tqchen commented on PR #783: URL: https://github.com/apache/tvm-ffi/pull/783#issuecomment-5655299937
## Aarch64 benchmark Common inline i64 operations on **Arm Neoverse-V2 (GB200 CPU), AArch64**, using GCC 13.3.0, C++17, `-O3 -DNDEBUG`. Values are median ns/op from seven samples in one run per mode, with identical fixtures, GCD algorithm and result-consumption barrier; exact expected values were validated before timing. Measured source snapshot: `d239d9be4b5a60740f2bbb38ca7f4cdac054c321`, with a Release runtime rebuilt from that source. These measurements precede the subsequent PR changes and remain qualified to that snapshot. `BigInt(no-builtin)` selects `TVM_FFI_BIGINT_FORCE_PORTABLE` for scalar checked-overflow helpers. Both ratios use the same unrounded Raw i64 control displayed in each row. Decimal precision does not imply measurement accuracy. | Operation | Raw i64 | BigInt | BigInt(no-builtin) | BigInt/raw | No-builtin/raw | |---|---:|---:|---:|---:|---:| | add | 0.405033 | 1.204716 | 0.958622 | 2.974x | 2.367x | | and | 0.405033 | 0.937263 | 0.794315 | 2.314x | 1.961x | | bool | 0.418583 | 0.516972 | 0.516846 | 1.235x | 1.235x | | div | 1.783821 | 2.382092 | 2.383186 | 1.335x | 1.336x | | eq | 0.434940 | 0.944828 | 0.662964 | 2.172x | 1.524x | | floordiv | 1.738411 | 2.222912 | 2.214855 | 1.279x | 1.274x | | floormod | 2.052742 | 2.223396 | 2.231701 | 1.083x | 1.087x | | from_double | 0.400024 | 1.416634 | 1.390755 | 3.541x | 3.477x | | gcd | 28.313606 | 68.613838 | 65.378715 | 2.423x | 2.309x | | ge | 0.419682 | 0.881351 | 0.663208 | 2.100x | 1.580x | | gt | 0.420898 | 0.942631 | 0.663330 | 2.240x | 1.576x | | le | 0.416138 | 0.940922 | 0.662479 | 2.261x | 1.592x | | lt | 0.435059 | 0.942997 | 0.662476 | 2.168x | 1.523x | | max | 0.420532 | 0.959721 | 0.959476 | 2.282x | 2.282x | | min | 0.435551 | 0.967899 | 0.959476 | 2.222x | 2.203x | | mod | 2.053841 | 2.348644 | 2.346935 | 1.144x | 1.143x | | mul | 0.324223 | 0.951786 | 4.926781 | 2.936x | 15.196x | | ne | 0.422363 | 0.662601 | 0.662842 | 1.569x | 1.569x | | neg | 0.364990 | 0.698368 | 0.697147 | 1.913x | 1.910x | | not | 0.364502 | 0.635380 | 0.635620 | 1.743x | 1.744x | | or | 0.405151 | 0.979130 | 0.794315 | 2.417x | 1.961x | | shl | 0.404907 | 2.849987 | 2.861462 | 7.039x | 7.067x | | shr | 0.370117 | 2.215343 | 2.215466 | 5.986x | 5.986x | | sub | 0.405029 | 1.184212 | 1.266605 | 2.924x | 3.127x | | to_double | 0.323730 | 0.324951 | 0.360107 | 1.004x | 1.112x | | truncdiv | 1.746834 | 2.375622 | 2.382946 | 1.360x | 1.364x | | truncmod | 2.053352 | 2.347424 | 2.349865 | 1.143x | 1.144x | | xor | 0.404667 | 0.981205 | 0.794193 | 2.425x | 1.963x | -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
