Skip to content

Conversation

@sbrantq
Copy link
Member

@sbrantq sbrantq commented Jan 4, 2026

No description provided.

Comment on lines +105 to +108
:param_a => ([0.0],),
:param_b => ([0.0],),
:ys_a => (ys_a,),
:ys_b => (ys_b,),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
:param_a => ([0.0],),
:param_b => ([0.0],),
:ys_a => (ys_a,),
:ys_b => (ys_b,),
:param_a => ([0.0],), :param_b => ([0.0],), :ys_a => (ys_a,), :ys_b => (ys_b,)

Comment on lines +171 to +174
:param_a => ([0.0],),
:param_b => ([0.0],),
:ys_a => (ys_a,),
:ys_b => (ys_b,),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
:param_a => ([0.0],),
:param_b => ([0.0],),
:ys_a => (ys_a,),
:ys_b => (ys_b,),
:param_a => ([0.0],), :param_b => ([0.0],), :ys_a => (ys_a,), :ys_b => (ys_b,)

function run_benchmarks(test::String)
results = BenchmarkResult[]

println("=" ^ 70)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
println("=" ^ 70)
println("="^70)


println("=" ^ 70)
println("Reactant Benchmark (matching test configuration)")
println("=" ^ 70)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
println("=" ^ 70)
println("="^70)

println("Reactant Benchmark (matching test configuration)")
println("=" ^ 70)
println("Reactant version: ", pkgversion(Reactant))
println("=" ^ 70)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
println("=" ^ 70)
println("="^70)

Comment on lines +269 to +271
arg_type = String
default = "reactant_results.json"
help = "Output JSON file path"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
arg_type = String
default = "reactant_results.json"
help = "Output JSON file path"
arg_type = String
default = "reactant_results.json"
help = "Output JSON file path"

Comment on lines +296 to +297
)
for r in results
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
)
for r in results
) for r in results

JSON3.pretty(io, output_data)
end

println("\nResults saved to $(args["output"])")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
println("\nResults saved to $(args["output"])")
return println("\nResults saved to $(args["output"])")

@@ -0,0 +1,829 @@
using ..Reactant: MLIR, Profiler


Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change

return permutedims(transposed, N:-1:1)
end

function to_row_major!(ptr::Ptr{T}, src::AbstractArray{T,N}, shape::NTuple{N,<:Integer}) where {T,N}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
function to_row_major!(ptr::Ptr{T}, src::AbstractArray{T,N}, shape::NTuple{N,<:Integer}) where {T,N}
function to_row_major!(
ptr::Ptr{T}, src::AbstractArray{T,N}, shape::NTuple{N,<:Integer}
) where {T,N}

Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reactant.jl Benchmarks

Details
Benchmark suite Current: dfb3701 Previous: af7379e Ratio
DGCNN [3, 128, 256]/reverse/CPU/DefaultAfterEnzyme 0.485428624 s 0.424363699 s 1.14
jacobi_2d [512, 512, 1024]/primal/CPU/Default_manual_vectorized 0.752954575 s 0.633263527 s 1.19
NewtonSchulz [4096 x 4096]/primal/CPU/Default 2.478280358 s 2.327525426 s 1.06
gesummv [4096]/primal/CPU/Default_manual_vectorized 0.078132365 s 0.06255775 s 1.25
bicg [2048, 4096]/primal/CPU/Default_manual_vectorized 0.001072467 s 0.000891146 s 1.20
NewtonSchulz [1024 x 1024]/primal/CPU/StructuredTensors 0.121403513 s 0.092695673 s 1.31
atax [2048]/primal/CPU/Default_manual_vectorized 0.000589148 s 0.000493948 s 1.19
covariance [2048, 2048]/primal/CPU/Default 0.051169315 s 0.048731277 s 1.05
NewtonSchulz [256 x 256]/primal/CPU/Default 0.012196126 s 0.009724805 s 1.25
NewtonSchulz [1024 x 1024]/primal/CPU/Default 0.0637323 s 0.050936962 s 1.25
syrk [2048]/primal/CPU/Julia 38.832992157 s 37.982631764000004 s 1.02
2mm [2048]/primal/CPU/Default_manual_vectorized 0.024038819 s 0.023694488 s 1.01
doitgen [256, 1024, 512]/primal/CPU/Default_manual_vectorized 0.078787224 s 0.07754438 s 1.02
DeepONet ([64, 1024], [1, 128])/reverse/CPU/DefaultAll 0.005995006 s 0.00464505 s 1.29
NewtonSchulz [4096 x 4096]/primal/CPU/StructuredTensors 3.772783064 s 3.455379657 s 1.09
2mm [2048]/primal/CPU/Default 0.019425076 s 0.018277235 s 1.06
doitgen [256, 1024, 512]/primal/CPU/Default 0.123162827 s 0.114673413 s 1.07
bicg [2048, 4096]/primal/CPU/Julia 0.064550434 s 0.058290076 s 1.11
NewtonSchulz [256 x 256]/primal/CPU/StructuredTensors 0.010799948 s 0.008257153 s 1.31
NewtonSchulz [1024 x 1024]/primal/CPU/StructuredTensors (Only Detection) 0.119268072 s 0.092303074 s 1.29
DGCNN [3, 128, 256]/primal/CPU/Default 0.104659866 s 0.097611261 s 1.07
jacobi_1d [2048, 1024]/primal/CPU/Default_manual_vectorized 0.006611932 s 0.006859018 s 0.96
covariance [2048, 2048]/primal/CPU/Default_manual_vectorized 0.027828516 s 0.026587214 s 1.05
atax [2048]/primal/CPU/Julia 0.029496233 s 0.027706914000000003 s 1.06
syrk [2048]/primal/CPU/Default_manual_vectorized 0.010675206 s 0.009644194 s 1.11
correlation [2048, 2048]/primal/CPU/Default 0.041718952 s 0.035708421 s 1.17
correlation [2048, 2048]/primal/CPU/Julia 23.091337811000002 s 23.291243788000003 s 0.99
DeepONet ([64, 1024], [1, 128])/primal/CPU/Default 0.001918288 s 0.001446377 s 1.33
atax [2048]/primal/CPU/Default 0.000619709 s 0.000425187 s 1.46
DGCNN [3, 128, 256]/reverse/CPU/DefaultBeforeEnzyme 0.680117477 s 0.595574363 s 1.14
NewtonSchulz [4096 x 4096]/primal/CPU/StructuredTensors (Only Detection) 3.786053713 s 3.450493243 s 1.10
correlation [2048, 2048]/primal/CPU/Default_manual_vectorized 0.045484381 s 0.03246424 s 1.40
3mm [256, 1024, 2048, 4096]/primal/CPU/Default 0.007912096 s 0.006308424 s 1.25
heat_3d [128, 128, 128, 256]/primal/CPU/Default 0.99123041 s 0.755898755 s 1.31
heat_3d [128, 128, 128, 256]/primal/CPU/Default_manual_vectorized 1.010480505 s 0.783056859 s 1.29
DGCNN [3, 128, 256]/reverse/CPU/DisableTransposeReshapeAfterEnzyme 0.512279861 s 0.451676152 s 1.13
gemmver [2048]/primal/CPU/Default_manual_vectorized 0.01683723 s 0.015788628 s 1.07
FNO [64, 64, 1, 4]/reverse/CPU/DefaultBeforeEnzyme 0.197478914 s 0.166279885 s 1.19
2mm [2048]/primal/CPU/Julia 84.32840578800001 s 63.11559082 s 1.34
mvt [4096]/primal/CPU/Default_manual_vectorized 0.008479294 s 0.008538767 s 0.99
DGCNN [3, 128, 256]/reverse/CPU/DisableTransposeReshapeBeforeEnzyme 0.510939249 s 0.440367595 s 1.16
gemmver [2048]/primal/CPU/Default 0.004261275 s 0.00341181 s 1.25
gemmver [2048]/primal/CPU/Julia 0.044262117000000004 s 0.034967203 s 1.27
DeepONet ([64, 1024], [1, 128])/reverse/CPU/DefaultBeforeEnzyme 0.005814048 s 0.005200358 s 1.12
gesummv [4096]/primal/CPU/Julia 0.354098036 s 0.37551715500000005 s 0.94
jacobi_1d [2048, 1024]/primal/CPU/Default 0.006925633 s 0.006994769 s 0.99
jacobi_2d [512, 512, 1024]/primal/CPU/Julia 1.7720443680000002 s 1.653038969 s 1.07
NewtonSchulz [1024 x 1024]/primal/CPU/Julia 0.13032732700000002 s 0.09745233900000001 s 1.34
FNO [64, 64, 1, 4]/reverse/CPU/DefaultAfterEnzyme 0.211459389 s 0.177849174 s 1.19
DGCNN [3, 128, 256]/reverse/CPU/DisableTransposeReshapeAll 0.508209178 s 0.439330977 s 1.16
fdtd_2d [1024, 2048, 256]/primal/CPU/Julia 28.847831237 s 27.420067098 s 1.05
DGCNN [3, 128, 256]/reverse/CPU/NoOpt 0.511653422 s 0.450452404 s 1.14
FNO [64, 64, 1, 4]/reverse/CPU/NoOpt 0.197437211 s 0.168566658 s 1.17
gemm [2048, 4096]/primal/CPU/Default 0.018042198 s 0.017328433 s 1.04
jacobi_1d [2048, 1024]/primal/CPU/Julia 0.000565873 s 0.00056232 s 1.01
mvt [4096]/primal/CPU/Julia 0.194915103 s 0.19613812900000002 s 0.99
bicg [2048, 4096]/primal/CPU/Default 0.001017764 s 0.000934318 s 1.09
doitgen [256, 1024, 512]/primal/CPU/Julia 436.49182763000005 s 370.671983606 s 1.18
gesummv [4096]/primal/CPU/Default 0.002883079 s 0.001542341 s 1.87
FNO [64, 64, 1, 4]/reverse/CPU/DefaultAll 0.210439718 s 0.181652715 s 1.16
jacobi_2d [512, 512, 1024]/primal/CPU/Default 0.702217603 s 0.680589291 s 1.03
DeepONet ([64, 1024], [1, 128])/primal/CPU/NoOpt 0.001957183 s 0.001436302 s 1.36
gemm [2048, 4096]/primal/CPU/Julia 284.99676956400003 s 290.73229743400003 s 0.98
fdtd_2d [1024, 2048, 256]/primal/CPU/Default 0.935118199 s 0.603319428 s 1.55
DGCNN [3, 128, 256]/primal/CPU/NoOpt 0.103482483 s 0.096035912 s 1.08
fdtd_2d [1024, 2048, 256]/primal/CPU/Default_manual_vectorized 1.767537681 s 1.183547831 s 1.49
covariance [2048, 2048]/primal/CPU/Julia 23.050343378 s 23.071833198 s 1.00
syr2k [2048]/primal/CPU/Default_manual_vectorized 0.020942579 s 0.019951166 s 1.05
DeepONet ([64, 1024], [1, 128])/reverse/CPU/NoOpt 0.005893717 s 0.004743481 s 1.24
DGCNN [3, 128, 256]/reverse/CPU/DefaultAll 0.441738534 s 0.38489178 s 1.15
3mm [256, 1024, 2048, 4096]/primal/CPU/Julia 15.526644485 s 14.518275856 s 1.07
NewtonSchulz [256 x 256]/primal/CPU/Julia 0.004707707 s 0.0037672100000000004 s 1.25
NewtonSchulz [256 x 256]/primal/CPU/StructuredTensors (Only Detection) 0.011088271 s 0.007750412 s 1.43
syr2k [2048]/primal/CPU/Julia 48.145183237000005 s 46.927868555 s 1.03
heat_3d [128, 128, 128, 256]/primal/CPU/Julia 12.129965253 s 11.601852481 s 1.05
mvt [4096]/primal/CPU/Default 0.0102247 s 0.008477991 s 1.21
NewtonSchulz [4096 x 4096]/primal/CPU/Julia 5.529777017000001 s 4.824705636 s 1.15
FNO [64, 64, 1, 4]/primal/CPU/NoOpt 0.085316761 s 0.075476623 s 1.13
DGCNN [3, 128, 256]/primal/CPU/DisableTransposeReshape 0.114418711 s 0.10505488 s 1.09
DeepONet ([64, 1024], [1, 128])/reverse/CPU/DefaultAfterEnzyme 0.005898681 s 0.004697247 s 1.26
FNO [64, 64, 1, 4]/primal/CPU/Default 0.083701387 s 0.073944389 s 1.13
gemm [2048, 4096]/primal/CPU/Default_manual_vectorized 0.024426193 s 0.021209428 s 1.15
syrk [2048]/primal/CPU/Default 0.010576437 s 0.009697487 s 1.09
3mm [256, 1024, 2048, 4096]/primal/CPU/Default_manual_vectorized 0.007996146 s 0.006325027 s 1.26
syr2k [2048]/primal/CPU/Default 0.020633268 s 0.019824948 s 1.04
NewtonSchulz [256 x 256]/primal/CUDA/Default 0.000580282 s 0.00060947 s 0.95
ViT tiny [256, 256, 3, 4]/reverse/CUDA/DisableScatterGatherPadAll 0.010277085 s 0.010707208 s 0.96
gemmver [2048]/primal/CUDA/Default_manual_vectorized 0.000057316 s 0.000059303 s 0.97
ViT tiny [256, 256, 3, 4]/reverse/CUDA/DisableTransposeReshapeAll 0.01041656 s 0.011085305 s 0.94
syr2k [2048]/primal/CUDA/Default 0.000499835 s 0.000498615 s 1.00
FNO [64, 64, 1, 4]/reverse/CUDA/DefaultAll 0.003090159 s 0.003321683 s 0.93
atax [2048]/primal/CUDA/Default_manual_vectorized 0.00002521 s 0.000025546 s 0.99
gesummv [4096]/primal/CUDA/Default_manual_vectorized 0.000206814 s 0.000205859 s 1.00
fdtd_2d [1024, 2048, 256]/primal/CUDA/Default_manual_vectorized 0.027729622 s 0.027651581 s 1.00
VGG11 bn=true [224, 224, 3, 4]/primal/CUDA/Default 0.001899408 s 0.001893666 s 1.00
gemm [2048, 4096]/primal/CUDA/Default 0.000444703 s 0.000444186 s 1.00
ViT tiny [256, 256, 3, 4]/primal/CUDA/NoOpt 0.003013352 s 0.003240339 s 0.93
3mm [256, 1024, 2048, 4096]/primal/CUDA/Default_manual_vectorized 0.00015642 s 0.000160521 s 0.97
jacobi_2d [512, 512, 1024]/primal/CUDA/Default_manual_vectorized 0.020318068 s 0.021911316 s 0.93
NewtonSchulz [1024 x 1024]/primal/CUDA/StructuredTensors 0.006608601 s 0.006489386 s 1.02
ViT tiny [256, 256, 3, 4]/primal/CUDA/Default 0.003470887 s 0.00301671 s 1.15
DGCNN [3, 128, 256]/primal/CUDA/DisableTransposeReshape 0.001240625 s 0.001250912 s 0.99
ViT tiny [256, 256, 3, 4]/reverse/CUDA/DisableScatterGatherAll 0.010280418 s 0.011100275 s 0.93
DGCNN [3, 128, 256]/reverse/CUDA/DisableTransposeReshapeAfterEnzyme 0.003385015 s 0.00339506 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/CUDA/DefaultAll 0.000627931 s 0.00062615 s 1.00
gemm [2048, 4096]/primal/CUDA/Default_manual_vectorized 0.000440581 s 0.000440148 s 1.00
DeepONet ([64, 1024], [1, 128])/primal/CUDA/Default 0.000223412 s 0.000244553 s 0.91
DGCNN [3, 128, 256]/reverse/CUDA/DisableTransposeReshapeAll 0.003348573 s 0.00334006 s 1.00
VGG11 bn=true [224, 224, 3, 4]/reverse/CUDA/DefaultAfterEnzyme 0.007068321 s 0.007041787 s 1.00
NewtonSchulz [4096 x 4096]/primal/CUDA/StructuredTensors (Only Detection) 0.108721176 s 0.108649577 s 1.00
covariance [2048, 2048]/primal/CUDA/Default_manual_vectorized 0.000259092 s 0.000260249 s 1.00
NewtonSchulz [4096 x 4096]/primal/CUDA/StructuredTensors 0.109063919 s 0.109083622 s 1.00
NewtonSchulz [4096 x 4096]/primal/CUDA/Default 0.040169628 s 0.040180272 s 1.00
correlation [2048, 2048]/primal/CUDA/Default 0.000297255 s 0.000296604 s 1.00
3mm [256, 1024, 2048, 4096]/primal/CUDA/Default 0.00015673 s 0.000162028 s 0.97
doitgen [256, 1024, 512]/primal/CUDA/Default_manual_vectorized 0.00189527 s 0.001879731 s 1.01
jacobi_1d [2048, 1024]/primal/CUDA/Default_manual_vectorized 0.024218085 s 0.024905851 s 0.97
VGG11 bn=true [224, 224, 3, 4]/reverse/CUDA/DefaultBeforeEnzyme 0.007132326 s 0.007106307 s 1.00
DGCNN [3, 128, 256]/primal/CUDA/NoOpt 0.001249434 s 0.001254809 s 1.00
covariance [2048, 2048]/primal/CUDA/Default 0.000307697 s 0.000307531 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/CUDA/DisablePadAll 0.010459878 s 0.010859268 s 0.96
VGG11 bn=true [224, 224, 3, 4]/reverse/CUDA/NoOpt 0.007169096 s 0.007146181 s 1.00
FNO [64, 64, 1, 4]/reverse/CUDA/DefaultAfterEnzyme 0.003114137 s 0.003153344 s 0.99
DGCNN [3, 128, 256]/reverse/CUDA/DefaultAfterEnzyme 0.003020316 s 0.00305861 s 0.99
DGCNN [3, 128, 256]/reverse/CUDA/DefaultAll 0.003565139 s 0.00354118 s 1.01
heat_3d [128, 128, 128, 256]/primal/CUDA/Default 0.013094224 s 0.012394229 s 1.06
mvt [4096]/primal/CUDA/Default_manual_vectorized 0.000109902 s 0.00010767 s 1.02
VGG11 bn=true [224, 224, 3, 4]/reverse/CUDA/DefaultAll 0.007060412 s 0.007036833 s 1.00
doitgen [256, 1024, 512]/primal/CUDA/Default 0.001940498 s 0.002242792 s 0.87
DGCNN [3, 128, 256]/primal/CUDA/Default 0.000962016 s 0.000973197 s 0.99
VGG11 bn=true [224, 224, 3, 4]/primal/CUDA/NoOpt 0.001914704 s 0.001918959 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/CUDA/NoOpt 0.000637746 s 0.000708686 s 0.90
syrk [2048]/primal/CUDA/Default_manual_vectorized 0.000255092 s 0.000254437 s 1.00
gemmver [2048]/primal/CUDA/Default 0.000042346 s 0.000045827 s 0.92
NewtonSchulz [256 x 256]/primal/CUDA/StructuredTensors 0.001412858 s 0.001408588 s 1.00
DGCNN [3, 128, 256]/reverse/CUDA/NoOpt 0.006580448 s 0.006560686 s 1.00
bicg [2048, 4096]/primal/CUDA/Default 0.000055537 s 0.000054067 s 1.03
DGCNN [3, 128, 256]/reverse/CUDA/DefaultBeforeEnzyme 0.006544176 s 0.006519929 s 1.00
bicg [2048, 4096]/primal/CUDA/Default_manual_vectorized 0.000055502 s 0.000055114 s 1.01
2mm [2048]/primal/CUDA/Default 0.000463714 s 0.000462026 s 1.00
jacobi_1d [2048, 1024]/primal/CUDA/Default 0.020452696 s 0.021464062 s 0.95
heat_3d [128, 128, 128, 256]/primal/CUDA/Default_manual_vectorized 0.013086938 s 0.012338928 s 1.06
jacobi_2d [512, 512, 1024]/primal/CUDA/Default 0.020331693 s 0.022080811 s 0.92
correlation [2048, 2048]/primal/CUDA/Default_manual_vectorized 0.000258795 s 0.000262917 s 0.98
DeepONet ([64, 1024], [1, 128])/reverse/CUDA/DefaultBeforeEnzyme 0.000610632 s 0.000616339 s 0.99
2mm [2048]/primal/CUDA/Default_manual_vectorized 0.000466847 s 0.000465271 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/CUDA/NoOpt 0.01056403 s 0.01215453 s 0.87
FNO [64, 64, 1, 4]/reverse/CUDA/DefaultBeforeEnzyme 0.003088813 s 0.003162558 s 0.98
NewtonSchulz [1024 x 1024]/primal/CUDA/Default 0.001920058 s 0.001920259 s 1.00
atax [2048]/primal/CUDA/Default 0.000025492 s 0.000027501 s 0.93
gesummv [4096]/primal/CUDA/Default 0.000106525 s 0.00010632 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/CUDA/DefaultAll 0.010372077 s 0.010928348 s 0.95
syr2k [2048]/primal/CUDA/Default_manual_vectorized 0.000499205 s 0.000497972 s 1.00
DGCNN [3, 128, 256]/reverse/CUDA/DisableTransposeReshapeBeforeEnzyme 0.014582225 s 0.014489654 s 1.01
FNO [64, 64, 1, 4]/reverse/CUDA/NoOpt 0.003262461 s 0.003309335 s 0.99
FNO [64, 64, 1, 4]/primal/CUDA/NoOpt 0.001101281 s 0.001129734 s 0.97
DeepONet ([64, 1024], [1, 128])/reverse/CUDA/DefaultAfterEnzyme 0.000595295 s 0.000618477 s 0.96
NewtonSchulz [256 x 256]/primal/CUDA/StructuredTensors (Only Detection) 0.001351699 s 0.001351742 s 1.00
DeepONet ([64, 1024], [1, 128])/primal/CUDA/NoOpt 0.000232152 s 0.000251493 s 0.92
mvt [4096]/primal/CUDA/Default 0.000110964 s 0.000108173 s 1.03
syrk [2048]/primal/CUDA/Default 0.000255944 s 0.000255388 s 1.00
fdtd_2d [1024, 2048, 256]/primal/CUDA/Default 0.023708669 s 0.024881436 s 0.95
NewtonSchulz [1024 x 1024]/primal/CUDA/StructuredTensors (Only Detection) 0.006484319 s 0.006665426 s 0.97
FNO [64, 64, 1, 4]/primal/CUDA/Default 0.001052756 s 0.001092638 s 0.96
ViT tiny [256, 256, 3, 4]/primal/TPU/Default 0.000217678 s 0.000217646 s 1.00
NewtonSchulz [4096 x 4096]/primal/TPU/Default 0.020848192 s 0.021263913 s 0.98
FNO [64, 64, 1, 4]/reverse/TPU/DefaultAll 0.003093538 s 0.003099974 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DisableTransposeReshapeAll 0.005005313 s 0.005005214 s 1.00
syrk [2048]/primal/TPU/Default_manual_vectorized 0.000031446 s 0.000031402 s 1.00
ViT tiny [256, 256, 3, 4]/primal/TPU/NoOpt 0.00058597 s 0.000586699 s 1.00
FNO [64, 64, 1, 4]/primal/TPU/NoOpt 0.00113824 s 0.001138299 s 1.00
gemm [2048, 4096]/primal/TPU/Default 0.000072761 s 0.000072392 s 1.01
3mm [256, 1024, 2048, 4096]/primal/TPU/Default_manual_vectorized 0.0000164 s 0.000016258 s 1.01
atax [2048]/primal/TPU/Default_manual_vectorized 0.00002416 s 0.0000241 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/DisableScatterGatherAll 0.001697264 s 0.001696212 s 1.00
syr2k [2048]/primal/TPU/Default 0.000058347 s 0.000057968 s 1.01
2mm [2048]/primal/TPU/Default_manual_vectorized 0.000075003 s 0.000074646 s 1.00
FNO [64, 64, 1, 4]/primal/TPU/Default 0.000953043 s 0.000952893 s 1.00
VGG11 bn=true [224, 224, 3, 4]/reverse/TPU/DefaultBeforeEnzyme 0.004180339 s 0.004179932 s 1.00
DeepONet ([64, 1024], [1, 128])/primal/TPU/Default 0.000006045 s 0.000006071 s 1.00
doitgen [256, 1024, 512]/primal/TPU/Default_manual_vectorized 0.001086287 s 0.001087234 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/DisableScatterGatherPadAll 0.001697602 s 0.001695937 s 1.00
doitgen [256, 1024, 512]/primal/TPU/Default 0.001649625 s 0.001575284 s 1.05
DGCNN [3, 128, 256]/primal/TPU/DisableTransposeReshape 0.00286477 s 0.002865261 s 1.00
FNO [64, 64, 1, 4]/reverse/TPU/DefaultAfterEnzyme 0.003085226 s 0.003085498 s 1.00
mvt [4096]/primal/TPU/Default 0.000045063 s 0.000045041 s 1.00
gemmver [2048]/primal/TPU/Default 0.000036457 s 0.000036488 s 1.00
FNO [64, 64, 1, 4]/reverse/TPU/NoOpt 0.002959479 s 0.002959606 s 1.00
syr2k [2048]/primal/TPU/Default_manual_vectorized 0.000057872 s 0.000057612 s 1.00
fdtd_2d [1024, 2048, 256]/primal/TPU/Default 0.018832917 s 0.018832993 s 1.00
covariance [2048, 2048]/primal/TPU/Default_manual_vectorized 0.000047597 s 0.000047469 s 1.00
heat_3d [128, 128, 128, 256]/primal/TPU/Default 0.261753595 s 0.26175347 s 1.00
NewtonSchulz [256 x 256]/primal/TPU/StructuredTensors (Only Detection) 0.000018253 s 0.000018206 s 1.00
gemmver [2048]/primal/TPU/Default_manual_vectorized 0.000037424 s 0.000037257 s 1.00
VGG11 bn=true [224, 224, 3, 4]/reverse/TPU/NoOpt 0.004053149 s 0.004052511 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DefaultBeforeEnzyme 0.004752273 s 0.004752431 s 1.00
bicg [2048, 4096]/primal/TPU/Default_manual_vectorized 0.000023445 s 0.000023471 s 1.00
NewtonSchulz [1024 x 1024]/primal/TPU/StructuredTensors 0.000208207 s 0.000207949 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/DisablePadAll 0.001697013 s 0.001695261 s 1.00
3mm [256, 1024, 2048, 4096]/primal/TPU/Default 0.000016344 s 0.000016257 s 1.01
jacobi_1d [2048, 1024]/primal/TPU/Default 0.008141185 s 0.008141321 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DefaultAfterEnzyme 0.004669998 s 0.00466929 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/DisableTransposeReshapeAll 0.001436771 s 0.00143419 s 1.00
syrk [2048]/primal/TPU/Default 0.000030377 s 0.000030317 s 1.00
correlation [2048, 2048]/primal/TPU/Default_manual_vectorized 0.000053242 s 0.000053085 s 1.00
DGCNN [3, 128, 256]/primal/TPU/NoOpt 0.002867674 s 0.002868033 s 1.00
NewtonSchulz [1024 x 1024]/primal/TPU/StructuredTensors (Only Detection) 0.000208403 s 0.000208155 s 1.00
VGG11 bn=true [224, 224, 3, 4]/primal/TPU/Default 0.000929502 s 0.000928859 s 1.00
NewtonSchulz [256 x 256]/primal/TPU/Default 0.000017408 s 0.000017369 s 1.00
jacobi_1d [2048, 1024]/primal/TPU/Default_manual_vectorized 0.005707395 s 0.005707716 s 1.00
FNO [64, 64, 1, 4]/reverse/TPU/DefaultBeforeEnzyme 0.00309309 s 0.003106801 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/TPU/DefaultBeforeEnzyme 0.000027356 s 0.000027143 s 1.01
VGG11 bn=true [224, 224, 3, 4]/reverse/TPU/DefaultAfterEnzyme 0.004180427 s 0.004179336 s 1.00
gesummv [4096]/primal/TPU/Default 0.000087504 s 0.000087411 s 1.00
covariance [2048, 2048]/primal/TPU/Default 0.000061508 s 0.000061237 s 1.00
gesummv [4096]/primal/TPU/Default_manual_vectorized 0.000087478 s 0.000087544 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/NoOpt 0.002020747 s 0.002019405 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/NoOpt 0.005226211 s 0.005226067 s 1.00
NewtonSchulz [1024 x 1024]/primal/TPU/Default 0.000207652 s 0.000207363 s 1.00
bicg [2048, 4096]/primal/TPU/Default 0.000023461 s 0.000023502 s 1.00
fdtd_2d [1024, 2048, 256]/primal/TPU/Default_manual_vectorized 0.027116985 s 0.027117358 s 1.00
jacobi_2d [512, 512, 1024]/primal/TPU/Default_manual_vectorized 0.022079447 s 0.022079266 s 1.00
jacobi_2d [512, 512, 1024]/primal/TPU/Default 0.026676316 s 0.026676173 s 1.00
gemm [2048, 4096]/primal/TPU/Default_manual_vectorized 0.00007272 s 0.00007251 s 1.00
ViT tiny [256, 256, 3, 4]/reverse/TPU/DefaultAll 0.00169746 s 0.001696136 s 1.00
NewtonSchulz [4096 x 4096]/primal/TPU/StructuredTensors (Only Detection) 0.020851029 s 0.021227325 s 0.98
2mm [2048]/primal/TPU/Default 0.00008655 s 0.000086339 s 1.00
NewtonSchulz [256 x 256]/primal/TPU/StructuredTensors 0.000018229 s 0.000018187 s 1.00
DGCNN [3, 128, 256]/primal/TPU/Default 0.002349459 s 0.002349255 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/TPU/DefaultAfterEnzyme 0.000027206 s 0.000027181 s 1.00
DeepONet ([64, 1024], [1, 128])/primal/TPU/NoOpt 0.000006206 s 0.000006192 s 1.00
atax [2048]/primal/TPU/Default 0.00002416 s 0.000024127 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DisableTransposeReshapeAfterEnzyme 0.005179481 s 0.005180006 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DefaultAll 0.004673207 s 0.004672524 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/TPU/NoOpt 0.000027446 s 0.00002718 s 1.01
VGG11 bn=true [224, 224, 3, 4]/reverse/TPU/DefaultAll 0.004180092 s 0.004178784 s 1.00
DeepONet ([64, 1024], [1, 128])/reverse/TPU/DefaultAll 0.000027243 s 0.000027181 s 1.00
VGG11 bn=true [224, 224, 3, 4]/primal/TPU/NoOpt 0.000866799 s 0.000866446 s 1.00
DGCNN [3, 128, 256]/reverse/TPU/DisableTransposeReshapeBeforeEnzyme 0.005005286 s 0.005004266 s 1.00
mvt [4096]/primal/TPU/Default_manual_vectorized 0.000045566 s 0.000045475 s 1.00
NewtonSchulz [4096 x 4096]/primal/TPU/StructuredTensors 0.020844957 s 0.021201078 s 0.98
heat_3d [128, 128, 128, 256]/primal/TPU/Default_manual_vectorized 0.26175361 s 0.261753456 s 1.00
correlation [2048, 2048]/primal/TPU/Default 0.000055815 s 0.000055623 s 1.00

This comment was automatically generated by workflow using github-action-benchmark.


if algorithm == :HMC
hmc_config_attr = @ccall MLIR.API.mlir_c.enzymeHMCConfigAttrGet(
MLIR.IR.context()::MLIR.API.MlirContext, trajectory_length::Float64, adapt_step_size::Bool, adapt_mass_matrix::Bool
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
MLIR.IR.context()::MLIR.API.MlirContext, trajectory_length::Float64, adapt_step_size::Bool, adapt_mass_matrix::Bool
MLIR.IR.context()::MLIR.API.MlirContext,
trajectory_length::Float64,
adapt_step_size::Bool,
adapt_mass_matrix::Bool,

max_tree_depth::Int64,
max_delta_energy::Float64,
adapt_step_size::Bool,
adapt_mass_matrix::Bool
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[JuliaFormatter] reported by reviewdog 🐶

Suggested change
adapt_mass_matrix::Bool
adapt_mass_matrix::Bool,

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants