Search results

Frontend batch_matmul lowering emits invalid ND reshape/slice for 4-D matmul operands

Status: Open.

#1233 In hw-native-sys/pypto;

· ndleslx opened on Apr 30, 2026

[Bug] a2a3 rejects row-broadcast weight slices lowered as 16x1 ColMajor Vec TLOAD from ND GM

Status: Open.

#1230 In hw-native-sys/pypto;

· high-cloud opened on Apr 30, 2026

[Bug] pto.texp op fails with 'expects src to use the row_major blayout' when data flows through GM

bugSomething isn't workingSomething isn't working

Status: Open.

Bug

#1229 In hw-native-sys/pypto;

· zhusy54 opened on Apr 30, 2026

[Bug] execute_compiled does not bucket tensor/scalar args, forces source-order constraint on orchestration signatures

bugSomething isn't workingSomething isn't working

Status: Open.

#1228 In hw-native-sys/pypto;

· zhangqi-chen opened on Apr 29, 2026

[Feature] Support pl.function calling another pl.function in the frontend

enhancementNew feature or requestNew feature or request

Status: Open.

#1221 In hw-native-sys/pypto;

· zhangqi-chen opened on Apr 29, 2026

[Bug] a2a3 FP32 matmul with b_trans=True gives wrong results for [16,16384] x [32,16384].T

bugSomething isn't workingSomething isn't working

Status: Open.

Bug

#1213 In hw-native-sys/pypto;

· high-cloud opened on Apr 28, 2026

[a2a3] Silent miscompile: column-strided slice of wide 2D tensor fed into pl.matmul produces wrong results (only row 0 correct)

bugSomething isn't workingSomething isn't working

Status: Open.

#1212 In hw-native-sys/pypto;

· zhaozhaozz opened on Apr 28, 2026

[Feature] Expose child_memory (persistent device weights) in pypto frontend, unified for L2/L3

enhancementNew feature or requestNew feature or request

Status: Open.

#1204 In hw-native-sys/pypto;

· lyfne123 opened on Apr 28, 2026

[Feature] Support target_memory on pl.slice (or fuse slice+move) for Mat→Left/Right path

enhancementNew feature or requestNew feature or request

Status: Open.

#1198 In hw-native-sys/pypto;

· zhangqi-chen opened on Apr 28, 2026

[Pass Bug] tpop ordering + cross-core move errors with 2-step unrolling and persistent Cube accumulators

bugSomething isn't workingSomething isn't working

Status: Open.

#1195 In hw-native-sys/pypto;

· bumble0918 opened on Apr 27, 2026

[Pass Bug] tpop ordering violation with pl.parallel in mixed pl.at block

bugSomething isn't workingSomething isn't working

Status: Open.

#1194 In hw-native-sys/pypto;

· bumble0918 opened on Apr 27, 2026

[Feature] Add orchestration-level collective communication operators to PyPTO compiler

enhancementNew feature or requestNew feature or request

Status: Open.

#1189 In hw-native-sys/pypto;

· bumble0918 opened on Apr 27, 2026