mirror of
https://github.com/morluto/rea
synced 2026-10-26 06:16:17 -04:00
morluto
5f702dd6cc
Normalize parsed facts at their owning boundaries across PE exports and certificates, ELF section names, GPU bundles, IL2CPP generic types, dyld slice selection, source maps, environment replay, and JavaScript mapped arguments. Preserve partial facts and explicit unsupported cases; clarify the corresponding documentation. Validation: 831 affected tests passed with one existing skip across 23 files; build, typecheck/lint, documentation, formatting, and 2133-file test discovery passed. Format controls are synthetic; real Unity, GPU, and ARM-device execution is unverified. Fixes #6709 Fixes #6688 Fixes #6719 Fixes #6718 Fixes #6704 Fixes #6703 Fixes #6692 Fixes #6723 Fixes #6687 Fixes #6726 Fixes #6696 Fixes #6691 Fixes #6727
88 lines
3.3 KiB
TypeScript
88 lines
3.3 KiB
TypeScript
import type { GpuObject } from "../domain/gpu/gpuCodeObjects.js";
|
|
import { assertArtifactReadActive } from "../artifacts/ArtifactFileOperation.js";
|
|
import {
|
|
gpuCount,
|
|
nextGpuContainerOffset,
|
|
rejectGpu,
|
|
type GpuSource,
|
|
} from "./GpuSource.js";
|
|
import { gpuObject } from "./GpuObject.js";
|
|
export async function* readCudaFatbinary(
|
|
source: GpuSource,
|
|
signal?: AbortSignal,
|
|
): AsyncGenerator<GpuObject> {
|
|
let base = 0;
|
|
while (base < source.size) {
|
|
const header = await source.read(base, 16);
|
|
if (header.readUInt32LE(0) !== 0xba55ed50)
|
|
rejectGpu("Unexpected bytes between CUDA fatbinary records");
|
|
if (header.readUInt16LE(4) !== 1 || header.readUInt16LE(6) < 16)
|
|
rejectGpu("Unsupported CUDA fatbinary header");
|
|
const start = base + header.readUInt16LE(6),
|
|
length = gpuCount(header.readBigUInt64LE(8));
|
|
if (start > source.size || length > source.size - start)
|
|
rejectGpu("CUDA fatbinary extent is truncated");
|
|
const end = start + length;
|
|
for (let position = start; position < end;) {
|
|
assertArtifactReadActive(signal);
|
|
if (end - position < 64) rejectGpu("CUDA entry header is truncated");
|
|
const h = await source.read(position, 64);
|
|
const kind = h.readUInt16LE(0),
|
|
version = h.readUInt16LE(2),
|
|
width = h.readUInt32LE(4);
|
|
const padded = gpuCount(h.readBigUInt64LE(8)),
|
|
encoded = h.readUInt32LE(16),
|
|
flags = h.readBigUInt64LE(40);
|
|
if (
|
|
width < 64 ||
|
|
width > end - position ||
|
|
padded > end - position - width ||
|
|
encoded > padded
|
|
)
|
|
rejectGpu("CUDA entry extent is invalid");
|
|
const compressed = (flags & 0xb000n) !== 0n;
|
|
const row = gpuObject(source, {
|
|
offset: position + width,
|
|
size: compressed ? encoded : padded,
|
|
kind: kind === 1 ? "ptx" : kind === 2 ? "cubin" : "unknown",
|
|
metadata: { offset: position, size: width },
|
|
});
|
|
const architectureSuffix =
|
|
(flags & 0x100000n) !== 0n
|
|
? "a"
|
|
: (flags & 0x200000n) !== 0n
|
|
? "f"
|
|
: "";
|
|
row.target = `sm_${h.readUInt32LE(28)}${architectureSuffix}`;
|
|
row.raw_flags = `0x${flags.toString(16)}`;
|
|
row.compression =
|
|
(flags & 0x8000n) !== 0n && (flags & 0x2000n) === 0n
|
|
? "zstd"
|
|
: (flags & 0x2000n) !== 0n && (flags & 0x8000n) === 0n
|
|
? "lz4"
|
|
: compressed
|
|
? "unsupported"
|
|
: "none";
|
|
row.compression_id = compressed ? `cuda-flags:${row.raw_flags}` : null;
|
|
row.decoded_size = compressed ? gpuCount(h.readBigUInt64LE(56)) : padded;
|
|
if (compressed && (encoded === 0 || row.decoded_size === 0))
|
|
rejectGpu("Compressed CUDA entry has an empty declared extent");
|
|
if (version !== 0x101) {
|
|
row.compression = "unsupported";
|
|
row.target = null;
|
|
row.decoded_size = null;
|
|
row.limitations.push(
|
|
`CUDA entry version 0x${version.toString(16)} is not decoded`,
|
|
);
|
|
}
|
|
if (row.compression === "unsupported")
|
|
row.limitations.push(
|
|
"This CUDA compression/version is not supported; original encoded bytes and target declarations remain located",
|
|
);
|
|
yield row;
|
|
position += width + padded;
|
|
}
|
|
base = await nextGpuContainerOffset(source, end);
|
|
}
|
|
}
|