);
                        }
                        close(p0[0]); close(p1[1]);
                                        /* following call never returns */
                        ray_pchild(p1[0], p0[1]);
                }
                if (r_proc[ray_pnprocs].pid < 0)
                        error(SYSTEM, "cannot fork child process");
                close(p1[0]); close(p0[1]);
                /*
                 * Close write stream on exec to avoid multiprocessing deadlock.
                 * No use in read stream without it, so set flag there as well.
                 */
                fcntl(p1[1], F_SETFD, FD_CLOEXEC);
                fcntl(p0[0], F_SETFD, FD_CLOEXEC);
                r_proc[ray_pnprocs].fd_send = p1[1];
                r_proc[ray_pnprocs].fd_recv = p0[0];
                r_proc[ray_pnprocs].npending = 0;
                ray_pnprocs++;
                ray_pnidle++;
        }
}


void
ray_pclose(             /* close one or more child processes */
        int     nsub
)
{
        static int      inclose = 0;
        RAY     res;
                                        /* check recursion */
        if (inclose)
                return;
        inclose++;
                                        /* check argument */
        if ((nsub <= 0) | (nsub > ray_pnprocs))
                nsub = ray_pnprocs;
                                        /* clear our ray queue */
        while (ray_presult(&res,0) > 0)
                ;
                                        /* clean up children */
        while (nsub--) {
                int     status;
                ray_pnprocs--;
                close(r_proc[ray_pnprocs].fd_recv);
                close(r_proc[ray_pnprocs].fd_send);
                if (waitpid(r_proc[ray_pnprocs].pid, &status, 0) < 0)
                        status = 127<<8;
                if (status) {
                        sprintf(errmsg,
                                "rendering process %d exited with code %d",
                                        r_proc[ray_pnprocs].pid, status>>8);
                        error(WARNING, errmsg);
                }
                ray_pnidle--;
        }
        inclose--;
}


void
quit(ec)                        /* make sure exit is called */
int     ec;
{
        if (ray_pnprocs > 0)    /* close children if any */
                ray_pclose(0);          
        exit(ec);
}
Revision:	2.22
Committed:	Sat Dec 12 05:20:10 2009 UTC (14 years, 4 months ago) by greg
Content type:	text/plain
Branch:	MAIN
Changes since 2.21:	+10 -11 lines
Log Message:	Created FIFO calls for ray multiprocessing
#	User	Rev	Content
1	greg	2.1	#ifndef lint
2	greg	2.22	static const char RCSid[] = "$Id: raypcalls.c,v 2.21 2009/12/12 00:03:42 greg Exp $";
3	greg	2.1	#endif
4			/*
5			* raypcalls.c - interface for parallel rendering using Radiance
6			*
7			* External symbols declared in ray.h
8			*/
9
10	greg	2.2	#include "copyright.h"
11	greg	2.1
12			/*
13			* These calls are designed similarly to the ones in raycalls.c,
14			* but allow for multiple rendering processes on the same host
15			* machine. There is no sense in specifying more child processes
16	greg	2.21	* than you have processor cores, but one child may help by allowing
17	greg	2.1	* asynchronous ray computation in an interactive program, and
18			* will protect the caller from fatal rendering errors.
19			*
20	greg	2.21	* You should first read and understand the header in raycalls.c,
21	greg	2.1	* as some things are explained there that are not repated here.
22			*
23			* The first step is opening one or more rendering processes
24			* with a call to ray_pinit(oct, nproc). Before calling fork(),
25			* ray_pinit() loads the octree and data structures into the
26	greg	2.13	* caller's memory, and ray_popen() synchronizes the ambient
27			* file, if any. Shared memory permits all sorts of queries
28	greg	2.21	* that wouldn't be possible otherwise without causing any real
29	greg	2.1	* memory overhead, since all the static data are shared
30	greg	2.21	* between processes. Rays are traced using a simple
31	greg	2.1	* queuing mechanism, explained below.
32			*
33	greg	2.14	* The ray queue buffers RAYQLEN rays before sending to
34	greg	2.21	* children, each of which may internally buffer RAYQLEN rays
35			* during evaluation. Rays are not returned in the order
36			* they are sent when multiple processes are open.
37	greg	2.14	*
38	greg	2.13	* Rays are queued and returned by a single
39	greg	2.1	* ray_pqueue() call. A ray_pqueue() return
40			* value of 0 indicates that no rays are ready
41			* and the queue is not yet full. A return value of 1
42			* indicates that a ray was returned, though it is probably
43			* not the one you just requested. Rays may be identified by
44			* the rno member of the RAY struct, which is incremented
45			* by the rayorigin() call, or may be set explicitly by
46			* the caller. Below is an example call sequence:
47			*
48			* myRay.rorg = ( ray origin point )
49			* myRay.rdir = ( normalized ray direction )
50			* myRay.rmax = ( maximum length, or zero for no limit )
51	greg	2.11	* rayorigin(&myRay, PRIMARY, NULL, NULL);
52	greg	2.1	* myRay.rno = ( my personal ray identifier )
53			* if (ray_pqueue(&myRay) == 1)
54			* { do something with results }
55			*
56			* Note the differences between this and the simpler ray_trace()
57			* call. In particular, the call may or may not return a value
58			* in the passed ray structure. Also, you need to call rayorigin()
59	greg	2.7	* yourself, which is normally called for you by ray_trace(). The
60			* benefit is that ray_pqueue() will trace rays faster in
61	greg	2.1	* proportion to the number of CPUs you have available on your
62			* system. If the ray queue is full before the call, ray_pqueue()
63			* will block until a result is ready so it can queue this one.
64	greg	2.3	* The global int ray_pnidle indicates the number of currently idle
65	greg	2.1	* children. If you want to check for completed rays without blocking,
66			* or get the results from rays that have been queued without
67			* queuing any new ones, the ray_presult() call is for you:
68			*
69			* if (ray_presult(&myRay, 1) == 1)
70			* { do something with results }
71			*
72			* If the second argument is 1, the call won't block when
73			* results aren't ready, but will immediately return 0.
74			* If the second argument is 0, the call will block
75			* until a value is available, returning 0 only if the
76			* queue is completely empty. A negative return value
77			* indicates that a rendering process died. If this
78	greg	2.21	* happens, ray_pclose(0) is automatically called to close
79	greg	2.3	* all child processes, and ray_pnprocs is set to zero.
80	greg	2.1	*
81			* If you just want to fill the ray queue without checking for
82	greg	2.3	* results, check ray_pnidle and call ray_psend():
83	greg	2.1	*
84	greg	2.3	* while (ray_pnidle) {
85	greg	2.1	* ( set up ray )
86			* ray_psend(&myRay);
87			* }
88			*
89	greg	2.7	* Note that it is a fatal error to call ra_psend() when
90			* ray_pnidle is zero. The ray_presult() and/or ray_pqueue()
91			* functions may be called subsequently to read back the results.
92	greg	2.1	*
93			* When you are done, you may call ray_pdone(1) to close
94			* all child processes and clean up memory used by Radiance.
95			* Any queued ray calculations will be awaited and discarded.
96			* As with ray_done(), ray_pdone(0) hangs onto data files
97			* and fonts that are likely to be used in subsequent renderings.
98	greg	2.21	* Whether you need to clean up memory or not, you should
99			* at least call ray_pclose(0) to await the child processes.
100	greg	2.1	*
101			* Warning: You cannot affect any of the rendering processes
102			* by changing global parameter values onece ray_pinit() has
103			* been called. Changing global parameters will have no effect
104			* until the next call to ray_pinit(), which restarts everything.
105			* If you just want to reap children so that you can alter the
106			* rendering parameters without reloading the scene, use the
107			* ray_pclose(0) and ray_popen(nproc) calls to close
108	greg	2.7	* then restart the child processes after the changes are made.
109	greg	2.1	*
110			* Note: These routines are written to coordinate with the
111			* definitions in raycalls.c, and in fact depend on them.
112			* If you want to trace a ray and get a result synchronously,
113	greg	2.13	* use the ray_trace() call to compute it in the parent process.
114	greg	2.7	* This will not interfere with any subprocess calculations,
115			* but beware that a fatal error may end with a call to quit().
116	greg	2.1	*
117			* Note: One of the advantages of using separate processes
118			* is that it gives the calling program some immunity from
119			* fatal rendering errors. As discussed in raycalls.c,
120			* Radiance tends to throw up its hands and exit at the
121			* first sign of trouble, calling quit() to return control
122	greg	2.7	* to the top level. Although you can avoid exit() with
123	greg	2.1	* your own longjmp() in quit(), the cleanup afterwards
124			* is always suspect. Through the use of subprocesses,
125			* we avoid this pitfall by closing the processes and
126			* returning a negative value from ray_pqueue() or
127			* ray_presult(). If you get a negative value from either
128			* of these calls, you can assume that the processes have
129	greg	2.21	* been cleaned up with a call to ray_pclose(), though you
130	greg	2.1	* will have to call ray_pdone() yourself if you want to
131	greg	2.7	* free memory. Obviously, you cannot continue rendering
132			* without risking further errors, but otherwise your
133			* process should not be compromised.
134	greg	2.1	*/
135
136	schorsch	2.6	#include "rtprocess.h"
137	greg	2.1	#include "ray.h"
138	schorsch	2.6	#include "ambient.h"
139	greg	2.18	#include <sys/types.h>
140			#include <sys/wait.h>
141	greg	2.1	#include "selcall.h"
142
143			#ifndef RAYQLEN
144	greg	2.13	#define RAYQLEN 12 /* # rays to send at once */
145	greg	2.1	#endif
146
147			#ifndef MAX_RPROCS
148			#if (FD_SETSIZE/2-4 < 64)
149			#define MAX_NPROCS (FD_SETSIZE/2-4)
150			#else
151			#define MAX_NPROCS 64 /* max. # rendering processes */
152			#endif
153			#endif
154
155			extern char shm_boundary; / boundary of shared memory */
156
157	greg	2.3	int ray_pnprocs = 0; /* number of child processes */
158			int ray_pnidle = 0; /* number of idle children */
159	greg	2.1
160			static struct child_proc {
161			int pid; /* child process id */
162			int fd_send; /* write to child here */
163			int fd_recv; /* read from child here */
164			int npending; /* # rays in process */
165	greg	2.21	RNUMBER rno[RAYQLEN]; /* working on these rays */
166	greg	2.1	} r_proc[MAX_NPROCS]; /* our child processes */
167
168			static RAY r_queue[2RAYQLEN]; / ray i/o buffer */
169			static int r_send_next; /* next send ray placement */
170			static int r_recv_first; /* position of first unreported ray */
171	greg	2.22	static int r_recv_next; /* next received ray placement */
172	greg	2.1
173			#define sendq_full() (r_send_next >= RAYQLEN)
174
175	schorsch	2.6	static int ray_pflush(void);
176	greg	2.13	static void ray_pchild(int fd_in, int fd_out);
177	greg	2.1
178	schorsch	2.6
179	greg	2.22	void
180	schorsch	2.6	ray_pinit( /* initialize ray-tracing processes */
181			char *otnm,
182			int nproc
183			)
184	greg	2.1	{
185			if (nobjects > 0) /* close old calculation */
186			ray_pdone(0);
187
188			ray_init(otnm); /* load the shared scene */
189
190			r_send_next = 0; /* set up queue */
191			r_recv_first = r_recv_next = RAYQLEN;
192
193			ray_popen(nproc); /* fork children */
194			}
195
196
197			static int
198	schorsch	2.6	ray_pflush(void) /* send queued rays to idle children */
199	greg	2.1	{
200			int nc, n, nw, i, sfirst;
201
202	schorsch	2.5	if ((ray_pnidle <= 0) \| (r_send_next <= 0))
203	greg	2.1	return(0); /* nothing we can send */
204
205			sfirst = 0; /* divvy up labor */
206	greg	2.3	nc = ray_pnidle;
207			for (i = ray_pnprocs; nc && i--; ) {
208	greg	2.1	if (r_proc[i].npending > 0)
209			continue; /* child looks busy */
210			n = (r_send_next - sfirst)/nc--;
211			if (!n)
212			continue;
213			/* smuggle set size in crtype */
214			r_queue[sfirst].crtype = n;
215			nw = writebuf(r_proc[i].fd_send, (char *)&r_queue[sfirst],
216			sizeof(RAY)*n);
217			if (nw != sizeof(RAY)*n)
218			return(-1); /* write error */
219			r_proc[i].npending = n;
220			while (n--) /* record ray IDs */
221			r_proc[i].rno[n] = r_queue[sfirst+n].rno;
222			sfirst += r_proc[i].npending;
223	greg	2.3	ray_pnidle--; /* now she's busy */
224	greg	2.1	}
225			if (sfirst != r_send_next)
226			error(CONSISTENCY, "code screwup in ray_pflush");
227			r_send_next = 0;
228			return(sfirst); /* return total # sent */
229			}
230
231
232	greg	2.22	void
233	schorsch	2.6	ray_psend( /* add a ray to our send queue */
234			RAY *r
235			)
236	greg	2.1	{
237			if (r == NULL)
238			return;
239			/* flush output if necessary */
240			if (sendq_full() && ray_pflush() <= 0)
241			error(INTERNAL, "ray_pflush failed in ray_psend");
242
243	greg	2.14	r_queue[r_send_next++] = *r;
244	greg	2.1	}
245
246
247	greg	2.22	int
248	schorsch	2.6	ray_pqueue( /* queue a ray for computation */
249			RAY *r
250			)
251	greg	2.1	{
252			if (r == NULL)
253			return(0);
254			/* check for full send queue */
255			if (sendq_full()) {
256	greg	2.19	RAY mySend = *r;
257	greg	2.1	/* wait for a result */
258	greg	2.19	if (ray_presult(r, 0) <= 0)
259			return(-1);
260	greg	2.1	/* put new ray in queue */
261	greg	2.14	r_queue[r_send_next++] = mySend;
262	greg	2.19	/* XXX r_send_next may now be > RAYQLEN */
263			return(1);
264	greg	2.1	}
265	greg	2.13	/* else add ray to send queue */
266	greg	2.14	r_queue[r_send_next++] = *r;
267	greg	2.1	/* check for returned ray... */
268			if (r_recv_first >= r_recv_next)
269			return(0);
270			/* ...one is sitting in queue */
271	greg	2.14	*r = r_queue[r_recv_first++];
272	greg	2.1	return(1);
273			}
274
275
276	greg	2.22	int
277	schorsch	2.6	ray_presult( /* check for a completed ray */
278			RAY *r,
279			int poll
280			)
281	greg	2.1	{
282			static struct timeval tpoll; /* zero timeval struct */
283			static fd_set readset, errset;
284			int n, ok;
285			register int pn;
286
287			if (r == NULL)
288			return(0);
289			/* check queued results first */
290			if (r_recv_first < r_recv_next) {
291	greg	2.14	*r = r_queue[r_recv_first++];
292	greg	2.1	return(1);
293			}
294	greg	2.3	n = ray_pnprocs - ray_pnidle; /* pending before flush? */
295	greg	2.1
296			if (ray_pflush() < 0) /* send new rays to process */
297			return(-1);
298			/* reset receive queue */
299			r_recv_first = r_recv_next = RAYQLEN;
300
301			if (!poll) /* count newly sent unless polling */
302	greg	2.3	n = ray_pnprocs - ray_pnidle;
303	greg	2.1	if (n <= 0) /* return if nothing to await */
304			return(0);
305	greg	2.16	if (!poll && ray_pnprocs == 1) /* one process -> skip select() */
306			FD_SET(r_proc[0].fd_recv, &readset);
307
308	greg	2.1	getready: /* any children waiting for us? */
309	greg	2.3	for (pn = ray_pnprocs; pn--; )
310	greg	2.1	if (FD_ISSET(r_proc[pn].fd_recv, &readset) \|\|
311			FD_ISSET(r_proc[pn].fd_recv, &errset))
312			break;
313	greg	2.22	/* call select() if we must */
314	greg	2.1	if (pn < 0) {
315			FD_ZERO(&readset); FD_ZERO(&errset); n = 0;
316	greg	2.3	for (pn = ray_pnprocs; pn--; ) {
317	greg	2.1	if (r_proc[pn].npending > 0)
318			FD_SET(r_proc[pn].fd_recv, &readset);
319			FD_SET(r_proc[pn].fd_recv, &errset);
320			if (r_proc[pn].fd_recv >= n)
321			n = r_proc[pn].fd_recv + 1;
322			}
323			/* find out who is ready */
324			while ((n = select(n, &readset, (fd_set *)NULL, &errset,
325			poll ? &tpoll : (struct timeval *)NULL)) < 0)
326			if (errno != EINTR) {
327			error(WARNING,
328			"select call failed in ray_presult");
329			ray_pclose(0);
330			return(-1);
331			}
332			if (n > 0) /* go back and get it */
333			goto getready;
334			return(0); /* else poll came up empty */
335			}
336			if (r_recv_next + r_proc[pn].npending > sizeof(r_queue)/sizeof(RAY))
337			error(CONSISTENCY, "buffer shortage in ray_presult()");
338
339			/* read rendered ray data */
340			n = readbuf(r_proc[pn].fd_recv, (char *)&r_queue[r_recv_next],
341			sizeof(RAY)*r_proc[pn].npending);
342			if (n > 0) {
343			r_recv_next += n/sizeof(RAY);
344			ok = (n == sizeof(RAY)*r_proc[pn].npending);
345			} else
346			ok = 0;
347			/* reset child's status */
348			FD_CLR(r_proc[pn].fd_recv, &readset);
349			if (n <= 0)
350			FD_CLR(r_proc[pn].fd_recv, &errset);
351			r_proc[pn].npending = 0;
352	greg	2.3	ray_pnidle++;
353	greg	2.1	/* check for rendering errors */
354			if (!ok) {
355			ray_pclose(0); /* process died -- clean up */
356			return(-1);
357			}
358			/* preen returned rays */
359			for (n = r_recv_next - r_recv_first; n--; ) {
360			register RAY *rp = &r_queue[r_recv_first + n];
361			rp->rno = r_proc[pn].rno[n];
362			rp->parent = NULL;
363			rp->newcset = rp->clipset = NULL;
364			rp->rox = NULL;
365			rp->slights = NULL;
366			}
367			/* return first ray received */
368	greg	2.13	*r = r_queue[r_recv_first++];
369	greg	2.1	return(1);
370			}
371
372
373	greg	2.22	void
374	schorsch	2.6	ray_pdone( /* reap children and free data */
375			int freall
376			)
377	greg	2.1	{
378			ray_pclose(0); /* close child processes */
379
380			if (shm_boundary != NULL) { /* clear shared memory boundary */
381			free((void *)shm_boundary);
382			shm_boundary = NULL;
383			}
384			ray_done(freall); /* free rendering data */
385			}
386
387
388			static void
389	schorsch	2.6	ray_pchild( /* process rays (never returns) */
390			int fd_in,
391			int fd_out
392			)
393	greg	2.1	{
394			int n;
395			register int i;
396	greg	2.15	/* flag child process for quit() */
397			ray_pnprocs = -1;
398	greg	2.1	/* read each ray request set */
399			while ((n = read(fd_in, (char *)r_queue, sizeof(r_queue))) > 0) {
400			int n2;
401	greg	2.12	if (n < sizeof(RAY))
402	greg	2.1	break;
403			/* get smuggled set length */
404	greg	2.12	n2 = sizeof(RAY)*r_queue[0].crtype - n;
405	greg	2.1	if (n2 < 0)
406			error(INTERNAL, "buffer over-read in ray_pchild");
407			if (n2 > 0) { /* read the rest of the set */
408	greg	2.12	i = readbuf(fd_in, (char *)r_queue + n, n2);
409			if (i != n2)
410	greg	2.1	break;
411			n += n2;
412			}
413	greg	2.12	n /= sizeof(RAY);
414	greg	2.1	/* evaluate rays */
415			for (i = 0; i < n; i++) {
416			r_queue[i].crtype = r_queue[i].rtype;
417			r_queue[i].parent = NULL;
418			r_queue[i].clipset = NULL;
419			r_queue[i].slights = NULL;
420	greg	2.21	r_queue[i].rlvl = 0;
421	greg	2.1	samplendx++;
422			rayclear(&r_queue[i]);
423			rayvalue(&r_queue[i]);
424			}
425			/* write back our results */
426			i = writebuf(fd_out, (char )r_queue, sizeof(RAY)n);
427			if (i != sizeof(RAY)*n)
428			error(SYSTEM, "write error in ray_pchild");
429			}
430			if (n)
431			error(SYSTEM, "read error in ray_pchild");
432			ambsync();
433			quit(0); /* normal exit */
434			}
435
436
437	greg	2.22	void
438	schorsch	2.6	ray_popen( /* open the specified # processes */
439			int nadd
440			)
441	greg	2.1	{
442			/* check if our table has room */
443	greg	2.3	if (ray_pnprocs + nadd > MAX_NPROCS)
444			nadd = MAX_NPROCS - ray_pnprocs;
445	greg	2.1	if (nadd <= 0)
446			return;
447	greg	2.13	ambsync(); /* load any new ambient values */
448	greg	2.20	if (shm_boundary == NULL) { /* first child process? */
449			preload_objs(); /* preload auxiliary data */
450			/* set shared memory boundary */
451			shm_boundary = (char *)malloc(16);
452			strcpy(shm_boundary, "SHM_BOUNDARY");
453			}
454	greg	2.13	fflush(NULL); /* clear pending output */
455	greg	2.1	while (nadd--) { /* fork each new process */
456			int p0[2], p1[2];
457			if (pipe(p0) < 0 \|\| pipe(p1) < 0)
458			error(SYSTEM, "cannot create pipe");
459	greg	2.3	if ((r_proc[ray_pnprocs].pid = fork()) == 0) {
460	greg	2.1	int pn; /* close others' descriptors */
461	greg	2.3	for (pn = ray_pnprocs; pn--; ) {
462	greg	2.1	close(r_proc[pn].fd_send);
463			close(r_proc[pn].fd_recv);
464			}
465			close(p0[0]); close(p1[1]);
466			/* following call never returns */
467			ray_pchild(p1[0], p0[1]);
468			}
469	greg	2.3	if (r_proc[ray_pnprocs].pid < 0)
470	greg	2.1	error(SYSTEM, "cannot fork child process");
471			close(p1[0]); close(p0[1]);
472	greg	2.9	/*
473			* Close write stream on exec to avoid multiprocessing deadlock.
474			* No use in read stream without it, so set flag there as well.
475			*/
476			fcntl(p1[1], F_SETFD, FD_CLOEXEC);
477			fcntl(p0[0], F_SETFD, FD_CLOEXEC);
478	greg	2.3	r_proc[ray_pnprocs].fd_send = p1[1];
479			r_proc[ray_pnprocs].fd_recv = p0[0];
480			r_proc[ray_pnprocs].npending = 0;
481			ray_pnprocs++;
482			ray_pnidle++;
483	greg	2.1	}
484			}
485
486
487	greg	2.22	void
488	schorsch	2.6	ray_pclose( /* close one or more child processes */
489			int nsub
490			)
491	greg	2.1	{
492			static int inclose = 0;
493			RAY res;
494			/* check recursion */
495			if (inclose)
496			return;
497			inclose++;
498			/* check argument */
499	schorsch	2.5	if ((nsub <= 0) \| (nsub > ray_pnprocs))
500	greg	2.3	nsub = ray_pnprocs;
501	greg	2.1	/* clear our ray queue */
502			while (ray_presult(&res,0) > 0)
503			;
504			/* clean up children */
505			while (nsub--) {
506			int status;
507	greg	2.3	ray_pnprocs--;
508			close(r_proc[ray_pnprocs].fd_recv);
509			close(r_proc[ray_pnprocs].fd_send);
510	greg	2.8	if (waitpid(r_proc[ray_pnprocs].pid, &status, 0) < 0)
511			status = 127<<8;
512	greg	2.1	if (status) {
513			sprintf(errmsg,
514			"rendering process %d exited with code %d",
515	greg	2.3	r_proc[ray_pnprocs].pid, status>>8);
516	greg	2.1	error(WARNING, errmsg);
517			}
518	greg	2.3	ray_pnidle--;
519	greg	2.1	}
520			inclose--;
521			}
522
523
524			void
525			quit(ec) /* make sure exit is called */
526			int ec;
527			{
528	greg	2.15	if (ray_pnprocs > 0) /* close children if any */
529			ray_pclose(0);
530	greg	2.1	exit(ec);
531			}