From 108457701c783df62033fd4f5f0f034b32dbc89e Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Tue, 12 Jun 2018 17:01:20 -0500
Subject: [PATCH 1/7] [ci skip] Add @record_time decorators to the sumpy FMM.

Depends on inducer/pytools!14.
---
 sumpy/fmm.py | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/sumpy/fmm.py b/sumpy/fmm.py
index ed2a1e4b..e9d4dd1d 100644
--- a/sumpy/fmm.py
+++ b/sumpy/fmm.py
@@ -34,7 +34,7 @@ from six.moves import zip
 import pyopencl as cl
 import pyopencl.array  # noqa
 
-from pytools import memoize_method
+from pytools import memoize_method, record_time
 
 from sumpy import (
         P2EFromSingleBox, P2EFromCSR,
@@ -263,9 +263,11 @@ class SumpyExpansionWrangler(object):
                     dtype=self.dtype)
                 for k in self.code.out_kernels])
 
+    @record_time("timing_data")
     def reorder_sources(self, source_array):
         return source_array.with_queue(self.queue)[self.tree.user_source_ids]
 
+    @record_time("timing_data")
     def reorder_potentials(self, potentials):
         from pytools.obj_array import is_obj_array, with_object_array_or_scalar
         assert is_obj_array(potentials)
@@ -297,6 +299,7 @@ class SumpyExpansionWrangler(object):
 
     # }}}
 
+    @record_time("timing_data")
     def form_multipoles(self,
             level_start_source_box_nrs, source_boxes,
             src_weights):
@@ -330,6 +333,7 @@ class SumpyExpansionWrangler(object):
 
         return mpoles
 
+    @record_time("timing_data")
     def coarsen_multipoles(self,
             level_start_source_parent_box_nrs,
             source_parent_boxes,
@@ -385,6 +389,7 @@ class SumpyExpansionWrangler(object):
 
         return mpoles
 
+    @record_time("timing_data")
     def eval_direct(self, target_boxes, source_box_starts,
             source_box_lists, src_weights):
         pot = self.output_zeros()
@@ -409,6 +414,7 @@ class SumpyExpansionWrangler(object):
 
         return pot
 
+    @record_time("timing_data")
     def multipole_to_local(self,
             level_start_target_box_nrs,
             target_boxes, src_box_starts, src_box_lists,
@@ -448,6 +454,7 @@ class SumpyExpansionWrangler(object):
 
         return local_exps
 
+    @record_time("timing_data")
     def eval_multipoles(self,
             target_boxes_by_source_level, source_boxes_by_level, mpole_exps):
         pot = self.output_zeros()
@@ -498,6 +505,7 @@ class SumpyExpansionWrangler(object):
 
         return pot
 
+    @record_time("timing_data")
     def form_locals(self,
             level_start_target_or_target_parent_box_nrs,
             target_or_target_parent_boxes, starts, lists, src_weights):
@@ -536,6 +544,7 @@ class SumpyExpansionWrangler(object):
 
         return local_exps
 
+    @record_time("timing_data")
     def refine_locals(self,
             level_start_target_or_target_parent_box_nrs,
             target_or_target_parent_boxes,
@@ -577,6 +586,7 @@ class SumpyExpansionWrangler(object):
 
         return local_exps
 
+    @record_time("timing_data")
     def eval_locals(self, level_start_target_box_nrs, target_boxes, local_exps):
         pot = self.output_zeros()
 
@@ -612,6 +622,7 @@ class SumpyExpansionWrangler(object):
 
         return pot
 
+    @record_time("timing_data")
     def finalize_potentials(self, potentials):
         return potentials
 
-- 
GitLab


From 936ce6f388c508424cf01660c7a593502e33b3e0 Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Thu, 21 Jun 2018 17:12:34 -0500
Subject: [PATCH 2/7] [ci skip] Implement timing data collection.

---
 sumpy/fmm.py     | 131 +++++++++++++++++++++++++++++++++++++----------
 test/test_fmm.py |  55 ++++++++++++++++++++
 2 files changed, 158 insertions(+), 28 deletions(-)

diff --git a/sumpy/fmm.py b/sumpy/fmm.py
index e9d4dd1d..2231bd7c 100644
--- a/sumpy/fmm.py
+++ b/sumpy/fmm.py
@@ -34,7 +34,7 @@ from six.moves import zip
 import pyopencl as cl
 import pyopencl.array  # noqa
 
-from pytools import memoize_method, record_time
+from pytools import memoize_method
 
 from sumpy import (
         P2EFromSingleBox, P2EFromCSR,
@@ -145,8 +145,34 @@ class SumpyExpansionWranglerCodeContainer(object):
 # }}}
 
 
+# {{{ timing future
+
+_SECONDS_PER_NANOSECOND = 1e-9
+
+
+class TimingFuture(object):
+
+    def __init__(self, events):
+        self.events = events
+
+    def __call__(self):
+        pyopencl.wait_for_events(self.events)
+
+        result = 0
+        for event in self.events:
+            result += event.profile.end - event.profile.start
+
+        return result * _SECONDS_PER_NANOSECOND
+
+# }}}
+
+
 # {{{ expansion wrangler
 
+class UnableToCollectTimingData(UserWarning):
+    pass
+
+
 class SumpyExpansionWrangler(object):
     """Implements the :class:`boxtree.fmm.ExpansionWranglerInterface`
     by using :mod:`sumpy` expansions/translations.
@@ -175,6 +201,7 @@ class SumpyExpansionWrangler(object):
         self.code = code_container
         self.queue = queue
         self.tree = tree
+        self.issued_timing_data_warning = False
 
         self.dtype = dtype
 
@@ -263,11 +290,9 @@ class SumpyExpansionWrangler(object):
                     dtype=self.dtype)
                 for k in self.code.out_kernels])
 
-    @record_time("timing_data")
     def reorder_sources(self, source_array):
         return source_array.with_queue(self.queue)[self.tree.user_source_ids]
 
-    @record_time("timing_data")
     def reorder_potentials(self, potentials):
         from pytools.obj_array import is_obj_array, with_object_array_or_scalar
         assert is_obj_array(potentials)
@@ -299,15 +324,35 @@ class SumpyExpansionWrangler(object):
 
     # }}}
 
-    @record_time("timing_data")
+    def _update_timing_data(self, description, timing_data, events):
+        if timing_data is None:
+            return
+
+        if not self.queue.properties & cl.command_queue_properties.PROFILING_ENABLE:
+            if not self.issued_timing_data_warning:
+                from warnings import warn
+                warn(
+                        "Profiling was not enabled in the command queue. "
+                        "Timing data will not be collected.",
+                        category=UnableToCollectTimingData,
+                        stacklevel=3)
+
+            self.issued_timing_data_warning = True
+            return
+
+        timing_data["description"] = description
+        timing_data["callback"] = TimingFuture(events)
+
     def form_multipoles(self,
             level_start_source_box_nrs, source_boxes,
-            src_weights):
+            src_weights, timing_data=None):
         mpoles = self.multipole_expansion_zeros()
 
         kwargs = self.extra_kwargs.copy()
         kwargs.update(self.box_source_list_kwargs())
 
+        events = []
+
         for lev in range(self.tree.nlevels):
             p2m = self.code.p2m(self.level_orders[lev])
             start, stop = level_start_source_box_nrs[lev:lev+2]
@@ -328,19 +373,21 @@ class SumpyExpansionWrangler(object):
                     rscale=level_to_rscale(self.tree, lev),
 
                     **kwargs)
+            events.append(evt)
 
             assert mpoles_res is mpoles_view
 
+        self._update_timing_data("form_multipoles", timing_data, events)
+
         return mpoles
 
-    @record_time("timing_data")
     def coarsen_multipoles(self,
             level_start_source_parent_box_nrs,
             source_parent_boxes,
-            mpoles):
+            mpoles, timing_data=None):
         tree = self.tree
 
-        evt = None
+        events = []
 
         # nlevels-1 is the last valid level index
         # nlevels-2 is the last valid level that could have children
@@ -382,16 +429,19 @@ class SumpyExpansionWrangler(object):
                     tgt_rscale=level_to_rscale(self.tree, target_level),
 
                     **self.kernel_extra_kwargs)
+            events.append(evt)
+
             assert mpoles_res is target_mpoles_view
 
-        if evt is not None:
-            mpoles.add_event(evt)
+        if events:
+            mpoles.add_event(events[-1])
+
+        self._update_timing_data("coarsen_multipoles", timing_data, events)
 
         return mpoles
 
-    @record_time("timing_data")
     def eval_direct(self, target_boxes, source_box_starts,
-            source_box_lists, src_weights):
+            source_box_lists, src_weights, timing_data=None):
         pot = self.output_zeros()
 
         kwargs = self.extra_kwargs.copy()
@@ -399,6 +449,8 @@ class SumpyExpansionWrangler(object):
         kwargs.update(self.box_source_list_kwargs())
         kwargs.update(self.box_target_list_kwargs())
 
+        events = []
+
         evt, pot_res = self.code.p2p()(self.queue,
                 target_boxes=target_boxes,
                 source_box_starts=source_box_starts,
@@ -407,20 +459,24 @@ class SumpyExpansionWrangler(object):
                 result=pot,
 
                 **kwargs)
+        events.append(evt)
 
         for pot_i, pot_res_i in zip(pot, pot_res):
             assert pot_i is pot_res_i
             pot_i.add_event(evt)
 
+        self._update_timing_data("eval_direct", timing_data, events)
+
         return pot
 
-    @record_time("timing_data")
     def multipole_to_local(self,
             level_start_target_box_nrs,
             target_boxes, src_box_starts, src_box_lists,
-            mpole_exps):
+            mpole_exps, timing_data=None):
         local_exps = self.local_expansion_zeros()
 
+        events = []
+
         for lev in range(self.tree.nlevels):
             start, stop = level_start_target_box_nrs[lev:lev+2]
             if start == stop:
@@ -451,20 +507,24 @@ class SumpyExpansionWrangler(object):
                     tgt_rscale=level_to_rscale(self.tree, lev),
 
                     **self.kernel_extra_kwargs)
+            events.append(evt)
+
+        self._update_timing_data("eval_direct", timing_data, events)
 
         return local_exps
 
-    @record_time("timing_data")
     def eval_multipoles(self,
-            target_boxes_by_source_level, source_boxes_by_level, mpole_exps):
+            target_boxes_by_source_level, source_boxes_by_level, mpole_exps,
+            timing_data=None):
         pot = self.output_zeros()
 
         kwargs = self.kernel_extra_kwargs.copy()
         kwargs.update(self.box_target_list_kwargs())
 
+        events = []
+
         wait_for = mpole_exps.events
 
-        has_evt = False
         for isrc_level, ssn in enumerate(source_boxes_by_level):
             if len(target_boxes_by_source_level[isrc_level]) == 0:
                 continue
@@ -491,29 +551,32 @@ class SumpyExpansionWrangler(object):
                     wait_for=wait_for,
 
                     **kwargs)
+            events.append(evt)
 
-            has_evt = True
             wait_for = [evt]
 
             for pot_i, pot_res_i in zip(pot, pot_res):
                 assert pot_i is pot_res_i
 
-        if has_evt:
+        if events:
             for pot_i in pot:
-                # Intentionally only adding the last event.
-                pot_i.add_event(evt)
+                pot_i.add_event(events[-1])
+
+        self._update_timing_data("eval_multipoles", timing_data, events)
 
         return pot
 
-    @record_time("timing_data")
     def form_locals(self,
             level_start_target_or_target_parent_box_nrs,
-            target_or_target_parent_boxes, starts, lists, src_weights):
+            target_or_target_parent_boxes, starts, lists, src_weights,
+            timing_data=None):
         local_exps = self.local_expansion_zeros()
 
         kwargs = self.extra_kwargs.copy()
         kwargs.update(self.box_source_list_kwargs())
 
+        events = []
+
         for lev in range(self.tree.nlevels):
             start, stop = \
                     level_start_target_or_target_parent_box_nrs[lev:lev+2]
@@ -539,16 +602,21 @@ class SumpyExpansionWrangler(object):
                     rscale=level_to_rscale(self.tree, lev),
 
                     **kwargs)
+            events.append(evt)
 
             assert result is target_local_exps_view
 
+        self._update_timing_data("form_locals", timing_data, events)
+
         return local_exps
 
-    @record_time("timing_data")
     def refine_locals(self,
             level_start_target_or_target_parent_box_nrs,
             target_or_target_parent_boxes,
-            local_exps):
+            local_exps, timing_data=None):
+
+        events = []
+
         for target_lev in range(1, self.tree.nlevels):
             start, stop = level_start_target_or_target_parent_box_nrs[
                     target_lev:target_lev+2]
@@ -579,20 +647,25 @@ class SumpyExpansionWrangler(object):
                     tgt_rscale=level_to_rscale(self.tree, target_lev),
 
                     **self.kernel_extra_kwargs)
+            events.append(evt)
 
             assert local_exps_res is target_local_exps_view
 
         local_exps.add_event(evt)
 
+        self._update_timing_data("refine_locals", timing_data, events)
+
         return local_exps
 
-    @record_time("timing_data")
-    def eval_locals(self, level_start_target_box_nrs, target_boxes, local_exps):
+    def eval_locals(self, level_start_target_box_nrs, target_boxes, local_exps,
+            timing_data=None):
         pot = self.output_zeros()
 
         kwargs = self.kernel_extra_kwargs.copy()
         kwargs.update(self.box_target_list_kwargs())
 
+        events = []
+
         for lev in range(self.tree.nlevels):
             start, stop = level_start_target_box_nrs[lev:lev+2]
             if start == stop:
@@ -616,13 +689,15 @@ class SumpyExpansionWrangler(object):
                     rscale=level_to_rscale(self.tree, lev),
 
                     **kwargs)
+            events.append(evt)
 
             for pot_i, pot_res_i in zip(pot, pot_res):
                 assert pot_i is pot_res_i
 
+        self._update_timing_data("eval_locals", timing_data, events)
+
         return pot
 
-    @record_time("timing_data")
     def finalize_potentials(self, potentials):
         return potentials
 
diff --git a/test/test_fmm.py b/test/test_fmm.py
index 0331db6c..5ec3498c 100644
--- a/test/test_fmm.py
+++ b/test/test_fmm.py
@@ -234,6 +234,61 @@ def test_sumpy_fmm(ctx_getter, knl, local_expn_class, mpole_expn_class):
     pconv_verifier()
 
 
+def test_sumpy_fmm_timing_data(ctx_getter):
+    logging.basicConfig(level=logging.INFO)
+
+    ctx = ctx_getter()
+    queue = cl.CommandQueue(
+            ctx,
+            properties=cl.command_queue_properties.PROFILING_ENABLE)
+
+    nsources = 500
+    dtype = np.float64
+
+    from boxtree.tools import (
+            make_normal_particle_array as p_normal)
+
+    knl = LaplaceKernel(2)
+    local_expn_class = VolumeTaylorLocalExpansion
+    mpole_expn_class = VolumeTaylorMultipoleExpansion
+    order = 1
+
+    sources = p_normal(queue, nsources, knl.dim, dtype, seed=15)
+
+    from boxtree import TreeBuilder
+    tb = TreeBuilder(ctx)
+
+    tree, _ = tb(queue, sources,
+            max_particles_in_box=30, debug=True)
+
+    from boxtree.traversal import FMMTraversalBuilder
+    tbuild = FMMTraversalBuilder(ctx)
+    trav, _ = tbuild(queue, tree, debug=True)
+
+    from pyopencl.clrandom import PhiloxGenerator
+    rng = PhiloxGenerator(ctx)
+    weights = rng.uniform(queue, nsources, dtype=np.float64)
+
+    out_kernels = [knl]
+
+    from functools import partial
+
+    from sumpy.fmm import SumpyExpansionWranglerCodeContainer
+    wcc = SumpyExpansionWranglerCodeContainer(
+            ctx,
+            partial(mpole_expn_class, knl),
+            partial(local_expn_class, knl),
+            out_kernels)
+
+    wrangler = wcc.get_wrangler(queue, tree, dtype,
+            fmm_level_to_order=lambda kernel, kernel_args, tree, lev: order)
+    from boxtree.fmm import drive_fmm
+
+    timing_data = {}
+    pot, = drive_fmm(trav, wrangler, weights, timing_data=timing_data)
+    assert timing_data
+
+
 def test_sumpy_fmm_exclude_self(ctx_getter):
     logging.basicConfig(level=logging.INFO)
 
-- 
GitLab


From cd467c415befc6bad4edcbe0cc4ac02309cbff35 Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Fri, 6 Jul 2018 17:46:18 -0500
Subject: [PATCH 3/7] Update for boxtree changes.

---
 requirements.txt |  2 +-
 setup.py         |  2 +-
 sumpy/fmm.py     | 16 +++++++++++-----
 test/test_fmm.py |  1 +
 4 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/requirements.txt b/requirements.txt
index 1e86e610..0acdd382 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,6 +3,6 @@ sympy==1.0
 git+https://github.com/inducer/pymbolic
 git+https://github.com/inducer/islpy
 git+https://github.com/inducer/pyopencl
-git+https://gitlab.tiker.net/inducer/boxtree
+git+https://gitlab.tiker.net/inducer/boxtree@timing-data
 git+https://github.com/inducer/loopy
 git+https://github.com/inducer/pyfmmlib
diff --git a/setup.py b/setup.py
index 8d085054..9457bbdc 100644
--- a/setup.py
+++ b/setup.py
@@ -94,7 +94,7 @@ setup(name="sumpy",
       install_requires=[
           "pytools>=2018.2",
           "loo.py>=2017.2",
-          "boxtree>=2013.1",
+          "boxtree>=2018.1",
           "pytest>=2.3",
           "six",
 
diff --git a/sumpy/fmm.py b/sumpy/fmm.py
index 2231bd7c..52748c7d 100644
--- a/sumpy/fmm.py
+++ b/sumpy/fmm.py
@@ -155,14 +155,20 @@ class TimingFuture(object):
     def __init__(self, events):
         self.events = events
 
-    def __call__(self):
+    @memoize_method
+    def get(self):
         pyopencl.wait_for_events(self.events)
 
         result = 0
         for event in self.events:
-            result += event.profile.end - event.profile.start
+            result += (
+                    (event.profile.end - event.profile.start)
+                    * _SECONDS_PER_NANOSECOND)
+        return result
 
-        return result * _SECONDS_PER_NANOSECOND
+    def __call__(self):
+        from boxtree.fmm import TimingResult
+        return TimingResult(wall_elapsed=self.get(), process_elapsed=None)
 
 # }}}
 
@@ -340,8 +346,8 @@ class SumpyExpansionWrangler(object):
             self.issued_timing_data_warning = True
             return
 
-        timing_data["description"] = description
-        timing_data["callback"] = TimingFuture(events)
+        timing_data.description = description
+        timing_data.callback = TimingFuture(events)
 
     def form_multipoles(self,
             level_start_source_box_nrs, source_boxes,
diff --git a/test/test_fmm.py b/test/test_fmm.py
index 5ec3498c..0c4406bf 100644
--- a/test/test_fmm.py
+++ b/test/test_fmm.py
@@ -286,6 +286,7 @@ def test_sumpy_fmm_timing_data(ctx_getter):
 
     timing_data = {}
     pot, = drive_fmm(trav, wrangler, weights, timing_data=timing_data)
+    print(timing_data)
     assert timing_data
 
 
-- 
GitLab


From faa80af925f30d600be343b2616e1f1e9f6d2568 Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Fri, 6 Jul 2018 18:08:22 -0500
Subject: [PATCH 4/7] Rename _update_timing_data to update_timing_data, as
 subclasses are expected to use it

---
 sumpy/fmm.py | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/sumpy/fmm.py b/sumpy/fmm.py
index 52748c7d..cd3ed0af 100644
--- a/sumpy/fmm.py
+++ b/sumpy/fmm.py
@@ -330,7 +330,7 @@ class SumpyExpansionWrangler(object):
 
     # }}}
 
-    def _update_timing_data(self, description, timing_data, events):
+    def update_timing_data(self, description, timing_data, events):
         if timing_data is None:
             return
 
@@ -383,7 +383,7 @@ class SumpyExpansionWrangler(object):
 
             assert mpoles_res is mpoles_view
 
-        self._update_timing_data("form_multipoles", timing_data, events)
+        self.update_timing_data("form_multipoles", timing_data, events)
 
         return mpoles
 
@@ -442,7 +442,7 @@ class SumpyExpansionWrangler(object):
         if events:
             mpoles.add_event(events[-1])
 
-        self._update_timing_data("coarsen_multipoles", timing_data, events)
+        self.update_timing_data("coarsen_multipoles", timing_data, events)
 
         return mpoles
 
@@ -471,7 +471,7 @@ class SumpyExpansionWrangler(object):
             assert pot_i is pot_res_i
             pot_i.add_event(evt)
 
-        self._update_timing_data("eval_direct", timing_data, events)
+        self.update_timing_data("eval_direct", timing_data, events)
 
         return pot
 
@@ -515,7 +515,7 @@ class SumpyExpansionWrangler(object):
                     **self.kernel_extra_kwargs)
             events.append(evt)
 
-        self._update_timing_data("eval_direct", timing_data, events)
+        self.update_timing_data("eval_direct", timing_data, events)
 
         return local_exps
 
@@ -568,7 +568,7 @@ class SumpyExpansionWrangler(object):
             for pot_i in pot:
                 pot_i.add_event(events[-1])
 
-        self._update_timing_data("eval_multipoles", timing_data, events)
+        self.update_timing_data("eval_multipoles", timing_data, events)
 
         return pot
 
@@ -612,7 +612,7 @@ class SumpyExpansionWrangler(object):
 
             assert result is target_local_exps_view
 
-        self._update_timing_data("form_locals", timing_data, events)
+        self.update_timing_data("form_locals", timing_data, events)
 
         return local_exps
 
@@ -659,7 +659,7 @@ class SumpyExpansionWrangler(object):
 
         local_exps.add_event(evt)
 
-        self._update_timing_data("refine_locals", timing_data, events)
+        self.update_timing_data("refine_locals", timing_data, events)
 
         return local_exps
 
@@ -700,7 +700,7 @@ class SumpyExpansionWrangler(object):
             for pot_i, pot_res_i in zip(pot, pot_res):
                 assert pot_i is pot_res_i
 
-        self._update_timing_data("eval_locals", timing_data, events)
+        self.update_timing_data("eval_locals", timing_data, events)
 
         return pot
 
-- 
GitLab


From 00d96cb4f4d79c6515247100071503dd8e40f140 Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Fri, 6 Jul 2018 18:13:09 -0500
Subject: [PATCH 5/7] Fix conda requirements.txt to point to timing-data branch
 of boxtree

---
 .test-conda-env-py3-requirements.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.test-conda-env-py3-requirements.txt b/.test-conda-env-py3-requirements.txt
index 45c20c1f..8c9f55d9 100644
--- a/.test-conda-env-py3-requirements.txt
+++ b/.test-conda-env-py3-requirements.txt
@@ -1,3 +1,3 @@
-git+https://gitlab.tiker.net/inducer/boxtree
+git+https://gitlab.tiker.net/inducer/boxtree@timing-data
 git+https://github.com/inducer/pymbolic
 git+https://github.com/inducer/loopy
-- 
GitLab


From d18375016671751d039bf3a8ecd37504193e4fd7 Mon Sep 17 00:00:00 2001
From: Matt Wala <wala1@illinois.edu>
Date: Mon, 9 Jul 2018 23:02:48 -0500
Subject: [PATCH 6/7] Update for boxtree changes

---
 sumpy/fmm.py     | 107 +++++++++++++++++++----------------------------
 test/test_fmm.py |   2 +-
 2 files changed, 43 insertions(+), 66 deletions(-)

diff --git a/sumpy/fmm.py b/sumpy/fmm.py
index cd3ed0af..17d52eed 100644
--- a/sumpy/fmm.py
+++ b/sumpy/fmm.py
@@ -150,13 +150,29 @@ class SumpyExpansionWranglerCodeContainer(object):
 _SECONDS_PER_NANOSECOND = 1e-9
 
 
-class TimingFuture(object):
+class UnableToCollectTimingData(UserWarning):
+    pass
+
+
+class SumpyTimingFuture(object):
 
-    def __init__(self, events):
+    def __init__(self, queue, events):
+        self.queue = queue
         self.events = events
 
     @memoize_method
-    def get(self):
+    def result(self):
+        from boxtree.fmm import TimingResult
+
+        if not self.queue.properties & cl.command_queue_properties.PROFILING_ENABLE:
+            from warnings import warn
+            warn(
+                    "Profiling was not enabled in the command queue. "
+                    "Timing data will not be collected.",
+                    category=UnableToCollectTimingData,
+                    stacklevel=3)
+            return TimingResult(wall_elapsed=None, process_elapsed=None)
+
         pyopencl.wait_for_events(self.events)
 
         result = 0
@@ -164,21 +180,20 @@ class TimingFuture(object):
             result += (
                     (event.profile.end - event.profile.start)
                     * _SECONDS_PER_NANOSECOND)
-        return result
 
-    def __call__(self):
-        from boxtree.fmm import TimingResult
-        return TimingResult(wall_elapsed=self.get(), process_elapsed=None)
+        return TimingResult(wall_elapsed=result, process_elapsed=None)
+
+    def done(self):
+        return all(
+                event.get_info(cl.event_info.COMMAND_EXECUTION_STATUS)
+                == cl.command_execution_status.COMPLETE
+                for event in self.events)
 
 # }}}
 
 
 # {{{ expansion wrangler
 
-class UnableToCollectTimingData(UserWarning):
-    pass
-
-
 class SumpyExpansionWrangler(object):
     """Implements the :class:`boxtree.fmm.ExpansionWranglerInterface`
     by using :mod:`sumpy` expansions/translations.
@@ -330,28 +345,9 @@ class SumpyExpansionWrangler(object):
 
     # }}}
 
-    def update_timing_data(self, description, timing_data, events):
-        if timing_data is None:
-            return
-
-        if not self.queue.properties & cl.command_queue_properties.PROFILING_ENABLE:
-            if not self.issued_timing_data_warning:
-                from warnings import warn
-                warn(
-                        "Profiling was not enabled in the command queue. "
-                        "Timing data will not be collected.",
-                        category=UnableToCollectTimingData,
-                        stacklevel=3)
-
-            self.issued_timing_data_warning = True
-            return
-
-        timing_data.description = description
-        timing_data.callback = TimingFuture(events)
-
     def form_multipoles(self,
             level_start_source_box_nrs, source_boxes,
-            src_weights, timing_data=None):
+            src_weights):
         mpoles = self.multipole_expansion_zeros()
 
         kwargs = self.extra_kwargs.copy()
@@ -383,14 +379,12 @@ class SumpyExpansionWrangler(object):
 
             assert mpoles_res is mpoles_view
 
-        self.update_timing_data("form_multipoles", timing_data, events)
-
-        return mpoles
+        return (mpoles, SumpyTimingFuture(self.queue, events))
 
     def coarsen_multipoles(self,
             level_start_source_parent_box_nrs,
             source_parent_boxes,
-            mpoles, timing_data=None):
+            mpoles):
         tree = self.tree
 
         events = []
@@ -442,12 +436,10 @@ class SumpyExpansionWrangler(object):
         if events:
             mpoles.add_event(events[-1])
 
-        self.update_timing_data("coarsen_multipoles", timing_data, events)
-
-        return mpoles
+        return (mpoles, SumpyTimingFuture(self.queue, events))
 
     def eval_direct(self, target_boxes, source_box_starts,
-            source_box_lists, src_weights, timing_data=None):
+            source_box_lists, src_weights):
         pot = self.output_zeros()
 
         kwargs = self.extra_kwargs.copy()
@@ -471,14 +463,12 @@ class SumpyExpansionWrangler(object):
             assert pot_i is pot_res_i
             pot_i.add_event(evt)
 
-        self.update_timing_data("eval_direct", timing_data, events)
-
-        return pot
+        return (pot, SumpyTimingFuture(self.queue, events))
 
     def multipole_to_local(self,
             level_start_target_box_nrs,
             target_boxes, src_box_starts, src_box_lists,
-            mpole_exps, timing_data=None):
+            mpole_exps):
         local_exps = self.local_expansion_zeros()
 
         events = []
@@ -515,13 +505,10 @@ class SumpyExpansionWrangler(object):
                     **self.kernel_extra_kwargs)
             events.append(evt)
 
-        self.update_timing_data("eval_direct", timing_data, events)
-
-        return local_exps
+        return (local_exps, SumpyTimingFuture(self.queue, events))
 
     def eval_multipoles(self,
-            target_boxes_by_source_level, source_boxes_by_level, mpole_exps,
-            timing_data=None):
+            target_boxes_by_source_level, source_boxes_by_level, mpole_exps):
         pot = self.output_zeros()
 
         kwargs = self.kernel_extra_kwargs.copy()
@@ -568,14 +555,11 @@ class SumpyExpansionWrangler(object):
             for pot_i in pot:
                 pot_i.add_event(events[-1])
 
-        self.update_timing_data("eval_multipoles", timing_data, events)
-
-        return pot
+        return (pot, SumpyTimingFuture(self.queue, events))
 
     def form_locals(self,
             level_start_target_or_target_parent_box_nrs,
-            target_or_target_parent_boxes, starts, lists, src_weights,
-            timing_data=None):
+            target_or_target_parent_boxes, starts, lists, src_weights):
         local_exps = self.local_expansion_zeros()
 
         kwargs = self.extra_kwargs.copy()
@@ -612,14 +596,12 @@ class SumpyExpansionWrangler(object):
 
             assert result is target_local_exps_view
 
-        self.update_timing_data("form_locals", timing_data, events)
-
-        return local_exps
+        return (local_exps, SumpyTimingFuture(self.queue, events))
 
     def refine_locals(self,
             level_start_target_or_target_parent_box_nrs,
             target_or_target_parent_boxes,
-            local_exps, timing_data=None):
+            local_exps):
 
         events = []
 
@@ -659,12 +641,9 @@ class SumpyExpansionWrangler(object):
 
         local_exps.add_event(evt)
 
-        self.update_timing_data("refine_locals", timing_data, events)
-
-        return local_exps
+        return (local_exps, SumpyTimingFuture(self.queue, [evt]))
 
-    def eval_locals(self, level_start_target_box_nrs, target_boxes, local_exps,
-            timing_data=None):
+    def eval_locals(self, level_start_target_box_nrs, target_boxes, local_exps):
         pot = self.output_zeros()
 
         kwargs = self.kernel_extra_kwargs.copy()
@@ -700,9 +679,7 @@ class SumpyExpansionWrangler(object):
             for pot_i, pot_res_i in zip(pot, pot_res):
                 assert pot_i is pot_res_i
 
-        self.update_timing_data("eval_locals", timing_data, events)
-
-        return pot
+        return (pot, SumpyTimingFuture(self.queue, events))
 
     def finalize_potentials(self, potentials):
         return potentials
diff --git a/test/test_fmm.py b/test/test_fmm.py
index 0c4406bf..71e3f044 100644
--- a/test/test_fmm.py
+++ b/test/test_fmm.py
@@ -234,7 +234,7 @@ def test_sumpy_fmm(ctx_getter, knl, local_expn_class, mpole_expn_class):
     pconv_verifier()
 
 
-def test_sumpy_fmm_timing_data(ctx_getter):
+def test_sumpy_fmm_timing_data_collection(ctx_getter):
     logging.basicConfig(level=logging.INFO)
 
     ctx = ctx_getter()
-- 
GitLab


From 54cef6b672d99639a4c0c5a9f26f73c04949ceab Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Andreas=20Kl=C3=B6ckner?= <inform@tiker.net>
Date: Wed, 11 Jul 2018 15:43:05 -0400
Subject: [PATCH 7/7] Point requirements.txt back at boxtree master

---
 .test-conda-env-py3-requirements.txt | 2 +-
 requirements.txt                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/.test-conda-env-py3-requirements.txt b/.test-conda-env-py3-requirements.txt
index 8c9f55d9..45c20c1f 100644
--- a/.test-conda-env-py3-requirements.txt
+++ b/.test-conda-env-py3-requirements.txt
@@ -1,3 +1,3 @@
-git+https://gitlab.tiker.net/inducer/boxtree@timing-data
+git+https://gitlab.tiker.net/inducer/boxtree
 git+https://github.com/inducer/pymbolic
 git+https://github.com/inducer/loopy
diff --git a/requirements.txt b/requirements.txt
index 3a159527..8e48bc1c 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,6 +3,6 @@ sympy==1.1.1
 git+https://github.com/inducer/pymbolic
 git+https://github.com/inducer/islpy
 git+https://github.com/inducer/pyopencl
-git+https://gitlab.tiker.net/inducer/boxtree@timing-data
+git+https://gitlab.tiker.net/inducer/boxtree
 git+https://github.com/inducer/loopy
 git+https://github.com/inducer/pyfmmlib
-- 
GitLab