search

openxla / xla

A machine learning compiler for GPUs, CPUs, and ML accelerators
Apache License 2.0
2.39k stars 356 forks source link

[xla:ffi] NFC: Use absl::InlinedVector to store dimensions #14261

Closed copybara-service[bot] closed 3 days ago

copybara-service[bot] commented 3 days ago

[xla:ffi] NFC: Use absl::InlinedVector to store dimensions

name old cpu/op new cpu/op delta BM_AddBufferArg/1 161ns ± 3% 145ns ± 2% -9.80% (p=0.000 n=20+19) BM_AddBufferArg/2 201ns ± 3% 176ns ± 2% -12.12% (p=0.000 n=20+20) BM_AddBufferArg/4 276ns ± 3% 237ns ± 2% -13.91% (p=0.000 n=20+20) BM_AddBufferArg/8 431ns ± 3% 359ns ± 2% -16.79% (p=0.000 n=20+20) BM_AddBufferArg/16 756ns ± 3% 617ns ± 3% -18.33% (p=0.000 n=20+20) BM_AddAttributes/1 274ns ± 3% 274ns ± 2% ~ (p=0.989 n=20+19) BM_AddAttributes/2 391ns ± 5% 391ns ± 5% ~ (p=0.925 n=20+20) BM_AddAttributes/4 629ns ± 6% 621ns ± 5% ~ (p=0.194 n=20+19) BM_AddAttributes/8 1.17µs ± 3% 1.17µs ± 5% ~ (p=0.445 n=20+20) BM_AddAttributes/16 2.43µs ± 4% 2.46µs ± 3% +1.06% (p=0.028 n=20+20) BM_UpdateCallFrame/1 91.7ns ± 4% 82.9ns ± 2% -9.62% (p=0.000 n=20+19) BM_UpdateCallFrame/2 102ns ± 4% 88ns ± 3% -14.33% (p=0.000 n=20+19) BM_UpdateCallFrame/4 124ns ± 3% 98ns ± 4% -20.81% (p=0.000 n=20+19) BM_UpdateCallFrame/8 164ns ± 3% 116ns ± 3% -29.33% (p=0.000 n=19+19) BM_UpdateCallFrame/16 279ns ± 5% 199ns ±23% -28.42% (p=0.000 n=19+20)