Adds GPU reduce dispatchers

[lukastruemper]

No description provided.

[daisytuner]

Daisytuner Report - python_npbench (zinnia)

@@                              Benchmarks                              @@
===========================================================================
  Benchmark              Speedup (Time) ±%           Energy      Regions   
===========================================================================
# adi                    
#   numpy                119.69 ms +0.2%             11.61 J     —         
#   docc-sequential      2.28x (52.38 ms) -1.8%      8.37 J      2         
#   docc-omp             2.07x (57.75 ms) +0.2%      8.87 J      2         
#   docc-cuda            2.53x (47.24 ms) +0.1%      4.64 J      6         
# atax                   
#   numpy                44.56 ms -2.4%              8.09 J      —         
#   docc-sequential      2.54x (17.55 ms) -0.0%      3.65 J      3         
#   docc-omp             2.01x (22.12 ms) +3.5%      4.70 J      3         
#   docc-cuda            2.40x (18.54 ms) +0.3%      1.84 J      7         
# gemm                   
#   numpy                103.75 ms +0.2%             20.71 J     —         
#   docc-sequential      4.11x (25.22 ms) -0.4%      7.29 J      1         
#   docc-omp             4.09x (25.36 ms) -0.1%      7.31 J      1         
#   docc-cuda            0.14x (752.27 ms) -0.0%     72.52 J     1         
# gesummv                
#   numpy                125.83 ms +2.7%             21.77 J     —         
#   docc-sequential      0.73x (171.94 ms) -1.4%     32.65 J     5         
#   docc-omp             1.67x (75.50 ms) -0.6%      16.00 J     5         
#   docc-cuda            7.03x (17.90 ms) -0.0%      1.85 J      11        
# gemver                 
#   numpy                84.71 ms -2.5%              15.36 J     —         
#   docc-sequential      9.82x (8.63 ms) +6.4%       1.73 J      7         
#   docc-omp             6.57x (12.89 ms) -4.7%      2.67 J      7         
#   docc-cuda            6.72x (12.61 ms) -0.2%      1.26 J      11        
# k2mm                   
#   numpy                93.80 ms +4.0%              19.21 J     —         
#   docc-sequential      0.69x (135.63 ms) +0.0%     24.40 J     3         
#   docc-omp             0.68x (137.62 ms) +1.9%     26.87 J     3         
#   docc-cuda            0.11x (820.43 ms) +0.0%     78.95 J     7         
# k3mm                   
#   numpy                73.96 ms -5.2%              17.59 J     —         
#   docc-sequential      0.24x (308.05 ms) -0.1%     50.09 J     4         
#   docc-omp             0.22x (330.08 ms) -0.2%     56.91 J     4         
#   docc-cuda            0.06x (1.30 s) +0.0%        125.16 J    8         
# mvt                    
+   numpy                42.15 ms -13.8%             7.64 J      —         
#   docc-sequential      2.30x (18.36 ms) +0.1%      3.84 J      2         
#   docc-omp             2.30x (18.35 ms) +0.2%      3.82 J      2         
#   docc-cuda            0.12x (361.31 ms) +0.1%     35.69 J     2         
# symm                   
#   numpy                63.83 ms +0.4%              6.20 J      —         
#   docc-sequential      4.40x (14.51 ms) -2.5%      3.10 J      3         
#   docc-omp             3.57x (17.90 ms) -1.6%      4.05 J      3         
# syr2k                  
#   numpy                75.12 ms -0.6%              7.25 J      —         
#   docc-sequential      2.10x (35.74 ms) -0.7%      7.17 J      1         
#   docc-omp             2.08x (36.08 ms) -0.6%      7.26 J      1         
#   docc-cuda            2.57x (29.27 ms) -5.1%      2.90 J      1         
# syrk                   
#   numpy                63.40 ms -0.0%              6.16 J      —         
#   docc-sequential      2.40x (26.41 ms) -0.8%      5.31 J      1         
#   docc-omp             2.41x (26.33 ms) -2.8%      5.31 J      1         
#   docc-cuda            2.85x (22.22 ms) -1.0%      2.18 J      1         
# trmm                   
#   numpy                70.81 ms -1.0%              6.88 J      —         
#   docc-sequential      4.92x (14.39 ms) +0.0%      3.00 J      3         
#   docc-omp             3.82x (18.53 ms) -5.3%      3.92 J      3         

[daisytuner]

Daisytuner Report - mlir_torch_models (chamomile)

@@                              Benchmarks                              @@
===========================================================================
  Benchmark              Speedup (Time) ±%           Energy      Regions   
===========================================================================
# resnet18               
#   torch                90.34 ms +3.6%              23.64 J     —         
#   torch-cuda           18.32 ms +1.6%              3.76 J      —         
#   docc-omp             0.13x (672.85 ms) +1.6%     215.95 J    45        
#   docc-cuda            0.37x (49.16 ms) +0.1%      9.49 J      101       
# segformer              
#   torch-cuda           37.87 ms -2.4%              7.50 J      —         
#   docc-cuda            0.24x (156.42 ms) +0.4%     27.61 J     404