date: 2026-09-15T09:38:00Z os: Darwin 25.2.0 arm64 macos: 26.2 (25C56) cpu: Apple M4 Pro logical cpus: 14 performance cores: 10 efficiency cores: 4 smt: none P-core L1d: 131072 B L2 (shared per cluster of 5): 16777216 B E-core L1d: 65536 B L2 (shared per cluster of 4): 4194304 B cache line: 128 B page: 16384 B memory: 24 GiB isa features: CRC32 FlagM FlagM2 FHM DotProd SHA3 RDM LSE SHA256 SHA512 SHA1 AES PMULL SB FRINTTS PACIMP LRCPC LRCPC2 FCMA JSCVT PAuth PAuth2 FPAC FPACCOMBINE DPB DPB2 BF16 I8MM WFxT RPRES ECV AFP LSE2 CSV2 CSV3 DIT FP16 BTI SME SME2 SME_F64F64 SME_I16I64 power: 'AC Power' load average at start: 7.02 7.46 7.78 frequency: not published by the vendor for this part; see estimated clock below compiler: Apple clang version 17.0.0 (clang-1700.4.4.1) estimated clock: 4.50 GHz (dependent 1-cycle add chain, 400000000 adds, min of 7 runs) bench: QUICK=0 REPS=default CLOCK_GHZ=4.50 --- records 4194304 record bytes 64 aos bytes 268435456 soa field bytes 16777216 soa fields 16 reps 31 (plus 1 warmup, discarded) evict sweep 134217728 bytes clock 4.50 GHz reference (double, from the struct copy): sum3 2096667.738 dot37 1048326.786 tolerance 1e-02 relative RESULT records 4194304 count RESULT aos_bytes 268435456 bytes RESULT soa_field_bytes 16777216 bytes RESULT evict_bytes 134217728 bytes RESULT reps 31 count RESULT sum3_reference 2096667.738 sum RESULT dot37_reference 1048326.786 sum RESULT tolerance 1e-02 relative RESULT clock 4.50 GHz cycles: each pass is converted with the clock sampled just before it (min of three dependent integer add chains of 2000000 adds), then the median of the conversions is taken sum3 aos_scalar min 0.650 median 0.669 max 0.918 cv 6.6% n 31 ns/rec same, in cycles (per-pass clock) min 2.933 median 3.017 max 4.142 cv 6.6% n 31 cycles/rec clock before each pass: min 4.417 median 4.513 max 4.513 GHz RESULT sum3_aos_scalar_bytes 64 B/rec RESULT sum3_aos_scalar_median 0.6685 ns/rec RESULT sum3_aos_scalar_min 0.6501 ns/rec RESULT sum3_aos_scalar_cv 6.61 percent RESULT sum3_aos_scalar_gbps 95.73 GB/s RESULT sum3_aos_scalar_cycles 3.017 cycles/rec RESULT sum3_aos_scalar_ghz 4.513 GHz RESULT sum3_aos_scalar_relerr 5.53e-05 relative sum3 aos_o2 min 0.603 median 0.625 max 0.647 cv 1.7% n 31 ns/rec same, in cycles (per-pass clock) min 2.719 median 2.819 max 2.918 cv 1.7% n 31 cycles/rec clock before each pass: min 4.511 median 4.512 max 4.513 GHz RESULT sum3_aos_o2_bytes 64 B/rec RESULT sum3_aos_o2_median 0.6247 ns/rec RESULT sum3_aos_o2_min 0.6025 ns/rec RESULT sum3_aos_o2_cv 1.69 percent RESULT sum3_aos_o2_gbps 102.44 GB/s RESULT sum3_aos_o2_cycles 2.819 cycles/rec RESULT sum3_aos_o2_ghz 4.512 GHz RESULT sum3_aos_o2_relerr 5.91e-07 relative sum3 soa_scalar min 0.554 median 0.555 max 0.592 cv 1.3% n 31 ns/rec same, in cycles (per-pass clock) min 2.475 median 2.505 max 2.670 cv 1.3% n 31 cycles/rec clock before each pass: min 4.465 median 4.512 max 4.513 GHz RESULT sum3_soa_scalar_bytes 4 B/rec RESULT sum3_soa_scalar_median 0.5552 ns/rec RESULT sum3_soa_scalar_min 0.5543 ns/rec RESULT sum3_soa_scalar_cv 1.26 percent RESULT sum3_soa_scalar_gbps 7.21 GB/s RESULT sum3_soa_scalar_cycles 2.505 cycles/rec RESULT sum3_soa_scalar_ghz 4.512 GHz RESULT sum3_soa_scalar_relerr 5.53e-05 relative sum3 soa_strict min 0.474 median 0.474 max 0.522 cv 1.9% n 31 ns/rec same, in cycles (per-pass clock) min 2.118 median 2.138 max 2.356 cv 1.9% n 31 cycles/rec clock before each pass: min 4.416 median 4.513 max 4.514 GHz RESULT sum3_soa_strict_bytes 4 B/rec RESULT sum3_soa_strict_median 0.4737 ns/rec RESULT sum3_soa_strict_min 0.4735 ns/rec RESULT sum3_soa_strict_cv 1.86 percent RESULT sum3_soa_strict_gbps 8.44 GB/s RESULT sum3_soa_strict_cycles 2.138 cycles/rec RESULT sum3_soa_strict_ghz 4.513 GHz RESULT sum3_soa_strict_relerr 5.53e-05 relative sum3 soa_o2 min 0.042 median 0.044 max 0.047 cv 2.1% n 31 ns/rec same, in cycles (per-pass clock) min 0.190 median 0.198 max 0.209 cv 1.8% n 31 cycles/rec clock before each pass: min 4.416 median 4.513 max 4.513 GHz RESULT sum3_soa_o2_bytes 4 B/rec RESULT sum3_soa_o2_median 0.0439 ns/rec RESULT sum3_soa_o2_min 0.0422 ns/rec RESULT sum3_soa_o2_cv 2.06 percent RESULT sum3_soa_o2_gbps 91.10 GB/s RESULT sum3_soa_o2_cycles 0.198 cycles/rec RESULT sum3_soa_o2_ghz 4.513 GHz RESULT sum3_soa_o2_relerr 5.91e-07 relative sum3 soa_neon min 0.045 median 0.046 max 0.054 cv 4.4% n 31 ns/rec same, in cycles (per-pass clock) min 0.201 median 0.206 max 0.242 cv 4.3% n 31 cycles/rec clock before each pass: min 4.417 median 4.513 max 4.514 GHz RESULT sum3_soa_neon_bytes 4 B/rec RESULT sum3_soa_neon_median 0.0456 ns/rec RESULT sum3_soa_neon_min 0.0445 ns/rec RESULT sum3_soa_neon_cv 4.44 percent RESULT sum3_soa_neon_gbps 87.72 GB/s RESULT sum3_soa_neon_cycles 0.206 cycles/rec RESULT sum3_soa_neon_ghz 4.513 GHz RESULT sum3_soa_neon_relerr 5.91e-07 relative dot37 aos_scalar min 0.778 median 0.787 max 1.162 cv 8.3% n 31 ns/rec same, in cycles (per-pass clock) min 3.200 median 3.547 max 5.243 cv 8.6% n 31 cycles/rec clock before each pass: min 3.924 median 4.512 max 4.514 GHz RESULT dot37_aos_scalar_bytes 64 B/rec RESULT dot37_aos_scalar_median 0.7871 ns/rec RESULT dot37_aos_scalar_min 0.7776 ns/rec RESULT dot37_aos_scalar_cv 8.34 percent RESULT dot37_aos_scalar_gbps 81.31 GB/s RESULT dot37_aos_scalar_cycles 3.547 cycles/rec RESULT dot37_aos_scalar_ghz 4.512 GHz RESULT dot37_aos_scalar_relerr 1.90e-03 relative dot37 aos_o2 min 0.639 median 0.658 max 0.752 cv 3.1% n 31 ns/rec same, in cycles (per-pass clock) min 2.883 median 2.969 max 3.392 cv 3.1% n 31 cycles/rec clock before each pass: min 4.415 median 4.513 max 4.513 GHz RESULT dot37_aos_o2_bytes 64 B/rec RESULT dot37_aos_o2_median 0.6583 ns/rec RESULT dot37_aos_o2_min 0.6388 ns/rec RESULT dot37_aos_o2_cv 3.10 percent RESULT dot37_aos_o2_gbps 97.23 GB/s RESULT dot37_aos_o2_cycles 2.969 cycles/rec RESULT dot37_aos_o2_ghz 4.513 GHz RESULT dot37_aos_o2_relerr 1.49e-04 relative dot37 soa_scalar min 0.745 median 0.765 max 0.893 cv 3.8% n 31 ns/rec same, in cycles (per-pass clock) min 3.362 median 3.449 max 3.758 cv 2.4% n 31 cycles/rec clock before each pass: min 3.924 median 4.512 max 4.513 GHz RESULT dot37_soa_scalar_bytes 8 B/rec RESULT dot37_soa_scalar_median 0.7647 ns/rec RESULT dot37_soa_scalar_min 0.7452 ns/rec RESULT dot37_soa_scalar_cv 3.79 percent RESULT dot37_soa_scalar_gbps 10.46 GB/s RESULT dot37_soa_scalar_cycles 3.449 cycles/rec RESULT dot37_soa_scalar_ghz 4.512 GHz RESULT dot37_soa_scalar_relerr 1.90e-03 relative dot37 soa_strict min 0.482 median 0.483 max 0.576 cv 3.4% n 31 ns/rec same, in cycles (per-pass clock) min 2.136 median 2.178 max 2.221 cv 0.7% n 31 cycles/rec clock before each pass: min 3.808 median 4.513 max 4.513 GHz RESULT dot37_soa_strict_bytes 8 B/rec RESULT dot37_soa_strict_median 0.4827 ns/rec RESULT dot37_soa_strict_min 0.4825 ns/rec RESULT dot37_soa_strict_cv 3.39 percent RESULT dot37_soa_strict_gbps 16.57 GB/s RESULT dot37_soa_strict_cycles 2.178 cycles/rec RESULT dot37_soa_strict_ghz 4.513 GHz RESULT dot37_soa_strict_relerr 1.90e-03 relative dot37 soa_o2 min 0.061 median 0.062 max 0.070 cv 2.6% n 31 ns/rec same, in cycles (per-pass clock) min 0.277 median 0.279 max 0.317 cv 2.5% n 31 cycles/rec clock before each pass: min 4.417 median 4.513 max 4.513 GHz RESULT dot37_soa_o2_bytes 8 B/rec RESULT dot37_soa_o2_median 0.0619 ns/rec RESULT dot37_soa_o2_min 0.0613 ns/rec RESULT dot37_soa_o2_cv 2.55 percent RESULT dot37_soa_o2_gbps 129.28 GB/s RESULT dot37_soa_o2_cycles 0.279 cycles/rec RESULT dot37_soa_o2_ghz 4.513 GHz RESULT dot37_soa_o2_relerr 9.57e-06 relative dot37 soa_neon min 0.062 median 0.062 max 0.070 cv 2.7% n 31 ns/rec same, in cycles (per-pass clock) min 0.279 median 0.280 max 0.300 cv 1.9% n 31 cycles/rec clock before each pass: min 4.289 median 4.512 max 4.513 GHz RESULT dot37_soa_neon_bytes 8 B/rec RESULT dot37_soa_neon_median 0.0621 ns/rec RESULT dot37_soa_neon_min 0.0617 ns/rec RESULT dot37_soa_neon_cv 2.72 percent RESULT dot37_soa_neon_gbps 128.87 GB/s RESULT dot37_soa_neon_cycles 0.280 cycles/rec RESULT dot37_soa_neon_ghz 4.512 GHz RESULT dot37_soa_neon_relerr 9.57e-06 relative checksums: every variant matched the double reference within 1e-02 relative