aprender-compute 0.31.2

High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading (was: trueno)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216

# NOTE: Workspace declaration removed — this crate is now a member of
# the root paiml/aprender workspace (APR-MONO consolidation).
# Sub-crates (aprender-gpu, aprender-quant, etc.) are also root workspace members.
# Old workspace.lints.clippy entries removed — root workspace handles linting.

[package]
name = "aprender-compute"
version.workspace = true
edition.workspace = true
rust-version = "1.89"
authors.workspace = true
license.workspace = true
description = "High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading (was: trueno)"
repository.workspace = true
readme = "README.md"
keywords = ["simd", "gpu", "wasm", "performance", "vectorization"]
categories = ["algorithms", "mathematics", "science"]
exclude = ["target/", ".profraw", ".profdata", ".vscode/", ".idea/", ".pmat", "proptest-regressions"]

[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--generate-link-to-definition"]

[lints.rust]
unsafe_code = "allow"


[dependencies]
thiserror = "2.0.17"
serde = { version = "1.0", features = ["derive"] }  # Hardware capability serialization (PMAT-447)
serde_json = "1.0"  # ML tuner data serialization (TUNER-010)
rayon = { version = "1.11", optional = true }
wgpu = { version = "27.0", optional = true }
pollster = { version = "0.4", optional = true }
bytemuck = { version = "1.24", features = ["derive"], optional = true }
futures-intrusive = { version = "0.5", optional = true }
# WASM async support for GPU
wasm-bindgen-futures = { version = "0.4", optional = true }
wasm-bindgen = { version = "0.2", optional = true }
# Native CUDA monitoring via trueno-gpu (TRUENO-SPEC-010)
trueno-gpu = { version = "0.31.2", path = "../aprender-gpu", optional = true, package = "aprender-gpu" }
# PMAT-336: Provable contracts compile-time enforcement
provable-contracts-macros = "0.3"
# P1a: Sovereign GEMM microkernel codegen (compile-time shape specialization)
trueno-gemm-codegen = { version = "0.31.2", path = "../aprender-gemm-codegen", package = "aprender-gemm-codegen" }
# Tracing/profiling support (renacer integration)
tracing = { version = "0.1", optional = true }
tracing-subscriber = { version = "0.3", features = ["env-filter", "fmt"], optional = true }
tracing-appender = { version = "0.2", optional = true }
dirs = { version = "5.0", optional = true }  # Hardware capability path detection (PMAT-447)
# TUI monitoring (TRUENO-SPEC-020) — presentar-terminal
crossterm = { version = "0.28", optional = true }
presentar-core = { version = "0.3", optional = true }
# Stress test reporting types compatible with renacer (TRUENO-SPEC-025)
# Note: renacer 0.9.1 has compilation issue; using compatible local types instead
dhat = { version = "0.3", optional = true }
trueno-quant = { version = "0.31.2", path = "../aprender-quant", package = "aprender-quant" }
num_cpus = "1.17.0"
anyhow = "1.0.100"
# Hardware capability detection (PMAT-447)
chrono = "0.4"
toml = "0.8"
# ML tuner integration (SHOWCASE-BRICK-001, Section 12)
# Uses aprender 0.24.0 RandomForest for learned optimization
aprender = { path = "../aprender-core", version = "0.31.2", package = "aprender-core", optional = true, default-features = false }
# Execution path graph (PAR-201, Section E.7)
trueno-graph = { version = "0.1.17", optional = true }
# TUI visualization for execution graphs (PAR-201)
presentar-terminal = { version = "0.3", optional = true }

[target.'cfg(target_arch = "wasm32")'.dependencies]
# Web-sys for WebGPU (auto-enabled when gpu-wasm feature is on)
web-sys = { version = "0.3", optional = true, features = ["console"] }

[target.'cfg(not(target_arch = "wasm32"))'.dependencies]
# Hostname detection only on native (not available on WASM)
hostname = "0.4"

[build-dependencies]
serde = { version = "1", features = ["derive"] }
serde_yaml_ng = "0.10"

[dev-dependencies]
proptest = "1.9"
criterion = { workspace = true }
tokio = { version = "1", features = ["macros", "rt"] }
serde_json = "1.0"
nalgebra = "0.34"  # For eigendecomposition benchmark comparison
# Simulation testing framework (TRUENO-SPEC-012)
simular = "0.2.0"
regex = "1.11"  # For PTX/WGSL pattern validation in falsification tests
# GPU edge-case test framework (TCE-001)
trueno-cuda-edge = { version = "0.31.2", path = "../aprender-cuda-edge", package = "aprender-cuda-edge" }
# Sparse and solver crates for contract tests
trueno-sparse = { version = "0.31.2", path = "../aprender-sparse", package = "aprender-sparse" }
trueno-solve = { version = "0.31.2", path = "../aprender-solve", package = "aprender-solve" }
ndarray = "0.17.2"
faer = "0.24"  # For GEMM benchmark comparison (fastest pure-Rust BLAS)
matrixmultiply = "0.3"  # Direct matrixmultiply crate benchmark (engine behind ndarray)

[features]
default = []
parallel = ["rayon"]
gpu = ["wgpu", "pollster", "bytemuck", "futures-intrusive"]
# GPU for WASM (WebGPU) - uses wasm-bindgen-futures instead of pollster
gpu-wasm = ["wgpu", "bytemuck", "futures-intrusive", "wasm-bindgen-futures", "wasm-bindgen", "web-sys"]
# Native CUDA monitoring via trueno-gpu (TRUENO-SPEC-010)
# Provides accurate device info (e.g., "NVIDIA GeForce RTX 4090") and real-time memory metrics
cuda-monitor = ["trueno-gpu/cuda"]
# Chaos engineering features (from renacer v0.4.1)
chaos-basic = []
chaos-network = ["chaos-basic"]
chaos-byzantine = ["chaos-basic"]
chaos-full = ["chaos-network", "chaos-byzantine"]
# Tracing/profiling (renacer integration)
tracing = ["dep:tracing"]
# Hardware capability detection with home directory support (PMAT-447)
hardware-detect = ["dep:dirs"]
# TUI monitoring (TRUENO-SPEC-020) with stress reporting and logging
tui-monitor = ["presentar-core", "presentar-terminal", "crossterm", "trueno-gpu", "dep:tracing", "tracing-subscriber", "tracing-appender", "dep:dirs"]
# CUDA support for TUI (real hardware monitoring)
cuda = ["trueno-gpu/cuda"]
# ML-based kernel selection and throughput prediction (SHOWCASE-BRICK-001)
# Uses aprender RandomForest for learned optimization
ml-tuner = ["aprender"]
# Execution path graph for profiling (PAR-201)
# Tracks PTX→kernel→brick call relationships, exports to CsrGraph
execution-graph = ["trueno-graph"]
# TUI visualization for execution graphs (PAR-201)
# Uses presentar-terminal Tree widget for hierarchical display
presentar-tui = ["presentar-terminal"]
# Heap profiling via dhat-rs (DHAT viewer: https://nnethercote.github.io/dh_view/dh_view.html)
dhat-heap = ["dep:dhat"]

[profile.release]
opt-level = 3
lto = true
codegen-units = 1
debug = true  # Enable debug symbols for profiling (flamegraph, perf)

[lib]
name = "trueno"
path = "src/lib.rs"

[[bench]]
name = "vector_ops"
harness = false

[[bench]]
name = "matrix_ops"
harness = false

[[bench]]
name = "gpu_ops"
path = "benches/gpu_ops/main.rs"
harness = false
required-features = ["gpu"]

[[bench]]
name = "gpu_reduction"
harness = false
required-features = ["gpu"]

[[bench]]
name = "async_gpu_ops"
harness = false
required-features = ["gpu"]

[[bench]]
name = "eigen_ops"
harness = false

[[bench]]
name = "tiling_ops"
harness = false

[[bench]]
name = "gemm_comparison"
harness = false

[[example]]
name = "gpu_batch_demo"
required-features = ["gpu"]

[[example]]
name = "gpu_tiled_reduction"
required-features = ["gpu"]

[[example]]
name = "wgpu_backward_demo"
required-features = ["gpu"]

[[example]]
name = "coop_gemm_bench"
required-features = ["gpu"]

[[example]]
name = "gpu_monitor_demo"
required-features = ["gpu", "tui-monitor"]

[package.metadata.release]
shared-version = true

[[package.metadata.release.pre-release-replacements]]
file = "CHANGELOG.md"
search = "## \\[Unreleased\\]"
replace = "## [{{version}}] - {{date}}"

# [workspace.dependencies] removed — root workspace handles shared deps

[profile.test]
opt-level = 1              # Slight optimization for faster test execution
incremental = true         # Reuse previous builds

[profile.dev]
panic = "abort"