@cannbot-plugin/cannbot-harness 0.0.0-stage → 0.1.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +104 -3
- package/THIRD_PARTY_NOTICES.md +11 -0
- package/assets/ops-direct-invoke/.cannbot-bundle +1 -0
- package/assets/ops-direct-invoke/.claude-plugin/plugin.json +21 -0
- package/assets/ops-direct-invoke/.codex-plugin/plugin.json +30 -0
- package/assets/ops-direct-invoke/AGENTS.md +93 -0
- package/assets/ops-direct-invoke/LICENSE +38 -0
- package/assets/ops-direct-invoke/QUICKSTART.md +44 -0
- package/assets/ops-direct-invoke/README.md +219 -0
- package/assets/ops-direct-invoke/SKILLS_LICENSE +38 -0
- package/assets/ops-direct-invoke/agents/ops-direct-invoke-architect.md +39 -0
- package/assets/ops-direct-invoke/agents/ops-direct-invoke-developer.md +39 -0
- package/assets/ops-direct-invoke/agents/ops-direct-invoke-verifier.md +38 -0
- package/assets/ops-direct-invoke/init.sh +8 -0
- package/assets/ops-direct-invoke/plugin-install.json +24 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/SKILL.md +88 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/evals/evals.json +207 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-arithmetic.md +399 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-atomic.md +192 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-buffer.md +364 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-cross-gen-fixpipe.md +28 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-cross-gen-migration.md +91 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-crosscore-sync.md +259 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-datacopy.md +405 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-gathermask.md +243 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-gmm.md +260 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-hccl-host.md +187 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-hcomm.md +229 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-host-runtime.md +129 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-loaddata.md +196 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-matmul.md +364 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-mrgsort.md +636 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-pipeline.md +206 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-precision.md +146 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-quickref.md +192 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-reduce-pattern.md +132 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-reduce.md +282 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-repeat-limits.md +144 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-restrictions.md +200 -0
- package/assets/ops-direct-invoke/skills/ascendc-api-best-practices/references/api-transpose.md +234 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/SKILL.md +245 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/block/block_mmad_glu.h +238 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/block/block_scheduler_group_matmul.h +106 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/block/weight_quant_matmul_block_mmad.h +362 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/block_epilogue_dequant_swiglu.h +265 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/block_epilogue_pertoken_quant_single_pass.h +252 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/block_epilogue_pertoken_quant_two_pass_chunked.h +313 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/block_epilogue_swiglu.h +171 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/cv_sync_constants.h +32 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/epilogue_fusion_membase.h +212 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/epilogue_fusion_regbase.h +324 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/online_softmax_cross_core_epilogue.h +396 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/epilogue/online_softmax_per_tile_epilogue.h +421 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/kernel/group_matmul_kernel_cv1_v2.h +150 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/kernel/group_matmul_kernel_glu_fused.h +345 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/kernel/weight_quant_matmul_kernel.h +611 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/policy/dispatch_policy.h +63 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/utils/common_utils.h +101 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/blaze_custom/utils/integral_constant.h +38 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/matmul/blaze_group_matmul_pertoken_quant_selector.h +119 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/matmul/blaze_group_matmul_tiling_data.h +38 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/matmul/blaze_matmul_tiling.h +457 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/matmul/blaze_matmul_tiling_data.h +50 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/mx/quant_matmul_mx_tiling.h +549 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/mx/quant_matmul_mx_tiling_data.h +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/weight_quant/weight_quant_tiling.h +428 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/assets/op_tiling/weight_quant/weight_quant_tiling_data.h +59 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/evals/evals.json +234 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/blaze-investigation/assembly-recovery-method.md +106 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/blaze-investigation/dependency-trace-method.md +140 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/blaze-investigation/design-contract-method.md +78 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/blaze-investigation/investigation-report-template.md +249 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/blaze-investigation/source-investigation-method.md +131 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/fundamentals/blaze-cmake-build-guide.md +184 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/fundamentals/blaze-framework-overview.md +403 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/fundamentals/blaze-matmul-layout.md +322 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/fundamentals/blaze-sync-patterns.md +814 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/fundamentals/tensor-api-reference.md +589 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/blaze-design-template.md +593 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/blaze-plan-template.md +391 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/glu-development.md +755 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/glu-precision-diagnosis.md +113 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/group-matmul-delta.md +299 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/per-token-quant-development.md +450 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/kernel-design/tiling-selection.md +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/launcher/launcher-development.md +346 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-elementwise-then-vector-pertoken-quant/cube-matmul-then-vector-elementwise-then-vector-pertoken-quant-design.md +196 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-elementwise-then-vector-pertoken-quant/cube-matmul-then-vector-elementwise-then-vector-pertoken-quant-development.md +104 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-glu/cube-matmul-then-vector-glu-design.md +150 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-glu/cube-matmul-then-vector-glu-development.md +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-glu-then-vector-pertoken-quant/cube-matmul-then-vector-glu-then-vector-pertoken-quant-design.md +189 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-glu-then-vector-pertoken-quant/cube-matmul-then-vector-glu-then-vector-pertoken-quant-development.md +52 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-pertoken-quant/cube-matmul-then-vector-pertoken-quant-design.md +138 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-pertoken-quant/cube-matmul-then-vector-pertoken-quant-development.md +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/cube-matmul-then-vector-softmax-reduce-design.md +205 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/cube-matmul-then-vector-softmax-reduce-development.md +99 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/cv-sync-and-two-phase-entry.md +156 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/cv1-v2-adaptation-for-softmax-reduce.md +331 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/online-softmax-reduce-epilogue-design.md +174 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/cube-matmul-then-vector-softmax-reduce/reduce-adaptation-guide.md +204 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/block-l0c2ub-extension.md +75 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/elementwise-broadcast-epilogue-fusion-design.md +280 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/elementwise-broadcast-epilogue-fusion-development.md +202 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/epilogue-membase-design.md +82 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/epilogue-regbase-design.md +95 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/fused-kernel-development.md +71 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/precision-diagnosis.md +86 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/elementwise-broadcast-epilogue-fusion/splitm-contract-and-debugging.md +86 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/index.md +14 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/scenario-extension-guide.md +84 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/layout-variant-boundary.md +76 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/pitfalls-and-diagnosis.md +124 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/prologue-ub-layout-design.md +135 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/prologue-vf-dequant-design.md +213 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/vc-sync-and-l1-layout-design.md +103 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/weight-dequant-prologue-fusion-design.md +289 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/weight-dequant-prologue-fusion-development.md +232 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/scenarios/weight-dequant-prologue-fusion/weight-quant-tiling-contract.md +83 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/troubleshooting/compile-troubleshooting.md +231 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/troubleshooting/optimized-runtime-hang.md +102 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/workflow/step1-project-setup.md +170 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/workflow/step2-blaze-investigation.md +192 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/workflow/step3-kernel-design.md +246 -0
- package/assets/ops-direct-invoke/skills/ascendc-blaze-best-practice/references/workflow/step4-operator-development.md +166 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/SKILL.md +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/evals/evals.json +123 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/references/crash_workflow.md +193 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/references/memcheck/README.md +137 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/references/memcheck/automated_workflow.md +418 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/references/memcheck/mssanitizer_guide.md +181 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/scripts/memcheck_input.json.template +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/scripts/parse_plog.py +191 -0
- package/assets/ops-direct-invoke/skills/ascendc-crash-debug/scripts/run_memcheck_pre.sh +696 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/SKILL.md +105 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/evals/evals.json +128 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/CMakeLists.txt +131 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/README.md +55 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_extension/add_custom_torch.cpp +82 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_extension/ops.h +23 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_extension/register.cpp +42 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_host/add_custom.asc +149 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_host/data_utils.h +98 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_kernel/add_custom_kernel.asc +136 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/op_kernel/add_custom_tiling.h +31 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/run.sh +86 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/scripts/gen_data.py +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/scripts/golden.py +36 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/scripts/test_torch.py +90 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/add_custom/scripts/verify_result.py +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kernel_launch_details.md +116 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/CMakeLists.txt +19 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/README.md +214 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/UPSTREAM.md +35 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/add_custom.cpp +99 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/add_custom_tiling.h +25 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/cmake/Modules/CMakeCCECompiler.cmake.in +15 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/cmake/Modules/CMakeCCEInformation.cmake +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/cmake/Modules/CMakeDetermineCCECompiler.cmake +119 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/cmake/Modules/CMakeTestCCECompiler.cmake +11 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/cmake/npu/CMakeLists.txt +36 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/data_utils.h +132 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/ext_lib/.gitattributes +1 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/input/.keep +0 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/main.cpp +108 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/output/.keep +0 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/run.sh +141 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/scripts/acl.json +1 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/scripts/gen_data.py +26 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_add_template/scripts/verify_result.py +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/kirin_vector_guide.md +172 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_guide.md +348 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/CMakeLists.txt +157 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/common/acl_utils.h +70 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/common/common_utils.h +84 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/common/io_utils.h +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_extension/matmul_blaze_torch.cpp +55 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_extension/ops.h +22 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_extension/register.cpp +22 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_kernel/include/blaze_custom/.gitkeep +0 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_kernel/matmul_blaze_kernel.cpp +33 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_kernel/matmul_blaze_kernel.h +99 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_tiling/matmul_constant.h +31 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_tiling/matmul_tiling_data.h +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/op_tiling/matmul_tiling_stub.h +77 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/run.sh +94 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/scripts/gen_data_mxfp4.py +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/scripts/gen_data_mxfp8.py +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/scripts/verify_result.py +132 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/test_matmul_blaze.cpp +145 -0
- package/assets/ops-direct-invoke/skills/ascendc-direct-invoke-template/references/matmul_blaze_template/test_matmul_blaze_torch.py +141 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/SKILL.md +56 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/evals/evals.json +191 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/aclnn-api-doc-template.md +368 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/detailed-design-template.md +388 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/iteration-plan-template.md +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/operator-readme-template.md +234 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/operator-readme-template_bak.md +165 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-gen/references/requirement-analysis-template.md +131 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/SKILL.md +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/evals/evals.json +197 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/references/api-index.md +207 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/references/compatibility.md +141 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/references/example-catalog.md +197 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/requirements.txt +2 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/__init__.py +6 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/ascend_content_fetcher.py +479 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/ascend_search_client.py +201 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/ascend_search_skill.py +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/clean_markdown.py +489 -0
- package/assets/ops-direct-invoke/skills/ascendc-docs-search/scripts/find_api_doc.sh +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/SKILL.md +123 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/evals/evals.json +211 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/references/asys_commands.md +64 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/references/env_config_guide.md +96 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/references/kirin_platform_guide.md +11 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/references/npu_commands.md +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/references/troubleshooting.md +177 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/scripts/_npu_info.py +763 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/scripts/check_env.sh +320 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/scripts/get_npu_arch.py +488 -0
- package/assets/ops-direct-invoke/skills/ascendc-env-check/scripts/npu_info.sh +228 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/SKILL.md +370 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/docs/CONTEXT.md +85 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/docs/MC2-CAPABILITY-PLAN.md +111 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/docs/adr/0001-mc2-skill-not-plugin.md +31 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/docs/adr/0002-decision-stack-path-registry.md +33 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/evals/evals.json +315 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/capability-declaration.md +97 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/codebase-analysis.md +34 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/fundamentals/architecture.md +444 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/fundamentals/communication.md +669 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/fundamentals/compute.md +587 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/fundamentals/fusion.md +550 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/design-template.md +313 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/dev-methodology.md +111 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/development-guide.md +445 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/host-and-testing.md +285 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/operator-anatomy.md +739 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/paradigm-mapping.md +93 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/operator-design/plan-template.md +235 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/review-checklist.md +96 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/compute-first-reduce-scatter/design.md +107 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/compute-first-reduce-scatter/development.md +272 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/get-all-gather-quant-matmul/design.md +93 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/get-all-gather-quant-matmul/development.md +70 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/index.md +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/put-all-gather-quant-matmul/design.md +83 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/put-all-gather-quant-matmul/development.md +62 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/put-all-to-all-quant-matmul/design.md +106 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/scenarios/put-all-to-all-quant-matmul/development.md +74 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/troubleshooting/failure-navigation.md +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/troubleshooting/optimization-playbook.md +143 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/workflow/step1-project-setup.md +52 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/workflow/step2-investigation.md +89 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/workflow/step3-design.md +69 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/workflow/step4-implementation.md +29 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/apace/workflow_integration.md +122 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/api-rules/dispatch-combine-contracts.md +60 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/api-rules/index.md +71 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/api-rules/mte-address-access.md +374 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/api-rules/sync-and-visibility.md +105 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/combine-reading-guide.md +102 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/common-reading-method.md +226 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/design-overview.md +558 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/dispatch-reading-guide.md +81 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/kernel-execution-units.md +189 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/reading/mega-moe-reading-guide.md +123 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/build-framework.md +177 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/change-routing.md +18 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/combine-dataflow.md +144 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/dispatch-dataflow.md +128 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/index.md +98 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/CMakeLists.txt +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/README.md +131 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/build.sh +56 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/include/moe_dispatch_base_compat.h +113 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/include/tiling_data.h +100 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/include/utils.h +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/kernel/moe_dispatch.h +524 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/kernel/mte_dispatch_comm.h +324 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/moe_dispatch.cpp +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/run.sh +61 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/run_multi_data_test.sh +80 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/scripts/verify_dispatch.py +248 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/moe_dispatch_direct_invoke_sample/test/test_moe_dispatch.cpp +679 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/sample-helper-map.md +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/samples/spec-template.md +78 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/tiling-scheme/double-buffer-protocol.md +69 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/tiling-scheme/index.md +111 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/tiling-scheme/multi-core-formulas.md +126 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/tiling-scheme/split-core-design.md +100 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/ascendc-api/moe-dispatch-combine/tiling-scheme/window-memory-layout.md +199 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/CMakeLists.txt +95 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/cmake/ascend.cmake +59 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/cmake/shmem.cmake +75 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/cmake/tensor_api.cmake +48 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/host_utils/acl_utils.h +73 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/host_utils/common_utils.h +181 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/host_utils/io_utils.h +175 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/kernel_utils/common_utils.h +99 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/kernel_utils/integral_constant.h +36 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/common/kernel_utils/tuple_utils.h +34 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/block/block_mmad.h +27 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/block/block_scheduler_policy.h +35 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/block/block_scheduler_utils.h +27 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/block/quant_matmul_mx_block_mmad_swat.h +392 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/block/quant_matmul_mx_block_scheduler_swat.h +290 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/kernel/all_to_all_comm_udma.h +255 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/kernel/all_to_all_matmul_impl.h +370 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/kernel/heavy_kernels.h +128 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/kernel/qbmm_mx_kernel.h +395 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/policy/dispatch_policy.h +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tile/copy_scale_l1_to_l0a.h +61 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tile/copy_scale_l1_to_l0b.h +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tile/pad_mx_kl1.h +138 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tile/tile_mmad_mx.h +34 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tiling/all_to_all_matmul_tiling_data.h +29 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tiling/quant_matmul_mx_tiling_swat.h +410 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tiling/quant_matmul_tiling_base.h +116 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tiling/quant_matmul_tiling_common.h +101 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/tiling/quant_matmul_tiling_data.h +61 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/include/utils/constant.h +106 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/run.sh +65 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/scripts/extract_perf.py +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/scripts/gen_data.py +139 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/scripts/verify_result.py +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/src/all_to_all_matmul.cpp +396 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/all_to_all_matmul/src/utils.h +168 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/codebase_map.md +342 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/comm_shmem.md +280 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/matmul_blaze.md +259 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/mc2_architecture.md +199 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/review-checklist.md +29 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/blaze-shmem/workflow_integration.md +215 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/CMakeLists.txt +559 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/build.sh +319 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/config.cmake +232 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/func.cmake +478 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/intf.cmake +4 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/intf_pub.cmake +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/modules/Findalog.cmake +105 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/scripts/check_version_compatible.py +94 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/cmake/scripts/prepare.sh +129 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/golden.py +27 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/run_test.sh +33 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/single_server_check_result.py +71 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/single_server_config.ini +17 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/single_server_gen_data.py +91 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/single_server_run.py +116 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/examples/test_golden_cpu.py +63 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/CMakeLists.txt +202 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/mc2_common_infershape.cpp +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/mc2_common_infershape.h +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/formulaic_tiling_datatype.h +161 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/hccl_formulaic_tiling.cpp +373 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/hccl_formulaic_tiling.h +150 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/hccl_performance.cpp +206 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/hccl_performance.h +118 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/hcom_topo_info.h +81 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/matmul_formulaic_tiling.cpp +338 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/matmul_formulaic_tiling.h +340 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/matmul_performance.cpp +272 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/matmul_performance.h +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/mc2_tiling_struct.h +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/mc2_tiling_utils.cpp +296 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_host/op_tiling/mc2_tiling_utils.h +269 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_kernel/mc2_matmul_block.h +235 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_kernel/mc2_matmul_block_l2cache.h +165 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_kernel/mc2_matmul_compute.h +218 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_kernel/mc2_nd_to_nz.h +579 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/op_kernel/mc2_tiling_struct.h +164 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/utils/mc2_hcom_topo_info.cpp +390 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/utils/mc2_hcom_topo_info.h +81 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/mc2_common/utils/ops_utils.h +77 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_api/aclnn_x_all_gather_matmul.cpp +65 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_api/aclnn_x_all_gather_matmul.h +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/arch22/x_all_gather_formulaic_tiling_a2a3.cpp +50 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/arch22/x_all_gather_formulaic_tiling_a2a3.h +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/arch22/x_all_gather_matmul_tiling_a2a3.cpp +72 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/arch22/x_all_gather_matmul_tiling_a2a3.h +31 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/x_all_gather_formulaic_tiling.cpp +82 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/x_all_gather_formulaic_tiling.h +55 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/x_all_gather_matmul_tiling_base.cpp +508 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/op_tiling/x_all_gather_matmul_tiling_base.h +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/x_all_gather_matmul_def.cpp +79 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_host/x_all_gather_matmul_proto.cpp +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_kernel/x_all_gather_matmul.cpp +96 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_kernel/x_all_gather_matmul_base.h +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_kernel/x_all_gather_matmul_full_mesh.h +230 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_kernel/x_all_gather_matmul_tiling.h +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/all_gather_matmul/x_all_gather_matmul/op_kernel/x_all_gather_matmul_tiling_key.h +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/codebase_map.md +281 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/comm_hccl.md +201 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/matmul_fusion.md +50 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/mc2_architecture.md +142 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/op_architecture.md +167 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/pipeline_tuning.md +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/review-checklist.md +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/foundations/hccl-matmul/workflow_integration.md +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/operators/collective-comm.md +145 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/requirement-analysis/classification.md +27 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/requirement-analysis/comm-path-decision.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/requirement-analysis/grill-protocol.md +202 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/requirement-analysis/quality-checklist.md +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/requirement-analysis/template.md +208 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/shared/pipeline_tuning.md +232 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/references/shared/profiling_mc2.md +278 -0
- package/assets/ops-direct-invoke/skills/ascendc-mc2-best-practice/scripts/fetch_apace.sh +96 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/SKILL.md +121 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/evals/evals.json +138 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/advanced_tiling.md +92 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/broadcast_design.md +160 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/code/broadcast_add_kernel.cpp +118 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/code/broadcast_add_tiling.cpp +126 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/code/broadcast_common.h +344 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/code/onedim_add_kernel.cpp +97 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/code/onedim_add_tiling.cpp +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/datacopypad_design.md +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/nddma_design.md +53 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/onedim_design.md +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/broadcast/ub_broadcast_design.md +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/common/core_shrink_design.md +126 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/common/datacopy_optimization_design.md +351 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/common/tail_block_design.md +268 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/common/ub_resident_design.md +342 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/guide.md +69 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_base.template +81 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_big_dim.template +217 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_cut_one_axis.template +336 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_cut_two_axis.template +781 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_n_last.template +340 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_small_shape.template +321 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_tensor_move.template +102 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_transdata_5hd.template +480 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_transdata_5hd_021axis.template +533 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/dav3510/transpose_with_gather.template +688 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/templates/transpose_tiling_data.template +192 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_big_dim.md +220 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_cut_one_axis.md +175 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_cut_two_axis.md +177 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_fusion_design.md +398 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_gather.md +191 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_n_last.md +276 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_small_shape.md +315 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_tensor_move.md +203 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_tiling_data.md +176 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_vconv_021.md +206 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/conversion/transpose_vconv_5hd.md +179 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/elementwise/double_buffer_design.md +288 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/elementwise/vector_efficiency_design.md +203 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/constant_folding_design.md +466 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/fullload_design.md +344 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/guide.md +16 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/mte2_preload_design.md +320 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/pingpong_design.md +254 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/scale_coalescing_design.md +282 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/small_m_wide_n_pipeline_design.md +137 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/streamk_design.md +374 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/matmul/swat_design.md +423 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/mc2/guide.md +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/mc2/local_matmul_design.md +130 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/mc2/pipeline_balancing_design.md +502 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/guide.md +178 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/kernel_utils.template +75 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ar_full_load.template +429 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ar_recompute.template +485 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ar_small_r.template +348 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ara_full_load.template +574 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ara_online.template +324 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_ara_recompute.template +436 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/dav310/softmax_v2_base.template +950 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/fused_attention_online_softmax_design.md +428 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ar_full_load.md +137 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ar_recompute.md +135 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ar_small_r.md +122 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ara_full_load.md +144 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ara_online.md +185 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_ara_recompute.md +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_tiling.template +460 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/softmax_v2_tiling_data.template +278 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/state_resident_design.md +337 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/reduce/templates/usage_guide.md +253 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/scalar/coding_principles.md +476 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/scalar/conv_scalar.md +165 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/scalar/guide.md +31 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/simt/optimization-guide.md +110 -0
- package/assets/ops-direct-invoke/skills/ascendc-performance-best-practices/references/sort/radix_sort.md +354 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/SKILL.md +361 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/evals/evals.json +161 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/ascendc-dumptensor-refs/api-reference.md +92 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/ascendc-dumptensor-refs/error-patterns.md +40 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/ascendc-dumptensor.md +205 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/best-practices.md +317 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/binary-search-debug.md +275 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/case-studies.md +459 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/common-traps.md +492 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/data-comparison.md +308 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/debug-summary-template.md +198 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/diagnosis-workflow.md +180 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/printf-debug.md +296 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/simt-precision-debug.md +76 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/references/tools-reference.md +364 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/scripts/error_analysis.py +148 -0
- package/assets/ops-direct-invoke/skills/ascendc-precision-debug/scripts/gen_boundary_test.py +169 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/SKILL.md +47 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/evals/evals.json +115 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/arithmetic_and_reduce.md +228 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/compute_api_membase_vs_regbase.md +107 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/datacopy_best_practices.md +139 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/index.md +38 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/pipeline_and_buffer.md +105 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/regbase_api_reference.md +303 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/regbase_api_sync.md +173 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/api/regbase_api_whitelist.md +142 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/index.md +24 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/regbase_build_notes.md +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/regbase_performance_practices.md +52 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/regbase_programming_notes.md +120 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/tiling_review_notes.md +86 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/dev-experience/working_vs_failed_dav_3510_cases.md +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/README.md +190 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_host/adam_apply_one_assign_def.cpp +102 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_host/arch35/adam_apply_one_assign_tiling_arch35.cpp +597 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_host/arch35/adam_apply_one_assign_tiling_arch35.h +82 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_kernel/adam_apply_one_assign.cpp +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_kernel.h +361 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_struct.h +27 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_tiling_struct.h +53 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-design/adam_apply_one_assign_analysis.md +143 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/examples-adam-design/adam_apply_one_assign_design.md +2865 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/CMakeLists.txt +328 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/README.md +163 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_graph/xdivy_graph_infer.cpp +38 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_graph/xdivy_proto.h +42 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_host/arch35/xdivy_tiling_arch35.cpp +601 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_host/arch35/xdivy_tiling_arch35.h +89 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_host/xdivy_def.cpp +60 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_host/xdivy_infershape.cpp +58 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_kernel/arch35/xdivy_kernel.h +331 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_kernel/arch35/xdivy_struct.h +28 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_kernel/arch35/xdivy_tiling_struct.h +57 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/op_kernel/xdivy.cpp +42 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/aclnn/CMakeLists.txt +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/aclnn/run.sh +56 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/aclnn/test_aclnn.cpp +354 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/geir/CMakeLists.txt +69 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/geir/run.sh +58 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/geir/test_geir.cpp +407 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/tiling/CMakeLists.txt +185 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/tiling/run.sh +61 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/example/xdivy_custom_package/tests/tiling/test_tiling.cpp +386 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/patterns.md +324 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-kernel-entrance.md +95 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-overview.md +252 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-compute-flow.md +214 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-dag-buffers.md +285 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-kernel-helpers.md +209 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-kernel-template.md +611 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-tiling-preprocess.md +545 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/broadcast/references/broadcast-template-standard/broadcast-standard-tiling.md +471 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/broadcast-rules.md +126 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/cast-rules.md +414 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/datacopypad-rules.md +185 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/nddma-rules.md +255 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/sync-and-consistency.md +314 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/value-depend-process.md +40 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/common/vf-programming-rules.md +150 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/dynamic_shape/knowledge.md +382 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/dynamic_shape/patterns.md +260 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/elewise/patterns.md +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/elewise/tiling.md +215 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/example-routes.yaml +17 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/general-methodology.md +94 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/mask_predicate/patterns.md +99 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/numerical_stable/knowledge.md +233 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/numerical_stable/patterns.md +390 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/CMakeLists.txt +289 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/README.md +148 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_graph/euclidean_norm_graph_infer.cpp +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_graph/euclidean_norm_proto.h +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_host/arch35/euclidean_norm_tiling_arch35.cpp +1040 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_host/arch35/euclidean_norm_tiling_arch35.h +28 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_host/euclidean_norm_def.cpp +86 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_host/euclidean_norm_infershape.cpp +205 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm.cpp +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_base.h +965 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_empty.h +153 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_group.h +396 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_tiling_data.h +74 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_tiling_key.h +32 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/aclnn/CMakeLists.txt +59 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/aclnn/run.sh +36 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/aclnn/test_aclnn.cpp +227 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/geir/CMakeLists.txt +84 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/geir/run.sh +48 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/geir/test_geir.cpp +297 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/tiling/CMakeLists.txt +143 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/tiling/run.sh +40 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/example/euclidean_norm_package/tests/tiling/test_tiling.cpp +318 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/patterns.md +527 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-binary-sum.md +260 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-highlevel-api.md +146 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-kernel-entrance.md +96 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-survival-node.md +74 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-base/reduction-binary-base-dag-buffers.md +612 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-base/reduction-binary-base-kernel-template.md +1261 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-base/reduction-binary-base-tiling.md +531 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-group/reduction-binary-group-dag-buffers.md +209 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-group/reduction-binary-group-kernel-template.md +588 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-binary-group/reduction-binary-group-tiling.md +203 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-empty/reduction-empty-dag-buffers.md +292 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-empty/reduction-empty-kernel-template.md +360 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-empty/reduction-empty-tiling.md +234 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-template-overview.md +238 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-tiling-preprocess.md +218 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-tuple-reduce.md +72 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/references/reduction-vf-fusion-rules.md +95 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/reduction/validate_gates.py +194 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/resolve_example.py +106 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/routes.yaml +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/stateful/patterns.md +146 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/variable_output/knowledge.md +366 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/paradigms/variable_output/patterns.md +203 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/api_misuse.md +50 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/common_traps.md +67 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/index.md +25 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/precision_failures.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/precision_guide.md +85 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/regbase_vs_membase_confusions.md +52 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/pitfalls/symptom_to_cause.md +38 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/reference-ops/open_source_operator_table.md +740 -0
- package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/regbase_development_guide.md +334 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/SKILL.md +107 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/evals/evals.json +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/compute-operators.md +34 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/data-transfer.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/debug-api.md +530 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/index-calculation.md +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/intra-core-shared.md +70 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/macro-api.md +65 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/multi-core-sync.md +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/overview.md +128 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/tensorlist-handling.md +92 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/thread-stride-pattern.md +49 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/vf-call.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references/vf-declaration.md +39 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-best-practices/references//347/274/226/350/257/221/344/270/216/350/277/220/350/241/214.md +158 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-tiling-design/SKILL.md +42 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-tiling-design/evals/evals.json +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-simt-tiling-design/references/guide.md +73 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/SKILL.md +316 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/evals/evals.json +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/references/constraint-examples.md +392 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/references/dependency-yaml-spec.md +625 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/references/empty-tensor-guide.md +236 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/references/parameter-definition.md +491 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/references/standards.md +11 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_empty_tensor_cases_derive.py +815 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_factor_values.py +1287 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_implicit_constraints.py +511 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_solver_config.py +618 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_test_cases.py +2800 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/generate_test_factors.py +762 -0
- package/assets/ops-direct-invoke/skills/ascendc-st-design/scripts/utils.py +1931 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/SKILL.md +91 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/evals/evals.json +72 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/broadcast/dynamic-ub-broadcast.md +87 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/broadcast/nddma-broadcast.md +199 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/broadcast/onedim.md +152 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/broadcast/patterns.md +165 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/broadcast/ub-broadcast.md +331 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/conversion/patterns.md +261 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/elewise/patterns.md +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/elewise/tiling.md +215 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/design/_governance.md +132 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/design/execution.md +469 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/design/resources.md +339 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/design/shape.md +271 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/foundation/composition.md +138 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/foundation/extension_points.md +173 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/foundation/fundamentals.md +245 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/foundation/roofline.md +227 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/implementation_ref.md +201 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/patterns.md +239 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/specialization/feature_flags.md +216 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/specialization/sparse_design.md +198 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/subfamilies/base_design.md +395 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/flashattention/subfamilies/quantization_design.md +233 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/gather.md +273 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/index_add.md +434 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/index_select.md +185 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/patterns.md +199 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/repeat_interleave.md +251 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/index/scatter.md +341 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/matmul/ascendc-api-gmm-tiling.md +313 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/matmul/ascendc-api-matmul-tiling.md +408 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/matmul/patterns.md +32 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/alg-dichotomy.md +56 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/alg-group-reduce.md +101 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/alg-welford.md +74 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/algorithms.md +35 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/ar-colsplit.md +149 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/ar-fullload.md +151 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/ara-fullload.md +228 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/ara-rowsplit.md +219 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/multi-axis-transform.md +83 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/multi-output-buffer.md +115 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/patterns.md +212 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/tiling-fields.md +94 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/reduction/with-index.md +247 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/scatter/patterns.md +827 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/sort/alg-two-level-mrgsort.md +681 -0
- package/assets/ops-direct-invoke/skills/ascendc-tiling-design/references/sort/patterns.md +249 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/SKILL.md +109 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/coverage_files_question.json +24 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_cmake.txt +46 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_inferdtype.cpp +139 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_inferdtype.csv +5 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_infershape.cpp +122 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_infershape.csv +5 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_param.h +319 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_tiling.cpp +154 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/csv-refactor/template_tiling.csv +5 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/question.json +100 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/template_report.md +112 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/assets/todo.json +24 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/evals/evals.json +198 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/coverage-guide/coverage-extraction.md +78 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/coverage-guide/test-implementation.md +136 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/custom/custom.md +5 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ops-transformer/csv-format-spec.md +263 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ops-transformer/csv-refactor-workflow.md +494 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ops-transformer/legacy-code-extraction.md +325 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ops-transformer/ops-transformer.md +75 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ut-guide/build_ut.md +65 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ut-guide/op-api-ut-guide.md +237 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ut-guide/op-host-ut-guide.md +240 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ut-guide/op-kernel-ut-guide.md +198 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/workflow/step1.md +43 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/workflow/step2.md +26 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/workflow/step3.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/workflow/step4.md +130 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/workflow/step5.md +5 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/scripts/extract_dtype_combinations.py +200 -0
- package/assets/ops-direct-invoke/skills/ascendc-ut-develop/scripts/generate_csv_template.py +270 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/SKILL.md +125 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/evals/evals.json +198 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/S1-input-collection.md +125 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/00-execution-order.md +58 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/01-variable-semantics.md +250 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/02-generate-mapper.md +218 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/S5_case_json_schema.md +373 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/S5_mapped_case_schema.md +131 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/case-mapper/s5_case_mapper_template.py +411 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/design-verifier/00-execution-order.md +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/00-execution-order.md +291 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/01-code-analyzer.md +121 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/02-interface-analyzer.md +417 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/03-param-derivation.md +45 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/04-test-design-template.md +231 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/source-analysis/05-network-search.md +154 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-a/00-overview.md +66 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-a/01-step1-tiling.md +114 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-a/02-step2-trace.md +81 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-a/03-step3-kernel.md +51 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-a/04-path-config-schema.md +342 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/00-execution-order.md +26 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/01-step1-merge.md +54 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/02-step2-group.md +44 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/03-step3-derive.md +416 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/04-step4-output.md +122 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/task-d/05-step5-gen-cases.md +156 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/ttk-converter/00-execution-order.md +108 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/ttk-converter/01-csv-common.md +33 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/ttk-converter/02-kernel-fields.md +64 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/ttk-converter/03-kernel-mode.md +142 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/references/workflow.md +193 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/_scout_common.py +53 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/assemble_dim_spec.py +232 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/build_param_def.py +134 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/build_path_list.py +520 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/format_json.py +135 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/gen_cases_template.py +196 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/generate_test_design.py +555 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/normalize_op_name.py +157 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/pick_dims.py +173 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/run_ttk_kernel_module.py +750 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/s2p0_verify.py +302 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/s3_task_d_gate.py +365 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/s5_append_empty.py +180 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/s5_check_mapper_outputs.py +292 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/s5_expand_high.py +208 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/scout_k.py +723 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/scout_t.py +983 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/ttk_generate_kernel_csv.py +346 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/ttk_precheck_env.py +248 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/ttk_validate_csv.py +315 -0
- package/assets/ops-direct-invoke/skills/ascendc-whitebox-design/scripts/update_path_list.py +143 -0
- package/assets/ops-direct-invoke/skills/cann-env-setup/SKILL.md +177 -0
- package/assets/ops-direct-invoke/skills/cann-env-setup/evals/evals.json +144 -0
- package/assets/ops-direct-invoke/skills/cann-env-setup/references/env_dependence.md +77 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/SKILL.md +51 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/overall-implementation.md +226 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/aclnn-adaptation.md +34 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/aicpu-load-balance-metadata.md +116 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/channel-buffer.md +36 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/compilation-launch.md +144 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/host-validation.md +34 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/layout-tail.md +53 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/synchronization.md +80 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/torch-registration.md +66 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/vf.md +144 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/patterns/workspace.md +67 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-best-practice/references/production-code.md +120 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/SKILL.md +25 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/core/methodology.md +104 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/references/readability-rules.md +280 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/steps/clause-review.md +59 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/steps/report-write.md +60 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/workflows/file-review.md +49 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-code-reviewer/workflows/pr-review.md +58 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/SKILL.md +97 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/references/checks-and-troubleshooting.md +96 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/references/env-selected-soc-schema.md +43 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/references/environment-schema.md +112 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/scripts/check_env.sh +34 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/scripts/env_check.py +946 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/scripts/probe_compile_stage.py +99 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/scripts/probe_kernel.py +33 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-env-check/tests/test_env_report.py +90 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/SKILL.md +60 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/references/pipeline-design.md +84 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/references/stage-cost-model.md +77 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/DESIGN-BRANCH.md.templ +119 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/INDEX.md.templ +34 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/API.md.templ +67 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/DevView.md.templ +67 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/HostLaunch.md.templ +119 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/InferShapeDtype.md.templ +56 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/Interface.md.templ +61 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/Kernel.md.templ +183 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/LaunchData.md.templ +126 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/Overview.md.templ +57 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/TemplatePartition.md.templ +69 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-op-design/templates/design/Validation.md.templ +80 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/SKILL.md +64 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/assets/case_contract.py +269 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/assets/test.py +571 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/references/cases.md +29 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/references/evidence.md +67 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/references/golden.md +12 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/build_blackbox_cases.py +536 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/build_tdd_cases.py +498 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/build_whitebox_cases.py +183 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/casebook.py +71 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/generator_support.py +95 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/prepare_golden.py +390 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/scripts/recipe_validation.py +144 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/tests/test_evidence_chain.py +339 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/tests/test_skills.py +547 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-test-generator/tests/test_synthesize_dtype.py +85 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-unit-planner/SKILL.md +35 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-unit-planner/references/decomposition-checks.md +46 -0
- package/assets/ops-direct-invoke/skills/cannbot-dsl-unit-planner/references/unit-format.md +77 -0
- package/assets/ops-direct-invoke/skills/npu-arch/SKILL.md +87 -0
- package/assets/ops-direct-invoke/skills/npu-arch/evals/evals.json +356 -0
- package/assets/ops-direct-invoke/skills/npu-arch/references/npu-arch-guide-triton.md +312 -0
- package/assets/ops-direct-invoke/skills/npu-arch/references/npu-arch-guide.md +454 -0
- package/assets/ops-direct-invoke/skills/npu-arch/references/npu-hardware-params.md +248 -0
- package/assets/ops-direct-invoke/skills/npu-arch/references/simt-arch-guide.md +303 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/SKILL.md +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/index.md +32 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/arithmetic_and_reduce.md +96 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/datacopy_best_practices.md +87 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/index.md +37 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/pipeline_and_buffer.md +87 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/precision_and_runtime.md +98 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/regbase_api_reference.md +88 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/regbase_api_sync.md +112 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/api/regbase_api_whitelist.md +121 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/complexity_and_route_selection.md +113 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/index.md +55 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/process_memory_cards.md +54 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/regbase_build_notes.md +87 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/regbase_kernel_case_notes.md +103 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/regbase_performance_practices.md +92 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/regbase_programming_notes.md +108 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/requirements_analysis_patterns.md +97 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/small_channel_transpose_notes.md +78 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/tiling_review_notes.md +87 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/dev-experience/working_vs_failed_950_cases.md +85 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/index.md +30 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/broadcast_patterns.md +132 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/index.md +48 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/inplace_operator_patterns.md +182 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/kernel_design_patterns.md +117 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/reduction_patterns.md +152 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/regbase_buffer_partitioning.md +122 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/regbase_kernel_dataflow_patterns.md +164 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/regbase_operator_patterns.md +138 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/regbase_sync_patterns.md +113 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/tiling_input_validation.md +267 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/patterns/tiling_patterns.md +124 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/api_misuse.md +161 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/common_traps.md +199 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/index.md +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/precision_failures.md +176 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/precision_guide.md +190 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/regbase_vs_membase_confusions.md +89 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/pitfalls/symptom_to_cause.md +58 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/knowledge/regbase_development_guide.md +290 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/design_sample/flash_attn-arch35-best-practices.md +80 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/design_sample/quant_flash_attn-arch35-best-practices.md +67 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/design_sample/sparse_flash_attention-arch35-best-practices.md +66 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/design_sample/sparse_flash_mla-arch35-best-practices.md +69 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/patterns.md +65 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/advanced.md +37 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/buffer.md +241 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/compute-flow.md +111 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/load-balancing.md +42 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/pipeline-example.md +173 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/pipeline.md +60 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/pre-knowledge.md +197 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/roofline.md +220 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/selection.md +32 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/stage-cost.md +63 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/trait-mla.md +36 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/trait-quantization.md +51 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/FA/references/trait-sparse.md +35 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/README.md +20 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/capability-mapping.md +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/dag-buffers.md +285 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/design-spec.md +24 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_host/adam_apply_one_assign_def.cpp +104 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_host/arch35/adam_apply_one_assign_tiling_arch35.cpp +572 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_host/arch35/adam_apply_one_assign_tiling_arch35.h +86 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_kernel/adam_apply_one_assign.cpp +46 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_kernel.h +343 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_struct.h +29 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-code/op_kernel/arch35/adam_apply_one_assign_tiling_struct.h +53 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-design/adam_apply_one_assign_analysis.md +143 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/examples-adam-design/adam_apply_one_assign_design.md +2840 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/CMakeLists.txt +328 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/README.md +162 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_graph/xdivy_graph_infer.cpp +40 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_graph/xdivy_proto.h +44 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_host/arch35/xdivy_tiling_arch35.cpp +582 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_host/arch35/xdivy_tiling_arch35.h +94 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_host/xdivy_def.cpp +62 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_host/xdivy_infershape.cpp +60 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_kernel/arch35/xdivy_kernel.h +337 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_kernel/arch35/xdivy_struct.h +30 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_kernel/arch35/xdivy_tiling_struct.h +57 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/op_kernel/xdivy.cpp +44 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/aclnn/CMakeLists.txt +67 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/aclnn/run.sh +57 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/aclnn/test_aclnn.cpp +353 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/geir/CMakeLists.txt +69 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/geir/run.sh +59 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/geir/test_geir.cpp +409 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/tiling/CMakeLists.txt +185 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/tiling/run.sh +62 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/example/xdivy_custom_package/tests/tiling/test_tiling.cpp +355 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/index.md +48 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/kernel-entrance.md +95 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/kernel-helpers.md +208 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/kernel-template.md +617 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/overview.md +250 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/pitfalls.md +50 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/tiling-preprocess.md +487 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/ascendc/tiling.md +374 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/dsl/capability-mapping.md +50 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/dsl/implementation-guide.md +52 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/adapters/dsl/index.md +33 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/capability-contract.md +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/compute-flow.md +216 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/index.md +23 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/liveness-fusion.md +200 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/scene.md +68 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/shape-preprocessing.md +209 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/splitting.md +158 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/sync-model.md +79 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/generic/verification.md +131 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/broadcast/patterns.md +62 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/broadcast-rules.md +126 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/cast-rules.md +414 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/datacopypad-rules.md +186 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/nddma-rules.md +257 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/sync-and-consistency.md +315 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/value-depend-process.md +40 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/common/vf-programming-rules.md +150 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/example-routes.yaml +31 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/general-methodology-dsl.md +11 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/general-methodology.md +94 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/binary-tree.md +260 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/capability-mapping.md +49 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/design-spec.md +26 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/CMakeLists.txt +289 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/README.md +147 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_graph/euclidean_norm_graph_infer.cpp +39 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_graph/euclidean_norm_proto.h +54 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_host/arch35/euclidean_norm_tiling_arch35.cpp +1027 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_host/arch35/euclidean_norm_tiling_arch35.h +28 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_host/euclidean_norm_def.cpp +85 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_host/euclidean_norm_infershape.cpp +211 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_apt.cpp +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_base.h +911 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_empty.h +141 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_group.h +381 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_tiling_data.h +65 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/op_kernel/arch35/euclidean_norm_tiling_key.h +30 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/aclnn/CMakeLists.txt +59 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/aclnn/run.sh +37 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/aclnn/test_aclnn.cpp +223 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/geir/CMakeLists.txt +84 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/geir/run.sh +49 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/geir/test_geir.cpp +296 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/tiling/CMakeLists.txt +143 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/tiling/run.sh +41 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/example/euclidean_norm_package/tests/tiling/test_tiling.cpp +320 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/highlevel-api.md +146 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/index.md +48 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/kernel-entrance.md +97 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/overview.md +237 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/survival-node.md +74 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-base/dag-buffers.md +612 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-base/kernel-template.md +1213 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-base/tiling.md +511 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-group/dag-buffers.md +209 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-group/kernel-template.md +577 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/binary-group/tiling.md +199 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/empty/dag-buffers.md +292 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/empty/kernel-template.md +349 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/templates/empty/tiling.md +230 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/tiling-preprocess.md +218 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/tuple-reduce.md +72 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/adapters/ascendc/vf-fusion-rules.md +95 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/axis-preprocessing.md +109 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/capability-contract.md +48 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/index.md +23 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/liveness-fusion.md +111 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/scene.md +92 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/splitting.md +230 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/sync-model.md +86 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/tree-reduction.md +160 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/generic/tuple-reduce.md +53 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/patterns.md +62 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/reduction/validate_gates.py +196 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/resolve_example.py +170 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/paradigms/routes.yaml +39 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/references/reference-ops/open_source_operator_table.md +715 -0
- package/assets/ops-direct-invoke/skills/ops-design-paradigms/tests/test_language_route.py +142 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/SKILL.md +249 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/assemble_cannbotdsl_design.py +156 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/assemble_workflow.py +571 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/build_implementation_subgraph.py +320 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/generate_workflow_guide.py +94 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/run_workflow.py +121 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/scripts/validate_unit_contracts.py +228 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//344/273/243/347/240/201/344/277/256/345/244/215.yaml +39 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//346/212/200/346/234/257/347/251/277/345/210/272.yaml +40 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//346/226/207/346/241/243/345/207/206/345/244/207.yaml +27 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//346/265/213/350/257/225/345/267/245/347/250/213/345/274/200/345/217/221.yaml +33 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//347/231/275/347/233/222/346/265/213/350/257/225/350/256/276/350/256/241.yaml +34 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//347/237/245/350/257/206/346/220/234/351/233/206.yaml +36 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//347/256/227/345/255/220/345/274/200/345/217/221.yaml +47 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/ascendc//351/273/221/347/233/222/346/265/213/350/257/225/350/256/276/350/256/241.yaml +27 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//344/272/244/344/273/230.yaml +15 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//344/273/243/347/240/201/346/243/200/350/247/206.yaml +18 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/210/233/345/273/272/345/267/245/344/275/234/345/214/272.yaml +17 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/215/225/345/205/203/344/273/273/345/212/241/346/213/206/345/210/206.yaml +22 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/215/225/345/205/203/345/256/236/347/216/260.yaml +27 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/215/225/345/205/203/351/233/206/346/210/220/345/244/215/351/252/214.yaml +19 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/256/236/347/216/260/344/277/256/345/244/215.yaml +21 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/256/236/347/216/260/346/243/200/350/247/206.yaml +20 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//345/256/236/347/216/260/351/227/250/347/246/201.yaml +17 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//346/200/247/350/203/275/344/274/230/345/214/226.yaml +22 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//346/226/207/346/241/243/347/224/237/346/210/220.yaml +15 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//346/226/271/346/241/210/350/256/276/350/256/241.yaml +32 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//346/265/213/350/257/225/347/224/237/346/210/220.yaml +29 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//347/216/257/345/242/203/345/260/261/347/273/252.yaml +25 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//347/262/276/345/272/246/351/252/214/350/257/201.yaml +25 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//350/247/204/346/240/274/347/224/237/346/210/220.yaml +18 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//351/234/200/346/261/202/345/210/206/346/236/220.yaml +16 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/cannbot-dsl//351/273/221/347/233/222/347/224/250/344/276/213/350/256/276/350/256/241.yaml +19 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/tasks/common/.gitkeep +0 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//344/273/243/347/240/201/346/243/200/350/247/206/346/212/245/345/221/212.md +70 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//345/212/237/350/203/275/351/252/214/346/224/266/346/212/245/345/221/212.md +80 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//346/265/213/350/257/225/346/211/247/350/241/214/350/256/260/345/275/225.md +73 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//347/231/275/347/233/222/346/265/213/350/257/225/350/256/276/350/256/241.md +37 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//347/237/245/350/257/206/346/220/234/351/233/206.md +25 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//351/252/214/346/224/266/346/212/245/345/221/212.md +39 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/templates//351/273/221/347/233/222/346/265/213/350/257/225/350/256/276/350/256/241.md +130 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/workflows/ascendc/basic.yaml +57 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/workflows/ascendc/feasibility.yaml +13 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/workflows/cannbot-dsl/op-dev.graph-preview.yaml +18 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke/workflows/cannbot-dsl/op-dev.yaml +142 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/SKILL.md +115 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/agents-entry.md +23 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/documentation.md +52 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/ops-direct-invoke.md +67 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/repo-skills.md +85 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/subagents.md +43 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/task-authoring.md +83 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/reference/tests.md +33 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/dsl_generation/test.py +499 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/support.py +122 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/test_all.py +59 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/variables/test.py +100 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/workflow_execution/test.py +134 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/workflow_generation/test.py +457 -0
- package/assets/ops-direct-invoke/skills/ops-direct-invoke-maintain/test/ut/workflow_guide/test.py +91 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/SKILL.md +22 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/evals/evals.json +120 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/benchmark_construction.md +77 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/float_compute.md +126 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/integer_compute.md +105 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/non_compute.md +81 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/quantization.md +49 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/random_generation.md +198 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/references/test_case_generation.md +61 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/scripts/bitwise_match.py +103 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/scripts/integer_compute_check.py +157 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/scripts/mixed_tolerance_check.py +178 -0
- package/assets/ops-direct-invoke/skills/ops-precision-standard/scripts/random_distribution_check.py +247 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/SKILL.md +185 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/evals/evals.json +196 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/SKILL.md +259 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/VERSION +10 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/evals/evals.json +33 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/references/collection-workflow.md +91 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/references/data-contract.md +63 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/references/environment-and-failure.md +82 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/references/validation-and-performance.md +63 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/references/visualization-contract.md +86 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/collect.py +2209 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/collect_discovery.py +987 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/collect_validate.py +607 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/environment_context.py +723 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/estimate_runtime.py +131 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/run_pipeline.py +339 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/runtime_guard.py +589 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/self_check.py +465 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize.py +1004 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize_common.py +258 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize_parsers.py +1555 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize_render.py +286 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize_resolvers.py +1224 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/scripts/visualize_source.py +1288 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/msopprof-visualization/templates/report_template.html +253 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/references/csv_fields_reference.md +260 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/references/msprof-guide.md +252 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/references/msprof-op-guide.md +241 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/references/optimization_quickref.md +265 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/scripts/msprof_perf_summary.py +2469 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/scripts/msprof_profile_run.sh +400 -0
- package/assets/ops-direct-invoke/skills/ops-profiling/scripts/perf_summary.py +448 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/SKILL.md +233 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/evals/evals.json +128 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/pc-source-mapping.md +76 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/performance-issues-aic.md +557 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/performance-issues-aiv.md +674 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/performance-issues-general.md +199 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/performance-issues-template.md +143 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/performance-metrics-reference.md +350 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/pipeline-bubble-analysis.md +269 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/simulator-advanced.md +174 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/references/troubleshooting.md +105 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/scripts/map_pc_to_source.py +741 -0
- package/assets/ops-direct-invoke/skills/ops-simulator/scripts/trace_bubble_analyzer.py +971 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/SKILL.md +350 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/evals/evals.json +114 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/README.md +71 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/add/spec.yaml +161 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/complex/spec.yaml +154 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/cumsum/spec.yaml +176 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/dropout/spec.yaml +170 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/fused_quant_matmul/spec.yaml +331 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/matmul/spec.yaml +201 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/nonzero/spec.yaml +174 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/reduce_sum/spec.yaml +151 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/reduce_sum_all/spec.yaml +135 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/reduce_sum_fixed/spec.yaml +143 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/softmax/spec.yaml +231 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/examples/split/spec.yaml +160 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/references/accumulation_order.md +63 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/references/error-codes.md +199 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/references/spec-cheatsheet.md +358 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/references/stage-rules.md +476 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/references/usage-scenarios.md +265 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/anti_pattern_registry.yaml +51 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/boundary_min_cases.yaml +215 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/category_enum.yaml +37 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/category_paradigm_map.yaml +145 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/chip_registry.yaml +82 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/error_code_enum.yaml +25 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/framework_oracle_registry.yaml +22 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/invariant_kind_registry.yaml +130 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/machine_check_kind_registry.yaml +84 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/paradigm_enum.yaml +48 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/primitive_whitelist.yaml +67 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/synthesize_pattern_registry.yaml +58 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/registries/tolerance_defaults.yaml +53 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/schemas/op-spec.json +762 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/check_registry_schema_sync.py +167 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/dump_rule_ids.py +168 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/__init__.py +8 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/_ast_sandbox.py +142 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/broadcast.py +270 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/dtype_eval.py +186 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/extreme_inputs_check.py +355 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/formula_eval.py +431 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/formula_oracle_equiv.py +603 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/invariant_exec.py +569 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/oracle_check.py +465 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/parser.py +71 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/promote.py +254 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/shape_eval.py +368 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/stages.py +995 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/evaluators/types.py +130 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/generate_spec.py +1571 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/scripts/validate_spec.py +1848 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/templates/spec.yaml.tmpl +100 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/conftest.py +21 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_attr_resolve.py +72 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_broadcast.py +85 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_doc_drift.py +45 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_dtype_eval.py +163 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_examples.py +54 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_extreme_inputs_check.py +181 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_formula_eval.py +154 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_formula_oracle_equiv.py +281 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_generate_spec.py +98 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_invariant_exec.py +235 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_main_pipeline.py +419 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_oracle.py +255 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_promote.py +127 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_registry_schema_sync.py +32 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_sandbox_escape.py +90 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_shape_eval.py +220 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_shape_literal.py +82 -0
- package/assets/ops-direct-invoke/skills/ops-spec-gen/tests/test_stage2_constraints.py +213 -0
- package/assets/ops-direct-invoke/skills/repo-build-guide/SKILL.md +10 -0
- package/assets/ops-direct-invoke/skills/repo-build-guide/references/build-guide.md +42 -0
- package/assets/ops-direct-invoke/skills/repo-coding-rules/SKILL.md +10 -0
- package/assets/ops-direct-invoke/skills/repo-coding-rules/references/red-lines.md +42 -0
- package/assets/ops-direct-invoke/skills/repo-env-check/SKILL.md +44 -0
- package/assets/ops-direct-invoke/skills/repo-env-check/references//347/216/257/345/242/203/344/277/241/346/201/257.md +42 -0
- package/assets/ops-direct-invoke/skills/repo-knowledge/SKILL.md +12 -0
- package/assets/ops-direct-invoke/skills/repo-knowledge/references/evaluation-contract.md +48 -0
- package/assets/ops-direct-invoke/skills/repo-op-templates/SKILL.md +15 -0
- package/assets/ops-direct-invoke/skills/repo-op-templates/references/operator-template.md +27 -0
- package/assets/ops-direct-invoke/skills/repo-op-templates/references/project-skeleton.md +41 -0
- package/assets/ops-direct-invoke/skills/repo-requirement/SKILL.md +39 -0
- package/assets/ops-direct-invoke/skills/repo-requirement/references/requirement-checklist.md +70 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/SKILL.md +17 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/references/blackbox-design.md +30 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/references/cannbench-integration.md +75 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/references/precision-and-perf.md +36 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/references/test-framework.md +75 -0
- package/assets/ops-direct-invoke/skills/repo-test-develop/references/whitebox-design.md +19 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/README.md +145 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/SKILL.md +84 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/approve_task.py +98 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/get_task.py +1133 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/init_status.py +131 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/link_agent_config.py +89 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/orchestrator.py +738 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/poll_workflow.py +392 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/task_graph.py +144 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/update_status.py +149 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/verdict_common.py +103 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/verdict_fail.py +22 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/scripts/verdict_pass.py +21 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/templates/basic_workflow.yaml +113 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_agent_exit.py +91 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_approval.py +490 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_link_agent_config.py +202 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_nested_subgraphs.py +337 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_orchestrator.py +1383 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_poll_workflow.py +371 -0
- package/assets/ops-direct-invoke/skills/workflow-orchestrator/tests/test_verdict.py +148 -0
- package/cordis.patch.yml +3 -0
- package/icon.svg +1 -0
- package/lib/client.js +384 -0
- package/lib/client.js.map +1 -0
- package/lib/index.js +20 -0
- package/locale/en.json +1 -0
- package/locale/zh.json +1 -0
- package/package.json +63 -3
- package/upstream.json +10 -0
|
@@ -0,0 +1,325 @@
|
|
|
1
|
+
# 从旧代码提取测试用例参数
|
|
2
|
+
|
|
3
|
+
本文档提供从硬编码测试代码提取参数到 CSV 格式的通用方法。
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## 旧代码结构识别
|
|
8
|
+
|
|
9
|
+
旧格式的测试代码通常包含以下结构:
|
|
10
|
+
|
|
11
|
+
```cpp
|
|
12
|
+
TEST_F(OpTest, test_case_0) {
|
|
13
|
+
gert::InfershapeContextPara para(
|
|
14
|
+
"OpName",
|
|
15
|
+
{
|
|
16
|
+
// 输入 Tensor 列表
|
|
17
|
+
{{{4, 13, 16, 512}, {4, 13, 16, 512}}, ge::DT_FLOAT16, ge::FORMAT_ND},
|
|
18
|
+
{{{}, {}}, ge::DT_FLOAT16, ge::FORMAT_ND}, // 空 Tensor
|
|
19
|
+
// ... 更多输入
|
|
20
|
+
},
|
|
21
|
+
{
|
|
22
|
+
// 输出 Tensor 列表
|
|
23
|
+
{{{4, 13, 16, 512}, {4, 13, 16, 512}}, ge::DT_FLOAT16, ge::FORMAT_ND},
|
|
24
|
+
// ... 更多输出
|
|
25
|
+
},
|
|
26
|
+
{
|
|
27
|
+
// 属性列表
|
|
28
|
+
{"num_heads", Ops::Transformer::AnyValue::CreateFrom<int64_t>(16)},
|
|
29
|
+
{"scale", Ops::Transformer::AnyValue::CreateFrom<float>(0.04166f)},
|
|
30
|
+
{"input_layout", Ops::Transformer::AnyValue::CreateFrom<std::string>("BSND")},
|
|
31
|
+
{"softmax_lse_flag", Ops::Transformer::AnyValue::CreateFrom<bool>(false)},
|
|
32
|
+
// ... 更多属性
|
|
33
|
+
}
|
|
34
|
+
);
|
|
35
|
+
|
|
36
|
+
std::vector<std::vector<int64_t>> expectOutputShape = {
|
|
37
|
+
{4, 13, 16, 512},
|
|
38
|
+
{0},
|
|
39
|
+
};
|
|
40
|
+
ExecuteTestCase(para, ge::GRAPH_SUCCESS, expectOutputShape);
|
|
41
|
+
}
|
|
42
|
+
```
|
|
43
|
+
|
|
44
|
+
---
|
|
45
|
+
|
|
46
|
+
## 提取参数通用方法
|
|
47
|
+
|
|
48
|
+
### 方法一:Tensor 参数提取
|
|
49
|
+
|
|
50
|
+
#### 旧代码格式
|
|
51
|
+
|
|
52
|
+
```cpp
|
|
53
|
+
{{{shape1, shape2, ...}, {shape1, shape2, ...}}, dtype, format}
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
#### 提取规则
|
|
57
|
+
|
|
58
|
+
| 旧代码组成部分 | 提取方法 | CSV 格式 |
|
|
59
|
+
|--------------|---------|---------|
|
|
60
|
+
| **shape** | 从第一个花括号提取 | 空格连接:`4 13 16 512` |
|
|
61
|
+
| **dtype** | 从 `ge::DT_XXX` 提取 | 去掉前缀:`DT_FLOAT16` → `FLOAT16` |
|
|
62
|
+
| **format** | 从 `ge::FORMAT_XXX` 提取 | 去掉前缀:`FORMAT_ND` → `ND` |
|
|
63
|
+
|
|
64
|
+
#### 不同状态的 Tensor 处理
|
|
65
|
+
|
|
66
|
+
| Tensor 状态 | 旧代码特征 | CSV 处理规则 |
|
|
67
|
+
|-----------|-----------|-------------|
|
|
68
|
+
| **正常 Tensor** | `{{{4, 13, ...}, {...}}, dtype, format}` | 正常填写 shape/dtype/format |
|
|
69
|
+
| **空 Tensor(Tiling)** | `{{{}, {}}, dtype, format}` | 三列都留空 |
|
|
70
|
+
| **空 Tensor(InferShape)** | `{{{}, {}}, dtype, format}` | shape 填 `0`,dtype/format 正常填写 |
|
|
71
|
+
| **空 Tensor(InferDataType)** | 不存在 shape 信息 | dtype 列留空 |
|
|
72
|
+
|
|
73
|
+
#### 示例对比
|
|
74
|
+
|
|
75
|
+
| 旧代码 | CSV(Tiling) | CSV(InferShape) |
|
|
76
|
+
|--------|--------------|------------------|
|
|
77
|
+
| `{{{4, 13, 16, 512}, {...}}, ge::DT_FLOAT16, ge::FORMAT_ND}` | `4 13 16 512,FLOAT16,ND` | `4 13 16 512,FLOAT16,ND` |
|
|
78
|
+
| `{{{}, {}}, ge::DT_FLOAT16, ge::FORMAT_ND}` | `,,` | `0,FLOAT16,ND` |
|
|
79
|
+
| `{{{32, 4096, 128}, {...}}, ge::DT_INT8, ge::FORMAT_ND}` | `32 4096 128,INT8,ND` | `32 4096 128,INT8,ND` |
|
|
80
|
+
|
|
81
|
+
---
|
|
82
|
+
|
|
83
|
+
### 方法二:属性值提取
|
|
84
|
+
|
|
85
|
+
#### 旧代码格式
|
|
86
|
+
|
|
87
|
+
```cpp
|
|
88
|
+
{"attr_name", Ops::Transformer::AnyValue::CreateFrom<type>(value)}
|
|
89
|
+
```
|
|
90
|
+
|
|
91
|
+
#### 提取规则
|
|
92
|
+
|
|
93
|
+
| 属性类型 | 旧代码特征 | CSV 写法 |
|
|
94
|
+
|---------|-----------|---------|
|
|
95
|
+
| **Int** | `CreateFrom<int64_t>(16)` | `16` |
|
|
96
|
+
| **Float** | `CreateFrom<float>(0.04166f)` | `0.04166`(去掉 f 后缀) |
|
|
97
|
+
| **String** | `CreateFrom<std::string>("BSND")` | `BSND`(去掉引号) |
|
|
98
|
+
| **Bool** | `CreateFrom<bool>(false)` | `0` 或 `1` |
|
|
99
|
+
|
|
100
|
+
#### 类型转换表
|
|
101
|
+
|
|
102
|
+
| C++ 类型 | CSV 写法 | 读取方法 |
|
|
103
|
+
|---------|---------|---------|
|
|
104
|
+
| `int64_t` | 数字 | `std::stoll` |
|
|
105
|
+
| `float` | 数字(可带小数) | `std::stof` |
|
|
106
|
+
| `std::string` | 字符串 | `ReadMap` |
|
|
107
|
+
| `bool` | `0` 或 `1` | `std::stoi` |
|
|
108
|
+
|
|
109
|
+
---
|
|
110
|
+
|
|
111
|
+
### 方法三:期望结果提取
|
|
112
|
+
|
|
113
|
+
#### 期望执行结果
|
|
114
|
+
|
|
115
|
+
| 旧代码 | CSV 写法 |
|
|
116
|
+
|-------|---------|
|
|
117
|
+
| `ge::GRAPH_SUCCESS` | `SUCCESS` |
|
|
118
|
+
| `ge::GRAPH_FAILED` | `FAILED` |
|
|
119
|
+
|
|
120
|
+
#### 期望输出 Shape(InferShape)
|
|
121
|
+
|
|
122
|
+
旧代码中通过 `expectOutputShape` 定义:
|
|
123
|
+
|
|
124
|
+
```cpp
|
|
125
|
+
std::vector<std::vector<int64_t>> expectOutputShape = {
|
|
126
|
+
{4, 13, 16, 512},
|
|
127
|
+
{0}, // 空 Tensor
|
|
128
|
+
};
|
|
129
|
+
```
|
|
130
|
+
|
|
131
|
+
转换为 CSV:
|
|
132
|
+
- 第一个输出:`4 13 16 512`(写入 `output0_shape` 列)
|
|
133
|
+
- 第二个输出:`0`(写入 `output1_shape` 列)
|
|
134
|
+
|
|
135
|
+
#### 期望 Tiling Key/DataHash(Tiling)
|
|
136
|
+
|
|
137
|
+
旧代码可能不显式定义,需要在重构后:
|
|
138
|
+
1. 先运行一次测试获取实际值
|
|
139
|
+
2. 或从已有测试结果中提取
|
|
140
|
+
|
|
141
|
+
---
|
|
142
|
+
|
|
143
|
+
## 完整转换示例
|
|
144
|
+
|
|
145
|
+
### 旧代码(InferShape)
|
|
146
|
+
|
|
147
|
+
```cpp
|
|
148
|
+
TEST_F(FusedInferAttentionScoreProto, infershape_0)
|
|
149
|
+
{
|
|
150
|
+
gert::InfershapeContextPara para(
|
|
151
|
+
"FusedInferAttentionScore",
|
|
152
|
+
{
|
|
153
|
+
{{{4, 13, 16, 512}, {4, 13, 16, 512}}, ge::DT_FLOAT16, ge::FORMAT_ND}, // query
|
|
154
|
+
{{{4, 10347, 1, 512}, {4, 10347, 1, 512}}, ge::DT_FLOAT16, ge::FORMAT_ND}, // key
|
|
155
|
+
{{{}, {}}, ge::DT_FLOAT16, ge::FORMAT_ND}, // pse_shift (空)
|
|
156
|
+
{{{4, 13, 10347}, {4, 13, 10347}}, ge::DT_INT8, ge::FORMAT_ND}, // atten_mask
|
|
157
|
+
},
|
|
158
|
+
{
|
|
159
|
+
{{{4, 13, 16, 512}, {4, 13, 16, 512}}, ge::DT_FLOAT16, ge::FORMAT_ND}, // attention_out
|
|
160
|
+
{{{}, {}}, ge::DT_FLOAT, ge::FORMAT_ND}, // softmax_lse (空)
|
|
161
|
+
},
|
|
162
|
+
{
|
|
163
|
+
{"num_heads", Ops::Transformer::AnyValue::CreateFrom<int64_t>(16)},
|
|
164
|
+
{"scale", Ops::Transformer::AnyValue::CreateFrom<float>(0.04166666666666666f)},
|
|
165
|
+
{"input_layout", Ops::Transformer::AnyValue::CreateFrom<std::string>("BSND")},
|
|
166
|
+
{"softmax_lse_flag", Ops::Transformer::AnyValue::CreateFrom<bool>(false)},
|
|
167
|
+
});
|
|
168
|
+
|
|
169
|
+
std::vector<std::vector<int64_t>> expectOutputShape = {
|
|
170
|
+
{4, 13, 16, 512},
|
|
171
|
+
{0},
|
|
172
|
+
};
|
|
173
|
+
ExecuteTestCase(para, ge::GRAPH_SUCCESS, expectOutputShape);
|
|
174
|
+
}
|
|
175
|
+
```
|
|
176
|
+
|
|
177
|
+
### 转换为 CSV
|
|
178
|
+
|
|
179
|
+
```csv
|
|
180
|
+
case_name,expectResult,query_shape,query_dtype,query_format,key_shape,key_dtype,key_format,pse_shift_shape,pse_shift_dtype,pse_shift_format,atten_mask_shape,atten_mask_dtype,atten_mask_format,attention_out_shape,attention_out_dtype,attention_out_format,softmax_lse_shape,softmax_lse_dtype,softmax_lse_format,num_heads,scale,input_layout,softmax_lse_flag,case_annotation
|
|
181
|
+
infershape_0,SUCCESS,4 13 16 512,FLOAT16,ND,4 10347 1 512,FLOAT16,ND,0,FLOAT16,ND,4 13 10347,INT8,ND,4 13 16 512,FLOAT16,ND,0,FLOAT,ND,16,0.04166666666666666,BSND,0,基础场景
|
|
182
|
+
```
|
|
183
|
+
|
|
184
|
+
---
|
|
185
|
+
|
|
186
|
+
## 大量 Tensor 的处理建议
|
|
187
|
+
|
|
188
|
+
当输入/输出 Tensor 数量 > 10 时,建议使用自动化方法:
|
|
189
|
+
|
|
190
|
+
### 建议 1:使用脚本生成 CSV 列名
|
|
191
|
+
|
|
192
|
+
```bash
|
|
193
|
+
python scripts/generate_csv_template.py {算子名}_def.cpp
|
|
194
|
+
```
|
|
195
|
+
|
|
196
|
+
脚本会自动:
|
|
197
|
+
- 解析输入/输出 Tensor 列表
|
|
198
|
+
- 解析属性列表
|
|
199
|
+
- 生成 CSV 列名行
|
|
200
|
+
- 生成参数结构体框架代码
|
|
201
|
+
|
|
202
|
+
### 建议 2:手动提取时按功能分组
|
|
203
|
+
|
|
204
|
+
将 Tensor 按功能分组排列,便于理解和填写:
|
|
205
|
+
|
|
206
|
+
```
|
|
207
|
+
基础输入组:query, key, value
|
|
208
|
+
量化组:dequant_scale1, quant_scale1
|
|
209
|
+
PagedAttention组:block_table, kv_padding_size
|
|
210
|
+
Optional组:pse_shift, antiquant_scale
|
|
211
|
+
```
|
|
212
|
+
|
|
213
|
+
### 建议 3:在注释中标注关键信息
|
|
214
|
+
|
|
215
|
+
```csv
|
|
216
|
+
case_annotation
|
|
217
|
+
基础场景:query + key + value
|
|
218
|
+
量化场景:启用 antiquant_scale + antiquant_offset
|
|
219
|
+
PagedAttention:启用 block_table
|
|
220
|
+
```
|
|
221
|
+
|
|
222
|
+
---
|
|
223
|
+
|
|
224
|
+
## 验证转换结果
|
|
225
|
+
|
|
226
|
+
转换完成后,需要验证:
|
|
227
|
+
|
|
228
|
+
1. **编译验证**
|
|
229
|
+
```bash
|
|
230
|
+
bash build.sh -u --ops={算子名}
|
|
231
|
+
```
|
|
232
|
+
|
|
233
|
+
2. **用例数量验证**
|
|
234
|
+
- 新旧用例数量应该一致
|
|
235
|
+
- 每个旧用例都应该对应一行 CSV
|
|
236
|
+
|
|
237
|
+
3. **参数正确性验证**
|
|
238
|
+
- 运行结果与旧代码一致
|
|
239
|
+
- 期望结果匹配
|
|
240
|
+
|
|
241
|
+
---
|
|
242
|
+
|
|
243
|
+
## 常见问题处理
|
|
244
|
+
|
|
245
|
+
### Q1: 旧代码中有多个 TEST_F,如何映射到 CSV?
|
|
246
|
+
|
|
247
|
+
**A:** 每个 `TEST_F` 对应一行 CSV,`case_name` 使用测试名称。
|
|
248
|
+
|
|
249
|
+
示例:
|
|
250
|
+
```cpp
|
|
251
|
+
TEST_F(OpTest, test_case_0) { ... }
|
|
252
|
+
TEST_F(OpTest, test_case_1) { ... }
|
|
253
|
+
```
|
|
254
|
+
|
|
255
|
+
转换为:
|
|
256
|
+
```csv
|
|
257
|
+
case_name,...
|
|
258
|
+
test_case_0,...
|
|
259
|
+
test_case_1,...
|
|
260
|
+
```
|
|
261
|
+
|
|
262
|
+
### Q2: 旧代码中有复杂的逻辑判断,如何处理?
|
|
263
|
+
|
|
264
|
+
**A:** 将逻辑判断的结果作为参数写入 CSV。
|
|
265
|
+
|
|
266
|
+
示例:
|
|
267
|
+
```cpp
|
|
268
|
+
if (condition) {
|
|
269
|
+
shape = {4, 13, 512};
|
|
270
|
+
} else {
|
|
271
|
+
shape = {1, 1};
|
|
272
|
+
}
|
|
273
|
+
```
|
|
274
|
+
|
|
275
|
+
转换时,根据 condition 的值,将对应 shape 写入 CSV。
|
|
276
|
+
|
|
277
|
+
### Q3: 旧代码中有循环生成的多个用例,如何处理?
|
|
278
|
+
|
|
279
|
+
**A:** 每个循环迭代对应一行 CSV。
|
|
280
|
+
|
|
281
|
+
示例:
|
|
282
|
+
```cpp
|
|
283
|
+
for (int i = 0; i < 3; i++) {
|
|
284
|
+
TEST_F(OpTest, test_i) { ... }
|
|
285
|
+
}
|
|
286
|
+
```
|
|
287
|
+
|
|
288
|
+
转换为:
|
|
289
|
+
```csv
|
|
290
|
+
case_name,...
|
|
291
|
+
test_0,...
|
|
292
|
+
test_1,...
|
|
293
|
+
test_2,...
|
|
294
|
+
```
|
|
295
|
+
|
|
296
|
+
### Q4: 如何处理精度损失?
|
|
297
|
+
|
|
298
|
+
**A:** 浮点属性建议使用旧代码中的精确值,避免转换过程中的精度损失。
|
|
299
|
+
|
|
300
|
+
示例:
|
|
301
|
+
```cpp
|
|
302
|
+
CreateFrom<float>(0.04166666666666666f) // 精确值
|
|
303
|
+
```
|
|
304
|
+
|
|
305
|
+
CSV:
|
|
306
|
+
```
|
|
307
|
+
0.04166666666666666 // 保留完整精度
|
|
308
|
+
```
|
|
309
|
+
|
|
310
|
+
---
|
|
311
|
+
|
|
312
|
+
## 迁移检查清单
|
|
313
|
+
|
|
314
|
+
完成迁移后,检查以下事项:
|
|
315
|
+
|
|
316
|
+
- [ ] 所有旧用例都已转换为 CSV 行
|
|
317
|
+
- [ ] case_name 与旧测试名称一致
|
|
318
|
+
- [ ] Tensor 参数(shape/dtype/format)正确
|
|
319
|
+
- [ ] 属性值正确(类型匹配)
|
|
320
|
+
- [ ] 期望结果正确(SUCCESS/FAILED)
|
|
321
|
+
- [ ] Optional Tensor 处理正确(按测试类型)
|
|
322
|
+
- [ ] 空 Tensor 处理正确
|
|
323
|
+
- [ ] 编译成功
|
|
324
|
+
- [ ] 所有用例通过
|
|
325
|
+
- [ ] 运行结果与旧代码一致
|
|
@@ -0,0 +1,75 @@
|
|
|
1
|
+
# ops-transformer 算子仓特性
|
|
2
|
+
|
|
3
|
+
## CSV 格式输入用例
|
|
4
|
+
|
|
5
|
+
transformer 仓在现有 UT 框架的基础上,封装了一层 CSV 导入工具。
|
|
6
|
+
|
|
7
|
+
使用 csv 格式输入的优势:
|
|
8
|
+
- 数据与代码分离,用例框架编译完成后,仅修改用例表格时无需重复编译
|
|
9
|
+
- 避免重复代码,对于多参数算子,可大幅缩减用例长度,缩小模型上下文
|
|
10
|
+
|
|
11
|
+
**对于 transformer 仓的算子,优先使用 CSV 格式创建用例。**
|
|
12
|
+
|
|
13
|
+
### 触发条件
|
|
14
|
+
|
|
15
|
+
- `repo_type` 为 **"ops-transformer"**
|
|
16
|
+
- 用户提及 **"CSV"**、**"CSV整改"**、**"CSV格式"**、**"CSV化"**
|
|
17
|
+
|
|
18
|
+
### 模板文件
|
|
19
|
+
|
|
20
|
+
模板文件位于 [assets/csv-refactor/](../../assets/csv-refactor/):
|
|
21
|
+
|
|
22
|
+
| 文件 | 用途 |
|
|
23
|
+
|-----|------|
|
|
24
|
+
| `template_param.h` | 参数结构体模板 |
|
|
25
|
+
| `template_tiling.cpp` | Tiling 测试模板 |
|
|
26
|
+
| `template_tiling.csv` | Tiling CSV 示例 |
|
|
27
|
+
| `template_infershape.cpp` | InferShape 测试模板 |
|
|
28
|
+
| `template_infershape.csv` | InferShape CSV 示例 |
|
|
29
|
+
| `template_inferdtype.cpp` | InferDataType 测试模板 |
|
|
30
|
+
| `template_inferdtype.csv` | InferDataType CSV 示例 |
|
|
31
|
+
| `template_cmake.txt` | CMakeLists.txt 模板 |
|
|
32
|
+
|
|
33
|
+
### 自动化工具
|
|
34
|
+
|
|
35
|
+
```bash
|
|
36
|
+
python scripts/generate_csv_template.py {op_name}_def.cpp
|
|
37
|
+
```
|
|
38
|
+
|
|
39
|
+
### 详细文档
|
|
40
|
+
|
|
41
|
+
- [CSV 重构流程详解](./csv-refactor-workflow.md)
|
|
42
|
+
- [CSV 格式规范](./csv-format-spec.md)
|
|
43
|
+
- [从旧代码提取参数](./legacy-code-extraction.md)
|
|
44
|
+
|
|
45
|
+
## 多架构支持
|
|
46
|
+
|
|
47
|
+
ops-transformer 仓支持多种芯片架构,需要特别注意:
|
|
48
|
+
|
|
49
|
+
### SoC 与架构对应关系
|
|
50
|
+
|
|
51
|
+
| SoC 参数 | 架构 | 编译示例 |
|
|
52
|
+
|---------|------|---------|
|
|
53
|
+
| `ascend910b` | arch32 | `--soc=ascend910b` |
|
|
54
|
+
| `ascend950` | arch35 | `--soc=ascend950` |
|
|
55
|
+
|
|
56
|
+
### 编译注意事项
|
|
57
|
+
|
|
58
|
+
**重要**:编译时必须指定 `--soc` 参数,否则对应架构的代码不会被编译!
|
|
59
|
+
|
|
60
|
+
```bash
|
|
61
|
+
# 编译 arch35 用例
|
|
62
|
+
bash build.sh --ophost_test --noexec --ops=<op_name> --soc=ascend950
|
|
63
|
+
|
|
64
|
+
# 编译 arch32 用例
|
|
65
|
+
bash build.sh --ophost_test --noexec --ops=<op_name> --soc=ascend910b
|
|
66
|
+
```
|
|
67
|
+
|
|
68
|
+
### 运行注意事项
|
|
69
|
+
|
|
70
|
+
运行测试时需要设置 `BUILD_PATH` 环境变量:
|
|
71
|
+
|
|
72
|
+
```bash
|
|
73
|
+
export BUILD_PATH=/path/to/ops-transformer/build
|
|
74
|
+
./build/tests/ut/framework_normal/op_host/transformer_op_host_ut --gtest_filter="*Arch35*"
|
|
75
|
+
```
|
|
@@ -0,0 +1,65 @@
|
|
|
1
|
+
# UT 编译运行指南
|
|
2
|
+
|
|
3
|
+
## 编译运行所有模块
|
|
4
|
+
|
|
5
|
+
```bash
|
|
6
|
+
bash build.sh -u --ops=$op_name --cov
|
|
7
|
+
```
|
|
8
|
+
|
|
9
|
+
## 编译部分模块
|
|
10
|
+
|
|
11
|
+
```bash
|
|
12
|
+
bash build.sh -u --opapi --ops=$op_name --cov
|
|
13
|
+
bash build.sh -u --ophost --ops=$op_name --cov
|
|
14
|
+
bash build.sh -u --opkernel --ops=$op_name --cov
|
|
15
|
+
```
|
|
16
|
+
|
|
17
|
+
## 查看编译帮助
|
|
18
|
+
|
|
19
|
+
```bash
|
|
20
|
+
bash build.sh -h
|
|
21
|
+
```
|
|
22
|
+
|
|
23
|
+
## 常用参数说明
|
|
24
|
+
|
|
25
|
+
| 参数 | 说明 |
|
|
26
|
+
|------|------|
|
|
27
|
+
| `-u` | 编译UT |
|
|
28
|
+
| `--ops=<op_name>` | 指定算子名称 |
|
|
29
|
+
| `--soc=<soc_version>` | 指定SOC版本 |
|
|
30
|
+
| `--cov` | 生成覆盖率报告 |
|
|
31
|
+
| `--opapi` | 仅编译op_api层 |
|
|
32
|
+
| `--ophost` | 仅编译op_host层 |
|
|
33
|
+
| `--opkernel` | 仅编译op_kernel层 |
|
|
34
|
+
|
|
35
|
+
## 环境变量
|
|
36
|
+
|
|
37
|
+
在 UT 运行前,设置环境变量`export ASCEND_SLOG_PRINT_TO_STDOUT=1`,将 UT 运行的 LOG 重定向到标准输出,方便定位问题
|
|
38
|
+
|
|
39
|
+
较难定位时,可以开启更低的 LOG 等级,输出更详细的信息:
|
|
40
|
+
|
|
41
|
+
| 等级 | 数值 |
|
|
42
|
+
|------|------|
|
|
43
|
+
| DEBUG | 0 |
|
|
44
|
+
| INFO | 1 |
|
|
45
|
+
| WARNING | 2 |
|
|
46
|
+
| ERROR(默认) | 3 |
|
|
47
|
+
|
|
48
|
+
`export ASCEND_GLOBAL_LOG_LEVEL=${数值}`
|
|
49
|
+
|
|
50
|
+
## 覆盖率报告位置
|
|
51
|
+
|
|
52
|
+
编译完成后,覆盖率报告位于:
|
|
53
|
+
```
|
|
54
|
+
<repo>/build/tests/ut/cov_report/cpp_utest/ops.info_filtered
|
|
55
|
+
```
|
|
56
|
+
|
|
57
|
+
## 查看覆盖率
|
|
58
|
+
|
|
59
|
+
```bash
|
|
60
|
+
# 查看摘要
|
|
61
|
+
lcov --summary <repo>/build/tests/ut/cov_report/cpp_utest/ops.info_filtered
|
|
62
|
+
|
|
63
|
+
# 查看未覆盖代码
|
|
64
|
+
lcov --list <repo>/build/tests/ut/cov_report/cpp_utest/ops.info_filtered | grep ":0"
|
|
65
|
+
```
|
package/assets/ops-direct-invoke/skills/ascendc-ut-develop/references/ut-guide/op-api-ut-guide.md
ADDED
|
@@ -0,0 +1,237 @@
|
|
|
1
|
+
# OP_API UT 详细指南
|
|
2
|
+
|
|
3
|
+
## 概述
|
|
4
|
+
|
|
5
|
+
测试 ACLNN 算子接口的正确性:
|
|
6
|
+
- GetWorkspaceSize 返回值
|
|
7
|
+
- 输入参数校验(dtype、shape、format、nullptr)
|
|
8
|
+
- 输出结果正确性
|
|
9
|
+
|
|
10
|
+
---
|
|
11
|
+
|
|
12
|
+
## 目录结构
|
|
13
|
+
|
|
14
|
+
```
|
|
15
|
+
<repo>/<category>/<op>/tests/ut/op_api/
|
|
16
|
+
├── CMakeLists.txt
|
|
17
|
+
├── test_aclnn_<op>.cpp # 主接口测试
|
|
18
|
+
├── test_aclnn_<op>_out.cpp # out变体测试(如有)
|
|
19
|
+
└── test_aclnn_<op>_inplace.cpp # inplace变体测试(如有)
|
|
20
|
+
```
|
|
21
|
+
|
|
22
|
+
---
|
|
23
|
+
|
|
24
|
+
## 核心组件
|
|
25
|
+
|
|
26
|
+
### TensorDesc
|
|
27
|
+
|
|
28
|
+
```cpp
|
|
29
|
+
// 基本构造
|
|
30
|
+
auto tensor = TensorDesc({3, 3, 3}, ACL_FLOAT, ACL_FORMAT_ND);
|
|
31
|
+
|
|
32
|
+
// 链式调用
|
|
33
|
+
auto tensor = TensorDesc({3, 3}, ACL_FLOAT, ACL_FORMAT_ND)
|
|
34
|
+
.ValueRange(-2.0, 2.0)
|
|
35
|
+
.Precision(0.0001, 0.0001);
|
|
36
|
+
|
|
37
|
+
// 非连续内存
|
|
38
|
+
auto tensor = TensorDesc({5, 4}, ACL_FLOAT, ACL_FORMAT_ND, {1, 5}, 0, {4, 5});
|
|
39
|
+
```
|
|
40
|
+
|
|
41
|
+
### ScalarDesc
|
|
42
|
+
|
|
43
|
+
```cpp
|
|
44
|
+
auto alpha = ScalarDesc(1.0f); // float
|
|
45
|
+
auto value = ScalarDesc(42); // int
|
|
46
|
+
auto flag = ScalarDesc(true); // bool
|
|
47
|
+
```
|
|
48
|
+
|
|
49
|
+
### OP_API_UT 宏
|
|
50
|
+
|
|
51
|
+
```cpp
|
|
52
|
+
// 单输入单输出
|
|
53
|
+
auto ut = OP_API_UT(aclnnAbs, INPUT(self), OUTPUT(out));
|
|
54
|
+
|
|
55
|
+
// 多输入带Scalar
|
|
56
|
+
auto ut = OP_API_UT(aclnnAdd, INPUT(self, other, alpha), OUTPUT(out));
|
|
57
|
+
|
|
58
|
+
// nullptr测试
|
|
59
|
+
auto ut = OP_API_UT(aclnnAbs, INPUT((aclTensor*)nullptr), OUTPUT(out));
|
|
60
|
+
```
|
|
61
|
+
|
|
62
|
+
---
|
|
63
|
+
|
|
64
|
+
## 测试用例示例
|
|
65
|
+
|
|
66
|
+
### 异常用例(最先编写)
|
|
67
|
+
|
|
68
|
+
```cpp
|
|
69
|
+
TEST_F(l2_abs_test, case_anullptr_input) {
|
|
70
|
+
auto out = TensorDesc({2, 2, 3}, ACL_FLOAT, ACL_FORMAT_ND);
|
|
71
|
+
auto ut = OP_API_UT(aclnnAbs, INPUT((aclTensor*)nullptr), OUTPUT(out));
|
|
72
|
+
uint64_t workspace_size = 0;
|
|
73
|
+
EXPECT_EQ(ut.TestGetWorkspaceSize(&workspace_size), ACLNN_ERR_PARAM_NULLPTR);
|
|
74
|
+
}
|
|
75
|
+
|
|
76
|
+
TEST_F(l2_abs_test, case_invalid_dtype) {
|
|
77
|
+
auto self = TensorDesc({3, 3, 3}, ACL_DOUBLE, ACL_FORMAT_ND);
|
|
78
|
+
auto out = TensorDesc(self);
|
|
79
|
+
auto ut = OP_API_UT(aclnnAbs, INPUT(self), OUTPUT(out));
|
|
80
|
+
uint64_t workspace_size = 0;
|
|
81
|
+
EXPECT_EQ(ut.TestGetWorkspaceSize(&workspace_size), ACLNN_ERR_PARAM_INVALID);
|
|
82
|
+
}
|
|
83
|
+
```
|
|
84
|
+
|
|
85
|
+
### 正常用例
|
|
86
|
+
|
|
87
|
+
```cpp
|
|
88
|
+
TEST_F(l2_abs_test, case_abs_for_float_type) {
|
|
89
|
+
auto self = TensorDesc({3, 3, 3}, ACL_FLOAT, ACL_FORMAT_ND).ValueRange(-2.0, 2.0);
|
|
90
|
+
auto out = TensorDesc(self).Precision(0.0001, 0.0001);
|
|
91
|
+
auto ut = OP_API_UT(aclnnAbs, INPUT(self), OUTPUT(out));
|
|
92
|
+
uint64_t workspace_size = 0;
|
|
93
|
+
EXPECT_EQ(ut.TestGetWorkspaceSize(&workspace_size), ACL_SUCCESS);
|
|
94
|
+
}
|
|
95
|
+
```
|
|
96
|
+
|
|
97
|
+
---
|
|
98
|
+
|
|
99
|
+
## CMakeLists.txt
|
|
100
|
+
|
|
101
|
+
```cmake
|
|
102
|
+
if(UT_TEST_ALL OR OP_API_UT)
|
|
103
|
+
add_modules_ut_sources(UT_NAME ${OP_API_MODULE_NAME} MODE PRIVATE DIR ${CMAKE_CURRENT_SOURCE_DIR})
|
|
104
|
+
endif()
|
|
105
|
+
```
|
|
106
|
+
|
|
107
|
+
---
|
|
108
|
+
|
|
109
|
+
## 常见返回值
|
|
110
|
+
|
|
111
|
+
| 返回值 | 说明 |
|
|
112
|
+
|-------|------|
|
|
113
|
+
| `ACL_SUCCESS` | 成功 |
|
|
114
|
+
| `ACLNN_ERR_PARAM_NULLPTR` | 参数为空指针 |
|
|
115
|
+
| `ACLNN_ERR_PARAM_INVALID` | 参数无效 |
|
|
116
|
+
|
|
117
|
+
---
|
|
118
|
+
|
|
119
|
+
## 编译命令
|
|
120
|
+
|
|
121
|
+
```bash
|
|
122
|
+
# 编译 op_api UT
|
|
123
|
+
bash build.sh -u --opapi --ops='<op_name>' --soc='<soc_version>'
|
|
124
|
+
|
|
125
|
+
# 编译并生成覆盖率
|
|
126
|
+
bash build.sh -u --opapi --ops='<op_name>' --soc='<soc_version>' --cov
|
|
127
|
+
```
|
|
128
|
+
|
|
129
|
+
---
|
|
130
|
+
|
|
131
|
+
## 测试用例编写顺序(TDD)
|
|
132
|
+
|
|
133
|
+
1. **异常用例**:nullptr 测试、无效 dtype 测试、shape 不匹配测试、超过 8 维测试
|
|
134
|
+
2. **正常用例**:所有支持的 dtype、所有支持的 format
|
|
135
|
+
3. **边界用例**:空 tensor、0 维 tensor
|
|
136
|
+
|
|
137
|
+
---
|
|
138
|
+
|
|
139
|
+
## Dtype 排列组合校验
|
|
140
|
+
|
|
141
|
+
### 概述
|
|
142
|
+
|
|
143
|
+
从算子定义文件(`xxx_def.cpp`)中提取输入 Tensor 支持的 dtype 列表,自动生成所有合法的 dtype 排列组合测试用例,确保:
|
|
144
|
+
|
|
145
|
+
- 所有支持的 dtype 组合都被覆盖
|
|
146
|
+
- 不支持的 dtype 组合被识别为异常用例
|
|
147
|
+
- 提升算子 dtype 分支的覆盖率
|
|
148
|
+
|
|
149
|
+
### 前置步骤
|
|
150
|
+
|
|
151
|
+
**1. 定位算子定义文件**
|
|
152
|
+
|
|
153
|
+
```bash
|
|
154
|
+
# 在算子目录下查找 def 文件
|
|
155
|
+
find ${op_path}/op_host -name "*_def.cpp"
|
|
156
|
+
```
|
|
157
|
+
|
|
158
|
+
**2. 提取 dtype 信息**
|
|
159
|
+
|
|
160
|
+
### Dtype 排列组合校验
|
|
161
|
+
|
|
162
|
+
#### DataType 形式
|
|
163
|
+
|
|
164
|
+
| 形式 | 语法 | 组合数 |
|
|
165
|
+
|------|------|--------|
|
|
166
|
+
| 数组 | `.DataType({...})` | 数组长度 |
|
|
167
|
+
| 固定 | `.DataTypeList({...})` | 不计入 |
|
|
168
|
+
|
|
169
|
+
**约束**:数组形式参数长度必须相同。
|
|
170
|
+
|
|
171
|
+
#### 校验流程
|
|
172
|
+
|
|
173
|
+
| 步骤 | 操作 | 命令/方法 | 通过条件 |
|
|
174
|
+
|------|------|---------|---------|
|
|
175
|
+
| 1 | 脚本提取 | `python scripts/extract_dtype_combinations.py ${def_file}` | 组合数 N > 0 |
|
|
176
|
+
| 2 | 补充用例 | 按模板补充 N 个 dtype 用例 | UT 文件用例数 M ≥ N |
|
|
177
|
+
| 3 | 覆盖率分析 | `lcov --list ops.info_filtered \| grep "dtype" \| grep ":0"` | 无未覆盖分支 |
|
|
178
|
+
| 4 | 实际测试 | `bash build.sh -u --opapi --ops=${op_name} --soc=${soc}` | N 个组合全部 PASS |
|
|
179
|
+
| 5 | 异常校验 | 构造 3 种非法 dtype 组合 | 返回 ACLNN_ERR_PARAM_INVALID |
|
|
180
|
+
|
|
181
|
+
**判定**:5 步全部通过 → dtype 校验完整。
|
|
182
|
+
|
|
183
|
+
**脚本输出**:
|
|
184
|
+
- `/tmp/${op_name}_dtype_combinations.json` - dtype 组合数据
|
|
185
|
+
- 测试用例模板(屏幕输出)- 直接补充到 UT 文件
|
|
186
|
+
|
|
187
|
+
#### 异常组合构造
|
|
188
|
+
|
|
189
|
+
| 异常类型 | 构造方法 | 预期结果 |
|
|
190
|
+
|---------|---------|---------|
|
|
191
|
+
| 跨位置组合 | query 用 DataType[0],key 用 DataType[1] | ACLNN_ERR_PARAM_INVALID |
|
|
192
|
+
| 未定义 dtype | 使用 def 文件中未出现的 dtype(如 DOUBLE) | ACLNN_ERR_PARAM_INVALID |
|
|
193
|
+
| dtype 不一致 | 同一位置不同 Tensor 使用不同 dtype | ACLNN_ERR_PARAM_INVALID |
|
|
194
|
+
|
|
195
|
+
**示例**:
|
|
196
|
+
|
|
197
|
+
```cpp
|
|
198
|
+
// 跨位置组合
|
|
199
|
+
TEST_F(l2_flash_attention_test, case_invalid_cross_position) {
|
|
200
|
+
auto query = TensorDesc({4, 13, 8192}, ACL_FLOAT16, ACL_FORMAT_ND); // 位置 0
|
|
201
|
+
auto key = TensorDesc({4, 13, 1024}, ACL_BF16, ACL_FORMAT_ND); // 位置 1
|
|
202
|
+
EXPECT_EQ(ut.TestGetWorkspaceSize(&workspace_size), ACLNN_ERR_PARAM_INVALID);
|
|
203
|
+
}
|
|
204
|
+
|
|
205
|
+
// 未定义 dtype
|
|
206
|
+
TEST_F(l2_flash_attention_test, case_invalid_unknown_dtype) {
|
|
207
|
+
auto query = TensorDesc({4, 13, 8192}, ACL_DOUBLE, ACL_FORMAT_ND); // DOUBLE 未定义
|
|
208
|
+
EXPECT_EQ(ut.TestGetWorkspaceSize(&workspace_size), ACLNN_ERR_PARAM_INVALID);
|
|
209
|
+
}
|
|
210
|
+
```
|
|
211
|
+
|
|
212
|
+
#### 异常处理
|
|
213
|
+
|
|
214
|
+
| 异常情况 | 原因 | 处理 |
|
|
215
|
+
|---------|------|------|
|
|
216
|
+
| lcov 有 `:0` | 代码存在 def 未定义的 dtype 分支 | 分析代码补充用例 |
|
|
217
|
+
| 测试有 FAIL | def 定义了不支持的 dtype | 删除组合或修正 def |
|
|
218
|
+
| 异常组合未返回错误码 | 算子校验逻辑缺失 | 修复算子代码 |
|
|
219
|
+
|
|
220
|
+
---
|
|
221
|
+
|
|
222
|
+
### Dtype 映射表
|
|
223
|
+
|
|
224
|
+
| ge::DataType | aclDataType | CSV 写法 |
|
|
225
|
+
|--------------|-------------|----------|
|
|
226
|
+
| `ge::DT_FLOAT` | `ACL_FLOAT` | `FLOAT` |
|
|
227
|
+
| `ge::DT_FLOAT16` | `ACL_FLOAT16` | `FLOAT16` |
|
|
228
|
+
| `ge::DT_BF16` | `ACL_BF16` | `BF16` |
|
|
229
|
+
| `ge::DT_INT8` | `ACL_INT8` | `INT8` |
|
|
230
|
+
| `ge::DT_INT32` | `ACL_INT32` | `INT32` |
|
|
231
|
+
| `ge::DT_INT64` | `ACL_INT64` | `INT64` |
|
|
232
|
+
| `ge::DT_BOOL` | `ACL_BOOL` | `BOOL` |
|
|
233
|
+
|
|
234
|
+
**补充查找**:若所需 dtype 未在表中列出,可查看 CANN 安装路径下的枚举源文件获取完整定义:
|
|
235
|
+
- `ge::DataType` / `AscendC::DataType`:`${ASCEND_HOME_PATH}/asc/include/basic_api/kernel_type.h`
|
|
236
|
+
- `aclDataType`:`${ASCEND_HOME_PATH}/include/acl/acl_base_rt.h`
|
|
237
|
+
|