def test_merge_several():
def merge(n, m):
r = -1
if n == 0:
if m == 0:
r = 0
elif m == 1:
r = 1
else:
r = 2
elif n == 1:
r = 4
else:
r = 6
return r
t = TranslationContext()
a = t.buildannotator()
a.build_types(merge, [int, int])
rtyper = t.buildrtyper()
rtyper.specialize()
graph = tgraphof(t, merge)
remove_same_as(graph)
merge_if_blocks(graph)
assert len(graph.startblock.exits) == 3
assert len(list(graph.iterblocks())) == 3
interp = LLInterpreter(rtyper)
for m in range(3):
res = interp.eval_graph(graph, [0, m])
assert res == m
res = interp.eval_graph(graph, [1, 0])
assert res == 4
res = interp.eval_graph(graph, [2, 0])
assert res == 6
def test_merge_several():
def merge(n, m):
r = -1
if n == 0:
if m == 0:
r = 0
elif m == 1:
r = 1
else:
r = 2
elif n == 1:
r = 4
else:
r = 6
return r
t = TranslationContext()
a = t.buildannotator()
a.build_types(merge, [int, int])
rtyper = t.buildrtyper()
rtyper.specialize()
graph = tgraphof(t, merge)
remove_same_as(graph)
merge_if_blocks(graph)
assert len(graph.startblock.exits) == 3
assert len(list(graph.iterblocks())) == 3
interp = LLInterpreter(rtyper)
for m in range(3):
res = interp.eval_graph(graph, [0, m])
assert res == m
res = interp.eval_graph(graph, [1, 0])
assert res == 4
res = interp.eval_graph(graph, [2, 0])
assert res == 6
def get_generated_c_source(fn, types):
"""Return the generated C source for fn."""
t = Translation(fn, types, backend="c")
t.annotate()
merge_if_blocks(t.driver.translator.graphs[0])
c_filename_path = t.source_c()
return t.driver.cbuilder.c_source_filename.join('..',
'rpython_translator_c_test.c').read()
def get_generated_c_source(fn, types):
"""Return the generated C source for fn."""
t = Translation(fn, types, backend="c")
t.annotate()
merge_if_blocks(t.driver.translator.graphs[0])
c_filename_path = t.source_c()
return t.driver.cbuilder.c_source_filename.join(
'..', 'rpython_translator_c_test.c').read()
def test_two_constants():
def fn():
r = range(10, 37, 4)
r.reverse()
return r[0]
t = TranslationContext()
a = t.buildannotator()
a.build_types(fn, [])
rtyper = t.buildrtyper()
rtyper.specialize()
backend_optimizations(t, merge_if_blocks=True)
graph = tgraphof(t, fn)
blocknum = len(list(graph.iterblocks()))
merge_if_blocks(graph)
assert blocknum == len(list(graph.iterblocks()))
def test_two_constants():
def fn():
r = range(10, 37, 4)
r.reverse()
return r[0]
t = TranslationContext()
a = t.buildannotator()
a.build_types(fn, [])
rtyper = t.buildrtyper()
rtyper.specialize()
backend_optimizations(t, merge_if_blocks=True)
graph = tgraphof(t, fn)
blocknum = len(list(graph.iterblocks()))
merge_if_blocks(graph)
assert blocknum == len(list(graph.iterblocks()))
def test_dont_merge():
def merge(n, m):
r = -1
if n == 0:
r += m
if n == 1:
r += 2 * m
else:
r += 6
return r
t = TranslationContext()
a = t.buildannotator()
a.build_types(merge, [int, int])
rtyper = t.buildrtyper()
rtyper.specialize()
graph = tgraphof(t, merge)
remove_same_as(graph)
blocknum = len(list(graph.iterblocks()))
merge_if_blocks(graph)
assert blocknum == len(list(graph.iterblocks()))
def test_dont_merge():
def merge(n, m):
r = -1
if n == 0:
r += m
if n == 1:
r += 2 * m
else:
r += 6
return r
t = TranslationContext()
a = t.buildannotator()
a.build_types(merge, [int, int])
rtyper = t.buildrtyper()
rtyper.specialize()
graph = tgraphof(t, merge)
remove_same_as(graph)
blocknum = len(list(graph.iterblocks()))
merge_if_blocks(graph)
assert blocknum == len(list(graph.iterblocks()))
def backend_optimizations(translator, graphs=None, secondary=False,
inline_graph_from_anywhere=False, **kwds):
# sensible keywords are
# raisingop2direct_call, inline_threshold, mallocs
# merge_if_blocks, constfold, heap2stack
# clever_malloc_removal, remove_asserts
config = translator.config.translation.backendopt.copy(as_default=True)
config.set(**kwds)
if graphs is None:
graphs = translator.graphs
for graph in graphs:
assert not hasattr(graph, '_seen_by_the_backend')
if config.print_statistics:
print "before optimizations:"
print_statistics(translator.graphs[0], translator, "per-graph.txt")
if config.raisingop2direct_call:
raisingop2direct_call(translator, graphs)
if config.remove_asserts:
constfold(config, graphs)
remove_asserts(translator, graphs)
if config.really_remove_asserts:
for graph in graphs:
removenoops.remove_debug_assert(graph)
# the dead operations will be killed by the remove_obvious_noops below
# remove obvious no-ops
def remove_obvious_noops():
for graph in graphs:
removenoops.remove_same_as(graph)
simplify.eliminate_empty_blocks(graph)
simplify.transform_dead_op_vars(graph, translator)
removenoops.remove_duplicate_casts(graph, translator)
if config.print_statistics:
print "after no-op removal:"
print_statistics(translator.graphs[0], translator)
remove_obvious_noops()
if config.inline or config.mallocs:
heuristic = get_function(config.inline_heuristic)
if config.inline:
threshold = config.inline_threshold
else:
threshold = 0
inline_malloc_removal_phase(config, translator, graphs,
threshold,
inline_heuristic=heuristic,
inline_graph_from_anywhere=inline_graph_from_anywhere)
constfold(config, graphs)
if config.clever_malloc_removal:
threshold = config.clever_malloc_removal_threshold
heuristic = get_function(config.clever_malloc_removal_heuristic)
log.inlineandremove("phase with threshold factor: %s" % threshold)
log.inlineandremove("heuristic: %s.%s" % (heuristic.__module__,
heuristic.__name__))
count = mallocprediction.clever_inlining_and_malloc_removal(
translator, graphs,
threshold = threshold,
heuristic=heuristic)
log.inlineandremove("removed %d simple mallocs in total" % count)
constfold(config, graphs)
if config.print_statistics:
print "after clever inlining and malloc removal"
print_statistics(translator.graphs[0], translator)
if config.storesink:
for graph in graphs:
storesink_graph(graph)
if config.profile_based_inline and not secondary:
threshold = config.profile_based_inline_threshold
heuristic = get_function(config.profile_based_inline_heuristic)
inline.instrument_inline_candidates(graphs, threshold)
counters = translator.driver_instrument_result(
config.profile_based_inline)
n = len(counters)
def call_count_pred(label):
if label >= n:
return False
return counters[label] > 250 # xxx introduce an option for this
inline_malloc_removal_phase(config, translator, graphs,
threshold,
inline_heuristic=heuristic,
call_count_pred=call_count_pred)
constfold(config, graphs)
if config.merge_if_blocks:
log.mergeifblocks("starting to merge if blocks")
for graph in graphs:
merge_if_blocks(graph, translator.config.translation.verbose)
if config.print_statistics:
print "after if-to-switch:"
print_statistics(translator.graphs[0], translator)
remove_obvious_noops()
for graph in graphs:
checkgraph(graph)
gilanalysis.analyze(graphs, translator)
def backend_optimizations(translator, graphs=None, secondary=False,
inline_graph_from_anywhere=False, **kwds):
# sensible keywords are
# inline_threshold, mallocs
# merge_if_blocks, constfold, heap2stack
# clever_malloc_removal, remove_asserts
# replace_we_are_jitted
config = translator.config.translation.backendopt.copy(as_default=True)
config.set(**kwds)
if graphs is None:
graphs = translator.graphs
for graph in graphs:
assert not hasattr(graph, '_seen_by_the_backend')
if config.print_statistics:
print "before optimizations:"
print_statistics(translator.graphs[0], translator, "per-graph.txt")
if config.replace_we_are_jitted:
for graph in graphs:
replace_we_are_jitted(graph)
if config.remove_asserts:
constfold(config, graphs)
remove_asserts(translator, graphs)
if config.really_remove_asserts:
for graph in graphs:
removenoops.remove_debug_assert(graph)
# the dead operations will be killed by the remove_obvious_noops below
# remove obvious no-ops
def remove_obvious_noops():
for graph in graphs:
removenoops.remove_same_as(graph)
simplify.eliminate_empty_blocks(graph)
simplify.transform_dead_op_vars(graph, translator)
removenoops.remove_duplicate_casts(graph, translator)
if config.print_statistics:
print "after no-op removal:"
print_statistics(translator.graphs[0], translator)
remove_obvious_noops()
if config.inline or config.mallocs:
heuristic = get_function(config.inline_heuristic)
if config.inline:
threshold = config.inline_threshold
else:
threshold = 0
inline_malloc_removal_phase(config, translator, graphs,
threshold,
inline_heuristic=heuristic,
inline_graph_from_anywhere=inline_graph_from_anywhere)
constfold(config, graphs)
if config.clever_malloc_removal:
threshold = config.clever_malloc_removal_threshold
heuristic = get_function(config.clever_malloc_removal_heuristic)
log.inlineandremove("phase with threshold factor: %s" % threshold)
log.inlineandremove("heuristic: %s.%s" % (heuristic.__module__,
heuristic.__name__))
count = mallocprediction.clever_inlining_and_malloc_removal(
translator, graphs,
threshold = threshold,
heuristic=heuristic)
log.inlineandremove("removed %d simple mallocs in total" % count)
constfold(config, graphs)
if config.print_statistics:
print "after clever inlining and malloc removal"
print_statistics(translator.graphs[0], translator)
if config.storesink:
for graph in graphs:
storesink_graph(graph)
if config.profile_based_inline and not secondary:
threshold = config.profile_based_inline_threshold
heuristic = get_function(config.profile_based_inline_heuristic)
inline.instrument_inline_candidates(graphs, threshold)
counters = translator.driver_instrument_result(
config.profile_based_inline)
n = len(counters)
def call_count_pred(label):
if label >= n:
return False
return counters[label] > 250 # xxx introduce an option for this
inline_malloc_removal_phase(config, translator, graphs,
threshold,
inline_heuristic=heuristic,
call_count_pred=call_count_pred)
constfold(config, graphs)
if config.merge_if_blocks:
log.mergeifblocks("starting to merge if blocks")
for graph in graphs:
merge_if_blocks(graph, translator.config.translation.verbose)
if config.print_statistics:
print "after if-to-switch:"
print_statistics(translator.graphs[0], translator)
remove_obvious_noops()
for graph in graphs:
checkgraph(graph)
gilanalysis.analyze(graphs, translator)