CPU FFT Benchmark Results

• Suite: cpu/fft
• Target profile: mac-cpu
• Suite file: benchmarks/cpu/fft.yaml
• Run metadata: data/results/mac-cpu/cpu/fft/20260923_205931/run.yaml
• Timestamp: 20260923_205931

Latest run: ./scripts/run_cpu_fft.sh 1 4.

This file is generated from sequential CPU FFT runs under data/results/mac-cpu/cpu/fft/20260923_205931.

• tenferro-rs commit: 8a1839febeb3c868a502e26397ca10761bbc568d

CPU Information

• Model: Apple M5 Max
• Vendor: Apple
• Logical CPUs: 18
• Physical CPUs: 18
• Sockets: 1
• Cores per socket: 18
• Threads per core: 1
• NUMA nodes: 1
• Performance levels: Super: 6 physical / 6 logical (L1i 192 KiB, L1d 128 KiB, L2 16 MiB, 6 CPUs/L2); Performance: 12 physical / 12 logical (L1i 128 KiB, L1d 64 KiB, L2 8 MiB, 6 CPUs/L2)
• Python platform: macOS-26.5.1-arm64-arm-64bit-Mach-O

Thread Environments

Threads: 1

• Run metadata: data/results/mac-cpu/cpu/fft/20260923_205931/run_t1.yaml
• OMP_NUM_THREADS: 1
• OMP_THREAD_LIMIT: 1
• OMP_DYNAMIC: FALSE
• RAYON_NUM_THREADS: 1
• OPENBLAS_NUM_THREADS: 1
• GOTO_NUM_THREADS: 1
• MKL_NUM_THREADS: 1
• VECLIB_MAXIMUM_THREADS: 1
• VECLIB_NUM_THREADS: 1
• NUMEXPR_NUM_THREADS: 1
• BLIS_NUM_THREADS: 1
• XLA_FLAGS: --xla_cpu_multi_thread_eigen=false intra_op_parallelism_threads=1

Threads: 4

• Run metadata: data/results/mac-cpu/cpu/fft/20260923_205931/run_t4.yaml
• OMP_NUM_THREADS: 4
• OMP_THREAD_LIMIT: 4
• OMP_DYNAMIC: FALSE
• RAYON_NUM_THREADS: 4
• OPENBLAS_NUM_THREADS: 4
• GOTO_NUM_THREADS: 4
• MKL_NUM_THREADS: 4
• VECLIB_MAXIMUM_THREADS: 4
• VECLIB_NUM_THREADS: 4
• NUMEXPR_NUM_THREADS: 4
• BLIS_NUM_THREADS: 4
• XLA_FLAGS: --xla_cpu_multi_thread_eigen=true intra_op_parallelism_threads=4

Timing Discipline

• Input tensors are created outside the timed region.
• Each timed call creates the output tensor.
• tenferro-rs immediate rows require BENCH_INCLUDE_SETUP_DIAGNOSTICS=1 and use the separate cpu/fft_setup suite.
• tenferro-rs read rows require BENCH_INCLUDE_SETUP_DIAGNOSTICS=1 because the API performs one-shot planning.
• tenferro-rs cached rows reuse a caller-owned FftExecutor across warmups and timed runs.
• tenferro-rs eager rows reuse one EagerRuntime and input EagerTensor; setup is outside timing.
• tenferro-rs trace rows construct and compile TracedTensorFftExt graphs outside timing and reuse the compiled program.
• PyTorch rows use torch.fft after warmup, allowing PyTorch internal planning/cache behavior.
• The primary fair comparison is cached FftExecutor versus warmed torch.fft; one-shot diagnostics are opt-in and never operation comparisons.
• This initial suite only measures 1D transforms to avoid row-major/column-major batched-axis layout artifacts.

Threads: 1 4

• CSV: data/results/mac-cpu/cpu/fft/20260923_205931/cpu_fft_t1_20260923_205931.csv
• CSV: data/results/mac-cpu/cpu/fft/20260923_205931/cpu_fft_t4_20260923_205931.csv
• Source table: data/results/mac-cpu/cpu/fft/20260923_205931/cpu_fft_20260923_205931.md

CPU FFT Benchmark Items

Median ± IQR (ms). Missing backends are shown as -.

Timing scope: input tensors are created outside the timed region; each timed call creates the FFT output tensor. The primary eager comparison is tenferro-rs FftExecutor cached versus warmed PyTorch torch.fft; traced rows reuse one compiled graph and one-shot rows are diagnostic. Rows are limited to 1D transforms so tenferro-rs column-major layout and PyTorch row-major layout do not change the measured transform axis.

Operation execution

Notes:

• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.318126
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.328916
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.334083
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.335250
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.339084
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.339333
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.349834
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.353166
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.470625
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.479250
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.482707
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.484083
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.493666
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.497125
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.510333
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=0.514292
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=10.479625
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=10.629417
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=10.667041
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=10.741876
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=13.106625
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=13.176875
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=13.386958
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=13.406833
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=4.682666
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=4.713959
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=4.770625
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=4.773417
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=6.157543
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=6.191875
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=6.270333
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=128; median_batch_ms=6.273917
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=10.068834
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=10.201417
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=10.284542
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=10.357710
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=10.864583
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=11.239375
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=11.411667
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=11.588667
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=3.432042
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=3.472376
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=3.483709
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=3.546959
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=4.333333
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=4.392083
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=4.509917
• pytorch-cpu: torch.fft warmup before measured runs; input allocation outside timed region; operations_per_sample=1; median_batch_ms=4.594208
• tenferro-fft-eager: API includes per-call planning or session entry; use shared cached executor for short operation timing
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=3.618000
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=3.632667
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=3.752666
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=3.917416
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=3.929792
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=4.148417
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=4.193833
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=4.218417
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.600208
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.616292
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.858084
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.943625
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.956250
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=5.994250
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=6.052417
• tenferro-fft-eager: EagerTensorFftExt with one reused eager runtime and input; operations_per_sample=1; median_batch_ms=6.600041
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.283875
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.306334
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.307542
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.316083
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.341584
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.378042
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.391667
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.421625
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.428041
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.436459
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.470250
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.488416
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.490792
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.581208
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.618791
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=0.679500
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=4.977292
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.002500
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.134208
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.290875
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.338584
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.368792
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.378250
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=5.387541
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=8.579750
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=8.743083
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=8.819084
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=8.857167
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=9.216875
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=9.253500
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=9.286875
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=128; median_batch_ms=9.286916
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=3.578292
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=3.625792
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=3.752625
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=3.945750
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=3.959250
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=4.094333
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=4.209958
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=4.247709
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=5.867708
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=5.900250
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=5.923042
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=5.954792
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=6.049333
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=6.405250
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=6.449458
• tenferro-fft-executor-cached: FftExecutor reused across warmups and measured runs; operations_per_sample=1; median_batch_ms=6.621333
• tenferro-fft-immediate: API includes per-call planning or session entry; use shared cached executor for short operation timing
• tenferro-fft-read: API includes per-call planning or session entry; use shared cached executor for short operation timing
• tenferro-fft-trace: API includes per-call planning or session entry; use shared cached executor for short operation timing
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=3.397334
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=3.420875
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=3.865500
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=4.123750
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=4.125375
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=4.177583
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=4.207083
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=4.310750
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=5.733667
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=5.773292
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=5.801834
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=5.861250
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=5.951125
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=6.177333
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=6.237875
• tenferro-fft-trace: TracedTensorFftExt graph compiled once; compiled program reused across runs; operations_per_sample=1; median_batch_ms=6.810583