예제 #1
0
파일: bitops.c 프로젝트: nanocritical/n
// Find the last non-zero word, last bit set in word has index
// [bit_fls(word)].
unsigned bitmap_fls(const uint32_t *bitmap, unsigned from, unsigned size) {
  unsigned from_word = BITMAP_WORD(from);
  unsigned index_last = BITMAP_WORD(size - 1);
  uint32_t last = bitmap[index_last] & BITMAP_MASK_LAST(size);

  if (index_last == from_word) {
    uint32_t mask = SLICE(~0u, (size - 1) % BITS_PER_WORD, from % BITS_PER_WORD)
        << (from % BITS_PER_WORD);
    last = last & mask;
  }

  if (last != 0)
    return index_last * BITS_PER_WORD + bit_fls(last);

  for (int w = index_last - 1; w >= (int)from_word; --w) {
    if (bitmap[w] != 0)
      return w * BITS_PER_WORD + bit_fls(bitmap[w]);
  }

  if (index_last != from_word && bitmap[from_word] != 0) {
    uint32_t first_mask = SLICE(~0u, BITS_PER_WORD - 1, from % BITS_PER_WORD)
        << (from % BITS_PER_WORD);
    uint32_t first_word = bitmap[from_word] & first_mask;
    if (first_word != 0)
      return from_word * BITS_PER_WORD + bit_fls(first_word);
  }

  return size;
}
예제 #2
0
void test_fls32()
{
    sput_fail_unless(bit_fls(0u)==-1, "Edge case: arg=0");
    sput_fail_unless(bit_fls(1u)==0,  "arg=1");
    sput_fail_unless(bit_fls(0xFFFFFFFFu)==31,  "arg=0xFFFFFFFF");
    sput_fail_unless(bit_fls(0x80000000u)==31,  "arg=0x80000000");

    for (uint32_t i=0; i<32; ++i)
    {
        sput_fail_unless(bit_fls(1u<<i)==i, "General Test0");
        sput_fail_unless(bit_fls(0xFFFFFFFFu>>i)==31-i, "General test1");
    }
}
예제 #3
0
void test_fls64()
{
    sput_fail_unless(bit_fls(0ull)==-1, "Edge case: arg=0");
    sput_fail_unless(bit_fls(1ull)==0,  "arg=1");
    sput_fail_unless(bit_fls(0xFFFFFFFFFFFFFFFFull)==63,  "arg=0xFFFFFFFF");
    sput_fail_unless(bit_ffs(0x8000000000000000ull)==63,  "arg=0x8000000000000000");

    for (uint64_t i=0; i<64; ++i)
    {
        sput_fail_unless(bit_fls(1ull<<i)==i, "General Test");
        sput_fail_unless(bit_fls(0xFFFFFFFFFFFFFFFFull>>i)==63-i, "General test1");
    }
}
예제 #4
0
파일: power_common.c 프로젝트: A1ve5/slurm
/* For a newly starting job, set "new_job_time" in each of it's nodes
 * NOTE: The job and node data structures must be locked on function entry */
extern void set_node_new_job(struct job_record *job_ptr,
			     struct node_record *node_record_table_ptr)
{
	int i, i_first, i_last;
	struct node_record *node_ptr;
	time_t now = time(NULL);

	if (!job_ptr || !job_ptr->node_bitmap) {
		error("%s: job_ptr node_bitmap is NULL", __func__);
		return;
	}

	i_first = bit_ffs(job_ptr->node_bitmap);
	if (i_first >= 0)
		i_last = bit_fls(job_ptr->node_bitmap);
	else
		i_last = i_first - 1;
	for (i = i_first; i <= i_last; i++) {
		if (!bit_test(job_ptr->node_bitmap, i))
			continue;
		node_ptr = node_record_table_ptr + i;
		if (node_ptr->power)
			node_ptr->power->new_job_time = now;
	}
}
예제 #5
0
파일: job_test.c 프로젝트: chrisdukey/slurm
/* Compute resource usage for the given job on all available resources
 *
 * IN: job_ptr     - pointer to the job requesting resources
 * IN: node_map    - bitmap of available nodes
 * IN/OUT: core_map    - bitmap of available cores
 * IN: cr_node_cnt - total number of nodes in the cluster
 * IN: cr_type     - resource type
 * OUT: cpu_cnt    - number of cpus that can be used by this job
 * IN: test_only   - ignore allocated memory check
 * RET SLURM_SUCCESS index of selected node or -1 if none
 */
static int _get_res_usage(struct job_record *job_ptr, bitstr_t *node_map,
			   bitstr_t *core_map, uint32_t cr_node_cnt,
			   struct node_use_record *node_usage,
			   uint16_t cr_type, uint16_t **cpu_cnt_ptr, 
			   bool test_only)
{
	uint16_t *cpu_cnt, max_cpu_cnt = 0, part_lln_flag = 0;
	int i, i_first, i_last;
	int node_inx = -1;

	if (cr_node_cnt != node_record_count) {
		error("select/serial: node count inconsistent with slurmctld");
		return SLURM_ERROR;
	}
	if (!job_ptr) {
		error("select/serial: NULL job pointer");
		return SLURM_ERROR;
	}

	if (job_ptr->part_ptr && (job_ptr->part_ptr->flags & PART_FLAG_LLN))
		part_lln_flag = 1;
	if (job_ptr->details && job_ptr->details->req_node_bitmap)
		bit_and(node_map, job_ptr->details->req_node_bitmap);
	cpu_cnt = xmalloc(cr_node_cnt * sizeof(uint16_t));
	i_first = bit_ffs(node_map);
	if (i_first >= 0)
		i_last  = bit_fls(node_map);
	else
		i_last = -2;
	for (i = i_first; i <= i_last; i++) {
		if (!bit_test(node_map, i))
			continue;
		cpu_cnt[i] = _can_job_run_on_node(job_ptr, core_map, i,
						  node_usage, cr_type,
						  test_only);
		if (!(cr_type & CR_LLN) && !part_lln_flag && cpu_cnt[i]) {
			bit_nclear(node_map, 0, (node_record_count - 1));
			bit_set(node_map, i);
			node_inx = i;
			break;	/* select/serial: only need one node */
		}
	}

	if ((cr_type & CR_LLN) || part_lln_flag) {
		for (i = i_first; i <= i_last; i++) {
			if (cpu_cnt[i] > max_cpu_cnt) {
				max_cpu_cnt = cpu_cnt[i];
				node_inx = i;
			}
		}

		if (node_inx >= 0) {
 			bit_nclear(node_map, 0, (node_record_count - 1));
			bit_set(node_map, node_inx);
 		}
 	}

	*cpu_cnt_ptr = cpu_cnt;
	return node_inx;
}
예제 #6
0
파일: node_conf.c 프로젝트: jsollom/slurm
/*
 * bitmap2node_name_sortable - given a bitmap, build a list of comma
 *	separated node names. names may include regular expressions
 *	(e.g. "lx[01-10]")
 * IN bitmap - bitmap pointer
 * IN sort   - returned sorted list or not
 * RET pointer to node list or NULL on error
 * globals: node_record_table_ptr - pointer to node table
 * NOTE: the caller must xfree the memory at node_list when no longer required
 */
char * bitmap2node_name_sortable (bitstr_t *bitmap, bool sort)
{
	int i, first, last;
	hostlist_t hl;
	char *buf;

	if (bitmap == NULL)
		return xstrdup("");

	first = bit_ffs(bitmap);
	if (first == -1)
		return xstrdup("");

	last  = bit_fls(bitmap);
	hl = hostlist_create("");
	for (i = first; i <= last; i++) {
		if (bit_test(bitmap, i) == 0)
			continue;
		hostlist_push(hl, node_record_table_ptr[i].name);
	}
	if (sort)
		hostlist_sort(hl);
	buf = hostlist_ranged_string_xmalloc(hl);
	hostlist_destroy(hl);

	return buf;
}
예제 #7
0
파일: job_test.c 프로젝트: BYUHPC/slurm
/* given an "avail" node_bitmap, return a corresponding "avail" core_bitmap */
bitstr_t *_make_core_bitmap(bitstr_t *node_map)
{
	uint32_t n, c, nodes, size;
	uint32_t coff;
	int i_first, i_last;

	nodes = bit_size(node_map);
	size = cr_get_coremap_offset(nodes);
	bitstr_t *core_map = bit_alloc(size);

	i_first = bit_ffs(node_map);
	if (i_first >= 0)
		i_last  = bit_fls(node_map);
	else
		i_last = -2;
	for (n = i_first, c = 0; n <= i_last; n++) {
		if (bit_test(node_map, n)) {
			coff = cr_get_coremap_offset(n + 1);
			while (c < coff) {
				bit_set(core_map, c++);
			}
		}
	}
	return core_map;
}
예제 #8
0
파일: power_common.c 프로젝트: A1ve5/slurm
/* For each running job, return power allocation/use information in a List
 * containing elements of type power_by_job_t.
 * NOTE: Job data structure must be locked on function entry
 * NOTE: Call list_delete() to free return value
 * NOTE: This function is currently unused. */
extern List get_job_power(List job_list,
			  struct node_record *node_record_table_ptr)
{
	struct node_record *node_ptr;
	struct job_record *job_ptr;
	ListIterator job_iterator;
	power_by_job_t *power_ptr;
	char jobid_buf[64] = "";
	int i, i_first, i_last;
	uint64_t debug_flag = slurm_get_debug_flags();
	List job_power_list = list_create(_job_power_del);
	time_t now = time(NULL);

	job_iterator = list_iterator_create(job_list);
	while ((job_ptr = (struct job_record *) list_next(job_iterator))) {
		if (!IS_JOB_RUNNING(job_ptr))
			continue;
		power_ptr = xmalloc(sizeof(power_by_job_t));
		power_ptr->job_id = job_ptr->job_id;
		power_ptr->start_time = job_ptr->start_time;
		list_append(job_power_list, power_ptr);
		if (!job_ptr->node_bitmap) {
			error("%s: %s node_bitmap is NULL", __func__,
			      jobid2fmt(job_ptr, jobid_buf, sizeof(jobid_buf)));
			continue;
		}
		i_first = bit_ffs(job_ptr->node_bitmap);
		if (i_first < 0)
			continue;
		i_last = bit_fls(job_ptr->node_bitmap);
		for (i = i_first; i <= i_last; i++) {
			if (!bit_test(job_ptr->node_bitmap, i))
				continue;
			node_ptr = node_record_table_ptr + i;
			if (node_ptr->power) {
				power_ptr->alloc_watts +=
					node_ptr->power->cap_watts;
			}
			if (node_ptr->energy) {
				power_ptr->used_watts +=
					node_ptr->energy->current_watts;
			}
		}
		if (debug_flag & DEBUG_FLAG_POWER) {
			info("%s: %s Age=%ld(sec) AllocWatts=%u UsedWatts=%u",
			     __func__,
			     jobid2fmt(job_ptr, jobid_buf, sizeof(jobid_buf)),
			     (long int) difftime(now, power_ptr->start_time),
			     power_ptr->alloc_watts, power_ptr->used_watts);
		}
	}
	list_iterator_destroy(job_iterator);

	return job_power_list;
}
예제 #9
0
void event_capture_init(event_capture_t* capture, uint64_t freq, uint32_t log2res)
{
    assert(capture);
    assert(freq);

    uint32_t log2 = bit_fls(freq);
    uint32_t quantShift = log2 > log2res ? log2-log2res : 0;

    capture->freq       = freq;
    capture->quantShift = quantShift;
    capture->maxPeriod  = (1ul<<TIME_BITS)<<quantShift;
}
예제 #10
0
/* power_job_reboot - Reboot compute nodes for a job from the head node */
extern int power_job_reboot(struct job_record *job_ptr)
{
	int rc = SLURM_SUCCESS;
	int i, i_first, i_last;
	struct node_record *node_ptr;
	bitstr_t *wake_node_bitmap = NULL;
	time_t now = time(NULL);
	char *nodes, *features = NULL;

	wake_node_bitmap = bit_alloc(node_record_count);
	i_first = bit_ffs(job_ptr->node_bitmap);
	i_last = bit_fls(job_ptr->node_bitmap);
	for (i = i_first; i <= i_last; i++) {
		if (!bit_test(job_ptr->node_bitmap, i))
			continue;
		node_ptr = node_record_table_ptr + i;
		resume_cnt++;
		resume_cnt_f++;
		node_ptr->node_state &= (~NODE_STATE_POWER_SAVE);
		node_ptr->node_state |=   NODE_STATE_POWER_UP;
		node_ptr->node_state |=   NODE_STATE_NO_RESPOND;
		bit_clear(power_node_bitmap, i);
		bit_clear(avail_node_bitmap, i);
		node_ptr->last_response = now + resume_timeout;
		bit_set(wake_node_bitmap,    i);
		bit_set(resume_node_bitmap,  i);
	}

	nodes = bitmap2node_name(wake_node_bitmap);
	if (nodes) {
#if _DEBUG
		info("power_save: reboot nodes %s", nodes);
#else
		verbose("power_save: reboot nodes %s", nodes);
#endif
		if (job_ptr->details && job_ptr->details->features)
			features = xlate_features(job_ptr->details->features);
		_run_prog(resume_prog, nodes, features);
		xfree(features);
	} else {
		error("power_save: bitmap2nodename");
		rc = SLURM_ERROR;
	}
	xfree(nodes);
	FREE_NULL_BITMAP(wake_node_bitmap);
	last_node_update = now;

	return rc;
}
예제 #11
0
파일: dist_tasks.c 프로젝트: SchedMD/slurm
/* Remove any specialized cores from those allocated to the job */
static void _clear_spec_cores(struct job_record *job_ptr,
			      bitstr_t *avail_core_bitmap)
{
	int first_node, last_node, i_node;
	int first_core, last_core, i_core;
	int alloc_node = -1, alloc_core = -1, size;
	job_resources_t *job_res = job_ptr->job_resrcs;
	multi_core_data_t *mc_ptr = NULL;

	if (job_ptr->details && job_ptr->details->mc_ptr)
		mc_ptr = job_ptr->details->mc_ptr;

	size = bit_size(job_res->core_bitmap);
	bit_nset(job_res->core_bitmap, 0, size - 1);

	first_node = bit_ffs(job_res->node_bitmap);
	if (first_node >= 0)
		last_node = bit_fls(job_res->node_bitmap);
	else
		last_node = first_node - 1;

	for (i_node = first_node; i_node <= last_node; i_node++) {
		if (!bit_test(job_res->node_bitmap, i_node))
			continue;
		job_res->cpus[++alloc_node] = 0;
		first_core = cr_get_coremap_offset(i_node);
		last_core  = cr_get_coremap_offset(i_node + 1) - 1;
		for (i_core = first_core; i_core <= last_core; i_core++) {
			alloc_core++;
			if (bit_test(avail_core_bitmap, i_core)) {
				uint16_t tpc = select_node_record[i_node].vpus;
				if (mc_ptr &&
				    (mc_ptr->threads_per_core != NO_VAL16) &&
				    (mc_ptr->threads_per_core < tpc))
					tpc = mc_ptr->threads_per_core;

				job_res->cpus[alloc_node] += tpc;
			} else {
				bit_clear(job_res->core_bitmap, alloc_core);
			}
		}
	}
}
예제 #12
0
파일: node_conf.c 프로젝트: HDOD/slurm
/*
 * bitmap2hostlist - given a bitmap, build a hostlist
 * IN bitmap - bitmap pointer
 * RET pointer to hostlist or NULL on error
 * globals: node_record_table_ptr - pointer to node table
 * NOTE: the caller must xfree the memory at node_list when no longer required
 */
hostlist_t bitmap2hostlist (bitstr_t *bitmap)
{
	int i, first, last;
	hostlist_t hl;

	if (bitmap == NULL)
		return NULL;

	first = bit_ffs(bitmap);
	if (first == -1)
		return NULL;

	last  = bit_fls(bitmap);
	hl = hostlist_create(NULL);
	for (i = first; i <= last; i++) {
		if (bit_test(bitmap, i) == 0)
			continue;
		hostlist_push_host(hl, node_record_table_ptr[i].name);
	}
	return hl;

}
예제 #13
0
static int glnvg__renderCreateTexture(void* uptr, int type, int w, int h, int imageFlags, const unsigned char* data)
{
    GLNVGcontext* gl = (GLNVGcontext*)uptr;
    GLNVGtexture* tex = glnvg__allocTexture(gl);

    if (tex == NULL) return 0;

    const bool formatRGBA8 = type == NVG_TEXTURE_RGBA;
    const bool useMipmaps = (imageFlags & NVG_IMAGE_GENERATE_MIPMAPS) != 0;
    glCreateTextures(GL_TEXTURE_2D, 1, &tex->tex);
    const uint32_t mipCount = useMipmaps ? bit_fls(core::max<uint32_t>(w, h)) : 1;
    glTextureStorage2D(tex->tex, mipCount, formatRGBA8 ? GL_RGBA8 : GL_R8, w, h);
    tex->width = w;
    tex->height = h;
    tex->type = type;


    if (data)
    {
        glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
        glPixelStorei(GL_UNPACK_ROW_LENGTH, tex->width);
        glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
        glPixelStorei(GL_UNPACK_SKIP_ROWS, 0);
        glTextureSubImage2D(tex->tex, 0, 0, 0, w, h, formatRGBA8 ? GL_RGBA : GL_RED, GL_UNSIGNED_BYTE, data);
        glPixelStorei(GL_UNPACK_ALIGNMENT, 4);
        glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
        glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
        glPixelStorei(GL_UNPACK_SKIP_ROWS, 0);
        // The new way to build mipmaps on GLES and GL3
        if (useMipmaps) {
            glGenerateTextureMipmap(tex->tex);
        }
    }

    glTextureParameteri(tex->tex, GL_TEXTURE_MIN_FILTER, useMipmaps ? GL_LINEAR_MIPMAP_LINEAR : GL_LINEAR);
    glTextureParameteri(tex->tex, GL_TEXTURE_MAG_FILTER, GL_LINEAR);

    return tex->id;
}
예제 #14
0
파일: job_test.c 프로젝트: kwangiit/SLURMPP
/* Compute resource usage for the given job on all available resources
 *
 * IN: job_ptr     - pointer to the job requesting resources
 * IN: node_map    - bitmap of available nodes
 * IN/OUT: core_map    - bitmap of available cores
 * IN: cr_node_cnt - total number of nodes in the cluster
 * IN: cr_type     - resource type
 * OUT: cpu_cnt    - number of cpus that can be used by this job
 * IN: test_only   - ignore allocated memory check
 * RET SLURM_SUCCESS index of selected node or -1 if none
 */
static int _get_res_usage(struct job_record *job_ptr, bitstr_t *node_map,
			   bitstr_t *core_map, uint32_t cr_node_cnt,
			   struct node_use_record *node_usage,
			   uint16_t cr_type, uint16_t **cpu_cnt_ptr, 
			   bool test_only)
{
	uint16_t *cpu_cnt;
	uint32_t n;
	int i_first, i_last;
	int node_inx = -1;

	if (cr_node_cnt != node_record_count) {
		error("select/serial: node count inconsistent with slurmctld");
		return SLURM_ERROR;
	}
	if (job_ptr->details && job_ptr->details->req_node_bitmap)
		bit_and(node_map, job_ptr->details->req_node_bitmap);
	cpu_cnt = xmalloc(cr_node_cnt * sizeof(uint16_t));
	i_first = bit_ffs(node_map);
	if (i_first >= 0)
		i_last  = bit_fls(node_map);
	else
		i_last = -2;
	for (n = i_first; n <= i_last; n++) {
		if (!bit_test(node_map, n))
			continue;
		cpu_cnt[n] = _can_job_run_on_node(job_ptr, core_map, n,
						  node_usage, cr_type,
						  test_only);
		if (cpu_cnt[n]) {
			bit_nclear(node_map, 0, (node_record_count - 1));
			bit_set(node_map, n);
			node_inx = n;
			break;	/* select/serial: only need one node */
		}
	}
	*cpu_cnt_ptr = cpu_cnt;
	return node_inx;
}
예제 #15
0
파일: gang.c 프로젝트: corburn/slurm
/* a helper function for _add_job_to_active when GS_SOCKET
 * a job has just been added to p_ptr->active_resmap, so set all cores of
 * each used socket to avoid activating another job on the same socket */
static void _fill_sockets(bitstr_t *job_nodemap, struct gs_part *p_ptr)
{
	uint32_t c, i;
	int n, first_bit, last_bit;

	if (!job_nodemap || !p_ptr || !p_ptr->active_resmap)
		return;
	first_bit = bit_ffs(job_nodemap);
	last_bit  = bit_fls(job_nodemap);
	if ((first_bit < 0) || (last_bit < 0))
		fatal("gang: _afill_sockets: nodeless job?");

	for (c = 0, n = 0; n < first_bit; n++) {
		c += _get_phys_bit_cnt(n);
	}
	for (n = first_bit; n <= last_bit; n++) {
		uint16_t s, socks, cps, cores_per_node;
		cores_per_node = _get_phys_bit_cnt(n);
		if (bit_test(job_nodemap, n) == 0) {
			c += cores_per_node;
			continue;
		}
		socks = _get_socket_cnt(n);
		cps = cores_per_node / socks;
		for (s = 0; s < socks; s++) {
			for (i = c; i < c+cps; i++) {
				if (bit_test(p_ptr->active_resmap, i))
					break;
			}
			if (i < c+cps) {
				/* set all bits on this used socket */
				bit_nset(p_ptr->active_resmap, c, c+cps-1);
			}
			c += cps;
		}
	}
}
예제 #16
0
/**
 * do_basil_reserve - create a BASIL reservation.
 * IN job_ptr - pointer to job which has just been allocated resources
 * RET 0 or error code, job will abort or be requeued on failure
 */
extern int do_basil_reserve(struct job_record *job_ptr)
{
	struct nodespec *ns_head = NULL;
	uint16_t mppwidth = 0, mppdepth, mppnppn;
	uint32_t mppmem = 0, node_min_mem = 0;
	uint32_t resv_id;
	int i, first_bit, last_bit;
	hostlist_t hl;
	long rc;
	char *user, batch_id[16];

	if (!job_ptr->job_resrcs || job_ptr->job_resrcs->nhosts == 0)
		return SLURM_SUCCESS;

	debug3("job #%u: %u nodes = %s, cpus=%u" , job_ptr->job_id,
		job_ptr->job_resrcs->nhosts,
		job_ptr->job_resrcs->nodes,
		job_ptr->job_resrcs->ncpus
	);

	if (job_ptr->job_resrcs->node_bitmap == NULL) {
		error("job %u node_bitmap not set", job_ptr->job_id);
		return SLURM_SUCCESS;
	}

	first_bit = bit_ffs(job_ptr->job_resrcs->node_bitmap);
	last_bit  = bit_fls(job_ptr->job_resrcs->node_bitmap);
	if (first_bit == -1 || last_bit == -1)
		return SLURM_SUCCESS;		/* no nodes allocated */

	mppdepth = MAX(1, job_ptr->details->cpus_per_task);
	mppnppn  = job_ptr->details->ntasks_per_node;

	/* mppmem */
	if (job_ptr->details->pn_min_memory & MEM_PER_CPU) {
		/* Only honour --mem-per-cpu if --ntasks has been given */
		if (job_ptr->details->num_tasks)
			mppmem = job_ptr->details->pn_min_memory & ~MEM_PER_CPU;
	} else if (job_ptr->details->pn_min_memory) {
		node_min_mem = job_ptr->details->pn_min_memory;
	}

	hl = hostlist_create("");
	if (hl == NULL)
		fatal("hostlist_create: malloc error");

	for (i = first_bit; i <= last_bit; i++) {
		struct node_record *node_ptr = node_record_table_ptr + i;
		uint32_t basil_node_id;

		if (!bit_test(job_ptr->job_resrcs->node_bitmap, i))
			continue;

		if (!node_ptr->name || node_ptr->name[0] == '\0')
			continue;	/* bad node */

		if (sscanf(node_ptr->name, "nid%05u", &basil_node_id) != 1)
			fatal("can not read basil_node_id from %s", node_ptr->name);

		if (ns_add_node(&ns_head, basil_node_id) != 0) {
			error("can not add node %s (nid%05u)", node_ptr->name,
			      basil_node_id);
			free_nodespec(ns_head);
			return SLURM_ERROR;
		}

		if (node_min_mem) {
			uint32_t node_cpus, node_mem;

			if (slurmctld_conf.fast_schedule) {
				node_cpus = node_ptr->config_ptr->cpus;
				node_mem  = node_ptr->config_ptr->real_memory;
			} else {
				node_cpus = node_ptr->cpus;
				node_mem  = node_ptr->real_memory;
			}
			/*
			 * ALPS 'Processing Elements per Node' value (aprun -N),
			 * which in slurm is --ntasks-per-node and 'mppnppn' in
			 * PBS: if --ntasks is specified, default to the number
			 * of cores per node (also the default for 'aprun -N').
			 */
			node_mem /= mppnppn ? mppnppn : node_cpus;

			mppmem = node_min_mem = MIN(node_mem, node_min_mem);
		}
	}

	/* mppwidth */
	for (i = 0; i < job_ptr->job_resrcs->nhosts; i++) {
		uint16_t node_tasks = job_ptr->job_resrcs->cpus[i] / mppdepth;

		if (mppnppn && mppnppn < node_tasks)
			node_tasks = mppnppn;
		mppwidth += node_tasks;
	}

	snprintf(batch_id, sizeof(batch_id), "%u", job_ptr->job_id);
	user = uid_to_string(job_ptr->user_id);
	rc   = basil_reserve(user, batch_id, mppwidth,
			     mppdepth, mppnppn, mppmem, ns_head);
	xfree(user);
	if (rc <= 0) {
		/* errno value will be resolved by select_g_job_begin() */
		errno = is_transient_error(rc) ? EAGAIN : ECONNABORTED;
		return SLURM_ERROR;
	}

	resv_id	= rc;
	if (_set_select_jobinfo(job_ptr->select_jobinfo->data,
			SELECT_JOBDATA_RESV_ID, &resv_id) != SLURM_SUCCESS) {
		/*
		 * This is a fatal error since it means we will not be able to
		 * confirm the reservation; no step will be able to run in it.
		 */
		error("job %u: can not set resId %u", job_ptr->job_id, resv_id);
		basil_release(resv_id);
		return SLURM_ERROR;
	}

	info("ALPS RESERVATION #%u, JobId %u: BASIL -n %d -N %d -d %d -m %d",
	     resv_id, job_ptr->job_id, mppwidth, mppnppn, mppdepth, mppmem);

	return SLURM_SUCCESS;
}
예제 #17
0
파일: grid.c 프로젝트: jtfrey/slurm
static int _grid_table_by_switch(button_processor_t *button_processor,
				 List node_list)
{
	int rc = SLURM_SUCCESS;
	int inx = 0, ii = 0;
	switch_record_bitmaps_t *sw_nodes_bitmaps_ptr = g_switch_nodes_maps;
#if TOPO_DEBUG
	/* engage if want original display below switched */
	ListIterator itr = list_iterator_create(node_list);
	sview_node_info_t *sview_node_info_ptr = NULL;
#endif
	button_processor->inx = &inx;
	for (ii=0; ii<g_topo_info_msg_ptr->record_count;
	     ii++, sw_nodes_bitmaps_ptr++) {
		int j = 0, first, last;
		if (g_topo_info_msg_ptr->topo_array[ii].level)
			continue;
		first = bit_ffs(sw_nodes_bitmaps_ptr->node_bitmap);
		if (first == -1)
			continue;
		last = bit_fls(sw_nodes_bitmaps_ptr->node_bitmap);
		button_processor->inx = &j;
		button_processor->force_row_break = false;
		for (j = first; j <= last; j++) {
			if (TOPO_DEBUG)
				g_print("allocated node = %s button# %d\n",
					g_node_info_ptr->node_array[j].name,
					j);
			if (!bit_test(sw_nodes_bitmaps_ptr->node_bitmap, j))
				continue;
			/* if (!working_sview_config.show_hidden) { */
			/* 	if (!check_part_includes_node(j)) */
			/* 		continue; */
			/* } */
			if (j == last)
				button_processor->force_row_break = true;
			if ((rc = _add_button_to_list(
				     &g_node_info_ptr->node_array[j],
				     button_processor)) != SLURM_SUCCESS)
				break;
			button_processor->force_row_break = false;
		}
		rc = _add_button_to_list(NULL, button_processor);
	}

#if TOPO_DEBUG
	/* engage this if want original display below
	 * switched grid */
	 button_processor->inx = &inx;
	 while ((sview_node_info_ptr = list_next(itr))) {
		 if ((rc = _add_button_to_list(
				sview_node_info_ptr->node_ptr,
	 			button_processor)) != SLURM_SUCCESS)
			 break;
	 	inx++;
	 }
	 list_iterator_destroy(itr);
#endif

	/* This is needed to get the correct width of the grid window.
	 * If it is not given then we get a really narrow window. */
	gtk_table_set_row_spacing(button_processor->table,
				  (*button_processor->coord_y)?
				  ((*button_processor->coord_y)-1):0, 1);

	return rc;

}
예제 #18
0
파일: power_save.c 프로젝트: edsw/slurm
/* power_job_reboot - Reboot compute nodes for a job from the head node */
extern int power_job_reboot(struct job_record *job_ptr)
{
	int rc = SLURM_SUCCESS;
	int i, i_first, i_last;
	struct node_record *node_ptr;
	bitstr_t *boot_node_bitmap = NULL;
	time_t now = time(NULL);
	char *nodes, *features = NULL;
	pid_t pid;

	boot_node_bitmap = node_features_reboot(job_ptr);
	if (boot_node_bitmap == NULL)
		return SLURM_SUCCESS;

	i_first = bit_ffs(boot_node_bitmap);
	if (i_first >= 0)
		i_last = bit_fls(boot_node_bitmap);
	else
		i_last = i_first - 1;
	for (i = i_first; i <= i_last; i++) {
		if (!bit_test(boot_node_bitmap, i))
			continue;
		node_ptr = node_record_table_ptr + i;
		resume_cnt++;
		resume_cnt_f++;
		node_ptr->node_state &= (~NODE_STATE_POWER_SAVE);
		node_ptr->node_state |=   NODE_STATE_POWER_UP;
		node_ptr->node_state |=   NODE_STATE_NO_RESPOND;
		bit_clear(power_node_bitmap, i);
		bit_clear(avail_node_bitmap, i);
		node_ptr->last_response = now + resume_timeout;
		bit_set(resume_node_bitmap,  i);
	}

	nodes = bitmap2node_name(boot_node_bitmap);
	if (nodes) {
		job_ptr->job_state |= JOB_CONFIGURING;
		job_ptr->wait_all_nodes = 1;
		if (job_ptr->details && job_ptr->details->features &&
		    node_features_g_user_update(job_ptr->user_id)) {
			features = node_features_g_job_xlate(
					job_ptr->details->features);
		}
		pid = _run_prog(resume_prog, nodes, features);
#if _DEBUG
		info("power_save: pid %d reboot nodes %s features %s",
		     (int) pid, nodes, features);
#else
		verbose("power_save: pid %d reboot nodes %s features %s",
			(int) pid, nodes, features);
#endif
		xfree(features);
	} else {
		error("power_save: bitmap2nodename");
		rc = SLURM_ERROR;
	}
	xfree(nodes);
	FREE_NULL_BITMAP(boot_node_bitmap);
	last_node_update = now;

	return rc;
}
예제 #19
0
/**
 * do_basil_reserve - create a BASIL reservation.
 * IN job_ptr - pointer to job which has just been allocated resources
 * RET 0 or error code, job will abort or be requeued on failure
 */
extern int do_basil_reserve(struct job_record *job_ptr)
{
    struct nodespec *ns_head = NULL;
    uint16_t mppwidth = 0, mppdepth, mppnppn;
    /* mppmem must be at least 1 for gang scheduling to work so
     * if you are wondering why gang scheduling isn't working you
     * should check your slurm.conf for DefMemPerNode */
    uint32_t mppmem = 0, node_min_mem = 0;
    uint32_t resv_id;
    int i, first_bit, last_bit;
    long rc;
    char *user, batch_id[16];
    struct basil_accel_param* bap;

    if (!job_ptr->job_resrcs || job_ptr->job_resrcs->nhosts == 0)
        return SLURM_SUCCESS;

    debug3("job #%u: %u nodes = %s, cpus=%u" , job_ptr->job_id,
           job_ptr->job_resrcs->nhosts,
           job_ptr->job_resrcs->nodes,
           job_ptr->job_resrcs->ncpus
          );

    if (job_ptr->job_resrcs->node_bitmap == NULL) {
        error("job %u node_bitmap not set", job_ptr->job_id);
        return SLURM_SUCCESS;
    }

    first_bit = bit_ffs(job_ptr->job_resrcs->node_bitmap);
    last_bit  = bit_fls(job_ptr->job_resrcs->node_bitmap);
    if (first_bit == -1 || last_bit == -1)
        return SLURM_SUCCESS;		/* no nodes allocated */

    mppdepth = MAX(1, job_ptr->details->cpus_per_task);
    mppnppn  = job_ptr->details->ntasks_per_node;

    /* mppmem */
    if (job_ptr->details->pn_min_memory & MEM_PER_CPU) {
        /* Only honour --mem-per-cpu if --ntasks has been given */
        if (job_ptr->details->num_tasks)
            mppmem = job_ptr->details->pn_min_memory & ~MEM_PER_CPU;
    } else if (job_ptr->details->pn_min_memory) {
        node_min_mem = job_ptr->details->pn_min_memory;
    }

    for (i = first_bit; i <= last_bit; i++) {
        struct node_record *node_ptr = node_record_table_ptr + i;
        uint32_t basil_node_id;

        if (!bit_test(job_ptr->job_resrcs->node_bitmap, i))
            continue;

        if (!node_ptr->name || node_ptr->name[0] == '\0')
            continue;	/* bad node */

        if (sscanf(node_ptr->name, "nid%05u", &basil_node_id) != 1)
            fatal("can not read basil_node_id from %s",
                  node_ptr->name);

        if (ns_add_node(&ns_head, basil_node_id, false) != 0) {
            error("can not add node %s (nid%05u)", node_ptr->name,
                  basil_node_id);
            free_nodespec(ns_head);
            return SLURM_ERROR;
        }

        if (node_min_mem) {
            uint32_t node_cpus, node_mem;
            int32_t tmp_mppmem;

            if (slurmctld_conf.fast_schedule) {
                node_cpus = node_ptr->config_ptr->cpus;
                node_mem  = node_ptr->config_ptr->real_memory;
            } else {
                node_cpus = node_ptr->cpus;
                node_mem  = node_ptr->real_memory;
            }
            /*
             * ALPS 'Processing Elements per Node' value (aprun -N),
             * which in slurm is --ntasks-per-node and 'mppnppn' in
             * PBS: if --ntasks is specified, default to the number
             * of cores per node (also the default for 'aprun -N').
             * On a heterogeneous system the nodes aren't
             * always the same so keep track of the lowest
             * mppmem and use it as the level for all
             * nodes (mppmem is 0 when coming in).
             */
            node_mem /= mppnppn ? mppnppn : node_cpus;
            tmp_mppmem = node_min_mem = MIN(node_mem, node_min_mem);

            /* If less than or equal to 0 make sure you
               have 1 at least since 0 means give all the
               memory to the job.
            */
            if (tmp_mppmem <= 0)
                tmp_mppmem = 1;

            if (mppmem)
                mppmem = MIN(mppmem, tmp_mppmem);
            else
                mppmem = tmp_mppmem;
        }
    }

    /* mppwidth */
    for (i = 0; i < job_ptr->job_resrcs->nhosts; i++) {
        uint16_t node_tasks = job_ptr->job_resrcs->cpus[i] / mppdepth;

        if (mppnppn && mppnppn < node_tasks)
            node_tasks = mppnppn;
        mppwidth += node_tasks;
    }

    snprintf(batch_id, sizeof(batch_id), "%u", job_ptr->job_id);
    user = uid_to_string(job_ptr->user_id);

    if (job_ptr->gres_list)
        bap = build_accel_param(job_ptr);
    else
        bap = NULL;

    rc   = basil_reserve(user, batch_id, mppwidth, mppdepth, mppnppn,
                         mppmem, ns_head, bap);
    xfree(user);
    if (rc <= 0) {
        /* errno value will be resolved by select_g_job_begin() */
        errno = is_transient_error(rc) ? EAGAIN : ECONNABORTED;
        return SLURM_ERROR;
    }

    resv_id	= rc;
    if (_set_select_jobinfo(job_ptr->select_jobinfo->data,
                            SELECT_JOBDATA_RESV_ID, &resv_id) != SLURM_SUCCESS) {
        /*
         * This is a fatal error since it means we will not be able to
         * confirm the reservation; no step will be able to run in it.
         */
        error("job %u: can not set resId %u", job_ptr->job_id, resv_id);
        basil_release(resv_id);
        return SLURM_ERROR;
    }
    if (mppmem)
        job_ptr->details->pn_min_memory = mppmem | MEM_PER_CPU;

    info("ALPS RESERVATION #%u, JobId %u: BASIL -n %d -N %d -d %d -m %d",
         resv_id, job_ptr->job_id, mppwidth, mppnppn, mppdepth, mppmem);

    return SLURM_SUCCESS;
}
예제 #20
0
int
main(int argc, char *argv[])
{
	note("Testing static decl");
	{
		bitstr_t bit_decl(bs, 65);
		/*bitstr_t *bsp = bs;*/

		bit_set(bs,9);
		bit_set(bs,14);
		TEST(bit_test(bs,9), "bit 9 set"); 
		TEST(!bit_test(bs,12), "bit 12 not set");
		TEST(bit_test(bs,14), "bit 14 set" );
		/*bit_free(bsp);*/	/* triggers TEST in bit_free - OK */
	}
	note("Testing basic vixie functions");
	{
		bitstr_t *bs = bit_alloc(16), *bs2;


		/*bit_set(bs, 42);*/ 	/* triggers TEST in bit_set - OK */
		bit_set(bs,9);
		bit_set(bs,14);
		TEST(bit_test(bs,9), "bit 9 set"); 
		TEST(!bit_test(bs,12), "bit 12 not set" );
		TEST(bit_test(bs,14), "bit 14 set");

		bs2 = bit_copy(bs);
		bit_fill_gaps(bs2);
		TEST(bit_ffs(bs2) == 9, "first bit set = 9 ");
		TEST(bit_fls(bs2) == 14, "last bit set = 14");
		TEST(bit_set_count(bs2) == 6, "bitstring");
		TEST(bit_test(bs2,12), "bitstring");
		TEST(bit_super_set(bs,bs2) == 1, "bitstring");
		TEST(bit_super_set(bs2,bs) == 0, "bitstring");

		bit_clear(bs,14);
		TEST(!bit_test(bs,14), "bitstring");

		bit_nclear(bs,9,14);
		TEST(!bit_test(bs,9), "bitstring");
		TEST(!bit_test(bs,12), "bitstring");
		TEST(!bit_test(bs,14), "bitstring");

		bit_nset(bs,9,14);
		TEST(bit_test(bs,9), "bitstring");
		TEST(bit_test(bs,12), "bitstring");
		TEST(bit_test(bs,14), "bitstring");

		TEST(bit_ffs(bs) == 9, "ffs");
		TEST(bit_ffc(bs) == 0, "ffc");
		bit_nset(bs,0,8);
		TEST(bit_ffc(bs) == 15, "ffc");

		bit_free(bs);
		/*bit_set(bs,9); */	/* triggers TEST in bit_set - OK */
	}
	note("Testing and/or/not");
	{
		bitstr_t *bs1 = bit_alloc(128);
		bitstr_t *bs2 = bit_alloc(128);

		bit_set(bs1, 100);
		bit_set(bs1, 104);
		bit_set(bs2, 100);

		bit_and(bs1, bs2);
		TEST(bit_test(bs1, 100), "and");
		TEST(!bit_test(bs1, 104), "and");

		bit_set(bs2, 110);
		bit_set(bs2, 111);
		bit_set(bs2, 112);
		bit_or(bs1, bs2);
		TEST(bit_test(bs1, 100), "or");
		TEST(bit_test(bs1, 110), "or");
		TEST(bit_test(bs1, 111), "or");
		TEST(bit_test(bs1, 112), "or");

		bit_not(bs1);
		TEST(!bit_test(bs1, 100), "not");
		TEST(bit_test(bs1, 12), "not");

		bit_free(bs1);
		bit_free(bs2);
	}

	note("testing bit selection");
	{
		bitstr_t *bs1 = bit_alloc(128), *bs2;
		bit_set(bs1, 21);
		bit_set(bs1, 100);
		bit_fill_gaps(bs1);
		bs2 = bit_pick_cnt(bs1,20);

		if (bs2) {
			TEST(bit_set_count(bs2) == 20, "pick");
			TEST(bit_ffs(bs2) == 21, "pick");
			TEST(bit_fls(bs2) == 40, "pick");
			bit_free(bs2);
		}
		else
			TEST(0, "alloc fail");

		bit_free(bs1);
	}
	note("Testing realloc");
	{
		bitstr_t *bs = bit_alloc(1);

		TEST(bit_ffs(bs) == -1, "bitstring");
		bit_set(bs,0);
		/*bit_set(bs, 1000);*/	/* triggers TEST in bit_set - OK */
		bs = bit_realloc(bs,1048576);
		bit_set(bs,1000);
		bit_set(bs,1048575);
		TEST(bit_test(bs, 0), "bitstring");
		TEST(bit_test(bs, 1000), "bitstring");
		TEST(bit_test(bs, 1048575), "bitstring");
		TEST(bit_set_count(bs) == 3, "bitstring");
		bit_clear(bs,0);
		bit_clear(bs,1000);
		TEST(bit_set_count(bs) == 1, "bitstring");
		TEST(bit_ffs(bs) == 1048575, "bitstring");
		bit_free(bs);
	}
	note("Testing bit_fmt");
	{
		char tmpstr[1024];
		bitstr_t *bs = bit_alloc(1024);

		TEST(!strcmp(bit_fmt(tmpstr,sizeof(tmpstr),bs), ""), "bitstring");
		bit_set(bs,42);
		TEST(!strcmp(bit_fmt(tmpstr,sizeof(tmpstr),bs), "42"), "bitstring");
		bit_set(bs,102);
		TEST(!strcmp(bit_fmt(tmpstr,sizeof(tmpstr),bs), "42,102"), "bitstring");
		bit_nset(bs,9,14);
		TEST(!strcmp(bit_fmt(tmpstr,sizeof(tmpstr), bs), 
					"9-14,42,102"), "bitstring");
	}

	note("Testing bit_nffc/bit_nffs");
	{
		bitstr_t *bs = bit_alloc(1024);

		bit_set(bs, 2);
		bit_set(bs, 6);
		bit_set(bs, 7);
		bit_nset(bs,12,1018); 

		TEST(bit_nffc(bs, 2) == 0, "bitstring");
		TEST(bit_nffc(bs, 3) == 3, "bitstring");
		TEST(bit_nffc(bs, 4) == 8, "bitstring");
		TEST(bit_nffc(bs, 5) == 1019, "bitstring");
		TEST(bit_nffc(bs, 6) == -1, "bitstring");

		TEST(bit_nffs(bs, 1) == 2, "bitstring");
		TEST(bit_nffs(bs, 2) == 6, "bitstring");
		TEST(bit_nffs(bs, 100) == 12, "bitstring");
		TEST(bit_nffs(bs, 1023) == -1, "bitstring");

		bit_free(bs);
	}

	note("Testing bit_unfmt");
	{
		bitstr_t *bs = bit_alloc(1024);
		bitstr_t *bs2 = bit_alloc(1024);
		char tmpstr[4096];

		bit_set(bs,1);
		bit_set(bs,3);
		bit_set(bs,30);
		bit_nset(bs,42,64);
		bit_nset(bs,97,1000);

		bit_fmt(tmpstr, sizeof(tmpstr), bs);
		TEST(bit_unfmt(bs2, tmpstr) != -1, "bitstring");
		TEST(bit_equal(bs, bs2), "bitstring");
	}

	totals();
	return failed;
}
예제 #21
0
파일: job_info.c 프로젝트: fafik23/slurm
/*
 * slurm_sprint_job_info - output information about a specific Slurm
 *	job based upon message as loaded using slurm_load_jobs
 * IN job_ptr - an individual job information record pointer
 * IN one_liner - print as a single line if true
 * RET out - char * containing formatted output (must be freed after call)
 *           NULL is returned on failure.
 */
extern char *
slurm_sprint_job_info ( job_info_t * job_ptr, int one_liner )
{
	int i, j, k;
	char time_str[32], *group_name, *user_name;
	char *gres_last = "", tmp1[128], tmp2[128];
	char *tmp6_ptr;
	char tmp_line[1024 * 128];
	char tmp_path[MAXPATHLEN];
	char *ionodes = NULL;
	uint16_t exit_status = 0, term_sig = 0;
	job_resources_t *job_resrcs = job_ptr->job_resrcs;
	char *out = NULL;
	time_t run_time;
	uint32_t min_nodes, max_nodes = 0;
	char *nodelist = "NodeList";
	bitstr_t *cpu_bitmap;
	char *host;
	int sock_inx, sock_reps, last;
	int abs_node_inx, rel_node_inx;
	int64_t nice;
	int bit_inx, bit_reps;
	uint64_t *last_mem_alloc_ptr = NULL;
	uint64_t last_mem_alloc = NO_VAL64;
	char *last_hosts;
	hostlist_t hl, hl_last;
	char select_buf[122];
	uint32_t cluster_flags = slurmdb_setup_cluster_flags();
	uint32_t threads;
	char *line_end = (one_liner) ? " " : "\n   ";

	if (cluster_flags & CLUSTER_FLAG_BG) {
		nodelist = "MidplaneList";
		select_g_select_jobinfo_get(job_ptr->select_jobinfo,
					    SELECT_JOBDATA_IONODES,
					    &ionodes);
	}

	/****** Line 1 ******/
	xstrfmtcat(out, "JobId=%u ", job_ptr->job_id);

	if (job_ptr->array_job_id) {
		if (job_ptr->array_task_str) {
			xstrfmtcat(out, "ArrayJobId=%u ArrayTaskId=%s ",
				   job_ptr->array_job_id,
				   job_ptr->array_task_str);
		} else {
			xstrfmtcat(out, "ArrayJobId=%u ArrayTaskId=%u ",
				   job_ptr->array_job_id,
				   job_ptr->array_task_id);
		}
	}
	xstrfmtcat(out, "JobName=%s", job_ptr->name);
	xstrcat(out, line_end);

	/****** Line 2 ******/
	user_name = uid_to_string((uid_t) job_ptr->user_id);
	group_name = gid_to_string((gid_t) job_ptr->group_id);
	xstrfmtcat(out, "UserId=%s(%u) GroupId=%s(%u) MCS_label=%s",
		   user_name, job_ptr->user_id, group_name, job_ptr->group_id,
		   (job_ptr->mcs_label==NULL) ? "N/A" : job_ptr->mcs_label);
	xfree(user_name);
	xfree(group_name);
	xstrcat(out, line_end);

	/****** Line 3 ******/
	nice = ((int64_t)job_ptr->nice) - NICE_OFFSET;
	xstrfmtcat(out, "Priority=%u Nice=%"PRIi64" Account=%s QOS=%s",
		   job_ptr->priority, nice, job_ptr->account, job_ptr->qos);
	if (slurm_get_track_wckey())
		xstrfmtcat(out, " WCKey=%s", job_ptr->wckey);
	xstrcat(out, line_end);

	/****** Line 4 ******/
	xstrfmtcat(out, "JobState=%s ", job_state_string(job_ptr->job_state));

	if (job_ptr->state_desc) {
		/* Replace white space with underscore for easier parsing */
		for (j=0; job_ptr->state_desc[j]; j++) {
			if (isspace((int)job_ptr->state_desc[j]))
				job_ptr->state_desc[j] = '_';
		}
		xstrfmtcat(out, "Reason=%s ", job_ptr->state_desc);
	} else
		xstrfmtcat(out, "Reason=%s ", job_reason_string(job_ptr->state_reason));

	xstrfmtcat(out, "Dependency=%s", job_ptr->dependency);
	xstrcat(out, line_end);

	/****** Line 5 ******/
	xstrfmtcat(out, "Requeue=%u Restarts=%u BatchFlag=%u Reboot=%u ",
		 job_ptr->requeue, job_ptr->restart_cnt, job_ptr->batch_flag,
		 job_ptr->reboot);
	if (WIFSIGNALED(job_ptr->exit_code))
		term_sig = WTERMSIG(job_ptr->exit_code);
	exit_status = WEXITSTATUS(job_ptr->exit_code);
	xstrfmtcat(out, "ExitCode=%u:%u", exit_status, term_sig);
	xstrcat(out, line_end);

	/****** Line 5a (optional) ******/
	if (job_ptr->show_flags & SHOW_DETAIL) {
		if (WIFSIGNALED(job_ptr->derived_ec))
			term_sig = WTERMSIG(job_ptr->derived_ec);
		else
			term_sig = 0;
		exit_status = WEXITSTATUS(job_ptr->derived_ec);
		xstrfmtcat(out, "DerivedExitCode=%u:%u", exit_status, term_sig);
		xstrcat(out, line_end);
	}

	/****** Line 6 ******/
	if (IS_JOB_PENDING(job_ptr))
		run_time = 0;
	else if (IS_JOB_SUSPENDED(job_ptr))
		run_time = job_ptr->pre_sus_time;
	else {
		time_t end_time;
		if (IS_JOB_RUNNING(job_ptr) || (job_ptr->end_time == 0))
			end_time = time(NULL);
		else
			end_time = job_ptr->end_time;
		if (job_ptr->suspend_time) {
			run_time = (time_t)
				(difftime(end_time, job_ptr->suspend_time)
				 + job_ptr->pre_sus_time);
		} else
			run_time = (time_t)
				difftime(end_time, job_ptr->start_time);
	}
	secs2time_str(run_time, time_str, sizeof(time_str));
	xstrfmtcat(out, "RunTime=%s ", time_str);

	if (job_ptr->time_limit == NO_VAL)
		xstrcat(out, "TimeLimit=Partition_Limit ");
	else {
		mins2time_str(job_ptr->time_limit, time_str, sizeof(time_str));
		xstrfmtcat(out, "TimeLimit=%s ", time_str);
	}

	if (job_ptr->time_min == 0)
		xstrcat(out, "TimeMin=N/A");
	else {
		mins2time_str(job_ptr->time_min, time_str, sizeof(time_str));
		xstrfmtcat(out, "TimeMin=%s", time_str);
	}
	xstrcat(out, line_end);

	/****** Line 7 ******/
	slurm_make_time_str(&job_ptr->submit_time, time_str, sizeof(time_str));
	xstrfmtcat(out, "SubmitTime=%s ", time_str);

	slurm_make_time_str(&job_ptr->eligible_time, time_str, sizeof(time_str));
	xstrfmtcat(out, "EligibleTime=%s", time_str);

	xstrcat(out, line_end);

	/****** Line 8 (optional) ******/
	if (job_ptr->resize_time) {
		slurm_make_time_str(&job_ptr->resize_time, time_str, sizeof(time_str));
		xstrfmtcat(out, "ResizeTime=%s", time_str);
		xstrcat(out, line_end);
	}

	/****** Line 9 ******/
	slurm_make_time_str(&job_ptr->start_time, time_str, sizeof(time_str));
	xstrfmtcat(out, "StartTime=%s ", time_str);

	if ((job_ptr->time_limit == INFINITE) &&
	    (job_ptr->end_time > time(NULL)))
		xstrcat(out, "EndTime=Unknown ");
	else {
		slurm_make_time_str(&job_ptr->end_time, time_str, sizeof(time_str));
		xstrfmtcat(out, "EndTime=%s ", time_str);
	}

	if (job_ptr->deadline) {
		slurm_make_time_str(&job_ptr->deadline, time_str, sizeof(time_str));
		xstrfmtcat(out, "Deadline=%s", time_str);
	} else {
		xstrcat(out, "Deadline=N/A");
	}

	xstrcat(out, line_end);

	/****** Line 10 ******/
	if (job_ptr->preempt_time == 0)
		xstrcat(out, "PreemptTime=None ");
	else {
		slurm_make_time_str(&job_ptr->preempt_time, time_str, sizeof(time_str));
		xstrfmtcat(out, "PreemptTime=%s ", time_str);
	}

	if (job_ptr->suspend_time) {
		slurm_make_time_str(&job_ptr->suspend_time, time_str, sizeof(time_str));
		xstrfmtcat(out, "SuspendTime=%s ", time_str);
	} else
		xstrcat(out, "SuspendTime=None ");

	xstrfmtcat(out, "SecsPreSuspend=%ld", (long int)job_ptr->pre_sus_time);
	xstrcat(out, line_end);

	/****** Line 11 ******/
	xstrfmtcat(out, "Partition=%s AllocNode:Sid=%s:%u",
		   job_ptr->partition, job_ptr->alloc_node, job_ptr->alloc_sid);
	xstrcat(out, line_end);

	/****** Line 12 ******/
	xstrfmtcat(out, "Req%s=%s Exc%s=%s", nodelist, job_ptr->req_nodes,
		   nodelist, job_ptr->exc_nodes);
	xstrcat(out, line_end);

	/****** Line 13 ******/
	xstrfmtcat(out, "%s=%s", nodelist, job_ptr->nodes);
	if (job_ptr->nodes && ionodes) {
		xstrfmtcat(out, "[%s]", ionodes);
		xfree(ionodes);
	}
	if (job_ptr->sched_nodes)
		xstrfmtcat(out, " Sched%s=%s", nodelist, job_ptr->sched_nodes);

	xstrcat(out, line_end);

	/****** Line 14 (optional) ******/
	if (job_ptr->batch_host) {
		xstrfmtcat(out, "BatchHost=%s", job_ptr->batch_host);
		xstrcat(out, line_end);
	}

	/****** Line 14a (optional) ******/
	if (job_ptr->fed_siblings) {
		xstrfmtcat(out, "FedOrigin=%s FedSiblings=%s",
			   job_ptr->fed_origin_str, job_ptr->fed_siblings_str);
		xstrcat(out, line_end);
	}

	/****** Line 15 ******/
	if (cluster_flags & CLUSTER_FLAG_BG) {
		select_g_select_jobinfo_get(job_ptr->select_jobinfo,
					    SELECT_JOBDATA_NODE_CNT,
					    &min_nodes);
		if ((min_nodes == 0) || (min_nodes == NO_VAL)) {
			min_nodes = job_ptr->num_nodes;
			max_nodes = job_ptr->max_nodes;
		} else if (job_ptr->max_nodes)
			max_nodes = min_nodes;
	} else if (IS_JOB_PENDING(job_ptr)) {
		min_nodes = job_ptr->num_nodes;
		max_nodes = job_ptr->max_nodes;
		if (max_nodes && (max_nodes < min_nodes))
			min_nodes = max_nodes;
	} else {
		min_nodes = job_ptr->num_nodes;
		max_nodes = 0;
	}

	_sprint_range(tmp_line, sizeof(tmp_line), min_nodes, max_nodes);
	xstrfmtcat(out, "NumNodes=%s ", tmp_line);
	_sprint_range(tmp_line, sizeof(tmp_line), job_ptr->num_cpus, job_ptr->max_cpus);
	xstrfmtcat(out, "NumCPUs=%s ", tmp_line);

	xstrfmtcat(out, "NumTasks=%u ", job_ptr->num_tasks);
	xstrfmtcat(out, "CPUs/Task=%u ", job_ptr->cpus_per_task);

	if (job_ptr->boards_per_node == (uint16_t) NO_VAL)
		xstrcat(out, "ReqB:S:C:T=*:");
	else
		xstrfmtcat(out, "ReqB:S:C:T=%u:", job_ptr->boards_per_node);

	if (job_ptr->sockets_per_board == (uint16_t) NO_VAL)
		xstrcat(out, "*:");
	else
		xstrfmtcat(out, "%u:", job_ptr->sockets_per_board);

	if (job_ptr->cores_per_socket == (uint16_t) NO_VAL)
		xstrcat(out, "*:");
	else
		xstrfmtcat(out, "%u:", job_ptr->cores_per_socket);

	if (job_ptr->threads_per_core == (uint16_t) NO_VAL)
		xstrcat(out, "*");
	else
		xstrfmtcat(out, "%u", job_ptr->threads_per_core);

	xstrcat(out, line_end);

	/****** Line 16 ******/
	/* Tres should already of been converted at this point from simple */
	xstrfmtcat(out, "TRES=%s",
		   job_ptr->tres_alloc_str ? job_ptr->tres_alloc_str
					   : job_ptr->tres_req_str);
	xstrcat(out, line_end);

	/****** Line 17 ******/
	if (job_ptr->sockets_per_node == (uint16_t) NO_VAL)
		xstrcat(out, "Socks/Node=* ");
	else
		xstrfmtcat(out, "Socks/Node=%u ", job_ptr->sockets_per_node);

	if (job_ptr->ntasks_per_node == (uint16_t) NO_VAL)
		xstrcat(out, "NtasksPerN:B:S:C=*:");
	else
		xstrfmtcat(out, "NtasksPerN:B:S:C=%u:", job_ptr->ntasks_per_node);

	if (job_ptr->ntasks_per_board == (uint16_t) NO_VAL)
		xstrcat(out, "*:");
	else
		xstrfmtcat(out, "%u:", job_ptr->ntasks_per_board);

	if ((job_ptr->ntasks_per_socket == (uint16_t) NO_VAL) ||
	    (job_ptr->ntasks_per_socket == (uint16_t) INFINITE))
		xstrcat(out, "*:");
	else
		xstrfmtcat(out, "%u:", job_ptr->ntasks_per_socket);

	if ((job_ptr->ntasks_per_core == (uint16_t) NO_VAL) ||
	    (job_ptr->ntasks_per_core == (uint16_t) INFINITE))
		xstrcat(out, "* ");
	else
		xstrfmtcat(out, "%u ", job_ptr->ntasks_per_core);

	if (job_ptr->core_spec == (uint16_t) NO_VAL)
		xstrcat(out, "CoreSpec=*");
	else if (job_ptr->core_spec & CORE_SPEC_THREAD)
		xstrfmtcat(out, "ThreadSpec=%d",
			   (job_ptr->core_spec & (~CORE_SPEC_THREAD)));
	else
		xstrfmtcat(out, "CoreSpec=%u", job_ptr->core_spec);

	xstrcat(out, line_end);

	if (job_resrcs && cluster_flags & CLUSTER_FLAG_BG) {
		if ((job_resrcs->cpu_array_cnt > 0) &&
		    (job_resrcs->cpu_array_value) &&
		    (job_resrcs->cpu_array_reps)) {
			int length = 0;
			xstrcat(out, "CPUs=");
			for (i = 0; i < job_resrcs->cpu_array_cnt; i++) {
				/* only print 60 characters worth of this record */
				if (length > 60) {
					/* skip to last CPU group entry */
					if (i < job_resrcs->cpu_array_cnt - 1) {
						continue;
					}
					/* add ellipsis before last entry */
					xstrcat(out, "...,");
				}

				length += xstrfmtcat(out, "%d", job_resrcs->cpus[i]);
				if (job_resrcs->cpu_array_reps[i] > 1) {
					length += xstrfmtcat(out, "*%d",
							     job_resrcs->cpu_array_reps[i]);
				}
				if (i < job_resrcs->cpu_array_cnt - 1) {
					xstrcat(out, ",");
					length++;
				}
			}
			xstrcat(out, line_end);
		}
	} else if (job_resrcs && job_resrcs->core_bitmap &&
		   ((last = bit_fls(job_resrcs->core_bitmap)) != -1)) {
		hl = hostlist_create(job_resrcs->nodes);
		if (!hl) {
			error("slurm_sprint_job_info: hostlist_create: %s",
			      job_resrcs->nodes);
			return NULL;
		}
		hl_last = hostlist_create(NULL);
		if (!hl_last) {
			error("slurm_sprint_job_info: hostlist_create: NULL");
			hostlist_destroy(hl);
			return NULL;
		}

		bit_inx = 0;
		i = sock_inx = sock_reps = 0;
		abs_node_inx = job_ptr->node_inx[i];

		gres_last = "";
		/* tmp1[] stores the current cpu(s) allocated */
		tmp2[0] = '\0';	/* stores last cpu(s) allocated */
		for (rel_node_inx=0; rel_node_inx < job_resrcs->nhosts;
		     rel_node_inx++) {

			if (sock_reps >=
			    job_resrcs->sock_core_rep_count[sock_inx]) {
				sock_inx++;
				sock_reps = 0;
			}
			sock_reps++;

			bit_reps = job_resrcs->sockets_per_node[sock_inx] *
				   job_resrcs->cores_per_socket[sock_inx];
			host = hostlist_shift(hl);
			threads = _threads_per_core(host);
			cpu_bitmap = bit_alloc(bit_reps * threads);
			for (j = 0; j < bit_reps; j++) {
				if (bit_test(job_resrcs->core_bitmap, bit_inx)){
					for (k = 0; k < threads; k++)
						bit_set(cpu_bitmap,
							(j * threads) + k);
				}
				bit_inx++;
			}
			bit_fmt(tmp1, sizeof(tmp1), cpu_bitmap);
			FREE_NULL_BITMAP(cpu_bitmap);
			/*
			 * If the allocation values for this host are not the
			 * same as the last host, print the report of the last
			 * group of hosts that had identical allocation values.
			 */
			if (xstrcmp(tmp1, tmp2) ||
			    ((rel_node_inx < job_ptr->gres_detail_cnt) &&
			     xstrcmp(job_ptr->gres_detail_str[rel_node_inx],
				     gres_last)) ||
			    (last_mem_alloc_ptr != job_resrcs->memory_allocated) ||
			    (job_resrcs->memory_allocated &&
			     (last_mem_alloc !=
			      job_resrcs->memory_allocated[rel_node_inx]))) {
				if (hostlist_count(hl_last)) {
					last_hosts =
						hostlist_ranged_string_xmalloc(
						hl_last);
					xstrfmtcat(out,
						   "  Nodes=%s CPU_IDs=%s "
						   "Mem=%"PRIu64" GRES_IDX=%s",
						   last_hosts, tmp2,
						   last_mem_alloc_ptr ?
						   last_mem_alloc : 0,
						    gres_last);
					xfree(last_hosts);
					xstrcat(out, line_end);

					hostlist_destroy(hl_last);
					hl_last = hostlist_create(NULL);
				}
				strcpy(tmp2, tmp1);
				if (rel_node_inx < job_ptr->gres_detail_cnt) {
					gres_last = job_ptr->
						    gres_detail_str[rel_node_inx];
				} else {
					gres_last = "";
				}
				last_mem_alloc_ptr = job_resrcs->memory_allocated;
				if (last_mem_alloc_ptr)
					last_mem_alloc = job_resrcs->
						memory_allocated[rel_node_inx];
				else
					last_mem_alloc = NO_VAL64;
			}
			hostlist_push_host(hl_last, host);
			free(host);

			if (bit_inx > last)
				break;

			if (abs_node_inx > job_ptr->node_inx[i+1]) {
				i += 2;
				abs_node_inx = job_ptr->node_inx[i];
			} else {
				abs_node_inx++;
			}
		}

		if (hostlist_count(hl_last)) {
			last_hosts = hostlist_ranged_string_xmalloc(hl_last);
			xstrfmtcat(out, "  Nodes=%s CPU_IDs=%s Mem=%"PRIu64" GRES_IDX=%s",
				 last_hosts, tmp2,
				 last_mem_alloc_ptr ? last_mem_alloc : 0,
				 gres_last);
			xfree(last_hosts);
			xstrcat(out, line_end);
		}
		hostlist_destroy(hl);
		hostlist_destroy(hl_last);
	}
	/****** Line 18 ******/
	if (job_ptr->pn_min_memory & MEM_PER_CPU) {
		job_ptr->pn_min_memory &= (~MEM_PER_CPU);
		tmp6_ptr = "CPU";
	} else
		tmp6_ptr = "Node";

	if (cluster_flags & CLUSTER_FLAG_BG) {
		convert_num_unit((float)job_ptr->pn_min_cpus, tmp1,
				 sizeof(tmp1), UNIT_NONE, NO_VAL,
				 CONVERT_NUM_UNIT_EXACT);
		xstrfmtcat(out, "MinCPUsNode=%s ", tmp1);
	} else {
		xstrfmtcat(out, "MinCPUsNode=%u ", job_ptr->pn_min_cpus);
	}

	convert_num_unit((float)job_ptr->pn_min_memory, tmp1, sizeof(tmp1),
			 UNIT_MEGA, NO_VAL, CONVERT_NUM_UNIT_EXACT);
	convert_num_unit((float)job_ptr->pn_min_tmp_disk, tmp2, sizeof(tmp2),
			 UNIT_MEGA, NO_VAL, CONVERT_NUM_UNIT_EXACT);
	xstrfmtcat(out, "MinMemory%s=%s MinTmpDiskNode=%s", tmp6_ptr, tmp1, tmp2);
	xstrcat(out, line_end);

	/****** Line ******/
	secs2time_str((time_t)job_ptr->delay_boot, tmp1, sizeof(tmp1));
	xstrfmtcat(out, "Features=%s DelayBoot=%s", job_ptr->features, tmp1);
	xstrcat(out, line_end);

	/****** Line ******/
	xstrfmtcat(out, "Gres=%s Reservation=%s",
		   job_ptr->gres, job_ptr->resv_name);
	xstrcat(out, line_end);

	/****** Line 20 ******/
	xstrfmtcat(out, "OverSubscribe=%s Contiguous=%d Licenses=%s Network=%s",
		   job_share_string(job_ptr->shared), job_ptr->contiguous,
		   job_ptr->licenses, job_ptr->network);
	xstrcat(out, line_end);

	/****** Line 21 ******/
	xstrfmtcat(out, "Command=%s", job_ptr->command);
	xstrcat(out, line_end);

	/****** Line 22 ******/
	xstrfmtcat(out, "WorkDir=%s", job_ptr->work_dir);

	if (cluster_flags & CLUSTER_FLAG_BG) {
		/****** Line 23 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_BG_ID);
		if (select_buf[0] != '\0') {
			xstrcat(out, line_end);
			xstrfmtcat(out, "Block_ID=%s", select_buf);
		}

		/****** Line 24 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_MIXED_SHORT);
		if (select_buf[0] != '\0') {
			xstrcat(out, line_end);
			xstrcat(out, select_buf);
		}

		/****** Line 26 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_LINUX_IMAGE);
		if (select_buf[0] != '\0') {
			xstrcat(out, line_end);
			xstrfmtcat(out, "CnloadImage=%s", select_buf);
		}
		/****** Line 27 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_MLOADER_IMAGE);
		if (select_buf[0] != '\0') {
			xstrcat(out, line_end);
			xstrfmtcat(out, "MloaderImage=%s", select_buf);
		}
		/****** Line 28 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_RAMDISK_IMAGE);
		if (select_buf[0] != '\0') {
			xstrcat(out, line_end);
			xstrfmtcat(out, "IoloadImage=%s", select_buf);
		}
	}

	/****** Line (optional) ******/
	if (job_ptr->admin_comment) {
		xstrcat(out, line_end);
		xstrfmtcat(out, "AdminComment=%s ", job_ptr->admin_comment);
	}

	/****** Line (optional) ******/
	if (job_ptr->comment) {
		xstrcat(out, line_end);
		xstrfmtcat(out, "Comment=%s ", job_ptr->comment);
	}

	/****** Line 30 (optional) ******/
	if (job_ptr->batch_flag) {
		xstrcat(out, line_end);
		slurm_get_job_stderr(tmp_path, sizeof(tmp_path), job_ptr);
		xstrfmtcat(out, "StdErr=%s", tmp_path);
	}

	/****** Line 31 (optional) ******/
	if (job_ptr->batch_flag) {
		xstrcat(out, line_end);
		slurm_get_job_stdin(tmp_path, sizeof(tmp_path), job_ptr);
		xstrfmtcat(out, "StdIn=%s", tmp_path);
	}

	/****** Line 32 (optional) ******/
	if (job_ptr->batch_flag) {
		xstrcat(out, line_end);
		slurm_get_job_stdout(tmp_path, sizeof(tmp_path), job_ptr);
		xstrfmtcat(out, "StdOut=%s", tmp_path);
	}

	/****** Line 33 (optional) ******/
	if (job_ptr->batch_script) {
		xstrcat(out, line_end);
		xstrcat(out, "BatchScript=\n");
		xstrcat(out, job_ptr->batch_script);
	}

	/****** Line 34 (optional) ******/
	if (job_ptr->req_switch) {
		char time_buf[32];
		xstrcat(out, line_end);
		secs2time_str((time_t) job_ptr->wait4switch, time_buf,
			      sizeof(time_buf));
		xstrfmtcat(out, "Switches=%u@%s\n", job_ptr->req_switch, time_buf);
	}

	/****** Line 35 (optional) ******/
	if (job_ptr->burst_buffer) {
		xstrcat(out, line_end);
		xstrfmtcat(out, "BurstBuffer=%s", job_ptr->burst_buffer);
	}

	/****** Line (optional) ******/
	if (job_ptr->burst_buffer_state) {
		xstrcat(out, line_end);
		xstrfmtcat(out, "BurstBufferState=%s",
			   job_ptr->burst_buffer_state);
	}

	/****** Line 36 (optional) ******/
	if (cpu_freq_debug(NULL, NULL, tmp1, sizeof(tmp1),
			   job_ptr->cpu_freq_gov, job_ptr->cpu_freq_min,
			   job_ptr->cpu_freq_max, NO_VAL) != 0) {
		xstrcat(out, line_end);
		xstrcat(out, tmp1);
	}

	/****** Line 37 ******/
	xstrcat(out, line_end);
	xstrfmtcat(out, "Power=%s", power_flags_str(job_ptr->power_flags));

	/****** Line 38 (optional) ******/
	if (job_ptr->bitflags) {
		xstrcat(out, line_end);
		if (job_ptr->bitflags & GRES_ENFORCE_BIND)
			xstrcat(out, "GresEnforceBind=Yes");
		if (job_ptr->bitflags & KILL_INV_DEP)
			xstrcat(out, "KillOInInvalidDependent=Yes");
		if (job_ptr->bitflags & NO_KILL_INV_DEP)
			xstrcat(out, "KillOInInvalidDependent=No");
		if (job_ptr->bitflags & SPREAD_JOB)
			xstrcat(out, "SpreadJob=Yes");
	}

	/****** END OF JOB RECORD ******/
	if (one_liner)
		xstrcat(out, "\n");
	else
		xstrcat(out, "\n\n");

	return out;
}
예제 #22
0
파일: job_info.c 프로젝트: IFCA/slurm
/*
 * slurm_sprint_job_info - output information about a specific Slurm
 *	job based upon message as loaded using slurm_load_jobs
 * IN job_ptr - an individual job information record pointer
 * IN one_liner - print as a single line if true
 * RET out - char * containing formatted output (must be freed after call)
 *           NULL is returned on failure.
 */
extern char *
slurm_sprint_job_info ( job_info_t * job_ptr, int one_liner )
{
	int i, j;
	char time_str[32], *group_name, *user_name;
	char tmp1[128], tmp2[128], tmp3[128], tmp4[128], tmp5[128], *tmp6_ptr;
	char tmp_line[512];
	char *ionodes = NULL;
	uint16_t exit_status = 0, term_sig = 0;
	job_resources_t *job_resrcs = job_ptr->job_resrcs;
	char *out = NULL;
	time_t run_time;
	uint32_t min_nodes, max_nodes = 0;
	char *nodelist = "NodeList";
	bitstr_t *core_bitmap;
	char *host;
	int sock_inx, sock_reps, last;
	int abs_node_inx, rel_node_inx;
	int bit_inx, bit_reps;
	uint32_t *last_mem_alloc_ptr = NULL;
	uint32_t last_mem_alloc = NO_VAL;
	char *last_hosts;
	hostlist_t hl, hl_last;
	char select_buf[122];
	uint32_t cluster_flags = slurmdb_setup_cluster_flags();

	if (cluster_flags & CLUSTER_FLAG_BG) {
		nodelist = "MidplaneList";
		select_g_select_jobinfo_get(job_ptr->select_jobinfo,
					    SELECT_JOBDATA_IONODES,
					    &ionodes);
	}

	/****** Line 1 ******/
	snprintf(tmp_line, sizeof(tmp_line), "JobId=%u ", job_ptr->job_id);
	out = xstrdup(tmp_line);
	if (job_ptr->array_job_id) {
		snprintf(tmp_line, sizeof(tmp_line), 
			 "ArrayJobId=%u ArrayTaskId=%u ",
			 job_ptr->array_job_id, job_ptr->array_task_id);
		xstrcat(out, tmp_line);
	}
	snprintf(tmp_line, sizeof(tmp_line), "Name=%s", job_ptr->name);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 2 ******/
	user_name = uid_to_string((uid_t) job_ptr->user_id);
	group_name = gid_to_string((gid_t) job_ptr->group_id);
	snprintf(tmp_line, sizeof(tmp_line),
		 "UserId=%s(%u) GroupId=%s(%u)",
		 user_name, job_ptr->user_id, group_name, job_ptr->group_id);
	xfree(user_name);
	xfree(group_name);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 3 ******/
	snprintf(tmp_line, sizeof(tmp_line),
		 "Priority=%u Account=%s QOS=%s",
		 job_ptr->priority, job_ptr->account, job_ptr->qos);
	xstrcat(out, tmp_line);
	if (slurm_get_track_wckey()) {
		snprintf(tmp_line, sizeof(tmp_line),
			 " WCKey=%s", job_ptr->wckey);
		xstrcat(out, tmp_line);
	}
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 4 ******/
	if (job_ptr->state_desc) {
		/* Replace white space with underscore for easier parsing */
		for (j=0; job_ptr->state_desc[j]; j++) {
			if (isspace((int)job_ptr->state_desc[j]))
				job_ptr->state_desc[j] = '_';
		}
		tmp6_ptr = job_ptr->state_desc;
	} else
		tmp6_ptr = job_reason_string(job_ptr->state_reason);
	snprintf(tmp_line, sizeof(tmp_line),
		 "JobState=%s Reason=%s Dependency=%s",
		 job_state_string(job_ptr->job_state), tmp6_ptr,
		 job_ptr->dependency);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 5 ******/
	snprintf(tmp_line, sizeof(tmp_line),
		 "Requeue=%u Restarts=%u BatchFlag=%u ",
		 job_ptr->requeue, job_ptr->restart_cnt, job_ptr->batch_flag);
	xstrcat(out, tmp_line);
	if (WIFSIGNALED(job_ptr->exit_code))
		term_sig = WTERMSIG(job_ptr->exit_code);
	exit_status = WEXITSTATUS(job_ptr->exit_code);
	snprintf(tmp_line, sizeof(tmp_line),
		 "ExitCode=%u:%u", exit_status, term_sig);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 5a (optional) ******/
	if (!(job_ptr->show_flags & SHOW_DETAIL))
		goto line6;
	if (WIFSIGNALED(job_ptr->derived_ec))
		term_sig = WTERMSIG(job_ptr->derived_ec);
	else
		term_sig = 0;
	exit_status = WEXITSTATUS(job_ptr->derived_ec);
	snprintf(tmp_line, sizeof(tmp_line),
		 "DerivedExitCode=%u:%u", exit_status, term_sig);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 6 ******/
line6:
	snprintf(tmp_line, sizeof(tmp_line), "RunTime=");
	xstrcat(out, tmp_line);
	if (IS_JOB_PENDING(job_ptr))
		run_time = 0;
	else if (IS_JOB_SUSPENDED(job_ptr))
		run_time = job_ptr->pre_sus_time;
	else {
		time_t end_time;
		if (IS_JOB_RUNNING(job_ptr) || (job_ptr->end_time == 0))
			end_time = time(NULL);
		else
			end_time = job_ptr->end_time;
		if (job_ptr->suspend_time) {
			run_time = (time_t)
				(difftime(end_time, job_ptr->suspend_time)
				 + job_ptr->pre_sus_time);
		} else
			run_time = (time_t)
				difftime(end_time, job_ptr->start_time);
	}
	secs2time_str(run_time, tmp1, sizeof(tmp1));
	sprintf(tmp_line, "%s ", tmp1);
	xstrcat(out, tmp_line);

	snprintf(tmp_line, sizeof(tmp_line), "TimeLimit=");
	xstrcat(out, tmp_line);
	if (job_ptr->time_limit == NO_VAL)
		sprintf(tmp_line, "Partition_Limit");
	else {
		mins2time_str(job_ptr->time_limit, tmp_line,
			      sizeof(tmp_line));
	}
	xstrcat(out, tmp_line);
	snprintf(tmp_line, sizeof(tmp_line), " TimeMin=");
	xstrcat(out, tmp_line);
	if (job_ptr->time_min == 0)
		sprintf(tmp_line, "N/A");
	else {
		mins2time_str(job_ptr->time_min, tmp_line,
			      sizeof(tmp_line));
	}
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 7 ******/
	slurm_make_time_str((time_t *)&job_ptr->submit_time, time_str,
			    sizeof(time_str));
	snprintf(tmp_line, sizeof(tmp_line), "SubmitTime=%s ", time_str);
	xstrcat(out, tmp_line);

	slurm_make_time_str((time_t *)&job_ptr->eligible_time, time_str,
			    sizeof(time_str));
	snprintf(tmp_line, sizeof(tmp_line), "EligibleTime=%s", time_str);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 8 (optional) ******/
	if (job_ptr->resize_time) {
		slurm_make_time_str((time_t *)&job_ptr->resize_time, time_str,
				    sizeof(time_str));
		snprintf(tmp_line, sizeof(tmp_line), "ResizeTime=%s", time_str);
		xstrcat(out, tmp_line);
		if (one_liner)
			xstrcat(out, " ");
		else
			xstrcat(out, "\n   ");
	}

	/****** Line 9 ******/
	slurm_make_time_str((time_t *)&job_ptr->start_time, time_str,
			    sizeof(time_str));
	snprintf(tmp_line, sizeof(tmp_line), "StartTime=%s ", time_str);
	xstrcat(out, tmp_line);

	snprintf(tmp_line, sizeof(tmp_line), "EndTime=");
	xstrcat(out, tmp_line);
	if ((job_ptr->time_limit == INFINITE) &&
	    (job_ptr->end_time > time(NULL)))
		sprintf(tmp_line, "Unknown");
	else {
		slurm_make_time_str ((time_t *)&job_ptr->end_time, time_str,
				     sizeof(time_str));
		sprintf(tmp_line, "%s", time_str);
	}
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 10 ******/
	if (job_ptr->preempt_time == 0)
		sprintf(tmp_line, "PreemptTime=None ");
	else {
		slurm_make_time_str((time_t *)&job_ptr->preempt_time,
				    time_str, sizeof(time_str));
		snprintf(tmp_line, sizeof(tmp_line), "PreemptTime=%s ",
			 time_str);
	}
	xstrcat(out, tmp_line);
	if (job_ptr->suspend_time) {
		slurm_make_time_str ((time_t *)&job_ptr->suspend_time,
				     time_str, sizeof(time_str));
	} else {
		strncpy(time_str, "None", sizeof(time_str));
	}
	snprintf(tmp_line, sizeof(tmp_line),
		 "SuspendTime=%s SecsPreSuspend=%ld",
		 time_str, (long int)job_ptr->pre_sus_time);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 11 ******/
	snprintf(tmp_line, sizeof(tmp_line),
		 "Partition=%s AllocNode:Sid=%s:%u",
		 job_ptr->partition, job_ptr->alloc_node, job_ptr->alloc_sid);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 12 ******/
	snprintf(tmp_line, sizeof(tmp_line), "Req%s=%s Exc%s=%s",
		 nodelist, job_ptr->req_nodes, nodelist, job_ptr->exc_nodes);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 13 ******/
	xstrfmtcat(out, "%s=", nodelist);
	xstrcat(out, job_ptr->nodes);
	if (job_ptr->nodes && ionodes) {
		snprintf(tmp_line, sizeof(tmp_line), "[%s]", ionodes);
		xstrcat(out, tmp_line);
		xfree(ionodes);
	}
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 14 (optional) ******/
	if (job_ptr->batch_host) {
		snprintf(tmp_line, sizeof(tmp_line), "BatchHost=%s",
			 job_ptr->batch_host);
		xstrcat(out, tmp_line);
		if (one_liner)
			xstrcat(out, " ");
		else
			xstrcat(out, "\n   ");
	}

	/****** Line 15 ******/
	if (cluster_flags & CLUSTER_FLAG_BG) {
		select_g_select_jobinfo_get(job_ptr->select_jobinfo,
					    SELECT_JOBDATA_NODE_CNT,
					    &min_nodes);
		if ((min_nodes == 0) || (min_nodes == NO_VAL)) {
			min_nodes = job_ptr->num_nodes;
			max_nodes = job_ptr->max_nodes;
		} else if (job_ptr->max_nodes)
			max_nodes = min_nodes;
	} else {
		min_nodes = job_ptr->num_nodes;
		max_nodes = job_ptr->max_nodes;
	}

	_sprint_range(tmp1, sizeof(tmp1), job_ptr->num_cpus, job_ptr->max_cpus);
	_sprint_range(tmp2, sizeof(tmp2), min_nodes, max_nodes);
	if (job_ptr->sockets_per_node == (uint16_t) NO_VAL)
		strcpy(tmp3, "*");
	else
		snprintf(tmp3, sizeof(tmp3), "%u", job_ptr->sockets_per_node);
	if (job_ptr->cores_per_socket == (uint16_t) NO_VAL)
		strcpy(tmp4, "*");
	else
		snprintf(tmp4, sizeof(tmp4), "%u", job_ptr->cores_per_socket);
	if (job_ptr->threads_per_core == (uint16_t) NO_VAL)
		strcpy(tmp5, "*");
	else
		snprintf(tmp5, sizeof(tmp5), "%u", job_ptr->threads_per_core);
	snprintf(tmp_line, sizeof(tmp_line),
		 "NumNodes=%s NumCPUs=%s CPUs/Task=%u ReqS:C:T=%s:%s:%s",
		 tmp2, tmp1, job_ptr->cpus_per_task, tmp3, tmp4, tmp5);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	if (!job_resrcs)
		goto line15;

	if (cluster_flags & CLUSTER_FLAG_BG) {
		if ((job_resrcs->cpu_array_cnt > 0) &&
		    (job_resrcs->cpu_array_value) &&
		    (job_resrcs->cpu_array_reps)) {
			int length = 0;
			xstrcat(out, "CPUs=");
			length += 10;
			for (i = 0; i < job_resrcs->cpu_array_cnt; i++) {
				if (length > 70) {
					/* skip to last CPU group entry */
					if (i < job_resrcs->cpu_array_cnt - 1) {
						continue;
					}
					/* add ellipsis before last entry */
					xstrcat(out, "...,");
					length += 4;
				}

				snprintf(tmp_line, sizeof(tmp_line), "%d",
					 job_resrcs->cpus[i]);
				xstrcat(out, tmp_line);
				length += strlen(tmp_line);
				if (job_resrcs->cpu_array_reps[i] > 1) {
					snprintf(tmp_line, sizeof(tmp_line),
						 "*%d",
						 job_resrcs->cpu_array_reps[i]);
					xstrcat(out, tmp_line);
					length += strlen(tmp_line);
				}
				if (i < job_resrcs->cpu_array_cnt - 1) {
					xstrcat(out, ",");
					length++;
				}
			}
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
		}
	} else {
		if (!job_resrcs->core_bitmap)
			goto line15;

		last  = bit_fls(job_resrcs->core_bitmap);
		if (last == -1)
			goto line15;

		hl = hostlist_create(job_ptr->nodes);
		if (!hl) {
			error("slurm_sprint_job_info: hostlist_create: %s",
			      job_ptr->nodes);
			return NULL;
		}
		hl_last = hostlist_create(NULL);
		if (!hl_last) {
			error("slurm_sprint_job_info: hostlist_create: NULL");
			hostlist_destroy(hl);
			return NULL;
		}

		bit_inx = 0;
		i = sock_inx = sock_reps = 0;
		abs_node_inx = job_ptr->node_inx[i];

/*	tmp1[] stores the current cpu(s) allocated	*/
		tmp2[0] = '\0';	/* stores last cpu(s) allocated */
		for (rel_node_inx=0; rel_node_inx < job_resrcs->nhosts;
		     rel_node_inx++) {

			if (sock_reps >=
			    job_resrcs->sock_core_rep_count[sock_inx]) {
				sock_inx++;
				sock_reps = 0;
			}
			sock_reps++;

			bit_reps = job_resrcs->sockets_per_node[sock_inx] *
				job_resrcs->cores_per_socket[sock_inx];

			core_bitmap = bit_alloc(bit_reps);
			for (j=0; j < bit_reps; j++) {
				if (bit_test(job_resrcs->core_bitmap, bit_inx))
					bit_set(core_bitmap, j);
				bit_inx++;
			}

			bit_fmt(tmp1, sizeof(tmp1), core_bitmap);
			FREE_NULL_BITMAP(core_bitmap);
			host = hostlist_shift(hl);
/*
 *		If the allocation values for this host are not the same as the
 *		last host, print the report of the last group of hosts that had
 *		identical allocation values.
 */
			if (strcmp(tmp1, tmp2) ||
			    (last_mem_alloc_ptr != job_resrcs->memory_allocated) ||
			    (job_resrcs->memory_allocated &&
			     (last_mem_alloc !=
			      job_resrcs->memory_allocated[rel_node_inx]))) {
				if (hostlist_count(hl_last)) {
					last_hosts = 
						hostlist_ranged_string_xmalloc(
						hl_last);
					snprintf(tmp_line, sizeof(tmp_line),
						 "  Nodes=%s CPU_IDs=%s Mem=%u",
						 last_hosts, tmp2,
						 last_mem_alloc_ptr ?
						 last_mem_alloc : 0);
					xfree(last_hosts);
					xstrcat(out, tmp_line);
					if (one_liner)
						xstrcat(out, " ");
					else
						xstrcat(out, "\n   ");

					hostlist_destroy(hl_last);
					hl_last = hostlist_create(NULL);
				}
				strcpy(tmp2, tmp1);
				last_mem_alloc_ptr = job_resrcs->memory_allocated;
				if (last_mem_alloc_ptr)
					last_mem_alloc = job_resrcs->
						memory_allocated[rel_node_inx];
				else
					last_mem_alloc = NO_VAL;
			}
			hostlist_push_host(hl_last, host);
			free(host);

			if (bit_inx > last)
				break;

			if (abs_node_inx > job_ptr->node_inx[i+1]) {
				i += 2;
				abs_node_inx = job_ptr->node_inx[i];
			} else {
				abs_node_inx++;
			}
		}

		if (hostlist_count(hl_last)) {
			last_hosts = hostlist_ranged_string_xmalloc(hl_last);
			snprintf(tmp_line, sizeof(tmp_line),
				 "  Nodes=%s CPU_IDs=%s Mem=%u",
				 last_hosts, tmp2,
				 last_mem_alloc_ptr ? last_mem_alloc : 0);
			xfree(last_hosts);
			xstrcat(out, tmp_line);
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
		}
		hostlist_destroy(hl);
		hostlist_destroy(hl_last);
	}
	/****** Line 15 ******/
line15:
	if (job_ptr->pn_min_memory & MEM_PER_CPU) {
		job_ptr->pn_min_memory &= (~MEM_PER_CPU);
		tmp6_ptr = "CPU";
	} else
		tmp6_ptr = "Node";

	if (cluster_flags & CLUSTER_FLAG_BG) {
		convert_num_unit((float)job_ptr->pn_min_cpus,
				 tmp1, sizeof(tmp1), UNIT_NONE);
		snprintf(tmp_line, sizeof(tmp_line), "MinCPUsNode=%s",	tmp1);
	} else {
		snprintf(tmp_line, sizeof(tmp_line), "MinCPUsNode=%u",
			 job_ptr->pn_min_cpus);
	}

	xstrcat(out, tmp_line);
	convert_num_unit((float)job_ptr->pn_min_memory, tmp1, sizeof(tmp1),
			 UNIT_MEGA);
	convert_num_unit((float)job_ptr->pn_min_tmp_disk, tmp2, sizeof(tmp2),
			 UNIT_MEGA);
	snprintf(tmp_line, sizeof(tmp_line),
		 " MinMemory%s=%s MinTmpDiskNode=%s",
		 tmp6_ptr, tmp1, tmp2);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 16 ******/
	snprintf(tmp_line, sizeof(tmp_line),
		 "Features=%s Gres=%s Reservation=%s",
		 job_ptr->features, job_ptr->gres, job_ptr->resv_name);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 17 ******/
	snprintf(tmp_line, sizeof(tmp_line),
		 "Shared=%s Contiguous=%d Licenses=%s Network=%s",
		 (job_ptr->shared == 0 ? "0" :
		  job_ptr->shared == 1 ? "1" : "OK"),
		 job_ptr->contiguous, job_ptr->licenses, job_ptr->network);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 18 ******/
	snprintf(tmp_line, sizeof(tmp_line), "Command=%s",
		 job_ptr->command);
	xstrcat(out, tmp_line);
	if (one_liner)
		xstrcat(out, " ");
	else
		xstrcat(out, "\n   ");

	/****** Line 19 ******/
	snprintf(tmp_line, sizeof(tmp_line), "WorkDir=%s",
		 job_ptr->work_dir);
	xstrcat(out, tmp_line);

	if (cluster_flags & CLUSTER_FLAG_BG) {
		/****** Line 20 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_BG_ID);
		if (select_buf[0] != '\0') {
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
			snprintf(tmp_line, sizeof(tmp_line),
				 "Block_ID=%s", select_buf);
			xstrcat(out, tmp_line);
		}

		/****** Line 21 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_MIXED_SHORT);
		if (select_buf[0] != '\0') {
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
			xstrcat(out, select_buf);
		}

		if (cluster_flags & CLUSTER_FLAG_BGL) {
			/****** Line 22 (optional) ******/
			select_g_select_jobinfo_sprint(
				job_ptr->select_jobinfo,
				select_buf, sizeof(select_buf),
				SELECT_PRINT_BLRTS_IMAGE);
			if (select_buf[0] != '\0') {
				if (one_liner)
					xstrcat(out, " ");
				else
					xstrcat(out, "\n   ");
				snprintf(tmp_line, sizeof(tmp_line),
					 "BlrtsImage=%s", select_buf);
				xstrcat(out, tmp_line);
			}
		}
		/****** Line 23 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_LINUX_IMAGE);
		if (select_buf[0] != '\0') {
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
			if (cluster_flags & CLUSTER_FLAG_BGL)
				snprintf(tmp_line, sizeof(tmp_line),
					 "LinuxImage=%s", select_buf);
			else
				snprintf(tmp_line, sizeof(tmp_line),
					 "CnloadImage=%s", select_buf);

			xstrcat(out, tmp_line);
		}
		/****** Line 24 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_MLOADER_IMAGE);
		if (select_buf[0] != '\0') {
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
			snprintf(tmp_line, sizeof(tmp_line),
				 "MloaderImage=%s", select_buf);
			xstrcat(out, tmp_line);
		}
		/****** Line 25 (optional) ******/
		select_g_select_jobinfo_sprint(job_ptr->select_jobinfo,
					       select_buf, sizeof(select_buf),
					       SELECT_PRINT_RAMDISK_IMAGE);
		if (select_buf[0] != '\0') {
			if (one_liner)
				xstrcat(out, " ");
			else
				xstrcat(out, "\n   ");
			if (cluster_flags & CLUSTER_FLAG_BGL)
				snprintf(tmp_line, sizeof(tmp_line),
					 "RamDiskImage=%s", select_buf);
			else
				snprintf(tmp_line, sizeof(tmp_line),
					 "IoloadImage=%s", select_buf);
			xstrcat(out, tmp_line);
		}
	}

	/****** Line 26 (optional) ******/
	if (job_ptr->comment) {
		if (one_liner)
			xstrcat(out, " ");
		else
			xstrcat(out, "\n   ");
		snprintf(tmp_line, sizeof(tmp_line), "Comment=%s ",
			 job_ptr->comment);
		xstrcat(out, tmp_line);
	}

	/****** Line 27 (optional) ******/
	if (job_ptr->batch_script) {
		if (one_liner)
			xstrcat(out, " ");
		else
			xstrcat(out, "\n   ");
		xstrcat(out, "BatchScript=\n");
		xstrcat(out, job_ptr->batch_script);
	}

	/****** Line 28 (optional) ******/
	if (job_ptr->req_switch) {
		char time_buf[32];
		if (one_liner)
			xstrcat(out, " ");
		else
			xstrcat(out, "\n   ");
		secs2time_str((time_t) job_ptr->wait4switch, time_buf,
			      sizeof(time_buf));
		snprintf(tmp_line, sizeof(tmp_line), "Switches=%u@%s\n",
			 job_ptr->req_switch, time_buf);
		xstrcat(out, tmp_line);
	}

	/****** Line 29 (optional) ******/
	if (one_liner)
		xstrcat(out, "\n");
	else
		xstrcat(out, "\n\n");

	return out;

}
예제 #23
0
파일: print.c 프로젝트: RSE-Cambridge/slurm
/* Translate bitmap representation from hex to decimal format, replacing
 * array_task_str. */
static void _xlate_task_str(slurmdb_job_rec_t *job_ptr)
{
	static int bitstr_len = -1;
	int buf_size, len;
	int i, i_first, i_last, i_prev, i_step = 0;
	bitstr_t *task_bitmap;
	char *in_buf = job_ptr->array_task_str;
	char *out_buf = NULL;

	if (!in_buf)
		return;

	if (strlen(in_buf) < 3 || in_buf[1] != 'x')
		return;

	i = strlen(in_buf);
	task_bitmap = bit_alloc(i * 4);
	bit_unfmt_hexmask(task_bitmap, in_buf);

	/* Check first for a step function */
	i_first = bit_ffs(task_bitmap);
	i_last  = bit_fls(task_bitmap);
	if (((i_last - i_first) > 10) &&
	    !bit_test(task_bitmap, i_first + 1)) {
		bool is_step = true;
		i_prev = i_first;
		for (i = i_first + 1; i <= i_last; i++) {
			if (!bit_test(task_bitmap, i))
				continue;
			if (i_step == 0) {
				i_step = i - i_prev;
			} else if ((i - i_prev) != i_step) {
				is_step = false;
				break;
			}
			i_prev = i;
		}
		if (is_step) {
			xstrfmtcat(out_buf, "%d-%d:%d",
				   i_first, i_last, i_step);
		}
	}

	if (bitstr_len > 0) {
		/* Print the first bitstr_len bytes of the bitmap string */
		buf_size = bitstr_len;
		out_buf = xmalloc(buf_size);
		bit_fmt(out_buf, buf_size, task_bitmap);
		len = strlen(out_buf);
		if (len > (buf_size - 3))
			for (i = 0; i < 3; i++)
				out_buf[buf_size - 2 - i] = '.';
	} else {
		/* Print the full bitmap's string representation.
		 * For huge bitmaps this can take roughly one minute,
		 * so let the client do the work */
		buf_size = bit_size(task_bitmap) * 8;
		while (1) {
			out_buf = xmalloc(buf_size);
			bit_fmt(out_buf, buf_size, task_bitmap);
			len = strlen(out_buf);
			if ((len > 0) && (len < (buf_size - 32)))
				break;
			xfree(out_buf);
			buf_size *= 2;
		}
	}

	if (job_ptr->array_max_tasks)
		xstrfmtcat(out_buf, "%c%u", '%', job_ptr->array_max_tasks);

	xfree(job_ptr->array_task_str);
	job_ptr->array_task_str = out_buf;
}
예제 #24
0
파일: job_test.c 프로젝트: BYUHPC/slurm
/*
 * Determine which of these nodes are usable by this job
 *
 * Remove nodes from the bitmap that don't have enough memory or gres to
 * support the job. 
 *
 * Return SLURM_ERROR if a required node can't be used.
 *
 * if node_state = NODE_CR_RESERVED, clear bitmap (if node is required
 *                                   then should we return NODE_BUSY!?!)
 *
 * if node_state = NODE_CR_ONE_ROW, then this node can only be used by
 *                                  another NODE_CR_ONE_ROW job
 *
 * if node_state = NODE_CR_AVAILABLE AND:
 *  - job_node_req = NODE_CR_RESERVED, then we need idle nodes
 *  - job_node_req = NODE_CR_ONE_ROW, then we need idle or non-sharing nodes
 */
static int _verify_node_state(struct part_res_record *cr_part_ptr,
			      struct job_record *job_ptr, bitstr_t * bitmap,
			      uint16_t cr_type,
			      struct node_use_record *node_usage,
			      enum node_cr_state job_node_req)
{
	struct node_record *node_ptr;
	uint32_t i, free_mem, gres_cpus, gres_cores, min_mem;
	int i_first, i_last;
	int core_start_bit, core_end_bit, cpus_per_core;
	List gres_list;

	if (job_ptr->details->pn_min_memory & MEM_PER_CPU)
		min_mem = job_ptr->details->pn_min_memory & (~MEM_PER_CPU);
	else
		min_mem = job_ptr->details->pn_min_memory;
	i_first = bit_ffs(bitmap);
	if (i_first >= 0)
		i_last  = bit_fls(bitmap);
	else
		i_last = -2;
	for (i = i_first; i <= i_last; i++) {
		if (!bit_test(bitmap, i))
			continue;
		node_ptr = select_node_record[i].node_ptr;
		core_start_bit = cr_get_coremap_offset(i);
		core_end_bit   = cr_get_coremap_offset(i+1) - 1;
		cpus_per_core  = select_node_record[i].cpus /
				 (core_end_bit - core_start_bit + 1);
		/* node-level memory check */
		if ((job_ptr->details->pn_min_memory) &&
		    (cr_type & CR_MEMORY)) {
			free_mem  = select_node_record[i].real_memory;
			free_mem -= node_usage[i].alloc_memory;
			if (free_mem < min_mem) {
				debug3("select/serial: node %s no mem %u < %u",
					select_node_record[i].node_ptr->name,
					free_mem, min_mem);
				goto clear_bit;
			}
		}

		/* node-level gres check */
		if (node_usage[i].gres_list)
			gres_list = node_usage[i].gres_list;
		else
			gres_list = node_ptr->gres_list;
		gres_cores = gres_plugin_job_test(job_ptr->gres_list,
						  gres_list, true,
						  NULL, 0, 0, job_ptr->job_id,
						  node_ptr->name);
		gres_cpus = gres_cores;
		if (gres_cpus != NO_VAL)
			gres_cpus *= cpus_per_core;
		if (gres_cpus == 0) {
			debug3("select/serial: node %s lacks gres",
			       node_ptr->name);
			goto clear_bit;
		}

		/* exclusive node check */
		if (node_usage[i].node_state >= NODE_CR_RESERVED) {
			debug3("select/serial: node %s in exclusive use",
			       node_ptr->name);
			goto clear_bit;

		/* non-resource-sharing node check */
		} else if (node_usage[i].node_state >= NODE_CR_ONE_ROW) {
			if ((job_node_req == NODE_CR_RESERVED) ||
			    (job_node_req == NODE_CR_AVAILABLE)) {
				debug3("select/serial: node %s non-sharing",
				       node_ptr->name);
				goto clear_bit;
			}
			/* cannot use this node if it is running jobs
			 * in sharing partitions */
			if (_is_node_busy(cr_part_ptr, i, 1,
					  job_ptr->part_ptr)) {
				debug3("select/serial: node %s sharing?",
				       node_ptr->name);
				goto clear_bit;
			}

		/* node is NODE_CR_AVAILABLE - check job request */
		} else {
			if (job_node_req == NODE_CR_RESERVED) {
				if (_is_node_busy(cr_part_ptr, i, 0,
						  job_ptr->part_ptr)) {
					debug3("select/serial:  node %s busy",
					       node_ptr->name);
					goto clear_bit;
				}
			} else if (job_node_req == NODE_CR_ONE_ROW) {
				/* cannot use this node if it is running jobs
				 * in sharing partitions */
				if (_is_node_busy(cr_part_ptr, i, 1, 
						  job_ptr->part_ptr)) {
					debug3("select/serial: node %s vbusy",
					       node_ptr->name);
					goto clear_bit;
				}
			}
		}
		continue;	/* node is usable, test next node */

clear_bit:	/* This node is not usable by this job */
		bit_clear(bitmap, i);
		if (job_ptr->details->req_node_bitmap &&
		    bit_test(job_ptr->details->req_node_bitmap, i)) {
			return SLURM_ERROR;
		}
	}

	return SLURM_SUCCESS;
}