mirror of
https://github.com/luxfi/zap.git
synced 2026-10-30 15:04:16 +00:00
zeekay
d14b104790
Three orthogonal additions, one wire format. v2 package — generic ad-hoc schema dispatch (~/work/lux/zap/v2): - README + IMPOSSIBILITY + BENCH_RESULTS document the v1↔v2 split. - builder/field/iter/list/pool/registry/schema/view + codegen package emit v1-equivalent fast paths from declarative .zap sources. - _compile_fail_test/ + examples/ cover negative and positive cases. - v2_test.go exercises the generic dispatch path. - doc.go in v1 marks v1 deprecated for *new* schema authoring; existing v1 buffers stay parseable by both v1 and v2. bufpool — pooled read slabs (bufpool.go + bufpool_test.go): - Message gains optional *bufRef; Release() returns the slab to its pool when sourced from the pooled read path, GC-managed otherwise. - ParseHeader returns (data, rootOffset) without allocating *Message, used by zapv2 to avoid two allocations per parse. - bench_read_test.go + node_quic_stream_test.go cover both paths. Transport allocator + UMA/RDMA cgo (transport/*): - Transport.Caps()/Buffer interface: capability + slab-typed buffer contract. Bytes() vs DevicePtr() decoupled so UMA and GPUDirect can share the same call surface. - DPDK PMD + GPUDirect RDMA + CUDA UMA real cgo backends gated behind build tags; stub paths remain default on darwin / non-Linux. - transport_test.go covers Caps reporting + Buffer lifecycle on the default transport. LLM.md + go.mod follow the additions.
42 lines
1.4 KiB
C
42 lines
1.4 KiB
C
// Copyright (C) 2025-2026, Lux Industries Inc. All rights reserved.
|
|
// See the file LICENSE for licensing terms.
|
|
//
|
|
// Host-side C API for the cudaMallocManaged-backed UMA pool. Headers are
|
|
// kept ASCII-clean so non-cgo builds can still include this file for
|
|
// static analysis.
|
|
#ifndef LUX_ZAP_TRANSPORT_UMA_CUDA_H
|
|
#define LUX_ZAP_TRANSPORT_UMA_CUDA_H
|
|
|
|
#include <stddef.h>
|
|
#include <stdint.h>
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
// luxzap_cuda_probe returns 0 if at least one CUDA device is usable on
|
|
// this host, otherwise a non-zero CUDA error code. Side-effect free
|
|
// beyond a single cudaGetDeviceCount call — safe to call before any
|
|
// other cudaMalloc*.
|
|
int luxzap_cuda_probe(void);
|
|
|
|
// luxzap_cuda_alloc_managed allocates `size` bytes via cudaMallocManaged
|
|
// with cudaMemAttachGlobal. Returns 0 on success; *out_ptr receives the
|
|
// unified pointer. Returns the cudaError_t on failure with *out_ptr=NULL.
|
|
int luxzap_cuda_alloc_managed(size_t size, void** out_ptr);
|
|
|
|
// luxzap_cuda_free releases memory previously returned by
|
|
// luxzap_cuda_alloc_managed. Safe to call with NULL.
|
|
int luxzap_cuda_free(void* ptr);
|
|
|
|
// luxzap_cuda_device_name fills `buf` with up to `buflen-1` chars of the
|
|
// device-0 name and NUL-terminates. Returns the number of bytes written
|
|
// (excluding the NUL), or 0 on error. Used only for the boot log line.
|
|
size_t luxzap_cuda_device_name(char* buf, size_t buflen);
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|
|
|
|
#endif // LUX_ZAP_TRANSPORT_UMA_CUDA_H
|