diff options
Diffstat (limited to 'benchmark/bm_registry.py')
| -rw-r--r-- | benchmark/bm_registry.py | 49 |
1 files changed, 29 insertions, 20 deletions
diff --git a/benchmark/bm_registry.py b/benchmark/bm_registry.py index e097e43..c67d94a 100644 --- a/benchmark/bm_registry.py +++ b/benchmark/bm_registry.py @@ -18,25 +18,12 @@ GR_ENV = { 'PYTHONPATH': os.path.join(INSTALL_PREFIX, 'gr/lib/python2.7/dist-packages:%s'%os.getenv('PYTHONPATH')), } -BENCHMARK_MATH_OPS = tokwargs( - wat='Benchmark GrExtras vs gr-blocks math blocks', - moar=''' - 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 Adder', args=['tb_grextras_math.py', '1e8', 'extras_add'], env=GRAS_ENV), - tokwargs(wat='gr-blocks Adder', args=['tb_grextras_math.py', '1e8', 'blocks_add'], env=GRAS_ENV), - tokwargs(wat='GrExtras Multiplier', args=['tb_grextras_math.py', '1e8', 'extras_mult'], env=GRAS_ENV), - tokwargs(wat='gr-blocks Multiplier', args=['tb_grextras_math.py', '1e8', 'blocks_mult'], env=GRAS_ENV), - ], - to_result = lambda t: 1e8/t -) - BENCHMARK_MANY_11_BLOCKS = tokwargs( wat='Benchmark the schedulers with many 1:1 ratio blocks', - moar='''TODO''', + moar='''\ +- Compare simultaneous 1:1 radio blocks in each scheduler. +- GRAS will use only the buffer pool allocator, +and every work will fully consume available buffers.''', tests = [ tokwargs(wat='Many 1:1 GRAS', args=['tb_many_1_to_1_blocks.py', '1e8'], env=GRAS_ENV), #tokwargs(wat='Many 1:1 GRAS - TPP', args=['tb_many_1_to_1_blocks.py', '1e8'], env=GRAS_ENV, envextra=tokwargs(GRAS_TPP='1')), @@ -47,7 +34,11 @@ BENCHMARK_MANY_11_BLOCKS = tokwargs( BENCHMARK_FILTER_BLOCK = tokwargs( wat='Benchmark the schedulers with a filter block', - moar='''TODO''', + 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.''', tests = [ tokwargs(wat='GRAS decim FIR', args=['tb_filter_block.py', '--num', '1e7', '--which', 'dfir'], env=GRAS_ENV), tokwargs(wat='GR decim FIR', args=['tb_filter_block.py', '--num', '1e7', '--which', 'dfir'], env=GR_ENV), @@ -57,9 +48,27 @@ BENCHMARK_FILTER_BLOCK = tokwargs( to_result = lambda t: 1e7/t ) +BENCHMARK_MATH_OPS = tokwargs( + wat='Benchmark GrExtras vs gr-blocks math 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 Adder', args=['tb_grextras_math.py', '1e8', 'extras_add'], env=GRAS_ENV), + tokwargs(wat='gr-blocks Adder', args=['tb_grextras_math.py', '1e8', 'blocks_add'], env=GRAS_ENV), + tokwargs(wat='GrExtras Multiplier', args=['tb_grextras_math.py', '1e8', 'extras_mult'], env=GRAS_ENV), + tokwargs(wat='gr-blocks Multiplier', args=['tb_grextras_math.py', '1e8', 'blocks_mult'], env=GRAS_ENV), + ], + to_result = lambda t: 1e8/t +) + BENCHMARK_DELAY_BLOCKS = tokwargs( wat='Benchmark GrExtras vs gr-core delay block', - moar='''TODO''', + moar='''\ +- Compare delay block implementations using GRAS. +- The GrExtras implementation uses zero-copy.''', tests = [ tokwargs(wat='GrExtras Delay\n(zero-copy)', args=['tb_grextras_delay.py', '1e8', 'extras_delay'], env=GRAS_ENV), tokwargs(wat='gr-core Delay\n(mem-copy)', args=['tb_grextras_delay.py', '1e8', 'core_delay'], env=GRAS_ENV), @@ -68,8 +77,8 @@ BENCHMARK_DELAY_BLOCKS = tokwargs( ) BENCHMARKS = ( - BENCHMARK_MATH_OPS, BENCHMARK_MANY_11_BLOCKS, BENCHMARK_FILTER_BLOCK, + BENCHMARK_MATH_OPS, BENCHMARK_DELAY_BLOCKS, ) |