torch-spyre/ktir-cpu

View on GitHub

A CPU interpreter and validator for KTIR (Kernel Tile IR), an MLIR-based tile-parallel IR for multi-core accelerator kernels.

Stars: 5
Forks: 12
Open beginner issues: 5
Indexed issues: 50
Avg merge: 8h 35m
Merged PRs (30d): 3

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 18, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: good first issue help wanted

Beginner friendly Everything

Open All

50 open issues indexed

linalg.generic cannot scatter into a composite output indexing map Open

Difficulty 4/5 3-5 days Newbie friendliness 58/100

torch-spyre/ktir-cpu#216 · Opened Sep 14, 2026
[Refactor] Unify indirect-access offset paths, add example tests Open

bug documentation good first issue

torch-spyre/ktir-cpu#212 · 1 assignee · Opened Aug 18, 2026
[Story] Decide whether three tile-wide ops priced zero are really free Open

enhancement

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#211 · 2 comments · Opened Aug 18, 2026
[Feature] Make "the simulator supports this kernel" a checkable statement Open

enhancement

torch-spyre/ktir-cpu#209 · 1 assignee · Opened Aug 17, 2026
[Feature] Add a write-time entry point for the review checks Open

enhancement

Difficulty 4/5 3-5 days Newbie friendliness 45/100

torch-spyre/ktir-cpu#206 · 1 comment · Opened Aug 10, 2026
[Bug] Scalar arith float binops always compute in f16, ignoring the declared result type Open

bug

Difficulty 3/5 1-2 days Newbie friendliness 72/100

torch-spyre/ktir-cpu#205 · Opened Aug 10, 2026
tensor_input_output_sizes() reports non-argument SSA names, and omits the real parameter Open

question

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

torch-spyre/ktir-cpu#204 · Opened Aug 10, 2026
[Bug] func.func parameter list truncated at the first ')': per-argument loc(...) drops parameters and mangles types Open

bug

Difficulty 3/5 1-2 days Newbie friendliness 55/100

torch-spyre/ktir-cpu#203 · Opened Aug 10, 2026
[Bug] Regex parser mis-bounds functions in a multi-function module: body and grid come from a later function Open

bug

Difficulty 4/5 3-5 days Newbie friendliness 35/100

torch-spyre/ktir-cpu#202 · Opened Aug 10, 2026
Inter-tile reduce diagnostic raises AttributeError instead of reporting the offending sets Open

Difficulty 2/5 Half a day Newbie friendliness 78/100

torch-spyre/ktir-cpu#198 · Opened Aug 5, 2026
[Feature] Establish a generalized benchmark suite Open

enhancement

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#196 · Opened Aug 4, 2026
Consider replacing regex parser with xDSL's MLIR text parser Open

question refactoring

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#187 · 1 reaction · Opened Jul 28, 2026
Clarify zero-padding semantics for OOB reads in regular and gather/scatter paths Open

documentation

torch-spyre/ktir-cpu#184 · 3 comments · 1 assignee · Opened Jul 28, 2026
Remove GIT_PAT usage from CI and docs Open

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

torch-spyre/ktir-cpu#178 · Opened Jul 23, 2026
[EPIC] Enhance Communication Ring Model Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#172 · 2 comments · Opened Jul 22, 2026
add multi-core FFN / SwiGLU to roofline plots in latency-demo notebook Open

Difficulty 3/5 1-2 days Newbie friendliness 64/100

torch-spyre/ktir-cpu#170 · Opened Jul 21, 2026
[latency] Measure cross-core reduce (inter_tile_reduce) impact using paged attention Open

latency

torch-spyre/ktir-cpu#162 · 1 comment · 1 assignee · Opened Jul 17, 2026
[Discussion] introduce mlir fixture files for (negative) tests Open

question

Difficulty 5/5 Over a week Newbie friendliness 25/100

torch-spyre/ktir-cpu#161 · Opened Jul 16, 2026
Create skill for Roofline Analysis Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#159 · Opened Jul 14, 2026
Create workflow skill for cross-core kernel/algorithm approaches Open

torch-spyre/ktir-cpu#158 · 5 assignees · Opened Jul 14, 2026
Best algorithms collection for perf estimates Open

torch-spyre/ktir-cpu#154 · 3 assignees · Opened Jul 9, 2026
hidden bug when parsing function header and attributes Open

Difficulty 2/5 1-3 hours Newbie friendliness 68/100

torch-spyre/ktir-cpu#151 · Opened Jul 6, 2026
[Feature] Remaining inter-tile comm ops: broadcast, reduce-scatter, per-tile sync, multi-cast load Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

torch-spyre/ktir-cpu#146 · Opened Jun 30, 2026
mlir_frontend path missing outs_operands and use_counts after #134 and #136 Open

bug good first issue

Difficulty 3/5 1-2 days Newbie friendliness 75/100

torch-spyre/ktir-cpu#139 · Opened Jun 24, 2026
Adding a rustlang port of ktir-cpu Open

torch-spyre/ktir-cpu#128 · 2 comments · 1 assignee · Opened Jun 17, 2026
Support for `linalg.matmul_transpose_b` vs support for `matmul` with an indexing map Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#119 · 2 comments · Opened Jun 16, 2026
comm: add efficient ReduceBackend variants + attribute-based dispatch Open

enhancement latency refactoring

Difficulty 4/5 3-5 days Newbie friendliness 52/100

torch-spyre/ktir-cpu#108 · Opened Jun 11, 2026
[Feature] HBM/LX memory bootstrap: explicit placement API and allocation-aware next_ptr Open

good first issue

Difficulty 4/5 3-5 days Newbie friendliness 55/100

torch-spyre/ktir-cpu#102 · Opened Jun 10, 2026
[Feature] Vectorized indirect coord build, reusing one subscript dispatch Open

refactoring

torch-spyre/ktir-cpu#98 · 1 comment · 1 assignee · Opened Jun 9, 2026
[Feature] MemoryOps.load/store accept ndarray coords to skip tolist()/tuple() conversion Open

refactoring

torch-spyre/ktir-cpu#97 · 2 comments · 1 assignee · Opened Jun 9, 2026
[Feature] Preserve BoxSet structure under permuted variables_space_order in indirect access Open

bug

torch-spyre/ktir-cpu#96 · 2 comments · 1 assignee · Opened Jun 9, 2026
[Feature] Strided sub-rectangle fast path in MemoryOps.load/store Open

refactoring

torch-spyre/ktir-cpu#95 · 2 comments · 1 assignee · Opened Jun 9, 2026
[Docs] Add latency estimation demo notebook Open

latency

torch-spyre/ktir-cpu#94 · 1 comment · 4 assignees · Opened Jun 5, 2026
test_ops.py: incomplete dtype coverage for arithmetic and math ops Open

torch-spyre/ktir-cpu#92 · 1 assignee · Opened Jun 5, 2026
ktdp.reduce / ktdp.coreid are non-spec ops in the regex executor (reconcile or remove) Open

refactoring

torch-spyre/ktir-cpu#88 · 1 assignee · Opened Jun 5, 2026
[Feature] Parallel GridExecutor to speed up multi-core simulation Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#78 · Opened Jun 2, 2026
[Docs] Fix misleading linalg.matmul comment in LatencyEstimator._estimate Open

good first issue refactoring

Difficulty 1/5 Under an hour Newbie friendliness 88/100

torch-spyre/ktir-cpu#75 · Opened Jun 2, 2026
[Question] cso direct-path: pointwise or sort-then-gather (access_tile_order) Open

question

torch-spyre/ktir-cpu#68 · 1 assignee · Opened May 28, 2026
[Bug] read_scattered cross-allocation safety: hard-guard the caller invariant Open

bug

Difficulty 4/5 3-5 days Newbie friendliness 38/100

torch-spyre/ktir-cpu#67 · Opened May 21, 2026
[Question] Coordinate collision semantics for indirect_store / scatter primitives Open

question

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#60 · 2 comments · Opened May 15, 2026
[Bug] MemoryOps contract is undefined for non-rectangular variables_space_set Open

bug

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#56 · Opened May 15, 2026
Add distributed memory view examples and test_examples coverage Open

good first issue

Difficulty 3/5 1-2 days Newbie friendliness 72/100

torch-spyre/ktir-cpu#55 · Opened May 15, 2026
[Question] Should ktdp.construct_indirect_access_tile compose with DistributedMemRef? Open

question

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#53 · 1 comment · Opened May 15, 2026
[Test] Add more dynamic-shape MLIR examples (varying dtypes, 2-D, multi-core) Open

good first issue

Difficulty 4/5 3-5 days Newbie friendliness 52/100

torch-spyre/ktir-cpu#47 · Opened May 13, 2026
@register should accept an optional attributes schema for runtime validation and documentation Open

question refactoring

Difficulty 4/5 3-5 days Newbie friendliness 55/100

torch-spyre/ktir-cpu#40 · Opened May 7, 2026
[Feature] HBM memory model: cross-load stick coalescing not modeled Open

question

Difficulty 4/5 3-5 days Newbie friendliness 42/100

torch-spyre/ktir-cpu#31 · 2 comments · Opened May 6, 2026
[Question] Out-of-Set Element Semantics for ktdp.load / ktdp.store Open

question

Difficulty 5/5 Over a week Newbie friendliness 30/100

torch-spyre/ktir-cpu#28 · 1 comment · Opened May 5, 2026
[CI] Install ktir-mlir-frontend Python wheels Open

enhancement

Difficulty 4/5 3-5 days Newbie friendliness 48/100

torch-spyre/ktir-cpu#7 · 1 reaction · Opened Apr 22, 2026
Transformer decoder op coverage: gap analysis and MLIR examples Open

good first issue help wanted

torch-spyre/ktir-cpu#3 · 2 assignees · Opened Apr 22, 2026
[RFC conformance] Add RFC-explicit non-ktdp ops Open

good first issue help wanted

Difficulty 5/5 Over a week Newbie friendliness 35/100

torch-spyre/ktir-cpu#2 · 2 comments · Opened Apr 22, 2026

torch-spyre/ktir-cpu

Get new issues in your inbox