Send whole sdfg with rpc

[NoraHagmeyer]

Description

Prepares removing the loop normalization and bringing down compile time for segformer with target cuda (with and without remote_tuning) .

• Replaces the RPCScheduler in the LoopSchedulingPass with an RPCSchedulingPass

• Sends the whole sdfg to the RPC server (still using a transformation, to be made more efficient)

• Changes the transer tuning rest endpoint to a newly added one

• Handles response of multiple region informations

• Removes the RPC Opt application

• Changes the result in the Opt Report (to be re-added later or changed to a better mechanism for tracking) regression already tracked through Re-enable rpc request tracking via Opt Report #908

• Possible conflicts with Reduce variability between compile outputs and transfertuning requests #880

Related Issues

• Closes #

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Refactor / internal change (no functional change)
• Documentation update

Quality Checklist

Please ensure the following are completed before requesting review:

• Runtime-compatible (no externally visible existing method has its parameters changed, was deleted or has its exception behavior changed)
• Compile-time compatible (new arguments & properties added have default values, but users need to recompile against the new version to work)

IR:

• Introduces new types into the SDFG IR
• Changed / added properties on SDFG IR Elements (LibraryNodes, ControlFlowNodes DataFlowNodes etc.)
  • New / modified types have serialization / deserialization support
  • Deserialization is backward compatible (new serialization can deserialize old format and represent it in the new format if applicable)
  • Serialized format is backward compatible (previous deserializer can read new serialized format without incorrectness (just lacking additional guarantees that are optional or did not exist before)
  • Serializer has an option to emit the previous format

Unit Tests

• New unit tests have been added for the changes.
• Existing unit tests pass locally.
• Edge cases and error paths are covered.

Integration Tests

• New integration tests have been added (or existing ones updated).
• Integration tests pass locally.
• Interactions with other components/services are verified.

Documentation

• Public APIs, options, and behavior changes are documented.
• README / docs site / inline comments updated where applicable.
• Changelog / release notes updated (if applicable).

Test Coverage

• Test coverage has not decreased as a result of this change.
• Coverage report reviewed and acceptable.
• Critical / new code paths are covered.

How Has This Been Tested?

RPC unit tests in combination with local rpc server
Transfer Tuning Integration tests in combination with local rpc server
Ran benchmarks for this code state

Additional Context

[daisytuner]

Daisytuner Report - python_npbench (zinnia)

@@                                            Benchmarks                                            @@
=======================================================================================================
  Benchmark              Speedup (Time) ±%           Energy      Regions   Types (M/F/R/O)   % Peak    
=======================================================================================================
# adi                    
#   numpy                118.15 ms -1.1%             11.73 J     —         —                 N/A       
#   docc-sequential      2.27x (52.12 ms) -1.1%      8.41 J      2         1M 1F             N/A       
#   docc-omp             2.05x (57.73 ms) +0.2%      9.17 J      2         1M 1F             N/A       
#   docc-cuda            2.43x (48.65 ms) +1.6%      4.86 J      6         1M 1F             N/A       
# atax                   
#   numpy                46.81 ms +9.8%              8.06 J      —         —                 N/A       
#   docc-sequential      2.67x (17.53 ms) +0.2%      3.66 J      3         1M                N/A       
#   docc-omp             2.05x (22.85 ms) +6.5%      5.04 J      3         1M                N/A       
#   docc-cuda            1.54x (30.37 ms) +0.2%      3.08 J      6         2M                N/A       
# gemm                   
#   numpy                99.87 ms +0.1%              19.12 J     —         —                 N/A       
#   docc-sequential      3.95x (25.25 ms) -0.4%      7.23 J      1         —                 N/A       
#   docc-omp             3.95x (25.26 ms) -0.5%      7.25 J      1         —                 N/A       
#   docc-cuda            0.13x (752.27 ms) -0.0%     74.38 J     1         —                 N/A       
# gesummv                
#   numpy                127.76 ms +8.6%             23.15 J     —         —                 N/A       
#   docc-sequential      0.74x (172.97 ms) -0.9%     32.81 J     5         3M                N/A       
#   docc-omp             1.76x (72.75 ms) +1.7%      15.38 J     5         3M                N/A       
#   docc-cuda            6.52x (19.59 ms) -0.0%      1.99 J      10        4M                N/A       
# gemver                 
-   numpy                102.23 ms +21.2%            18.35 J     —         —                 N/A       
#   docc-sequential      12.40x (8.24 ms) -1.0%      1.88 J      7         3M                N/A       
#   docc-omp             7.59x (13.48 ms) -7.1%      2.97 J      7         3M                N/A       
#   docc-cuda            8.28x (12.35 ms) +0.1%      1.29 J      12        6M                N/A       
# k2mm                   
#   numpy                101.44 ms -1.5%             20.34 J     —         —                 N/A       
#   docc-sequential      0.75x (135.21 ms) +0.1%     24.57 J     3         1M                N/A       
#   docc-omp             0.75x (135.48 ms) -4.4%     26.08 J     3         1M                N/A       
#   docc-cuda            0.12x (820.34 ms) -0.0%     80.96 J     7         1M                N/A       
# k3mm                   
#   numpy                77.48 ms -0.9%              17.87 J     —         —                 N/A       
#   docc-sequential      0.25x (309.31 ms) +0.6%     50.77 J     4         1M                N/A       
#   docc-omp             0.23x (331.11 ms) -0.2%     56.98 J     4         1M                N/A       
#   docc-cuda            0.06x (1.30 s) -0.0%        128.22 J    8         1M                N/A       
# mvt                    
+   numpy                31.18 ms -29.5%             5.72 J      —         —                 N/A       
#   docc-sequential      1.70x (18.38 ms) +0.2%      3.84 J      2         —                 N/A       
#   docc-omp             1.70x (18.38 ms) -0.7%      3.84 J      2         —                 N/A       
#   docc-cuda            1.02x (30.49 ms) +0.2%      3.07 J      1         1M                N/A       
# symm                   
#   numpy                64.00 ms +1.0%              6.34 J      —         —                 N/A       
#   docc-sequential      4.29x (14.93 ms) +2.6%      3.17 J      3         2M 1F             N/A       
#   docc-omp             3.58x (17.86 ms) -2.8%      3.96 J      3         2M 1F             N/A       
# syr2k                  
#   numpy                74.12 ms -0.0%              7.39 J      —         —                 N/A       
#   docc-sequential      2.05x (36.14 ms) +2.1%      7.28 J      1         1F                N/A       
#   docc-omp             2.07x (35.85 ms) +0.2%      7.28 J      1         1F                N/A       
#   docc-cuda            2.50x (29.66 ms) -0.1%      2.98 J      1         1F                N/A       
# syrk                   
#   numpy                62.26 ms -0.9%              6.16 J      —         —                 N/A       
#   docc-sequential      2.32x (26.83 ms) +2.3%      5.32 J      1         1F                N/A       
#   docc-omp             2.32x (26.84 ms) +2.4%      5.33 J      1         1F                N/A       
#   docc-cuda            2.79x (22.27 ms) +0.3%      2.28 J      1         1F                N/A       
# trmm                   
#   numpy                71.09 ms -0.4%              7.11 J      —         —                 N/A       
#   docc-sequential      4.94x (14.40 ms) +0.3%      3.02 J      3         2M 1F             N/A       
#   docc-omp             3.93x (18.07 ms) -2.5%      3.89 J      3         2M 1F             N/A       

[daisytuner]

Daisytuner Report - mlir_torch_models (chamomile)

@@                                            Benchmarks                                            @@
=======================================================================================================
  Benchmark              Speedup (Time) ±%           Energy      Regions   Types (M/F/R/O)   % Peak    
=======================================================================================================
# resnet18               
#   torch                90.68 ms +3.2%              22.22 J     —         —                 N/A       
#   torch-cuda           18.45 ms -1.4%              3.50 J      —         —                 N/A       
#   docc-omp             0.13x (674.58 ms) -2.3%     208.31 J    45        44M               N/A       
#   docc-cuda            0.36x (51.36 ms) +0.1%      9.38 J      101       72M               N/A       
# segformer              
#   torch-cuda           39.48 ms +2.6%              7.53 J      —         —                 N/A       
#   docc-cuda            0.28x (139.02 ms) +1.3%     23.16 J     435       337M              N/A       

[Moehre2]

Not sure if this is intended or slipped through... Therefore, approving.

The merge-base changed after approval.

The merge-base changed after approval.