Lib

pub const engine = @import("engine.zig")

Top-level runtime handle and tier-selection helpers.

Other zinc_rt submodules go through this surface to obtain the active execution tier (CPU reference, T1 PM4, T2 UMQ, Metal, etc.).

pub const ir_op = @import("ir/op.zig")

IR opcode definitions used by the M0/M1 decode graph.

Op kinds are stable identifiers shared by the emitter and the validator.

pub const ir_graph = @import("ir/graph.zig")

IR graph container — nodes plus their op/argument metadata.

`forward_zinc_rt` emits one of these per token before lowering to the tier the engine selected.

pub const kmd = @import("kmd.zig")

Kernel-mode-driver glue.

Houses helpers and constants shared by the ring backends (UMQ user queues, KFD compute queues) so they describe submissions in one place.

pub const ring = @import("ring/mod.zig")

Common ring-submission surface — what every concrete ring backend (CPU, UMQ, KFD, CS) implements so the engine can speak to them uniformly.

pub const cpu_ring = @import("ring/cpu.zig")

Reference CPU ring used by the scalar M1 forward path.

Executes ops synchronously on the host so we have a ground-truth correctness oracle for the GPU rings.

pub const umq = @import("ring/umq.zig")

T2 user-mode queue ring backed by `DRM_IOCTL_AMDGPU_USERQ`.

The "blessed" direct path when amdgpu firmware admits compute user queues.

pub const kfd = @import("ring/kfd.zig")

T1 PM4 ring backed by `/dev/kfd` (the ROCm/tinygrad ABI).

Fallback when UMQ is rejected; talks PM4 packets to the compute scheduler directly.

pub const cs = @import("ring/cs.zig")

Generic libdrm compute-stream submission used by the PM4 backends.

Wraps context create/submit so the ring backends share one BO/syncobj path.

pub const pm4_packet = @import("ring/packet.zig")

PM4 packet builders (NOPs, indirect-buffer dispatch, COPY_DATA, fences).

Consumed by both the UMQ and KFD rings to produce wire-compatible command-buffer bytes.

pub const kernels = @import("isa/cpu_zig/mod.zig")

Scalar CPU kernels (dequantization, matvec, softmax, etc.) used by the reference forward path and by GPU-tier correctness checks.

pub const fast_pool = @import("fast_pool.zig")

Tiny fixed-size worker pool used to fan out per-layer / per-expert work in the scalar M1 decode path without paying allocator/scheduler overhead.