- ops-npu-bench: time and energy of the corpus and of an int_synth graph on the device and on the CPU, every device result compared with the interpreter's first; the package's energy counter (a vendor counter) and the device runtime's own estimate (a model), each named as what it is. - wai-ops-npu.yml: on the linux-amd runner, the probes, the corpus, the twelve kernel edits and the width test on the device, with the device declared and required; on linux-arm64, the crate builds and an absent device is SKIPPED without WAI_OPS_NPU_REQUIRE and a failure with it. - The corpus test requires the vectors it runs to be exactly those npu.jsonl lists in the subset. Co-Authored-By: Claude Opus 5.5 <[email protected]>
34 lines
1.1 KiB
TOML
34 lines
1.1 KiB
TOML
# `wai-ops/1-npu` on neural processors: the integer operator set's arithmetic
|
|
# operators as kernels for one compute core of a tile array, built at run time
|
|
# with the array's open toolchain and run through its device runtime, exact for
|
|
# every graph of the accelerator subset (extensions/integer-operators.md §9.2,
|
|
# §9.5).
|
|
#
|
|
# A SEPARATE crate with its own lockfile and no dependency but `wai-ops`, so
|
|
# that `wai-ops` stays `no_std` and `wai-rs` never sees a device runtime. The
|
|
# device is found, and the program and host side are built, only on Linux with
|
|
# the device present; everywhere else the crate builds and its device tests
|
|
# say SKIPPED.
|
|
[package]
|
|
name = "wai-ops-npu"
|
|
version = "0.1.0"
|
|
edition = "2024"
|
|
rust-version = "1.97"
|
|
license = "Apache-2.0"
|
|
publish = false
|
|
description = "WAI integer operator set wai-ops/1-npu on neural processors: tile-array kernels exact for the accelerator subset"
|
|
|
|
[lib]
|
|
name = "wai_ops_npu"
|
|
path = "src/lib.rs"
|
|
|
|
[dependencies]
|
|
wai-ops = { path = "../wai-ops" }
|
|
|
|
# Time and energy of graphs on the device and on the CPU (src/bin/bench.rs).
|
|
[[bin]]
|
|
name = "ops-npu-bench"
|
|
path = "src/bin/bench.rs"
|
|
|
|
[profile.release]
|
|
debug = false
|