daisytuner / docc / 28158800507
62%

DEFAULT BRANCH: main
Ran
Jobs 6
Files 493
Run time 2min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 61.644% (+0.06%) from 61.582%
28158800507

push

github

web-flow 
MapFusionByDomain (#771)

 + New Map fusion caches data about iteration domain and map candidates
 + only matches up iteration domain exactly, per loop level.
 + Can support fusing non-leaf stacks of loops (stack ends where the shallower stack stops being perfectly nested & parallel)
 + new Element::replace for bulk replacements
 + New PatternMatcher visitor supports descending into replaced or modified nodes to allow for single-pass nested loop fusings
 + LoopAnalysis can now be kept up-to-date with changes done by Map-fusion
 + unit tests for the updating of LoopAnalysis
 * updated LoopAnalysis to be easier to keep up-to-date with changes. LoopTree is no longer ordered, if you want to iterate in pre-order, use the specific method for that
 + convenience StructuredSDFGBuilder.remove_from_parent()
 + RedundantLoadElim pass to skip reading from memory locations that have just been written (same block). Fusing no longer needs to do this
     RedundantLoadElimination does a simple check for other writes to the same structure. Can skip writes if redundant or not modify, if their are writes to different indices
* Updated verifiers to match new fusion
~ moved verifier checks behind correctness checks in npbench harness. Its more critical if we do not even get the expected results
* Added MapFusionByDomain also to loop-norm stage (currently inactive, causes more kernels that currently cannot be safely offloaded to CUDA.
---------

Co-authored-by: Lukas Truemper <lukas.truemper@outlook.de>

771 of 1186 new or added lines in 55 files covered. (65.01%)

6 existing lines in 6 files now uncovered.

38302 of 62134 relevant lines covered (61.64%)

987.24 hits per line

Uncovered Changes

Lines Coverage ∆ File
175
65.82
 opt/src/passes/map_fusion_by_domain_pass.cpp
22
79.95
 -2.47% sdfg/src/data_flow/library_nodes/math/tensor/einsum_node.cpp
17
15.7
 -1.3% sdfg/src/data_flow/library_nodes/math/tensor/spatial_tensor_node.cpp
16
82.59
 7.16% sdfg/src/analysis/loop_analysis.cpp
14
72.99
 -5.19% sdfg/src/data_flow/pointer_metadata.cpp
12
49.62
 -1.55% sdfg/src/data_flow/library_nodes/math/blas/batched_gemm_node.cpp
10
75.94
 -2.45% sdfg/src/structured_control_flow/structured_loop.cpp
9
92.37
 opt/src/passes/redundant_load_elimination_pass.cpp
9
33.19
 -1.36% sdfg/src/data_flow/library_nodes/math/tensor/tensor_layout.cpp
8
45.75
 -0.81% sdfg/src/data_flow/library_nodes/math/blas/gemm_node.cpp
8
0.0
 0.0% sdfg/src/data_flow/library_nodes/math/tensor/broadcast_node.cpp
7
79.78
 -1.65% opt/src/targets/cuda/cuda_map_dispatcher.cpp
7
38.96
 -1.86% sdfg/src/data_flow/library_nodes/call_node.cpp
6
75.17
 -0.27% sdfg/src/builder/structured_sdfg_builder.cpp
6
57.89
 -3.22% sdfg/src/structured_control_flow/while.cpp
6
78.32
 -0.25% sdfg/src/symbolic/utils.cpp
5
61.9
 -8.37% sdfg/include/sdfg/data_flow/pointer_metadata.h
5
43.54
 -1.07% sdfg/src/data_flow/library_nodes/math/blas/dot_node.cpp
5
67.38
 -0.61% sdfg/src/data_flow/library_nodes/math/tensor/conv_node.cpp
5
57.46
 -0.93% sdfg/src/data_flow/library_nodes/math/tensor/elementwise_node.cpp
5
83.91
 -1.64% sdfg/src/data_flow/library_nodes/math/tensor/reduce_node.cpp
5
0.0
 0.0% sdfg/src/data_flow/library_nodes/math/tensor/transpose_node.cpp
5
76.0
 -10.36% sdfg/src/structured_control_flow/if_else.cpp
4
45.71
 -0.44% sdfg/src/data_flow/library_nodes/math/tensor/matmul_node.cpp
4
0.0
 0.0% sdfg/src/data_flow/library_nodes/stdlib/calloc.cpp
4
37.21
 -1.82% sdfg/src/data_flow/library_nodes/stdlib/memset.cpp
3
40.0
 opt/include/sdfg/passes/map_fusion_by_domain_pass.h
3
0.0
 0.0% sdfg/src/data_flow/library_nodes/metadata_node.cpp
3
28.3
 -1.7% sdfg/src/data_flow/library_nodes/stdlib/alloca.cpp
3
47.54
 -2.46% sdfg/src/data_flow/library_nodes/stdlib/free.cpp
3
57.53
 -2.47% sdfg/src/data_flow/library_nodes/stdlib/malloc.cpp
3
39.74
 -1.59% sdfg/src/data_flow/library_nodes/stdlib/memcpy.cpp
3
45.59
 -2.1% sdfg/src/data_flow/library_nodes/stdlib/memmove.cpp
3
0.0
 0.0% sdfg/src/data_flow/library_nodes/stdlib/trap.cpp
3
0.0
 0.0% sdfg/src/data_flow/library_nodes/stdlib/unreachable.cpp
2
60.0
 -2.5% sdfg/src/data_flow/library_nodes/barrier_local_node.cpp
1
0.0
 opt/include/sdfg/passes/redundant_load_elimination_pass.h
1
0.0
 0.0% sdfg/src/data_flow/library_nodes/invoke_node.cpp
1
67.24
 -0.58% sdfg/src/data_flow/library_nodes/load_const_node.cpp
1
13.26
 -0.04% sdfg/src/data_flow/library_nodes/math/cmath/cmath_node.cpp
1
44.88
 -0.18% sdfg/src/data_flow/library_nodes/math/tensor/batchnorm_node.cpp
1
0.0
 0.0% sdfg/src/data_flow/library_nodes/stdlib/assert.cpp
1
31.25
 0.0% sdfg/src/structured_control_flow/return.cpp

Coverage Regressions

Lines Coverage ∆ File
1
0.0
 0.0% sdfg/include/sdfg/data_flow/library_nodes/math/tensor/broadcast_node.h
1
82.59
 7.16% sdfg/src/analysis/loop_analysis.cpp
1
43.54
 -1.07% sdfg/src/data_flow/library_nodes/math/blas/dot_node.cpp
1
57.53
 -2.47% sdfg/src/data_flow/library_nodes/stdlib/malloc.cpp
1
72.99
 -5.19% sdfg/src/data_flow/pointer_metadata.cpp
1
76.0
 -10.36% sdfg/src/structured_control_flow/if_else.cpp
Jobs
ID Job ID Ran Files Coverage
1 llvm - 28158800507.1 26
33.12
 GitHub Action Run
2 python-3.12 - 28158800507.2 15
73.07
 GitHub Action Run
3 python-3.13 - 28158800507.3 15
73.07
 GitHub Action Run
4 python-3.14 - 28158800507.4 15
73.07
 GitHub Action Run
5 sdfg - 28158800507.5 452
65.3
 GitHub Action Run
6 python-3.11 - 28158800507.6 15
73.07
 GitHub Action Run
Source Files on build 28158800507