[ir] Make statements final, and change `IRNode::as` to use `static_cast` #7896

bobcao3 · 2023-04-26T01:25:53Z

In theory and in small scale experiments with different compilers, this does result in less calls to dynamic_cast and potentially cheaper dynamic_cast & virtual functions. However on Windows the benefits seem to be within run-to-run variance at least on my machine.

Walkthrough

`🤖 Generated by Copilot at 310a4e3`

Replace dynamic_cast with static_cast for IRNode::as method to improve performance (link)
Add final keyword to all statement classes in taichi/ir/statements.h to prevent unintended inheritance (link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link, link)

netlify · 2023-04-26T01:26:03Z

✅ Deploy Preview for docsite-preview canceled.

Name	Link
🔨 Latest commit	`5e0a34e`
🔍 Latest deploy log	https://app.netlify.com/sites/docsite-preview/deploys/64487de2bce23700080d2e59

for more information, see https://pre-commit.ci

bobcao3 · 2023-04-26T01:55:58Z

/benchmark

jim19930609

Unless there's significant performance gain, I would suggest that we keep the type check TI_ASSERT(is<T>());.

static_cast to Derived class pointers doesn't perform any check and is easy to cause undefined behavior if not used with caution.

taichi-gardener · 2023-04-26T02:15:56Z

Benchmark Report

Baseline: v1.5.0
Current: 5e0a34e4222e6da8abc0b07afa51aa8e4e13f61f

Item	Current	Change
c_api_library_size:android@arch=x64	1960320.000	$\textcolor{gray}{\textsf{ 0.00\%}}$
c_api_library_size:linux@arch=x64	52474888.000	$\textcolor{gray}{\textsf{ 0.00\%}}$
c_api_library_size:wall_time@arch=x64	0.0177	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=cuda,n=134217728	826.719	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=cuda,n=16777216	808.394	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=cuda,n=33554432	818.861	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=cuda,n=536870912	818.053	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=vulkan,n=134217728	688.889	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=vulkan,n=16777216	694.280	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=vulkan,n=33554432	708.785	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:GB/s@arch=vulkan,n=536870912	651.315	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=cuda,n=134217728	0.605	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=cuda,n=16777216	0.0773	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=cuda,n=33554432	0.153	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=cuda,n=536870912	2.445	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=vulkan,n=134217728	0.726	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=vulkan,n=16777216	0.09	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=vulkan,n=33554432	0.176	$\textcolor{gray}{\textsf{ 0.00\%}}$
fill:wall_time@arch=vulkan,n=536870912	3.071	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=2,n_grid=128	2067.345	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=2,n_grid=256	1210.752	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=2,n_grid=32	2119.636	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=2,n_grid=64	2126.721	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=3,n_grid=128	13.852	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=3,n_grid=256	0.333	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=3,n_grid=32	468.748	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=cuda,dim=3,n_grid=64	92.409	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=2,n_grid=128	1415.555	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=2,n_grid=256	953.419	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=2,n_grid=32	2203.425	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=2,n_grid=64	2236.451	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=3,n_grid=128	20.088	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=3,n_grid=256	0.354	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=3,n_grid=32	944.886	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:fps@arch=vulkan,dim=3,n_grid=64	121.122	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=2,n_grid=128	423392168.593	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=2,n_grid=256	991848129.001	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=2,n_grid=32	27131344.642	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=2,n_grid=64	108888091.572	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=3,n_grid=128	181556803.801	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=3,n_grid=256	34924451.290	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=3,n_grid=32	95999583.015	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=cuda,dim=3,n_grid=64	151403596.190	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=2,n_grid=128	289905703.604	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=2,n_grid=256	781041065.492	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=2,n_grid=32	28203835.939	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=2,n_grid=64	114506311.519	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=3,n_grid=128	263294523.345	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=3,n_grid=256	37114474.846	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=3,n_grid=32	193512621.704	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:particles/s@arch=vulkan,dim=3,n_grid=64	198446251.496	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=2,n_grid=128	0.484	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=2,n_grid=256	0.826	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=2,n_grid=32	0.472	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=2,n_grid=64	0.47	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=3,n_grid=128	72.193	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=3,n_grid=256	3002.412	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=3,n_grid=32	2.133	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=cuda,dim=3,n_grid=64	10.821	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=2,n_grid=128	0.706	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=2,n_grid=256	1.049	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=2,n_grid=32	0.454	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=2,n_grid=64	0.447	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=3,n_grid=128	49.782	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=3,n_grid=256	2825.248	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=3,n_grid=32	1.058	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm:wall_time@arch=vulkan,dim=3,n_grid=64	8.256	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=128	0.0922	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=256	0.0921	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=32	0.091	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=64	0.0925	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=128	0.202	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=256	0.204	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=32	0.201	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=64	0.201	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=128	0.0212	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=256	0.0266	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=32	0.0214	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=64	0.0215	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=128	0.0367	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=256	0.0365	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=32	0.0427	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=64	0.0363	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=128	92.229	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=256	92.075	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=32	90.994	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=64	92.479	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=128	202.352	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=256	203.797	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=32	201.266	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=64	200.872	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=128	21.173	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=256	26.642	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=32	21.397	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=64	21.548	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=128	36.658	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=256	36.533	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=32	42.722	$\textcolor{gray}{\textsf{ 0.00\%}}$
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=64	36.302	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=128,variant=CacheBlock	0.783	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=128,variant=Naive	0.782	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=256,variant=CacheBlock	2.977	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=256,variant=Naive	2.726	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=262144,variant=CacheBlock	1047.359	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=262144,variant=Naive	526.561	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=512,variant=CacheBlock	11.185	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=cuda,n=512,variant=Naive	7.711	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=vulkan,n=128,variant=Naive	0.402	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=vulkan,n=256,variant=Naive	1.210	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=vulkan,n=262144,variant=Naive	359.861	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:bips@arch=vulkan,n=512,variant=Naive	3.180	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=128,variant=CacheBlock	0.0209	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=128,variant=Naive	0.021	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=256,variant=CacheBlock	0.022	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=256,variant=Naive	0.024	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=262144,variant=CacheBlock	65.612	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=262144,variant=Naive	130.506	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=512,variant=CacheBlock	0.0234	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=cuda,n=512,variant=Naive	0.034	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=vulkan,n=128,variant=Naive	0.0408	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=vulkan,n=256,variant=Naive	0.0542	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=vulkan,n=262144,variant=Naive	190.961	$\textcolor{gray}{\textsf{ 0.00\%}}$
nbody:wall_time@arch=vulkan,n=512,variant=Naive	0.0824	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=1024	753.313	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=2048	778.279	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=256	46.529	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=4096	815.515	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=1024	721.281	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=2048	771.020	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=256	45.686	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=4096	799.939	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=1024	514.921	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=2048	606.787	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=256	46.340	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=4096	643.080	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=1024	740.469	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=2048	776.373	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=256	46.580	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=4096	813.678	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=1024	730.783	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=2048	776.433	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=256	46.218	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=4096	812.121	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=1024	776.054	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=2048	745.159	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=256	51.987	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=4096	818.705	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=1024	570.670	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=2048	749.802	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=256	49.972	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=4096	821.555	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=1024	444.688	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=2048	575.649	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=256	48.452	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=4096	609.640	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=1024	644.072	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=2048	743.627	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=256	50.121	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=4096	817.045	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=1024	629.983	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=2048	744.279	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=256	50.977	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=4096	817.569	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=1024	125.552	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=2048	129.713	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=256	7.755	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=4096	135.919	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=1024	15387.322	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=2048	16448.430	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=256	974.640	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=4096	17065.363	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=1024	21969.967	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=2048	25889.584	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=256	1977.163	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=4096	27438.091	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=1024	3949.166	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=2048	4140.654	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=256	248.428	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=4096	4339.616	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=1024	7795.019	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=2048	8281.950	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=256	492.988	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=4096	8662.629	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=1024	129.342	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=2048	124.193	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=256	8.664	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=4096	136.451	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=1024	12174.299	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=2048	15995.769	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=256	1066.061	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=4096	17526.507	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=1024	18973.365	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=2048	24561.023	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=256	2067.285	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=4096	26011.324	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=1024	3435.049	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=2048	3966.012	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=256	267.312	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=4096	4357.573	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=1024	6719.819	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=2048	7938.979	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=256	543.754	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=4096	8720.734	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=1024	0.0167	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=2048	0.0647	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=256	0.0169	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=4096	0.247	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=1024	0.0174	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=2048	0.0653	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=256	0.0172	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=4096	0.252	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=1024	0.0244	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=2048	0.0829	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=256	0.017	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=4096	0.313	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=1024	0.017	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=2048	0.0648	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=256	0.0169	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=4096	0.247	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=1024	0.0172	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=2048	0.0648	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=256	0.017	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=4096	0.248	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=1024	0.0162	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=2048	0.0675	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=256	0.0151	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=4096	0.246	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=1024	0.022	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=2048	0.0671	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=256	0.0157	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=4096	0.245	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=1024	0.0283	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=2048	0.0874	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=256	0.0162	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=4096	0.33	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=1024	0.0195	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=2048	0.0677	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=256	0.0157	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=4096	0.246	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=1024	0.02	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=2048	0.0676	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=256	0.0154	$\textcolor{gray}{\textsf{ 0.00\%}}$
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=4096	0.246	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=1024,arch=cuda	282.343	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=1024,arch=vulkan	160.411	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=256,arch=cuda	25.745	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=256,arch=vulkan	10.208	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=4096,arch=cuda	360.692	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:GB/s@N=4096,arch=vulkan	358.051	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=1024,arch=cuda	33657.965	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=1024,arch=vulkan	19122.506	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=256,arch=cuda	49105.421	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=256,arch=vulkan	19470.291	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=4096,arch=cuda	2687.366	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:fps@N=4096,arch=vulkan	2667.686	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=1024,arch=cuda	0.0297	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=1024,arch=vulkan	0.0523	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=256,arch=cuda	0.0204	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=256,arch=vulkan	0.0514	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=4096,arch=cuda	0.372	$\textcolor{gray}{\textsf{ 0.00\%}}$
stencil:wall_time@N=4096,arch=vulkan	0.375	$\textcolor{gray}{\textsf{ 0.00\%}}$

bobcao3 · 2023-04-26T04:25:51Z

Unless there's significant performance gain, I would suggest that we keep the type check TI_ASSERT(is<T>());.

static_cast to Derived class pointers doesn't perform any check and is easy to cause undefined behavior if not used with caution.

perf gain is minimal probably because modern CPUs getting too good at prediction. I think a potential way to do this is to use system assert statement instead of TI_ASSERT, which gets voided in Release build. This could go in junction with shipping both release mode and debug mode binary

jim19930609 · 2023-04-26T05:54:33Z

Unless there's significant performance gain, I would suggest that we keep the type check TI_ASSERT(is<T>());.
static_cast to Derived class pointers doesn't perform any check and is easy to cause undefined behavior if not used with caution.

perf gain is minimal probably because modern CPUs getting too good at prediction. I think a potential way to do this is to use system assert statement instead of TI_ASSERT, which gets voided in Release build. This could go in junction with shipping both release mode and debug mode binary

Agreed!

jim19930609 · 2023-04-26T05:56:38Z

Benchmark Report

Baseline: v1.5.0 Current: 5e0a34e4222e6da8abc0b07afa51aa8e4e13f61f

Item Baseline Current Change
c_api_library_size:android@arch=x64 0.000 1960320.000 0.00%
c_api_library_size:linux@arch=x64 0.000 52474888.000 0.00%
c_api_library_size:wall_time@arch=x64 0.000 0.0177 0.00%
fill:GB/s@arch=cuda,n=134217728 0.000 826.719 0.00%
fill:GB/s@arch=cuda,n=16777216 0.000 808.394 0.00%
fill:GB/s@arch=cuda,n=33554432 0.000 818.861 0.00%
fill:GB/s@arch=cuda,n=536870912 0.000 818.053 0.00%
fill:GB/s@arch=vulkan,n=134217728 0.000 688.889 0.00%
fill:GB/s@arch=vulkan,n=16777216 0.000 694.280 0.00%
fill:GB/s@arch=vulkan,n=33554432 0.000 708.785 0.00%
fill:GB/s@arch=vulkan,n=536870912 0.000 651.315 0.00%
fill:wall_time@arch=cuda,n=134217728 0.000 0.605 0.00%
fill:wall_time@arch=cuda,n=16777216 0.000 0.0773 0.00%
fill:wall_time@arch=cuda,n=33554432 0.000 0.153 0.00%
fill:wall_time@arch=cuda,n=536870912 0.000 2.445 0.00%
fill:wall_time@arch=vulkan,n=134217728 0.000 0.726 0.00%
fill:wall_time@arch=vulkan,n=16777216 0.000 0.09 0.00%
fill:wall_time@arch=vulkan,n=33554432 0.000 0.176 0.00%
fill:wall_time@arch=vulkan,n=536870912 0.000 3.071 0.00%
mpm:fps@arch=cuda,dim=2,n_grid=128 0.000 2067.345 0.00%
mpm:fps@arch=cuda,dim=2,n_grid=256 0.000 1210.752 0.00%
mpm:fps@arch=cuda,dim=2,n_grid=32 0.000 2119.636 0.00%
mpm:fps@arch=cuda,dim=2,n_grid=64 0.000 2126.721 0.00%
mpm:fps@arch=cuda,dim=3,n_grid=128 0.000 13.852 0.00%
mpm:fps@arch=cuda,dim=3,n_grid=256 0.000 0.333 0.00%
mpm:fps@arch=cuda,dim=3,n_grid=32 0.000 468.748 0.00%
mpm:fps@arch=cuda,dim=3,n_grid=64 0.000 92.409 0.00%
mpm:fps@arch=vulkan,dim=2,n_grid=128 0.000 1415.555 0.00%
mpm:fps@arch=vulkan,dim=2,n_grid=256 0.000 953.419 0.00%
mpm:fps@arch=vulkan,dim=2,n_grid=32 0.000 2203.425 0.00%
mpm:fps@arch=vulkan,dim=2,n_grid=64 0.000 2236.451 0.00%
mpm:fps@arch=vulkan,dim=3,n_grid=128 0.000 20.088 0.00%
mpm:fps@arch=vulkan,dim=3,n_grid=256 0.000 0.354 0.00%
mpm:fps@arch=vulkan,dim=3,n_grid=32 0.000 944.886 0.00%
mpm:fps@arch=vulkan,dim=3,n_grid=64 0.000 121.122 0.00%
mpm:particles/s@arch=cuda,dim=2,n_grid=128 0.000 423392168.593 0.00%
mpm:particles/s@arch=cuda,dim=2,n_grid=256 0.000 991848129.001 0.00%
mpm:particles/s@arch=cuda,dim=2,n_grid=32 0.000 27131344.642 0.00%
mpm:particles/s@arch=cuda,dim=2,n_grid=64 0.000 108888091.572 0.00%
mpm:particles/s@arch=cuda,dim=3,n_grid=128 0.000 181556803.801 0.00%
mpm:particles/s@arch=cuda,dim=3,n_grid=256 0.000 34924451.290 0.00%
mpm:particles/s@arch=cuda,dim=3,n_grid=32 0.000 95999583.015 0.00%
mpm:particles/s@arch=cuda,dim=3,n_grid=64 0.000 151403596.190 0.00%
mpm:particles/s@arch=vulkan,dim=2,n_grid=128 0.000 289905703.604 0.00%
mpm:particles/s@arch=vulkan,dim=2,n_grid=256 0.000 781041065.492 0.00%
mpm:particles/s@arch=vulkan,dim=2,n_grid=32 0.000 28203835.939 0.00%
mpm:particles/s@arch=vulkan,dim=2,n_grid=64 0.000 114506311.519 0.00%
mpm:particles/s@arch=vulkan,dim=3,n_grid=128 0.000 263294523.345 0.00%
mpm:particles/s@arch=vulkan,dim=3,n_grid=256 0.000 37114474.846 0.00%
mpm:particles/s@arch=vulkan,dim=3,n_grid=32 0.000 193512621.704 0.00%
mpm:particles/s@arch=vulkan,dim=3,n_grid=64 0.000 198446251.496 0.00%
mpm:wall_time@arch=cuda,dim=2,n_grid=128 0.000 0.484 0.00%
mpm:wall_time@arch=cuda,dim=2,n_grid=256 0.000 0.826 0.00%
mpm:wall_time@arch=cuda,dim=2,n_grid=32 0.000 0.472 0.00%
mpm:wall_time@arch=cuda,dim=2,n_grid=64 0.000 0.47 0.00%
mpm:wall_time@arch=cuda,dim=3,n_grid=128 0.000 72.193 0.00%
mpm:wall_time@arch=cuda,dim=3,n_grid=256 0.000 3002.412 0.00%
mpm:wall_time@arch=cuda,dim=3,n_grid=32 0.000 2.133 0.00%
mpm:wall_time@arch=cuda,dim=3,n_grid=64 0.000 10.821 0.00%
mpm:wall_time@arch=vulkan,dim=2,n_grid=128 0.000 0.706 0.00%
mpm:wall_time@arch=vulkan,dim=2,n_grid=256 0.000 1.049 0.00%
mpm:wall_time@arch=vulkan,dim=2,n_grid=32 0.000 0.454 0.00%
mpm:wall_time@arch=vulkan,dim=2,n_grid=64 0.000 0.447 0.00%
mpm:wall_time@arch=vulkan,dim=3,n_grid=128 0.000 49.782 0.00%
mpm:wall_time@arch=vulkan,dim=3,n_grid=256 0.000 2825.248 0.00%
mpm:wall_time@arch=vulkan,dim=3,n_grid=32 0.000 1.058 0.00%
mpm:wall_time@arch=vulkan,dim=3,n_grid=64 0.000 8.256 0.00%
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=128 0.000 0.0922 0.00%
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=256 0.000 0.0921 0.00%
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=32 0.000 0.091 0.00%
mpm_compile:compile_time@arch=cuda,dim=2,n_grid=64 0.000 0.0925 0.00%
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=128 0.000 0.202 0.00%
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=256 0.000 0.204 0.00%
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=32 0.000 0.201 0.00%
mpm_compile:compile_time@arch=cuda,dim=3,n_grid=64 0.000 0.201 0.00%
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=128 0.000 0.0212 0.00%
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=256 0.000 0.0266 0.00%
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=32 0.000 0.0214 0.00%
mpm_compile:compile_time@arch=vulkan,dim=2,n_grid=64 0.000 0.0215 0.00%
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=128 0.000 0.0367 0.00%
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=256 0.000 0.0365 0.00%
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=32 0.000 0.0427 0.00%
mpm_compile:compile_time@arch=vulkan,dim=3,n_grid=64 0.000 0.0363 0.00%
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=128 0.000 92.229 0.00%
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=256 0.000 92.075 0.00%
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=32 0.000 90.994 0.00%
mpm_compile:wall_time@arch=cuda,dim=2,n_grid=64 0.000 92.479 0.00%
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=128 0.000 202.352 0.00%
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=256 0.000 203.797 0.00%
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=32 0.000 201.266 0.00%
mpm_compile:wall_time@arch=cuda,dim=3,n_grid=64 0.000 200.872 0.00%
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=128 0.000 21.173 0.00%
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=256 0.000 26.642 0.00%
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=32 0.000 21.397 0.00%
mpm_compile:wall_time@arch=vulkan,dim=2,n_grid=64 0.000 21.548 0.00%
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=128 0.000 36.658 0.00%
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=256 0.000 36.533 0.00%
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=32 0.000 42.722 0.00%
mpm_compile:wall_time@arch=vulkan,dim=3,n_grid=64 0.000 36.302 0.00%
nbody:bips@arch=cuda,n=128,variant=CacheBlock 0.000 0.783 0.00%
nbody:bips@arch=cuda,n=128,variant=Naive 0.000 0.782 0.00%
nbody:bips@arch=cuda,n=256,variant=CacheBlock 0.000 2.977 0.00%
nbody:bips@arch=cuda,n=256,variant=Naive 0.000 2.726 0.00%
nbody:bips@arch=cuda,n=262144,variant=CacheBlock 0.000 1047.359 0.00%
nbody:bips@arch=cuda,n=262144,variant=Naive 0.000 526.561 0.00%
nbody:bips@arch=cuda,n=512,variant=CacheBlock 0.000 11.185 0.00%
nbody:bips@arch=cuda,n=512,variant=Naive 0.000 7.711 0.00%
nbody:bips@arch=vulkan,n=128,variant=Naive 0.000 0.402 0.00%
nbody:bips@arch=vulkan,n=256,variant=Naive 0.000 1.210 0.00%
nbody:bips@arch=vulkan,n=262144,variant=Naive 0.000 359.861 0.00%
nbody:bips@arch=vulkan,n=512,variant=Naive 0.000 3.180 0.00%
nbody:wall_time@arch=cuda,n=128,variant=CacheBlock 0.000 0.0209 0.00%
nbody:wall_time@arch=cuda,n=128,variant=Naive 0.000 0.021 0.00%
nbody:wall_time@arch=cuda,n=256,variant=CacheBlock 0.000 0.022 0.00%
nbody:wall_time@arch=cuda,n=256,variant=Naive 0.000 0.024 0.00%
nbody:wall_time@arch=cuda,n=262144,variant=CacheBlock 0.000 65.612 0.00%
nbody:wall_time@arch=cuda,n=262144,variant=Naive 0.000 130.506 0.00%
nbody:wall_time@arch=cuda,n=512,variant=CacheBlock 0.000 0.0234 0.00%
nbody:wall_time@arch=cuda,n=512,variant=Naive 0.000 0.034 0.00%
nbody:wall_time@arch=vulkan,n=128,variant=Naive 0.000 0.0408 0.00%
nbody:wall_time@arch=vulkan,n=256,variant=Naive 0.000 0.0542 0.00%
nbody:wall_time@arch=vulkan,n=262144,variant=Naive 0.000 190.961 0.00%
nbody:wall_time@arch=vulkan,n=512,variant=Naive 0.000 0.0824 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=1024 0.000 753.313 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=2048 0.000 778.279 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=256 0.000 46.529 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=1,n=4096 0.000 815.515 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=1024 0.000 721.281 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=2048 0.000 771.020 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=256 0.000 45.686 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=128,n=4096 0.000 799.939 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=1024 0.000 514.921 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=2048 0.000 606.787 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=256 0.000 46.340 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=256,n=4096 0.000 643.080 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=1024 0.000 740.469 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=2048 0.000 776.373 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=256 0.000 46.580 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=32,n=4096 0.000 813.678 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=1024 0.000 730.783 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=2048 0.000 776.433 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=256 0.000 46.218 0.00%
nested-saxpy:gbs@arch=cuda,len_coeff=64,n=4096 0.000 812.121 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=1024 0.000 776.054 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=2048 0.000 745.159 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=256 0.000 51.987 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=1,n=4096 0.000 818.705 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=1024 0.000 570.670 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=2048 0.000 749.802 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=256 0.000 49.972 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=128,n=4096 0.000 821.555 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=1024 0.000 444.688 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=2048 0.000 575.649 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=256 0.000 48.452 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=256,n=4096 0.000 609.640 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=1024 0.000 644.072 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=2048 0.000 743.627 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=256 0.000 50.121 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=32,n=4096 0.000 817.045 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=1024 0.000 629.983 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=2048 0.000 744.279 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=256 0.000 50.977 0.00%
nested-saxpy:gbs@arch=vulkan,len_coeff=64,n=4096 0.000 817.569 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=1024 0.000 125.552 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=2048 0.000 129.713 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=256 0.000 7.755 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=1,n=4096 0.000 135.919 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=1024 0.000 15387.322 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=2048 0.000 16448.430 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=256 0.000 974.640 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=128,n=4096 0.000 17065.363 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=1024 0.000 21969.967 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=2048 0.000 25889.584 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=256 0.000 1977.163 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=256,n=4096 0.000 27438.091 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=1024 0.000 3949.166 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=2048 0.000 4140.654 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=256 0.000 248.428 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=32,n=4096 0.000 4339.616 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=1024 0.000 7795.019 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=2048 0.000 8281.950 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=256 0.000 492.988 0.00%
nested-saxpy:gflops@arch=cuda,len_coeff=64,n=4096 0.000 8662.629 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=1024 0.000 129.342 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=2048 0.000 124.193 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=256 0.000 8.664 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=1,n=4096 0.000 136.451 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=1024 0.000 12174.299 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=2048 0.000 15995.769 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=256 0.000 1066.061 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=128,n=4096 0.000 17526.507 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=1024 0.000 18973.365 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=2048 0.000 24561.023 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=256 0.000 2067.285 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=256,n=4096 0.000 26011.324 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=1024 0.000 3435.049 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=2048 0.000 3966.012 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=256 0.000 267.312 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=32,n=4096 0.000 4357.573 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=1024 0.000 6719.819 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=2048 0.000 7938.979 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=256 0.000 543.754 0.00%
nested-saxpy:gflops@arch=vulkan,len_coeff=64,n=4096 0.000 8720.734 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=1024 0.000 0.0167 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=2048 0.000 0.0647 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=256 0.000 0.0169 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=1,n=4096 0.000 0.247 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=1024 0.000 0.0174 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=2048 0.000 0.0653 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=256 0.000 0.0172 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=128,n=4096 0.000 0.252 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=1024 0.000 0.0244 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=2048 0.000 0.0829 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=256 0.000 0.017 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=256,n=4096 0.000 0.313 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=1024 0.000 0.017 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=2048 0.000 0.0648 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=256 0.000 0.0169 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=32,n=4096 0.000 0.247 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=1024 0.000 0.0172 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=2048 0.000 0.0648 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=256 0.000 0.017 0.00%
nested-saxpy:wall_time@arch=cuda,len_coeff=64,n=4096 0.000 0.248 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=1024 0.000 0.0162 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=2048 0.000 0.0675 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=256 0.000 0.0151 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=1,n=4096 0.000 0.246 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=1024 0.000 0.022 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=2048 0.000 0.0671 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=256 0.000 0.0157 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=128,n=4096 0.000 0.245 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=1024 0.000 0.0283 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=2048 0.000 0.0874 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=256 0.000 0.0162 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=256,n=4096 0.000 0.33 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=1024 0.000 0.0195 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=2048 0.000 0.0677 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=256 0.000 0.0157 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=32,n=4096 0.000 0.246 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=1024 0.000 0.02 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=2048 0.000 0.0676 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=256 0.000 0.0154 0.00%
nested-saxpy:wall_time@arch=vulkan,len_coeff=64,n=4096 0.000 0.246 0.00%
stencil:GB/s@N=1024,arch=cuda 0.000 282.343 0.00%
stencil:GB/s@N=1024,arch=vulkan 0.000 160.411 0.00%
stencil:GB/s@N=256,arch=cuda 0.000 25.745 0.00%
stencil:GB/s@N=256,arch=vulkan 0.000 10.208 0.00%
stencil:GB/s@N=4096,arch=cuda 0.000 360.692 0.00%
stencil:GB/s@N=4096,arch=vulkan 0.000 358.051 0.00%
stencil:fps@N=1024,arch=cuda 0.000 33657.965 0.00%
stencil:fps@N=1024,arch=vulkan 0.000 19122.506 0.00%
stencil:fps@N=256,arch=cuda 0.000 49105.421 0.00%
stencil:fps@N=256,arch=vulkan 0.000 19470.291 0.00%
stencil:fps@N=4096,arch=cuda 0.000 2687.366 0.00%
stencil:fps@N=4096,arch=vulkan 0.000 2667.686 0.00%
stencil:wall_time@N=1024,arch=cuda 0.000 0.0297 0.00%
stencil:wall_time@N=1024,arch=vulkan 0.000 0.0523 0.00%
stencil:wall_time@N=256,arch=cuda 0.000 0.0204 0.00%
stencil:wall_time@N=256,arch=vulkan 0.000 0.0514 0.00%
stencil:wall_time@N=4096,arch=cuda 0.000 0.372 0.00%
stencil:wall_time@N=4096,arch=vulkan 0.000 0.375 0.00%

Diff of 0% seems pretty weird. Is there anything wrong with our baseline? @feisuzhu

Make statements final, and change IRNode::as to use static_cast

310a4e3

bobcao3 requested review from ailzhang, k-ye and jim19930609 April 26, 2023 01:25

[pre-commit.ci] auto fixes from pre-commit.com hooks

5e0a34e

for more information, see https://pre-commit.ci

jim19930609 reviewed Apr 26, 2023

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[ir] Make statements final, and change `IRNode::as` to use `static_cast` #7896

[ir] Make statements final, and change `IRNode::as` to use `static_cast` #7896

bobcao3 commented Apr 26, 2023 •

edited by ghost

netlify bot commented Apr 26, 2023 •

edited

bobcao3 commented Apr 26, 2023

jim19930609 left a comment

taichi-gardener commented Apr 26, 2023

bobcao3 commented Apr 26, 2023

jim19930609 commented Apr 26, 2023

jim19930609 commented Apr 26, 2023

Benchmark Report

[ir] Make statements final, and change IRNode::as to use static_cast #7896

Are you sure you want to change the base?

[ir] Make statements final, and change IRNode::as to use static_cast #7896

Conversation

bobcao3 commented Apr 26, 2023 • edited by ghost

Walkthrough

🤖 Generated by Copilot at 310a4e3

netlify bot commented Apr 26, 2023 • edited

✅ Deploy Preview for docsite-preview canceled.

bobcao3 commented Apr 26, 2023

jim19930609 left a comment

Choose a reason for hiding this comment

taichi-gardener commented Apr 26, 2023

Benchmark Report

bobcao3 commented Apr 26, 2023

jim19930609 commented Apr 26, 2023

jim19930609 commented Apr 26, 2023

Benchmark Report

[ir] Make statements final, and change `IRNode::as` to use `static_cast` #7896

[ir] Make statements final, and change `IRNode::as` to use `static_cast` #7896

bobcao3 commented Apr 26, 2023 •

edited by ghost

`🤖 Generated by Copilot at 310a4e3`

netlify bot commented Apr 26, 2023 •

edited