4 files changed, 70 insertions, 31 deletions
diff --git a/Theron b/Theron
-Subproject bca3174ae410767c283742cebbb996b745fc40e
+Subproject b2ddbe8b835d6378a1b06ab536f961d7a08912d
diff --git a/benchmark/bm_registry.py b/benchmark/bm_registry.py
index 7c9bb1c..af79a50 100644
--- a/benchmark/bm_registry.py
+++ b/benchmark/bm_registry.py
@@ -6,7 +6,7 @@ def tokwargs(**kwargs): return kwargs
 
 INSTALL_PREFIX = '/opt/usr'
 
-DURATION = '1.7' #seconds
+DURATION = '1.3' #seconds
 
 GRAS_ENV = {
     'PATH': os.path.join(INSTALL_PREFIX, 'gras/bin:$PATH:%s'%os.getenv('PATH')),
@@ -27,8 +27,8 @@ BENCHMARK_MANY_11_BLOCKS = tokwargs(
 - GRAS will use only the buffer pool allocator,
 and every work will fully consume available buffers.''',
     tests = [
-        tokwargs(wat='GRAS',     args=['tb_many_1_to_1_blocks.py', DURATION], env=GRAS_ENV),
-        tokwargs(wat='GR',       args=['tb_many_1_to_1_blocks.py', DURATION], env=GR_ENV),
+        tokwargs(wat='GRAS',     args=['tb_many_1_to_1_blocks.py', DURATION], env=GRAS_ENV, expand=True),
+        tokwargs(wat='GRSS',     args=['tb_many_1_to_1_blocks.py', DURATION], env=GR_ENV),
     ],
 )
 
@@ -39,40 +39,58 @@ BENCHMARK_MANY_RATE_BLOCKS = tokwargs(
 - GRAS will use only the buffer pool allocator,
 and every work will fully consume available buffers.''',
     tests = [
-        tokwargs(wat='GRAS',     args=['tb_many_rate_changes.py', '--dur', DURATION], env=GRAS_ENV),
-        tokwargs(wat='GR',       args=['tb_many_rate_changes.py', '--dur', DURATION], env=GR_ENV),
+        tokwargs(wat='GRAS',     args=['tb_many_rate_changes.py', '--dur', DURATION], env=GRAS_ENV, expand=True),
+        tokwargs(wat='GRSS',     args=['tb_many_rate_changes.py', '--dur', DURATION], env=GR_ENV),
     ],
 )
 
-BENCHMARK_FILTER_BLOCK = tokwargs(
-    wat='Benchmark the schedulers with a filter block',
+BENCHMARK_DFIR_BLOCK = tokwargs(
+    wat='Benchmark the schedulers with a FIR block',
     moar='''\
 - Compare filter blocks in each scheduler.
-- Shows both schedulers using circular buffer.
-- The decimating FIR filter is compared.
-- The rational resampler filter is compared.''',
+- Shows both schedulers using circular buffer.''',
     tests = [
-        tokwargs(wat='GRAS decim FIR',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'dfir'], env=GRAS_ENV),
-        tokwargs(wat='GR decim FIR',       args=['tb_filter_block.py', '--dur', DURATION, '--which', 'dfir'], env=GR_ENV),
-        tokwargs(wat='GRAS resampler',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'resamp'], env=GRAS_ENV),
-        tokwargs(wat='GR resampler',       args=['tb_filter_block.py', '--dur', DURATION, '--which', 'resamp'], env=GR_ENV),
+        tokwargs(wat='GRAS',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'dfir'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='GRSS',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'dfir'], env=GR_ENV),
     ],
 )
 
-BENCHMARK_MATH_OPS = tokwargs(
-    wat='Benchmark GrExtras vs gr-blocks math blocks',
+BENCHMARK_RESAMP_BLOCK = tokwargs(
+    wat='Benchmark the schedulers with a resampler block',
+    moar='''\
+- Compare filter blocks in each scheduler.
+- Shows both schedulers using circular buffer.''',
+    tests = [
+        tokwargs(wat='GRAS',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'resamp'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='GRSS',     args=['tb_filter_block.py', '--dur', DURATION, '--which', 'resamp'], env=GR_ENV),
+    ],
+)
+
+BENCHMARK_ADD_OPS = tokwargs(
+    wat='Benchmark GrExtras vs gr-blocks adder blocks',
+    moar='''\
+- Compare math block implementations using GRAS.
+- All blocks are using vector optimization.
+- GrExtras math blocks avoid an unnecessary memcpy.
+- GrExtras math blocks enable automatic bufer in-placing.''',
+    tests = [
+        tokwargs(wat='GrExtras\n(GRAS)',        args=['tb_grextras_math.py', DURATION, 'extras_add'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-blocks\n(GRAS)',       args=['tb_grextras_math.py', DURATION, 'blocks_add'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-blocks\n(GRSS)',       args=['tb_grextras_math.py', DURATION, 'blocks_add'], env=GR_ENV),
+    ],
+)
+
+BENCHMARK_MULT_OPS = tokwargs(
+    wat='Benchmark GrExtras vs gr-blocks multiplier blocks',
     moar='''\
 - Compare math block implementations using GRAS.
 - All blocks are using vector optimization.
 - GrExtras math blocks avoid an unnecessary memcpy.
 - GrExtras math blocks enable automatic bufer in-placing.''',
     tests = [
-        tokwargs(wat='GrExtras Add\n(GRAS)',        args=['tb_grextras_math.py', DURATION, 'extras_add'], env=GRAS_ENV),
-        tokwargs(wat='gr-blocks Add\n(GRAS)',       args=['tb_grextras_math.py', DURATION, 'blocks_add'], env=GRAS_ENV),
-        tokwargs(wat='gr-blocks Add\n(GR)',         args=['tb_grextras_math.py', DURATION, 'blocks_add'], env=GR_ENV),
-        tokwargs(wat='GrExtras Mult\n(GRAS)',       args=['tb_grextras_math.py', DURATION, 'extras_mult'], env=GRAS_ENV),
-        tokwargs(wat='gr-blocks Mult\n(GRAS)',      args=['tb_grextras_math.py', DURATION, 'blocks_mult'], env=GRAS_ENV),
-        tokwargs(wat='gr-blocks Mult\n(GR)',        args=['tb_grextras_math.py', DURATION, 'blocks_mult'], env=GR_ENV),
+        tokwargs(wat='GrExtras\n(GRAS)',       args=['tb_grextras_math.py', DURATION, 'extras_mult'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-blocks\n(GRAS)',      args=['tb_grextras_math.py', DURATION, 'blocks_mult'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-blocks\n(GRSS)',      args=['tb_grextras_math.py', DURATION, 'blocks_mult'], env=GR_ENV),
     ],
 )
 
@@ -82,16 +100,18 @@ BENCHMARK_DELAY_BLOCKS = tokwargs(
 - Compare delay block implementations using GRAS.
 - The GrExtras implementation uses zero-copy.''',
     tests = [
-        tokwargs(wat='GrExtras Delay\n(GRAS)',          args=['tb_grextras_delay.py', DURATION, 'extras_delay'], env=GRAS_ENV),
-        tokwargs(wat='gr-core Delay\n(GRAS)',           args=['tb_grextras_delay.py', DURATION, 'core_delay'], env=GRAS_ENV),
-        tokwargs(wat='gr-core Delay\n(GR)',             args=['tb_grextras_delay.py', DURATION, 'core_delay'], env=GR_ENV),
+        tokwargs(wat='GrExtras\n(GRAS)',          args=['tb_grextras_delay.py', DURATION, 'extras_delay'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-core\n(GRAS)',           args=['tb_grextras_delay.py', DURATION, 'core_delay'], env=GRAS_ENV, expand=True),
+        tokwargs(wat='gr-core\n(GRSS)',           args=['tb_grextras_delay.py', DURATION, 'core_delay'], env=GR_ENV),
     ],
 )
 
 BENCHMARKS = (
     BENCHMARK_MANY_11_BLOCKS,
     BENCHMARK_MANY_RATE_BLOCKS,
-    BENCHMARK_FILTER_BLOCK,
-    BENCHMARK_MATH_OPS,
+    BENCHMARK_DFIR_BLOCK,
+    BENCHMARK_RESAMP_BLOCK,
+    BENCHMARK_ADD_OPS,
+    BENCHMARK_MULT_OPS,
     BENCHMARK_DELAY_BLOCKS,
 )
diff --git a/benchmark/run_benchmarks.py b/benchmark/run_benchmarks.py
index 8cb3503..779aa6a 100644
--- a/benchmark/run_benchmarks.py
+++ b/benchmark/run_benchmarks.py
@@ -15,7 +15,7 @@ cpu_count = multiprocessing.cpu_count()
 
 from bm_registry import BENCHMARKS
 
-NUM_RUNS_PER_TEST = 5
+NUM_RUNS_PER_TEST = 3
 
 BAD_BOOST_KILL_DURATION = 5.0 #seconds
 
@@ -45,6 +45,25 @@ def run_a_single_one(args, env):
     raise Exception, 'no result found!'
     #return t1-t0
 
+def expand_tests(bm):
+    for run in bm['tests']:
+        if run.has_key('expand') and run['expand']:
+            import copy
+            new_run = copy.deepcopy(run)
+            new_run['wat'] += '\n(Block)'
+            new_run['env']['GRAS_YIELD'] = 'BLOCKING'
+            yield new_run
+            new_run = copy.deepcopy(run)
+            new_run['wat'] += '\n(Spin)'
+            new_run['env']['GRAS_YIELD'] = 'STRONG'
+            yield new_run
+            new_run = copy.deepcopy(run)
+            new_run['wat'] += '\n(TPB)'
+            new_run['env']['GRAS_YIELD'] = 'BLOCKING'
+            new_run['env']['GRAS_TPP'] = '1'
+            yield new_run
+        else: yield run
+
 def do_a_benchmark(bm):
     title = bm['wat']
     print '#'*(len(title)+25)
@@ -53,7 +72,7 @@ def do_a_benchmark(bm):
     result_means = list()
     result_stddevs = list()
     test_names = list()
-    for run in bm['tests']:
+    for run in expand_tests(bm):
         test_name = run['wat']
         print '-'*(len(test_name)+25)
         print '-- running test:', test_name.replace('\n', ' ')
diff --git a/lib/block_actor.cpp b/lib/block_actor.cpp
index 750e52d..fe7f79e 100644
--- a/lib/block_actor.cpp
+++ b/lib/block_actor.cpp
@@ -14,7 +14,7 @@ ThreadPoolConfig::ThreadPoolConfig(void)
     thread_count = std::max(size_t(2), thread_count);
     node_mask = 0;
     processor_mask = 0xffffffff;
-    yield_strategy = "STRONG";
+    yield_strategy = "BLOCKING";
 
     //environment variable override
     const char * gras_yield = getenv("GRAS_YIELD");
@@ -44,7 +44,7 @@ ThreadPool::ThreadPool(const ThreadPoolConfig &config)
     );
 
     if (config.yield_strategy.empty()) params.mYieldStrategy = Theron::YIELD_STRATEGY_STRONG;
-    //else if (config.yield_strategy == "BLOCKING") params.mYieldStrategy = Theron::YIELD_STRATEGY_BLOCKING;
+    else if (config.yield_strategy == "BLOCKING") params.mYieldStrategy = Theron::YIELD_STRATEGY_BLOCKING;
     else if (config.yield_strategy == "POLITE") params.mYieldStrategy = Theron::YIELD_STRATEGY_POLITE;
     else if (config.yield_strategy == "STRONG") params.mYieldStrategy = Theron::YIELD_STRATEGY_STRONG;
     else if (config.yield_strategy == "AGGRESSIVE") params.mYieldStrategy = Theron::YIELD_STRATEGY_AGGRESSIVE;