Merge pull request #878 from dongahn/more-min-max

resource: beefing up min/max count-based matching

resource/policies/base/matcher.cpp

@@ -9,7 +9,7 @@
 \*****************************************************************************/
 
 #include <algorithm>
-
+#include <limits>
 #include "policies/base/matcher.hpp"
 
 namespace Flux {
@@ -140,6 +140,12 @@ unsigned int matcher_util_api_t::calc_count (
     return count;
 }
 
+unsigned int matcher_util_api_t::calc_effective_max (
+    const Flux::Jobspec::Resource &resource) const
+{
+    return calc_count (resource, std::numeric_limits<unsigned int>::max ());
+}
+
 int matcher_util_api_t::register_resource_pair (const std::string &subsystem,
                                                 std::string &r_pair)
 {

resource/policies/base/matcher.hpp

@@ -104,6 +104,15 @@ class matcher_util_api_t
     unsigned int calc_count (const Flux::Jobspec::Resource &resource,
                              unsigned int qual_cnt) const;
 
+    /*! Calculate the effective max count that should be allocated.
+     *
+     *  \param resource  resource section of the jobspec.
+     *  \return          Effective max count; 0 if an error is encountered.
+     */
+    unsigned int calc_effective_max (
+                     const Flux::Jobspec::Resource &resource) const;
+
+
     /*! Set prune filters based on spec. The spec should comply with
      *  the following format:
      *  <HL-resource1:LL-resource1[,HL-resource2:LL-resource2...]...]>

resource/traversers/dfu_impl.cpp

@@ -238,7 +238,7 @@ int dfu_impl_t::match (vtx_t u, const std::vector<Resource> &resources,
                 for (auto &c_resource : resource.with)
                     if (c_resource.type == "slot") {
                         *slot_resource = &c_resource;
-                        *nslots = c_resource.count.min;
+                        *nslots = m_match->calc_effective_max (c_resource);
                     }
             }
             matched = true;
@@ -248,7 +248,7 @@ int dfu_impl_t::match (vtx_t u, const std::vector<Resource> &resources,
             if (matched == true)
                 goto ret;
             *slot_resource = &resource;
-            *nslots = resource.count.min; // slot is limited to only min
+            *nslots = m_match->calc_effective_max (resource);
             matched = true;
         }
     }
@@ -422,7 +422,9 @@ bool dfu_impl_t::is_enough (const subsystem_t &subsystem,
                [&] (const Resource &resource) {
                    unsigned int total = dfu.total_count (subsystem,
                                                          resource.type);
-                   unsigned int required = multiplier * resource.count.min;
+                   unsigned int required = multiplier
+                                           * m_match->calc_effective_max (
+                                                          resource);
                    return total >= required;
                });
 }
@@ -434,7 +436,8 @@ int dfu_impl_t::new_sat_types (const subsystem_t &subsystem,
 {
     for (auto &resource : resources) {
         unsigned int total = dfu.total_count (subsystem, resource.type);
-        unsigned int required = multiplier * resource.count.min;
+        unsigned int required = multiplier
+                                * m_match->calc_effective_max (resource);
         bool sat = total >= required;
         if (sat && sat_types.find (resource.type) == sat_types.end ()) {
             auto ret = sat_types.insert (resource.type);

t/Makefile.am

@@ -77,6 +77,7 @@ TESTS = \
     t3028-resource-grow.t \
     t3029-resource-prune.t \
     t3030-resource-multi.t \
+    t3031-resource-minmax2.t \
     t3300-system-dontblock.t \
     t3301-system-latestart.t \
     t4000-match-params.t \

t/data/resource/commands/min_max2/cmds01.in

@@ -0,0 +1,2 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test001.yaml
+quit

t/data/resource/commands/min_max2/cmds02.in

@@ -0,0 +1,2 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test002.yaml
+quit

t/data/resource/commands/min_max2/cmds03.in

@@ -0,0 +1,2 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test003.yaml
+quit

t/data/resource/commands/min_max2/cmds04.in

@@ -0,0 +1,2 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test004.yaml
+quit

t/data/resource/commands/min_max2/cmds05.in

@@ -0,0 +1,7 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test005.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test006.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test006.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test006.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test006.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test006.yaml
+quit

t/data/resource/commands/min_max2/cmds06.in

@@ -0,0 +1,4 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test007.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test008.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test008.yaml
+quit

t/data/resource/commands/min_max2/cmds08.in

@@ -0,0 +1,2 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test009.yaml
+quit

t/data/resource/commands/min_max2/cmds09.in

@@ -0,0 +1,3 @@
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test005.yaml
+match allocate @TEST_SRCDIR@/data/resource/jobspecs/min_max2/test009.yaml
+quit

t/data/resource/commands/min_max2/sc

@@ -0,0 +1,55 @@
+#cmds012="${cmd_dir}/cmds02.in"
+#test012_desc="min=1/max=99999 on core works (pol=first)"
+#test_expect_success "${test012_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds012} > cmds012 &&
+#    ${query} -L ${grugs} -S CA -P first -t 012.R.out < cmds012 &&
+#    test_cmp 012.R.out ${exp_dir}/012.R.out
+#'
+#
+#cmds013="${cmd_dir}/cmds03.in"
+#test013_desc="min=1/max=18 on core min=1/max=2 on gpu works (pol=first)"
+#test_expect_success "${test013_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds013} > cmds013 &&
+#    ${query} -L ${grugs} -S CA -P first -t 013.R.out < cmds013 &&
+#    test_cmp 013.R.out ${exp_dir}/013.R.out
+#'
+#
+#cmds014="${cmd_dir}/cmds04.in"
+#test014_desc="min=1/max=99999 on core min=1/max=99999 on gpu works (pol=first)"
+#test_expect_success "${test014_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds014} > cmds014 &&
+#    ${query} -L ${grugs} -S CA -P first -t 014.R.out < cmds014 &&
+#    test_cmp 014.R.out ${exp_dir}/014.R.out
+#'
+#
+#cmds015="${cmd_dir}/cmds05.in"
+#test015_desc="min=1/max=18 on core results in count=4 (pol=first)"
+#test_expect_success "${test015_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds015} > cmds015 &&
+#    ${query} -L ${grugs} -S CA -P first -t 015.R.out < cmds015 &&
+#    test_cmp 015.R.out ${exp_dir}/015.R.out
+#'
+#
+#cmds016="${cmd_dir}/cmds06.in"
+#test016_desc="A node exclusive request (node=1) works (pol=first)"
+#test_expect_success "${test016_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds016} > cmds016 &&
+#    ${query} -L ${grugs} -S CA -P first -t 016.R.out < cmds016 &&
+#    test_cmp 016.R.out ${exp_dir}/016.R.out
+#'
+#
+#cmds018="${cmd_dir}/cmds08.in"
+#test018_desc="A multi-level min/max works (pol=first)"
+#test_expect_success "${test018_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds018} > cmds018 &&
+#    ${query} -L ${grugs} -S CA -P first -t 018.R.out < cmds018 &&
+#    test_cmp 018.R.out ${exp_dir}/018.R.out
+#'
+#
+#cmds019="${cmd_dir}/cmds09.in"
+#test019_desc="A multi-level min/max works with an existing allocation (pol=first)"
+#test_expect_success "${test019_desc}" '
+#    sed "s~@TEST_SRCDIR@~${SHARNESS_TEST_SRCDIR}~g" ${cmds019} > cmds019 &&
+#    ${query} -L ${grugs} -S CA -P first -t 019.R.out < cmds019 &&
+#    test_cmp 019.R.out ${exp_dir}/019.R.out
+#'

t/data/resource/expected/min_max/011.R.out

@@ -1,3 +1,22 @@
+      ---------------core0[1:x]
+      ---------------core1[1:x]
+      ---------------core2[1:x]
+      ---------------core3[1:x]
+      ---------------core4[1:x]
+      ---------------core5[1:x]
+      ---------------core6[1:x]
+      ---------------core7[1:x]
+      ---------------core8[1:x]
+      ---------------core9[1:x]
+      ---------------core10[1:x]
+      ---------------core11[1:x]
+      ---------------core12[1:x]
+      ---------------core13[1:x]
+      ---------------core14[1:x]
+      ---------------core15[1:x]
+      ---------------core16[1:x]
+      ---------------core17[1:x]
+      ------------socket0[1:x]
       ---------------core18[1:x]
       ---------------core19[1:x]
       ---------------core20[1:x]
@@ -2196,8 +2215,8 @@ INFO: =============================
       ---------------core16[1:x]
       ---------------core17[1:x]
       ------------socket0[1:x]
-      ---------node107[1:s]
-      ------rack5[1:s]
+      ---------node89[1:s]
+      ------rack4[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=3
@@ -2223,14 +2242,52 @@ INFO: =============================
       ---------------core16[1:x]
       ---------------core17[1:x]
       ------------socket0[1:x]
-      ---------node89[1:s]
-      ------rack4[1:s]
+      ---------------core18[1:x]
+      ---------------core19[1:x]
+      ---------------core20[1:x]
+      ---------------core21[1:x]
+      ---------------core22[1:x]
+      ---------------core23[1:x]
+      ---------------core24[1:x]
+      ---------------core25[1:x]
+      ---------------core26[1:x]
+      ---------------core27[1:x]
+      ---------------core28[1:x]
+      ---------------core29[1:x]
+      ---------------core30[1:x]
+      ---------------core31[1:x]
+      ---------------core32[1:x]
+      ---------------core33[1:x]
+      ---------------core34[1:x]
+      ---------------core35[1:x]
+      ------------socket1[1:x]
+      ---------node71[1:s]
+      ------rack3[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=4
 INFO: RESOURCES=ALLOCATED
 INFO: SCHEDULED AT=Now
 INFO: =============================
+      ---------------core0[1:x]
+      ---------------core1[1:x]
+      ---------------core2[1:x]
+      ---------------core3[1:x]
+      ---------------core4[1:x]
+      ---------------core5[1:x]
+      ---------------core6[1:x]
+      ---------------core7[1:x]
+      ---------------core8[1:x]
+      ---------------core9[1:x]
+      ---------------core10[1:x]
+      ---------------core11[1:x]
+      ---------------core12[1:x]
+      ---------------core13[1:x]
+      ---------------core14[1:x]
+      ---------------core15[1:x]
+      ---------------core16[1:x]
+      ---------------core17[1:x]
+      ------------socket0[1:x]
       ---------------core18[1:x]
       ---------------core19[1:x]
       ---------------core20[1:x]
@@ -2250,8 +2307,8 @@ INFO: =============================
       ---------------core34[1:x]
       ---------------core35[1:x]
       ------------socket1[1:x]
-      ---------node89[1:s]
-      ------rack4[1:s]
+      ---------node70[1:s]
+      ------rack3[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=5

t/data/resource/expected/min_max/012.R.out

@@ -1,3 +1,22 @@
+      ---------------core0[1:x]
+      ---------------core1[1:x]
+      ---------------core2[1:x]
+      ---------------core3[1:x]
+      ---------------core4[1:x]
+      ---------------core5[1:x]
+      ---------------core6[1:x]
+      ---------------core7[1:x]
+      ---------------core8[1:x]
+      ---------------core9[1:x]
+      ---------------core10[1:x]
+      ---------------core11[1:x]
+      ---------------core12[1:x]
+      ---------------core13[1:x]
+      ---------------core14[1:x]
+      ---------------core15[1:x]
+      ---------------core16[1:x]
+      ---------------core17[1:x]
+      ------------socket0[1:x]
       ---------------core18[1:x]
       ---------------core19[1:x]
       ---------------core20[1:x]
@@ -2196,8 +2215,8 @@ INFO: =============================
       ---------------core16[1:x]
       ---------------core17[1:x]
       ------------socket0[1:x]
-      ---------node107[1:s]
-      ------rack5[1:s]
+      ---------node89[1:s]
+      ------rack4[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=3
@@ -2223,14 +2242,52 @@ INFO: =============================
       ---------------core16[1:x]
       ---------------core17[1:x]
       ------------socket0[1:x]
-      ---------node89[1:s]
-      ------rack4[1:s]
+      ---------------core18[1:x]
+      ---------------core19[1:x]
+      ---------------core20[1:x]
+      ---------------core21[1:x]
+      ---------------core22[1:x]
+      ---------------core23[1:x]
+      ---------------core24[1:x]
+      ---------------core25[1:x]
+      ---------------core26[1:x]
+      ---------------core27[1:x]
+      ---------------core28[1:x]
+      ---------------core29[1:x]
+      ---------------core30[1:x]
+      ---------------core31[1:x]
+      ---------------core32[1:x]
+      ---------------core33[1:x]
+      ---------------core34[1:x]
+      ---------------core35[1:x]
+      ------------socket1[1:x]
+      ---------node71[1:s]
+      ------rack3[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=4
 INFO: RESOURCES=ALLOCATED
 INFO: SCHEDULED AT=Now
 INFO: =============================
+      ---------------core0[1:x]
+      ---------------core1[1:x]
+      ---------------core2[1:x]
+      ---------------core3[1:x]
+      ---------------core4[1:x]
+      ---------------core5[1:x]
+      ---------------core6[1:x]
+      ---------------core7[1:x]
+      ---------------core8[1:x]
+      ---------------core9[1:x]
+      ---------------core10[1:x]
+      ---------------core11[1:x]
+      ---------------core12[1:x]
+      ---------------core13[1:x]
+      ---------------core14[1:x]
+      ---------------core15[1:x]
+      ---------------core16[1:x]
+      ---------------core17[1:x]
+      ------------socket0[1:x]
       ---------------core18[1:x]
       ---------------core19[1:x]
       ---------------core20[1:x]
@@ -2250,8 +2307,8 @@ INFO: =============================
       ---------------core34[1:x]
       ---------------core35[1:x]
       ------------socket1[1:x]
-      ---------node89[1:s]
-      ------rack4[1:s]
+      ---------node70[1:s]
+      ------rack3[1:s]
       ---resv_test0[1:s]
 INFO: =============================
 INFO: JOBID=5