Reduction fusion testing. - #5

Closed

charithaintc wants to merge 68 commits into

mainfrom

test_fusion_approach

Commits on May 20, 2026

[KB] Update level1 & level2 status (#155)
rengolin
authored
8842681

Commits on May 22, 2026

[KB] Enable dynamo plugin for KB tests (#157)
rengolin
authored
a7e5ff5

Commits on May 26, 2026

[KB] Load MLIR runtime on CPU plugin (#158)
rengolin
authored
4c4174f

Commits on May 27, 2026

Commits on May 29, 2026

[Torch-MLIR] Update version (#165)
rengolin
authored
911d54a

Commits on May 31, 2026

[KB] Level1 size updates for smoke test (#166)
rengolin
authored
cccc0f3

Commits on Jun 1, 2026

Commits on Jun 2, 2026

[Cleanup] Examples cleanup (#173)
rengolin
authored
02dd5f4
[runner] Dump object file (#174)
adam-smnk
authored
ba50ac7
[Cleanup] Remove unused workload module (#172)
rengolin
authored
e36be4d
[ingress][torch] Add optional model_datatype argument to import_from_file (#169)
tkarna
authored
45f5473
[KB] Update examples in README
rengolin
committed
3a55932
Update README.md with current status
rengolin
authored
aa3b065
[xegpu] Fix GA matmul tuner (#175)
tkarna
authored
293b533
[KB] Use --dtype instead of --bf16 (#176)
rengolin
authored
2064122
Batch matmul pipeline (#177)
rengolin
authored
2a3282c

Commits on Jun 3, 2026

[KB] Add dump-assembly option to kernel-bench (#179)
rengolin
authored
d85a49b
[ingress][torch] Use Runner in PyTorch backend (#178)
adam-smnk
authored
03118d9
[KB] Fix default pipeline name (#180)
adam-smnk
authored
6e645a3
[KB] Avoid process migration when benchmarking (#183)
rengolin
authored
7cba4c1
Update README.md
rengolin
authored
c28d14c
[ingress][torch][runner] PyTorch backend - dump object file (#181)
adam-smnk
authored
c61a06e
[Descriptor] Allow variable replacement in descriptor files (#182)
rengolin
authored
d50a962

Commits on Jun 8, 2026

[xegpu] Add transpose A/B support in mlp schedule (#184)
tkarna
authored
c005d21
[KB] Matvec schedule - medium perf (#185)
rengolin
authored
794e215

Commits on Jun 10, 2026

[examples][xegpu] Initialize matmul inputs with random floats (#187)
tkarna
authored
6e2218a

Commits on Jun 11, 2026

[xegpu] Add cost model based matrix multiplication tuner (#188)
tkarna
authored
931b039
[KB] Add element-wise kernel (#190)
rengolin
authored
9c740a1

Commits on Jun 15, 2026

Commits on Jun 16, 2026

[KB] bf16 element-wise is done at F32 on CPU
rengolin
committed
a1e6d30
Linalg/XeGPU Layer norm example (#186)

nbpatel
and
rengolin
authored
74bc5b8

Commits on Jun 17, 2026

Commits on Jun 22, 2026

Bump PyTorch version (#197)
adam-smnk
authored
9161613

Commits on Jun 25, 2026

Commits on Jun 26, 2026

[tools] Fix KB model type conversion (#201)

rengolin
and
adam-smnk
authored
04e5ba8

Commits on Jun 29, 2026

[xegpu] Elementwise schedule and refactoring (#200)
tkarna
authored
72840c0

Commits on Jun 30, 2026

[KB] Small CI test config (#204)
adam-smnk
authored
46bd97c
[transform_ext][schedule] Tiling - optionally filter out non-tileable ops (#205)
adam-smnk
authored
a5f745c
[xegpu] Add KernelBench routine (#203)
tkarna
authored
4891f82

Commits on Jul 1, 2026

[schedule][transform][x86] AMX move data offsets to subviews (#206)

adam-smnk
and
jopperm
authored
7577818
[CI][xegpu] Lighter kernel bench CI tests (#210)
tkarna
authored
198060c
[utils] Enable OS AMX support (#207)
adam-smnk
authored
6f14f3a

Commits on Jul 2, 2026

[KernelBench] Add XeGPU support in kernel bench tool (#209)
tkarna
authored
7e3f0cc
[KB][x86] AMX lowering pipelines (#211)

adam-smnk
and
rengolin
authored
a6b7dd3

Commits on Jul 3, 2026

[ingress][torch] Suppress future warnings (#208)
adam-smnk
authored
b2b8b65

Commits on Jul 6, 2026

[schedule][KB] Parallelize execution using OpenMP (#212)
adam-smnk
authored
2f3d4bb
[KB] Default to torch-compile (#213)
rengolin
authored
2bb8bb8

Commits on Jul 7, 2026

Commits on Jul 8, 2026

save changes
charithaintc
committed
301292c

Commits on Jul 20, 2026

add schedule changes for SLM promotion
charithaintc
committed
be3c0b1
inner and outer tiling in attention
charithaintc
committed
feddd9c

Commits on Jul 21, 2026

add vector and buffer
charithaintc
committed
6ee359d
fix large buffer issue
charithaintc
committed
bf3735c

Commits on Jul 22, 2026

try to fold unit dims
charithaintc
committed
b9a2db7
save progress
charithaintc
committed
52eb8f1