← CPU results

cpu/overhead_smoke

Legacy report: ingestion provenance is unrecorded; no raw artifact is inferred.

Non-gating trend: eight steady-state operations and two runtime session-entry diagnostics are shown separately in the original report below. Allocation counts come from a separate pooled pass.

Original report

CPU Overhead Smoke (ns/op and allocs/op)

• Suite: cpu/overhead_smoke
• Target profile: zen5-cpu
• Raw run: data/results/zen5-cpu/cpu/overhead_smoke/20261028_011419
• tenferro-rs commit: 8fec1ca70852b930c092e7177c30bb8af7594a57
• Collection command: BENCHMARK_TARGET_PROFILE=zen5-cpu TENFERRO_CPU_FEATURES=blas-mkl /workspaces/tenferro-benchmark/scripts/run_overhead_smoke.sh

Trend view for tenferro-rs #2178 (tenferro-benchmark #197); nothing here gates. Every row runs on one run thread; the prepared-runtime diagnostics set the CPU backend thread budget explicitly. allocs/op counts process-wide Rust allocator events over 64 calls after pool warmup, dropping outputs between calls. These observed counts include Rust workers and exclude foreign BLAS/C allocations; compare matching fixtures, thread budgets and execution scopes. ns/op is the median of retained-output batches, with destruction after the clock stops. Shared nightly runner timings are reference indicators only.

Steady-state operations

Inputs, entered reusable sessions, initialization, tracing and preparation are outside timing. Only the declared operation and intrinsic output allocation or pooled checkout are timed.

Runtime session-entry diagnostics

These rows time public Runtime::run_prepared, including its internal session entry and exit. Runtime construction, inputs, tracing and preparation remain outside timing. They are diagnostic public-path latency, not shared-session operation-performance evidence. Their allocation counters also include internal session entry. They are intentionally separate from the steady-state table.

Numerical checks: 10/10 rows passed against independent host loops before counting and sampling.

Prepared-runtime region execution (diagnostics)

Whether the runtime executed the traced region as one fused command or declined to per-op dispatch (cumulative over validation, the allocator pass and timed batches). run_prepared enters a runtime-internal session per call, and its entry is included in both diagnostic ns/op and allocs/op.

CPU Information

• Model: AMD Ryzen AI 9 HX 470 w/ Radeon 890M
• Vendor: AuthenticAMD
• Logical CPUs: 24
• Sockets: 1
• Cores per socket: 12
• Threads per core: 2
• NUMA nodes: 1
• Python platform: Linux-7.0.0-38-generic-x86_64-with-glibc2.39
• L3 domains (shared CPUs): 0-3,12-15; 4-11,16-23
• CPU boost: enabled

Run metadata

Full metadata: data/results/zen5-cpu/cpu/overhead_smoke/20261028_011419/run.yaml

timestamp: '2026-10-28T01:14:19Z'
tenferro_rs:
  path: /workspaces/tenferro-benchmark/extern/tenferro-rs
  commit: 8fec1ca70852b930c092e7177c30bb8af7594a57
  dirty: false
  features:
  - blas-mkl
  resolved_path: /workspaces/tenferro-benchmark/extern/tenferro-rs
blas:
  implementation: mkl
  version: 2026.0.2
  root: /opt/intel/oneapi/mkl/latest
  library: /opt/intel/oneapi/mkl/latest/lib/libmkl_rt.so
environment:
  hostname: bcfefbe95606
  os: Linux-7.0.0-38-generic-x86_64-with-glibc2.39
  arch: x86_64
  cpu: x86_64
  affinity:
    pinning: taskset
    requested_measured_cpus: '0'
    observer_cpus: 4-11,16-23
    override_of: null
    thread_budget: '1'
    measured_cpus: '0'
    l3_domains:
    - 0-3,12-15
    - 4-11,16-23
    cpu_boost: enabled
    scaling_governor: performance
    scaling_driver: amd-pstate-epp
  env:
    OMP_NUM_THREADS: '1'
    OMP_THREAD_LIMIT: '1'
    OMP_DYNAMIC: 'FALSE'
    RAYON_NUM_THREADS: '1'
    BENCHMARK_TARGET_PROFILE: zen5-cpu
    BENCHMARK_COMMIT: 3f31f38e0491084b9414035754a3809ef0f5ae94
    OPENBLAS_ROOT: /opt/openblas
    MKLROOT: /opt/intel/oneapi/mkl/latest
    OPENBLAS_NUM_THREADS: '1'
    GOTO_NUM_THREADS: '1'
    MKL_NUM_THREADS: '1'
    MKL_DYNAMIC: 'FALSE'
    VECLIB_MAXIMUM_THREADS: '1'
    VECLIB_NUM_THREADS: '1'
    NUMEXPR_NUM_THREADS: '1'
    BLIS_NUM_THREADS: '1'
    PJRT_NPROC: '1'
    XLA_FLAGS: --xla_cpu_multi_thread_eigen=false --xla_cpu_experimental_ynn_fusion_type=
      intra_op_parallelism_threads=1
    JULIA_NUM_THREADS: '1'
    PUBLICATION_GATE_PROFILE: full
    USE_CUDA: '0'
    BENCHMARK_BUILD_PROFILE: release
    RUSTC_VERSION: rustc 1.99.1 (58cf8f9c3 2026-10-13)
    BENCH_INSTANCE: null
collection:
  command: 'BENCHMARK_TARGET_PROFILE=zen5-cpu TENFERRO_CPU_FEATURES=blas-mkl /workspaces/tenferro-benchmark/scripts/run_overhead_smoke.sh '
  timing_scope: mixed; see timing_scopes_by_case
  timing_scopes_by_case:
    typed_elemwise_f64_add_bcast_n16: steady_state_host_api
    typed_elemwise_f64_add_bcast_8x8x8: steady_state_host_api
    typed_elemwise_f64_add_permuted_view_8x8x8: steady_state_host_api
    permute_materialize_f64_r4_4x4x4x4: steady_state_host_api
    prepared_mul_add_f64_budget1_32x32x32_session_entry_diagnostic: public_path_latency_including_internal_session_entry
    prepared_mul_add_f64_budget2_32x32x32_session_entry_diagnostic: public_path_latency_including_internal_session_entry
    reduce_axis_f64_8x8x8: steady_state_host_api
    concatenate_axis1_f64_8x8x8: steady_state_host_api
    einsum_prepared_f64_n2: steady_state_host_api
    einsum_prepared_f64_n4: steady_state_host_api
  setup_policy:
    steady_state_host_api: inputs, entered session, initialization, tracing and preparation
      outside timing
    public_path_latency_including_internal_session_entry: runtime/input/tracing/preparation
      setup outside timing; run_prepared internal session entry/exit inside timing;
      diagnostic only
  output_policy: outputs retained until the clock stops; destruction outside timing
  allocation_scope: 'separate pooled pass: 64 calls after warmup, outputs dropped
    between calls'
  gating: none (trend view)