Benchmark coverage and runbooks#
This page is the operational index for QuantEM.GPU performance work. It keeps required configurations visible even when they have never run, imports every retained atomic measurement from the current evidence index, and maps each open gate to a repository-owned runbook.
The registry separates three concepts:
a coverage gate defines the exact module, platform, computer, geometry, dtype, cache state, timing boundary, parity, and memory evidence required;
a retained measurement records what actually ran, including p50, p95, maximum, resident bytes, measured peaks, swap, device, date, and revision;
a runbook gives a stable command, preflight, required environment, and artifacts for repeating or closing one gate.
Status is not support by implication
A pending row is planned work. A partial row retains useful evidence but does not satisfy the full gate. A portable parity command never becomes physical timing evidence, and a prepared reopen never becomes a cold-source claim.
Agent entry point#
Start by validating the registry, then ask it for the next open gate. Do not guess a command from an old experiment page.
python scripts/benchmark_registry.py validate
python scripts/benchmark_registry.py next --limit 10
python scripts/benchmark_registry.py next --computer "MacBook Air (M2, 8 GB)"
python scripts/benchmark_registry.py next --platform "Python MPS"
python scripts/benchmark_registry.py show io.mps.apple-m5-max-128gb.bin2.cold-original
python scripts/benchmark_registry.py command io.mps.apple-m5-max-128gb.bin2.cold-original
Use --performance-entrypoint-only to hide gates whose current repository
entry point is parity-only:
python scripts/benchmark_registry.py next --performance-entrypoint-only
Before a physical run, create the experiment manifest and RUNS.md row in the
private evidence archive, as described in
Continuous profiling. The command
output intentionally does not launch hardware on its own: device ownership,
fixture identity, cold-source control, and output locations must be resolved
first.
Promotion rule#
A measured row needs all of the following under one comparison key:
the exact source and implementation revisions;
computer, accelerator, runtime, and device ownership;
source and selected scan geometry, detector geometry, bin, crop, and dtype;
explicit cold, warm, prepared, resident, or saved-result state;
run-level records with p50, p95, maximum, and the wall-clock boundary;
logical resident bytes, accelerator allocation/peak, process or browser-tree peak, total-device peak when available, pressure, and swap;
independent scientific parity and output fingerprints; and
a retained manifest, raw artifact, and terminal
RUNS.mdstatus.
If a field is unavailable, the row remains partial or pending. Calculated payloads are useful planning data but never replace measured peaks.
Current qualified load measurements#
Only explicitly designated current rows appear here. Historical, superseded, refuted, and unmeasured configurations remain in the complete tables below rather than being silently deleted.
Platform |
Computer |
State |
Selected scan |
Source detector |
Detector bin |
Output detector |
Source dtype |
Staging dtype |
Resident dtype |
Scientific gate |
Cache/process state |
Wall boundary |
Samples |
p50 |
p95 |
Maximum |
Logical resident |
Accelerator/driver peak |
Process/tree peak |
Process physical-footprint peak |
Swap delta |
Parity |
Device tested |
Date tested |
Revision |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
one same-process warmup; operating-system source pages uncontrolled; no eviction performed; fresh returned destination released after each trial |
public io.load return after MPS synchronization; full-volume hash and release excluded |
7 |
0.406624 s |
0.428164 s |
0.428164 s |
18.000 GiB |
18.442 GiB |
18.692 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
fresh QH5 index root and destination; macOS F_NOCACHE on source hashing and indexed descriptors; immutable source already audited |
catalog, pipeline compilation, plan, complete exact private-resident volume, seven products, metadata, and provenance |
7 |
0.577793 s |
0.900979 s |
0.900979 s |
18.000 GiB |
18.571 GiB |
0.874 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.225984 s |
0.231192 s |
0.231192 s |
4.500 GiB |
5.087 GiB |
0.657 GiB |
5.336 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, audit creation, first allocation, pipeline compilation, application presentation, and post-boundary hashes excluded |
6 |
0.200036 s |
0.204694 s |
0.204694 s |
1.125 GiB |
1.712 GiB |
1.039 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
exact |
original compressed source reread every visit; existing QH5 index and source-bound packing layout; exact DPC sums reused after first load; OS pages uncontrolled |
indexed source open through synchronous complete packed-resident return; catalog and independent full-count audit separate |
7 |
1.496660 s |
2.086119 s |
2.086119 s |
18.000 GiB |
n/a |
n/a |
2.531 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-09-06 |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.678703 s |
0.697179 s |
0.697179 s |
4.500 GiB |
5.087 GiB |
0.662 GiB |
5.271 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.640942 s |
0.651931 s |
0.651931 s |
1.125 GiB |
1.712 GiB |
1.018 GiB |
2.319 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
prepared immutable block indexes; explicitly warm source pages; fresh browser target per retained run; 699 pageouts; zero swap growth; no dropped runs |
navigation through scientifically usable exact resident output and diagnostic frame checksums; exhaustive full-volume hash and application E2E excluded |
7 |
1.358000 s |
1.594000 s |
1.594000 s |
18.000 GiB |
n/a |
6.500 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
prepared indexes; source pages unspecified; CPU exact-reference validation only |
public CPU reference load; full-volume hash excluded |
1 |
32.788039 s |
32.788039 s |
32.788039 s |
18.000 GiB |
n/a |
19.141 GiB |
n/a |
0 B |
Pass |
Apple M5 Max CPU; 128 GB unified memory |
2026-08-22 |
|
Dtype and residency coverage#
Each row is one atomic source, staging, and resident dtype contract. Exact rows preserve scientific counts; browse-only rows are explicit saturating representations and cannot satisfy an exact gate.
Platform |
Computer |
State |
Selected scan |
Source detector |
Scan bin |
Detector bin |
Output detector |
Crop |
Source dtype |
Staging dtype |
Resident dtype |
Scientific gate |
Precision contract |
Implementation basis |
Next gate or reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint8 |
uint8 |
uint8 |
exact |
native integer counts |
src/quantem/gpu/io/hdf5/cuda/kernels/bslz4.cu shuf_8_batched unshuffles one-byte bitshuffle/LZ4 blocks, including the partial final block; tests/hardware/test_bitshuffle_uint8.py (QEM_TEST_BACKEND=cuda) matches h5py for full, partial and multi-block frames. No real uint8 acquisition is retained. |
Load a real full-volume native uint8 bitshuffle/LZ4 acquisition through public CUDA io.load and retain its complete-volume hash, source/resident provenance and peak memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint16 |
exact |
native integer counts |
The CUDA decoder has a dedicated uint16 bitshuffle path and existing unit parity; the current registry has no uncontended full-volume native uint16 distribution. |
Run the public CUDA load entry point on an uncontended device and retain a complete-volume uint16 hash plus source/staging/resident provenance and peak allocation. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint32 |
uint32 |
uint32 |
exact |
native integer counts |
src/quantem/gpu/io/encoded.py holds a uint32 source as exact uint16 codes when every count fits 16 bits and rejects it otherwise; the encoded resident has no uint32 form. |
CUDA io.load has no uint32 resident: uint32 counts that fit 16 bits are held exactly as uint16, larger ones are refused. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint8 |
exact |
source-identity-bound complete value-range audit |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no uint8 output. |
CUDA io.load has no uint8 output dtype. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint8 |
browse-only |
explicit saturation to 255 |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no saturating uint8 browse output. |
CUDA io.load has no saturating uint8 browse output. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint32 |
exact |
lossless integer widening |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no uint32 widening. |
CUDA io.load has no uint16-to-uint32 widening. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint16 |
exact |
source-identity-bound uint8 staging with exact uint16 reconstruction |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no audited uint8 staging. |
Audited uint8 staging with uint16 resident reconstruction is not implemented for CUDA. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint8 |
uint8 |
uint8 |
exact |
native integer counts |
src/quantem/gpu/io/hdf5/mps/kernels/bslz4.msl shuf_8_batched unshuffles one-byte bitshuffle/LZ4 blocks, including the partial final block; tests/hardware/test_bitshuffle_uint8.py (QEM_TEST_BACKEND=mps) matches h5py for full, partial and multi-block frames. No real uint8 acquisition is retained. |
Load a real full-volume native uint8 bitshuffle/LZ4 acquisition through public Python MPS io.load and retain its complete-volume hash, source/resident provenance and peak memory. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint16 |
exact |
native integer counts |
The MPS uint16 decoder and retained full-volume canonical hash preserve native counts at detector bin 1. |
n/a |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint32 |
uint32 |
uint32 |
exact |
native integer counts |
src/quantem/gpu/io/encoded.py holds a uint32 source as exact uint16 codes when every count fits 16 bits and rejects it otherwise; the encoded resident has no uint32 form. |
Python MPS io.load has no uint32 resident: uint32 counts that fit 16 bits are held exactly as uint16, larger ones are refused. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint8 |
exact |
source-identity-bound complete value-range audit |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no uint8 output. |
Python MPS io.load has no uint8 output dtype. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint8 |
browse-only |
explicit saturation to 255 |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no saturating uint8 browse output. |
Python MPS io.load has no saturating uint8 browse output. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint32 |
exact |
lossless integer widening |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no uint32 widening. |
Python MPS io.load has no uint16-to-uint32 widening. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint16 |
exact |
source-identity-bound uint8 staging with exact uint16 reconstruction |
src/quantem/gpu/io/load.py accepts dtype only as native (or scaled_uint16 for float32 sources): io.load keeps stored counts and has no audited uint8 staging. |
Audited uint8 staging with uint16 resident reconstruction is not implemented for Python MPS. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint8 |
uint8 |
uint8 |
exact |
native integer counts |
Native4DSTEMIO can describe one-byte indexed sources, but Metal4DSTEMIndexedLoader.swift requires a uint16 source and uint16 staging at the public integrated boundary. |
The integrated indexed Swift/Metal loader does not admit a native uint8 source. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint16 |
exact |
native integer counts |
Metal4DSTEMIndexedLoader.swift supports the uint16 fallback stage, while retained optimized measurements used separately registered audited uint8 staging. |
Run a source whose complete audit does not authorize uint8 staging and retain exact uint16 staging-to-resident parity and allocation evidence. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint32 |
uint32 |
uint32 |
exact |
native integer counts |
NativeHDF5Bridge.swift describes one- and two-byte indexed sources, and Metal4DSTEMIndexedLoader.swift requires uint16 input. |
The native indexed Swift/Metal source contract does not admit uint32 HDF5 detector values. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint8 |
exact |
source-identity-bound complete value-range audit |
Metal4DSTEMExactBinner.provenance accepts only uint16 or uint32 output, although it can use an audited uint8 staging buffer. |
The exact native Swift/Metal load contract cannot publish a uint8 resident scientific volume. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint8 |
browse-only |
explicit saturation to 255 |
The indexed exact loader exposes uint16 or uint32 scientific output and has no separate saturating browse-resident API. |
Explicit saturating uint8 resident output is not implemented by the public native Swift/Metal load boundary. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint32 |
exact |
lossless integer widening |
Metal4DSTEMExactBinner has tested uint16-to-uint32 kernels and provenance, but Metal4DSTEMIndexedBinnedLoad fixes integrated resident output to uint16. |
Expose uint32 output through the public indexed loader and cache contract, then retain complete load, metadata, and physical parity. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint16 |
exact |
source-identity-bound uint8 staging with exact uint16 reconstruction |
Native4DSTEMValueRangeAudit binds source identity and maximum counts; the retained controlled full-native run used audited uint8 staging and reproduced exact uint16 volume and product hashes. |
n/a |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint8 |
uint8 |
uint8 |
exact |
native integer counts |
h5reader.ts, bslz4.ts, and local-h5.ts recognize matching native uint8 source, decode, and resident modes; no retained full-volume physical uint8 fixture proves the complete path. |
Run an exact physical hardware-browser full-volume uint8 source through local HDF5 decode and retain output hash, allocation, and source/decode/resident provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint16 |
exact |
native integer counts |
The physical WebGPU path retained exact native uint16 resident checks and seven runs, but device allocation and timed complete-volume hashing remain incomplete. |
Add WebGPU device-allocation telemetry and retain an exact complete-volume hash inside the timed boundary before promoting the integrated load gate. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint32 |
uint32 |
uint32 |
exact |
native integer counts |
h5reader.ts, bslz4.ts, and local-h5.ts implement matching native uint32 source/decode/resident modes; physical full-volume proof is absent. |
Run a native uint32 HDF5 source on physical hardware WebGPU and retain the complete resident hash, browser/device memory, and source/decode/resident provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint8 |
exact |
source-identity-bound complete value-range audit |
bslz4.ts has an audit-dependent low8 kernel, but local-h5.ts does not bind a complete audit identity to the returned uint8 resident state. |
Replace the experimental low8 global flag with a typed source-identity-bound audit in the local-HDF5 public result, then retain exact full-volume parity. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint8 |
browse-only |
explicit saturation to 255 |
bslz4.ts decodes all uint16 planes and saturates to 255; source tests distinguish this from the experimental low8 audit path. |
Run the fused full-plane clip8 path on physical hardware with values above 255 and retain output hash, browser/device memory, and browse-only provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint32 |
exact |
lossless integer widening |
local-h5.ts requires decode dtype uint32 to match a uint32 HDF5 source and rejects a uint16 source request. |
The WebGPU local-HDF5 public path does not widen uint16 source values into uint32 resident storage. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint16 |
exact |
source-identity-bound uint8 staging with exact uint16 reconstruction |
local-h5.ts keeps decode and resident integer mode matched for native paths and rejects mismatched uint32 requests; it exposes no uint8-stage-to-uint16 reconstruction contract. |
Audited uint8 staging followed by exact uint16 resident reconstruction is not implemented in the public WebGPU local-HDF5 path. |
CPU reference |
Portable CI runner |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint8 |
uint8 |
uint8 |
exact |
native integer counts |
src/quantem/gpu/io/hdf5/cpu.py returns the HDF5 native dtype unchanged at detector bin 1; a complete public-API uint8 fixture artifact is not retained. |
Add a repository fixture that proves public io.load preserves every native uint8 count, shape, order, and provenance field at detector bin 1. |
CPU reference |
Portable CI runner |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint16 |
exact |
native integer counts |
The CPU reference returns native uint16 counts at detector bin 1 and has retained physical reference probes; the portable registry gate is not a complete public-API artifact. |
Promote only after the portable public API fixture retains complete-volume uint16 hash, shape, order, mask, and metadata parity. |
CPU reference |
Portable CI runner |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint32 |
uint32 |
uint32 |
exact |
native integer counts |
The CPU reference preserves the HDF5 native dtype at detector bin 1; public load defaults may auto narrow uint32, so the exact uint32 request must be explicit and tested. |
Add a public-API native uint32 fixture that disables advisory auto narrowing and proves complete count, dtype, shape, and metadata parity. |
CPU reference |
Portable CI runner |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint8 |
exact |
source-identity-bound complete value-range audit |
src/quantem/gpu/io/load.py refuses every dtype request on the CPU reference, which returns stored counts and has no uint8 output. |
CPU reference io.load has no uint8 output dtype. |
CPU reference |
Portable CI runner |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint8 |
browse-only |
explicit saturation to 255 |
src/quantem/gpu/io/load.py refuses every dtype request on the CPU reference, which returns stored counts and has no saturating uint8 browse output. |
CPU reference io.load has no saturating uint8 browse output. |
CPU reference |
Portable CI runner |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint16 |
uint32 |
exact |
lossless integer widening |
src/quantem/gpu/io/load.py refuses every dtype request on the CPU reference, which returns stored counts and has no uint32 widening. |
CPU reference io.load has no uint16-to-uint32 widening. |
CPU reference |
Portable CI runner |
Not supported |
512 × 512 |
192 × 192 |
1 |
1 |
192 × 192 |
none |
uint16 |
uint8 |
uint16 |
exact |
source-identity-bound uint8 staging with exact uint16 reconstruction |
src/quantem/gpu/io/load.py refuses every dtype request on the CPU reference, which returns stored counts and has no audited uint8 staging. |
Audited uint8 staging with uint16 resident reconstruction is not implemented for CPU reference. |
Detector-product coverage#
Each row fixes one platform, reproducible computer, detector bin, and exact product-suite boundary. Partial diagnostics retain their sample and memory context, but only fully measured gates display p50, p95, and maximum timing.
Platform |
Computer |
State |
Selected scan |
Source detector |
Detector bin |
Output detector |
Resident dtype |
Cache/process state |
Wall boundary |
Samples |
p50 |
p95 |
Maximum |
Logical resident |
Accelerator/driver peak |
Process/tree peak |
Swap delta |
Parity |
Device tested |
Date tested |
Revision |
Next gate or reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
warm operating-system source pages; unspecified; empty screening-result cache |
quantem.gpu.screening.prepare package wall through exact complete six-array screening result in balanced A-B-B-A-B-A-A-B comparison |
6 |
n/a |
n/a |
n/a |
n/a |
2.084 GiB |
1.958 GiB |
0 B |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
Run the current public ScreeningResult on an uncontended CUDA device with benchmark_screening.py –require-exact-full-suite; retain all nine hashes, uint64 total/ABF/ADF dtype and shape, p50/p95/maximum, and complete memory telemetry. |
Python MPS |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
not admissible |
blocked exact-residency contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
The exact 18 GiB resident input exceeds the computer’s 8 GB unified memory before decoder, allocator, products, and operating-system headroom. Next: Implement and prove an exact bounded-streaming product path that never requires an 18 GiB resident input before attempting this physical row. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
exact resident detector-bin-1 MPS input; source pages uncontrolled or warm; no prepared product result; five pageouts; zero swapout growth |
single first-execution diagnostic only; synchronized detector-product publication was 0.499909 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
27.510 GiB |
20.436 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
Repeat the exact product suite without pageout growth for synchronized p50, p95, and maximum timing plus complete memory telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
resident exact full-native MPS input |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run only after an uncontended pressure preflight; use a single exact full-native smoke with a hard stop on swap/pageout growth before considering repeated product timing. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
not admissible |
blocked exact-residency contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
The exact 18 GiB private-resident input exceeds the computer’s 8 GB unified memory before decoder, Metal, products, and operating-system headroom. Next: Implement and prove an exact bounded-streaming or mapped product path that never requires an 18 GiB private-resident input before attempting this physical row. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
resident exact detector-bin-2 Metal input |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run the complete product suite from the explicit exact detector-bin-2 resident volume with full parity, synchronized timing, pressure, swap, and peak memory. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
resident exact detector-bin-4 Metal input |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run the complete exact detector-bin-4 product suite on the current clean revision with full hashes, synchronized timing, pressure, compressed memory, swap, and peak memory; detector bin 2 remains the locked application acceptance policy. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
Not supported |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
not applicable |
unsupported load-plan contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not applicable |
n/a |
n/a |
n/a |
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
prepared immutable QH5 indexes; source pages unspecified |
complete exact resident volume, products, metadata, calibration, and provenance; full private readback/hash excluded |
6 |
0.313870 s |
0.318865 s |
0.318865 s |
18.000 GiB |
18.571 GiB |
0.639 GiB |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.225984 s |
0.231192 s |
0.231192 s |
4.500 GiB |
5.087 GiB |
0.657 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, audit creation, first allocation, pipeline compilation, application presentation, and post-boundary hashes excluded |
6 |
0.200036 s |
0.204694 s |
0.204694 s |
1.125 GiB |
1.712 GiB |
1.039 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
not applicable |
unsupported load-plan contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not applicable |
n/a |
n/a |
n/a |
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
prepared QH5 indexes; F_NOCACHE source descriptors; operating-system source-page state unspecified |
package exact indexed load plus exact full-scan products; app UI, index creation, and full-volume hash excluded |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
18.571 GiB |
0.624 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
Repeat uncontended full-native publication with complete product hashes, p50/p95/max, driver allocation, RSS, pressure, and swap. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.678703 s |
0.697179 s |
0.697179 s |
4.500 GiB |
5.087 GiB |
0.662 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.640942 s |
0.651931 s |
0.651931 s |
1.125 GiB |
1.712 GiB |
1.018 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
Not supported |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
not applicable |
unsupported load-plan contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not applicable |
n/a |
n/a |
n/a |
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
not admissible |
blocked exact-residency contract |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
The exact 18 GiB resident input exceeds the computer’s 8 GB total memory before browser, adapter, decoder, products, and operating-system headroom. Next: Implement and prove an exact bounded-streaming browser product path that never requires an 18 GiB resident input before attempting this physical row. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
prepared QH5 indexes; warm source pages; exact full-native uint16 resident input; fresh browser target per retained run |
complete internal GPU product suite from exact resident input through awaited readbacks; parity hashing, sorting, and encoding excluded |
7 |
0.482000 s |
0.483600 s |
0.483600 s |
18.000 GiB |
n/a |
6.492 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
n/a |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
resident exact full-native hardware-WebGPU input |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run a guarded physical hardware-browser admission smoke, then retain the complete native product suite with full parity, timing, browser/device memory, pressure, and swap if admission remains safe. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
Production detector binning currently produces a float32 resident volume and cannot satisfy the exact-integer input contract. Next: Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
CPU reference |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
resident exact full-native CPU input |
complete reference product suite |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run the independent full-native CPU reference product suite only when host CPU, storage, and memory ownership are uncontended; retain all hashes, timing, and peak RSS. |
CPU reference |
MacBook Air (M2, 8 GB) |
! Blocked |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
not admissible |
blocked reference product suite |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Not run |
n/a |
n/a |
n/a |
The exact 18 GiB resident input exceeds the computer’s 8 GB unified memory before products and operating-system headroom. Next: Implement and prove an exact bounded-streaming CPU reference product path that never requires an 18 GiB resident input before attempting this physical row. |
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
exact full-native CPU-resident input after a warm operating-system page-cache source load |
single independent detector-product traversal was 31.078040 seconds; source load excluded; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
36.450 GiB |
n/a |
Pass |
Apple M5 Max CPU; 128 GB unified memory |
2026-08-19 |
|
Add ABF, retain all product hashes, and repeat the exact physical CPU reference traversal for p50/p95/maximum and complete process-memory telemetry. |
CPU reference |
MacBook Pro (M5, 24 GB) |
○ Pending |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
resident exact full-native CPU input |
complete reference product suite |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Run only after an uncontended memory preflight; retain one exact full-native CPU reference smoke, complete product hashes, pressure, swap, and peak RSS before repetition. |
CPU reference |
Portable CI runner |
◐ Partial |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
resident reference input |
complete reference product suite |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pending |
n/a |
n/a |
n/a |
Keep exact reference tests current; no CPU speed claim is required. |
Coverage summary#
State |
Gate count |
|---|---|
✓ Measured |
18 |
◐ Partial |
34 |
○ Pending |
73 |
! Blocked |
35 |
× Refuted |
4 |
Not supported |
27 |
↺ Superseded |
0 |
Platform |
Tracked gates |
|---|---|
CUDA |
17 |
Python MPS |
37 |
Native Swift/Metal |
56 |
WebGPU |
58 |
CPU reference |
23 |
Platform and computer coverage#
Each row identifies one reproducible hardware configuration. Counts describe tracked cells, including explicit unsupported contracts; a pending value remains a test to run. Load, admission, memory, and performance gates are multiplied across compatible computers because hardware changes the result. Platform-wide correctness or unsupported contracts are recorded once instead of creating misleading duplicate hardware rows.
Platform |
Computer |
Tracked cells |
Measured |
Partial |
Pending |
Blocked |
Refuted |
Unsupported |
|---|---|---|---|---|---|---|---|---|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
17 |
2 |
5 |
5 |
0 |
0 |
5 |
Python MPS |
MacBook Air (M2, 8 GB) |
9 |
0 |
0 |
7 |
2 |
0 |
0 |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
18 |
2 |
3 |
6 |
0 |
2 |
5 |
Python MPS |
MacBook Pro (M5, 24 GB) |
10 |
0 |
1 |
8 |
0 |
1 |
0 |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
14 |
0 |
0 |
11 |
2 |
0 |
1 |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
27 |
8 |
5 |
7 |
0 |
0 |
7 |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
15 |
4 |
3 |
7 |
0 |
0 |
1 |
WebGPU |
MacBook Air (M2, 8 GB) |
15 |
0 |
0 |
4 |
11 |
0 |
0 |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
27 |
2 |
4 |
7 |
9 |
1 |
4 |
WebGPU |
MacBook Pro (M5, 24 GB) |
16 |
0 |
0 |
7 |
9 |
0 |
0 |
CPU reference |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
2 |
0 |
0 |
2 |
0 |
0 |
0 |
CPU reference |
MacBook Air (M2, 8 GB) |
2 |
0 |
0 |
0 |
2 |
0 |
0 |
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
2 |
0 |
2 |
0 |
0 |
0 |
0 |
CPU reference |
MacBook Pro (M5, 24 GB) |
2 |
0 |
0 |
2 |
0 |
0 |
0 |
CPU reference |
Portable CI runner |
15 |
0 |
11 |
0 |
0 |
0 |
4 |
Required coverage gates#
Every row is one exact scientific and device configuration. A pending row is work to do, not an implicit failure and not evidence that a backend is supported on that device.
Platform |
Computer |
State |
Module |
Operation |
Selected scan |
Source detector |
Detector bin |
Output detector |
Source dtype |
Staging dtype |
Resident dtype |
Scientific gate |
Cache/process state |
Wall boundary |
Priority |
Runbook |
Next gate |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
CoM, DPC, and iDPC |
CoM row/column, rotation, and iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident CUDA inputs |
complete synchronized product publication |
2 |
|
Add a stable physical timing entry point and retain complete full-precision parity, synchronized timing, and memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm streamed screening source |
complete exact product suite |
2 |
|
Run the current public ScreeningResult on an uncontended CUDA device with benchmark_screening.py –require-exact-full-suite; retain all nine hashes, uint64 total/ABF/ADF dtype and shape, p50/p95/maximum, and complete memory telemetry. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident CUDA maps |
synchronized display-ready outputs |
3 |
|
Add a stable CUDA display-kernel harness for first execution and warm publication with parity and memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident CUDA maps |
synchronized Fourier output |
3 |
|
Add a stable CUDA FFT harness for first execution and warm publication with parity and memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
I/O and load |
cold arbitrary-source full-native load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold arbitrary source; control not yet qualified |
source open through exact resident output |
1 |
|
Run controlled cold-source discovery/read/decode/load distributions without disturbing unrelated GPU owners. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
I/O and load |
warm-source full-native resident load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm source; historical p50 exists but complete current distribution is missing |
first usable exact resident output |
2 |
|
Re-run the current clean revision for seven trials with p95, maximum, process RSS, total-card peak, and exact hashes. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
I/O and load |
native uint16 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Run the public CUDA load entry point on an uncontended device and retain a complete-volume uint16 hash plus source/staging/resident provenance and peak allocation. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
I/O and load |
native uint8 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint8 |
uint8 |
uint8 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Load a real full-volume native uint8 bitshuffle/LZ4 acquisition through public CUDA io.load and retain its complete-volume hash, source/resident provenance and peak memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
I/O and load |
audited uint8 staging to exact uint16 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
Audited uint8 staging with uint16 resident reconstruction is not implemented for CUDA. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
I/O and load |
native uint32 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint32 |
uint32 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
CUDA io.load has no uint32 resident: uint32 counts that fit 16 bits are held exactly as uint16, larger ones are refused. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
I/O and load |
exact uint16 to uint32 widening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact widened resident output |
4 |
|
CUDA io.load has no uint16-to-uint32 widening. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
I/O and load |
audited lossless uint16 to uint8 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint8 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
CUDA io.load has no uint8 output dtype. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
Not supported |
I/O and load |
explicit saturating uint16 to uint8 browse residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint8 |
browse-only |
source and prepared state declared by the run |
complete saturating browse resident output |
4 |
|
CUDA io.load has no saturating uint8 browse output. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
one-pass exact screening products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm source; empty result cache |
complete six-array screening result build |
2 |
|
n/a |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
prepared immutable screening-result reopen |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable screening-result cache; fresh-process reopen |
complete saved-result reopen; source build excluded |
3 |
|
n/a |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
2 |
|
Add a stable physical calibration harness and retain complete fit timing, parameters, repeatability, parity, and memory. |
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
◐ Partial |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Add a stable CUDA performance entry point and retain physical full-size timing, memory, phase/loss parity, and hashes. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
CoM, DPC, and iDPC |
CoM row/column, rotation, and iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS inputs |
complete synchronized product publication |
2 |
|
Run the physical MPS entry point with complete full-precision parity, synchronized timing, pressure, swap, and peak memory. |
Python MPS |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admissible |
blocked exact-residency contract |
4 |
|
Implement and prove an exact bounded-streaming product path that never requires an 18 GiB resident input before attempting this physical row. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS maps |
synchronized display-ready outputs |
3 |
|
Run first-execution and warm MPS display-kernel distributions with full parity, synchronized publication, pressure, swap, and peak memory. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS maps |
synchronized Fourier output |
3 |
|
Run first-execution and warm MPS FFT distributions with full parity, synchronized publication, pressure, swap, and peak memory. |
Python MPS |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
full-native resident admission |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admitted on the 8 GB resident path |
admission decision before allocation |
1 |
|
Keep this fail-closed until a bounded exact streaming result can preserve the full logical tensor without materializing an 18 GiB resident destination. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
Screening |
one-pass exact screening products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
declared source-page state; empty result cache |
first usable and exact complete screening result build |
1 |
|
Run bounded-staging MPS screening construction with full-source product parity, source-stage timing, pressure, swap, and p50/p95/max. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
Screening |
prepared immutable screening-result reopen |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable screening-result cache |
complete saved-result reopen; source build excluded |
2 |
|
Run a fresh-process prepared-result reopen distribution with complete product hashes, cache bytes, RSS, compressed memory, swap, and p50/p95/max. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
2 |
|
After the bounded object smoke passes, run the frozen fit with deterministic repeatability, fitted-parameter parity, pressure, swap, and memory. |
Python MPS |
MacBook Air (M2, 8 GB) |
○ Pending |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Run one bounded smoke, then the frozen 512 workflow with full phase/loss parity, pressure, swap, memory, and repeat distributions. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
× Refuted |
CoM, DPC, and iDPC |
CoM row/column, rotation, and iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident full-native MPS inputs |
complete synchronized product publication |
1 |
|
Diagnose the full-native iDPC disagreement without weakening the frozen 2e-5 maximum-error gate, then rerun the complete synchronized suite with p50/p95/maximum and memory telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact full-native MPS input |
complete product suite publication |
2 |
|
Repeat the exact product suite without pageout growth for synchronized p50, p95, and maximum timing plus complete memory telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS maps |
synchronized display-ready outputs |
3 |
|
Add a stable MPS display-kernel harness for first execution and warm publication with parity and memory. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
× Refuted |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident full-precision MPS map |
synchronized Fourier output |
1 |
|
Diagnose the Torch-MPS FFT disagreement against the frozen float32 reference without weakening tolerances, then rerun first-execution and warm FFT distributions with complete parity and memory telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
cold original compressed-source first encounter |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold original compressed source; control not yet qualified |
source open through exact complete resident output |
1 |
|
Run seven independently controlled cold-source trials with stage timing, complete parity, and full memory telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
native uint16 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
n/a |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
native uint8 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint8 |
uint8 |
uint8 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Load a real full-volume native uint8 bitshuffle/LZ4 acquisition through public Python MPS io.load and retain its complete-volume hash, source/resident provenance and peak memory. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
post-warmup exact package load with fresh returned destination |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
one same-process warmup; operating-system source pages uncontrolled; fresh returned destination released after each trial |
public io.load return after MPS synchronization; full-volume hash and release excluded |
3 |
|
n/a |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
exact warm package load with explicit destination reuse |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled source pages; explicit resident destination recycled |
public package load return after final Metal completion |
3 |
|
Repeat the recycled-destination lifecycle with full-volume canonical-layout parity and complete release telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
audited uint8 staging to exact uint16 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
Audited uint8 staging with uint16 resident reconstruction is not implemented for Python MPS. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
native uint32 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint32 |
uint32 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
Python MPS io.load has no uint32 resident: uint32 counts that fit 16 bits are held exactly as uint16, larger ones are refused. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
exact uint16 to uint32 widening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact widened resident output |
4 |
|
Python MPS io.load has no uint16-to-uint32 widening. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
audited lossless uint16 to uint8 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint8 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
Python MPS io.load has no uint8 output dtype. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
explicit saturating uint16 to uint8 browse residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint8 |
browse-only |
source and prepared state declared by the run |
complete saturating browse resident output |
4 |
|
Python MPS io.load has no saturating uint8 browse output. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Screening |
one-pass exact screening products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
declared source-page state; empty result cache |
first usable and exact complete screening result build |
2 |
|
Run complete MPS screening construction with full product parity, source-stage timing, memory, and p50/p95/max. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Screening |
prepared immutable screening-result reopen |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable screening-result cache |
complete saved-result reopen; source build excluded |
3 |
|
Run a fresh-process prepared-result reopen distribution with complete product hashes, cache bytes, RSS, and p50/p95/max. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
1 |
|
Run the frozen 200-trial plus Nelder-Mead workflow on physical MPS with deterministic repeatability and full memory telemetry. |
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Run native 512 evidence with the frozen reference engine, full parity, memory, and repeat distributions. |
Python MPS |
MacBook Pro (M5, 24 GB) |
◐ Partial |
CoM, DPC, and iDPC |
CoM row/column, rotation, and iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS inputs |
complete synchronized product publication |
2 |
|
Repeat the complete public DPC pipeline without display contention or pageout growth for synchronized p50, p95, maximum, and complete memory telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact full-native MPS input |
complete product suite publication |
1 |
|
Run only after an uncontended pressure preflight; use a single exact full-native smoke with a hard stop on swap/pageout growth before considering repeated product timing. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS maps |
synchronized display-ready outputs |
3 |
|
Run first-execution and warm transform, histogram, and color distributions with frozen parity and complete memory telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
× Refuted |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident MPS maps |
synchronized Fourier output |
1 |
|
Diagnose the Torch-MPS FFT disagreement against the frozen float32 reference without weakening tolerances, then rerun first-execution and warm FFT distributions with complete parity and memory telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
cold original compressed-source first encounter |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold original compressed source; no prepared result |
source open through exact complete resident output |
1 |
|
After the bounded admission smoke passes, run independently controlled cold-source trials with full-native parity, pressure, paging, and swap telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
post-warmup exact package load with fresh returned destination |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
one same-process warmup; source pages declared by run; fresh destination |
public io.load return after backend synchronization |
2 |
|
Run one bounded admission smoke before seven retained trials; sample driver allocation, process high-water, pressure, and swap without recycling the destination. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Screening |
one-pass exact screening products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
declared source-page state; empty result cache |
first usable and exact complete screening result build |
2 |
|
Run bounded-staging MPS screening construction with full-source product parity, source-stage timing, memory, and p50/p95/max. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Screening |
prepared immutable screening-result reopen |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable screening-result cache |
complete saved-result reopen; source build excluded |
3 |
|
Run a fresh-process prepared-result reopen distribution with complete product hashes, cache bytes, RSS, and p50/p95/max. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
1 |
|
Run the frozen 200-trial plus Nelder-Mead workflow with deterministic repeatability, fitted-parameter parity, and full memory telemetry. |
Python MPS |
MacBook Pro (M5, 24 GB) |
○ Pending |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Run the frozen 512 workflow with full phase/loss parity, memory, and repeat distributions. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
CoM, DPC, and iDPC |
rotation and Fourier iDPC from resident CoM maps |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident float32 CoM row and column maps |
synchronized rotation and iDPC publication |
2 |
|
Run the current native Metal path with frozen full-precision parity, synchronized timing, pressure, swap, and peak memory. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-2 Metal input |
complete product suite publication |
1 |
|
Run the complete product suite from the explicit exact detector-bin-2 resident volume with full parity, synchronized timing, pressure, swap, and peak memory. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-4 Metal input |
complete product suite publication |
2 |
|
Run the complete exact detector-bin-4 product suite on the current clean revision with full hashes, synchronized timing, pressure, compressed memory, swap, and peak memory; detector bin 2 remains the locked application acceptance policy. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admissible |
blocked exact-residency contract |
4 |
|
Implement and prove an exact bounded-streaming or mapped product path that never requires an 18 GiB private-resident input before attempting this physical row. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
Not supported |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported load-plan contract |
4 |
|
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized display-ready outputs |
3 |
|
Run the current Metal display-kernel distribution with full parity, first-versus-warm timing, pressure, swap, and peak memory. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized Fourier output |
3 |
|
Run the current Metal FFT distribution with full parity, first-versus-warm timing, pressure, swap, and peak memory. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
full-native resident admission |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admitted on the 8 GB resident path |
admission decision before allocation |
1 |
|
Keep this fail-closed until an exact bounded residency contract preserves full logical coverage without allocating the 18 GiB destination. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
I/O and load |
cold original compressed-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared-result substitution |
source open through exact complete 4.5 GiB working volume and products |
1 |
|
Run coordinated physical 8 GB trials with full scan, no crop, exact uint16 sums, pressure, RSS, compressed memory, and swap. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
I/O and load |
prepared-index exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable index; source pages declared per run |
exact complete 4.5 GiB working volume and products |
1 |
|
Run repeated uncontended physical 8 GB prepared-index loads and prove admission, parity, memory, and recovery. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
I/O and load |
cold original compressed-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared-result substitution |
source open through exact complete working volume and products |
2 |
|
Run pressure-guarded physical cold-source trials with exact sums, full geometry/calibration provenance, RSS, compressed memory, and swap. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
I/O and load |
prepared-index exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable index; source pages declared per run |
exact complete working volume and products |
2 |
|
Run repeated uncontended physical prepared-index loads with exact parity, memory, cancellation, and recovery evidence. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
2 |
|
After the bounded object smoke passes, run the full fit with fitted-parameter parity, deterministic repeatability, pressure, swap, memory, and timing. |
Native Swift/Metal |
MacBook Air (M2, 8 GB) |
○ Pending |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Run one bounded smoke, then the current native 512 workflow with exact phase/loss parity, pressure, swap, memory, and repeat distributions. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
CoM, DPC, and iDPC |
rotation and Fourier iDPC from resident CoM maps |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident float32 CoM row and column maps |
synchronized rotation and iDPC publication |
3 |
|
Run the current native Metal path with frozen full-precision parity, synchronized timing, first-versus-warm states, and peak memory. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused |
exact indexed load, detector sum, and seven native-resolution products |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused |
exact indexed load, detector sum, and seven native-resolution products |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared exact resident input |
complete exact resident volume and seven products |
3 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported load-plan contract |
4 |
|
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized display-ready outputs |
3 |
|
Register the current Metal display-kernel distribution with exact revision, device, parity, peak memory, and first-versus-warm states. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized Fourier output |
3 |
|
Register the current Metal FFT distribution with exact revision, device, parity, peak memory, and first-versus-warm states. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
audit-free arbitrary-source cold first encounter |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold arbitrary source; no audit or index prepared |
source selection through exact complete resident output and products |
1 |
|
Measure discovery, value-range audit, index creation, load, products, and peak memory from an unaudited source without relabeling controlled F_NOCACHE evidence. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
cold arbitrary-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold arbitrary source; no audit or index prepared |
source selection through exact complete working output and products |
1 |
|
Measure discovery, value-range audit, index creation, exact sum-bin-2 load, products, and peak memory from an unaudited source. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
controlled uncached exact private-resident load and products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
fresh QH5 index root and destination; F_NOCACHE; already-audited immutable source |
catalog through exact complete resident volume and seven products |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
prepared-index exact private-resident load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
cold arbitrary-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold arbitrary source; no audit or index prepared |
source selection through exact complete working output and products |
2 |
|
Measure discovery, value-range audit, index creation, exact sum-bin-4 load, products, and peak memory from an unaudited source. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
audited uint8 staging to exact uint16 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
3 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
native uint16 exact residency with uint16 staging |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Run a source whose complete audit does not authorize uint8 staging and retain exact uint16 staging-to-resident parity and allocation evidence. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
exact uint16 to uint32 widening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact widened resident output |
3 |
|
Expose uint32 output through the public indexed loader and cache contract, then retain complete load, metadata, and physical parity. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
prepared-index exact private-resident load and products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable QH5 indexes; source pages unspecified |
complete exact resident volume, products, metadata, calibration, and provenance |
3 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
prepared-index exact private-resident load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, audit creation, first allocation, pipeline compilation, application presentation, and post-boundary hashes excluded |
3 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
native uint32 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint32 |
uint32 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
The native indexed Swift/Metal source contract does not admit uint32 HDF5 detector values. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
native uint8 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint8 |
uint8 |
uint8 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
The integrated indexed Swift/Metal loader does not admit a native uint8 source. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
audited lossless uint16 to uint8 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint8 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
The exact native Swift/Metal load contract cannot publish a uint8 resident scientific volume. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
explicit saturating uint16 to uint8 browse residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint8 |
browse-only |
source and prepared state declared by the run |
complete saturating browse resident output |
4 |
|
Explicit saturating uint8 resident output is not implemented by the public native Swift/Metal load boundary. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
exact detector-bin-8 resident load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported contract test |
5 |
|
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; bin 8 must fail closed. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
Screening |
prepared screening workflow |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported contract test |
5 |
|
The backend parity contract currently marks the prepared screening workflow not implemented for native Swift/Metal. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Selective loading |
64 by 64 evidence-selective rectangle |
64 × 64 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source-page state declared |
exact selected resident output |
2 |
|
Implement and expose the backend-neutral bulk selection contract, then prove source-range selectivity, order, metadata, parity, bytes, memory, and timing. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Selective loading |
arbitrary evidence-selective positions with order and duplicates |
n/a |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source-page state declared |
exact ordered selected resident output |
2 |
|
Implement the canonical selector contract without UI policy, then prove evidence-selective IO, order, duplicates, parity, bytes, memory, and timing. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
2 |
|
Register a current clean full-fit distribution with fitted-parameter parity, repeatability, cache policy, and memory. |
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Register a current clean physical 512 distribution with exact phase/loss parity, cache policy, and memory. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
CoM, DPC, and iDPC |
rotation and Fourier iDPC from resident CoM maps |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident float32 CoM row and column maps |
synchronized rotation and iDPC publication |
3 |
|
Repeat the complete CoM row/column, rotation, and iDPC suite with full-precision parity, synchronized timing, and memory telemetry. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared exact resident input |
complete exact resident volume and seven products |
1 |
|
Repeat uncontended full-native publication with complete product hashes, p50/p95/max, driver allocation, RSS, pressure, and swap. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused |
exact indexed load, detector sum, and seven native-resolution products |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused |
exact indexed load, detector sum, and seven native-resolution products |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
Not supported |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported load-plan contract |
4 |
|
Metal4DSTEMLoadPlan currently supports detector bins 1, 2, and 4 only; detector bin 8 must fail closed. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized display-ready outputs |
3 |
|
Run the current Metal display-kernel distribution with exact revision, full parity, first-versus-warm timing, and peak memory. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident Metal maps |
synchronized Fourier output |
3 |
|
Run the current Metal FFT distribution with exact revision, full parity, first-versus-warm timing, and peak memory. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
cold original compressed-source full-native load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
source selection through exact complete resident output and products |
1 |
|
Run arbitrary-source cold discovery, audit, index, full-native load, products, and complete memory/page-fault telemetry on physical MacBook Pro (M5, 24 GB). |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
I/O and load |
prepared-index full-native exact load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE descriptors; exact private destination; pressure gate stopped repetition |
exact indexed load and full-scan products; app and separately timed hashes excluded |
1 |
|
Eliminate or safely bound the observed swapout before attempting a repeated full-native distribution; preserve the exact full-volume and product contract. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
cold original compressed-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
source selection through exact complete working output and products |
1 |
|
Run arbitrary-source cold discovery, audit, index creation, exact sum-bin-2 load, products, and complete memory/page-fault telemetry. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-2 load with destination reuse |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private destination reused |
exact indexed detector sum-bin-2 load and native-resolution products; app and separately timed hashes excluded |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
cold original compressed-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
source selection through exact complete working output and products |
2 |
|
Run arbitrary-source cold discovery, audit, index creation, exact sum-bin-4 load, products, and complete memory/page-fault telemetry. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-4 fused load with destination reuse |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
2 |
|
n/a |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete 200-trial plus Nelder-Mead fit |
2 |
|
Run the current full fit with fitted-parameter parity, deterministic repeatability, cache policy, memory, and timing. |
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
○ Pending |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared BF evidence |
complete synchronized object, phase, and loss |
2 |
|
Run the current native 512 workflow with exact phase/loss parity, cache policy, memory, and repeat distributions. |
WebGPU |
MacBook Air (M2, 8 GB) |
○ Pending |
CoM, DPC, and iDPC |
DPC row/column and optimized-rotation iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident exact hardware-WebGPU state |
awaited GPU readback for each published product |
2 |
|
Run the physical hardware-browser path with frozen full-precision parity, awaited GPU timing, pressure, swap, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admissible |
blocked exact-residency contract |
4 |
|
Implement and prove an exact bounded-streaming browser product path that never requires an 18 GiB resident input before attempting this physical row. |
WebGPU |
MacBook Air (M2, 8 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized display-ready publication |
3 |
|
Run first-execution and warm hardware-browser display-kernel distributions with parity, presentation timing, pressure, swap, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized Fourier publication |
3 |
|
Run first-execution and warm hardware-browser FFT distributions with parity, presentation timing, pressure, swap, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
full-native resident admission |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admitted on the 8 GB resident path |
admission decision before allocation |
1 |
|
Keep this fail-closed until exact bounded residency can preserve the full logical tensor without an 18 GiB browser/device destination. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete 4.5 GiB working output |
1 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
prepared-index exact detector-bin-2 hardware load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact complete 4.5 GiB working output |
1 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete working output |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
prepared-index exact detector-bin-4 hardware load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact complete working output |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-8 load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete working output |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
prepared-index exact detector-bin-8 hardware load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact complete working output |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Air (M2, 8 GB) |
○ Pending |
Selective loading |
64 by 64 prepared-shard rectangle |
64 × 64 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes and frame-span manifest |
exact selected resident output; harness setup excluded |
2 |
|
Run the physical browser selector with full selected-tensor parity, source bytes, pressure, swap, browser/device memory, and timing; retain the intra-shard range-read gap explicitly. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
CoM, DPC, and iDPC |
DPC row/column and optimized-rotation iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident exact full-native WebGPU state |
awaited GPU readback for each published product |
3 |
|
n/a |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact hardware-WebGPU input; prepared QH5 indexes and warm source pages declared by measurement |
complete internal GPU product suite through awaited readbacks; parity hashing excluded |
2 |
|
n/a |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized display-ready publication |
3 |
|
Add a hardware-browser display-kernel entry point with parity, presentation, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized Fourier publication |
3 |
|
Add a hardware-browser FFT entry point with parity, presentation, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
cold original compressed-source hardware load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete resident output |
1 |
|
Run seven independently controlled arbitrary-source cold browser loads with full parity and memory. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
1 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-2 hardware load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume; no float32 resident substitution |
1 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
prepared-index hardware-WebGPU full-native load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; warm source pages; fresh browser target per retained run |
navigation through exact full-native resident output and diagnostic frame checksums |
2 |
|
Add total WebGPU device-allocation telemetry when the runtime exposes it; keep cold-source and application-E2E trials as separately labeled gates. Repeated harness wall is not yet at or below one second. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-4 hardware load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume; no float32 resident substitution |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
native uint16 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Add WebGPU device-allocation telemetry and retain an exact complete-volume hash inside the timed boundary before promoting the integrated load gate. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
native uint32 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint32 |
uint32 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Run a native uint32 HDF5 source on physical hardware WebGPU and retain the complete resident hash, browser/device memory, and source/decode/resident provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
native uint8 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint8 |
uint8 |
uint8 |
exact |
source and prepared state declared by the run |
complete exact resident output |
3 |
|
Run an exact physical hardware-browser full-volume uint8 source through local HDF5 decode and retain output hash, allocation, and source/decode/resident provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
audited lossless uint16 to uint8 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint8 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
3 |
|
Replace the experimental low8 global flag with a typed source-identity-bound audit in the local-HDF5 public result, then retain exact full-volume parity. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-8 load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-8 hardware load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume; no float32 resident substitution |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
audited uint8 staging to exact uint16 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
Audited uint8 staging followed by exact uint16 resident reconstruction is not implemented in the public WebGPU local-HDF5 path. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
I/O and load |
exact uint16 to uint32 widening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact widened resident output |
4 |
|
The WebGPU local-HDF5 public path does not widen uint16 source values into uint32 resident storage. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
I/O and load |
explicit saturating uint16 to uint8 browse residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint8 |
browse-only |
source and prepared state declared by the run |
complete saturating browse resident output |
4 |
|
Run the fused full-plane clip8 path on physical hardware with values above 255 and retain output hash, browser/device memory, and browse-only provenance. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
Screening |
prepared screening workflow |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not applicable |
unsupported contract test |
5 |
|
The backend parity contract currently marks the prepared screening workflow not implemented for WebGPU. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Selective loading |
64 by 64 prepared-shard rectangle |
64 × 64 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes and frame-span manifest |
exact selected resident output; harness setup excluded |
2 |
|
Add full selected-tensor parity and true intra-shard range reads; current evidence uses qualified frame probes and shard-level selection. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
○ Pending |
Selective loading |
arbitrary evidence-selective positions with order and duplicates |
n/a |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source-page state declared |
exact ordered selected resident output |
2 |
|
Implement ordered/duplicate sparse selectors and true range reads, then run full physical parity, bytes, memory, and timing. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
× Refuted |
Single-sideband ptychography |
object, phase, and loss |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared detector-bin-2 evidence |
complete physical browser object, phase, and loss |
1 |
|
Restore frozen phase parity before any performance promotion; retain the current mismatch as refuted. |
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
Not supported |
Single-sideband ptychography |
200 trials plus Nelder-Mead |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
not applicable |
unsupported contract test |
5 |
|
The frozen backend contract does not implement WebGPU calibration fitting. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
CoM, DPC, and iDPC |
DPC row/column and optimized-rotation iDPC |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident exact hardware-WebGPU state |
awaited GPU readback for each published product |
2 |
|
Run the physical hardware-browser path with frozen full-precision parity, awaited GPU timing, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact full-native hardware-WebGPU input |
complete product suite publication |
1 |
|
Run a guarded physical hardware-browser admission smoke, then retain the complete native product suite with full parity, timing, browser/device memory, pressure, and swap if admission remains safe. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-2 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-2 accumulation and residency, then run the complete hardware product suite with full parity, timing, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-4 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-4 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
resident exact detector-bin-8 hardware-WebGPU input required |
complete product suite publication |
1 |
|
Integrate exact-integer detector-bin-8 accumulation and residency, then run the complete hardware product suite with full parity, timing, pressure, swap, and browser/device memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
Display kernels |
transform, histogram, and color |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized display-ready publication |
3 |
|
Run first-execution and warm hardware-browser display-kernel distributions with parity, presentation timing, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
FFT |
Fourier transform |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident hardware-WebGPU maps |
synchronized Fourier publication |
3 |
|
Run first-execution and warm hardware-browser FFT distributions with parity, presentation timing, browser-tree memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
cold original compressed-source full-native hardware load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete resident output |
1 |
|
After bounded admission passes, run controlled arbitrary-source cold browser trials with full parity and complete memory/page-fault telemetry. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-2 load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
1 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
prepared-index exact full-native hardware load |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact full-native resident output |
2 |
|
Run one bounded admission smoke before a repeated exact production-path distribution with browser-tree and adapter memory. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-2 hardware load |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-4 load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
2 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-4 hardware load |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
cold original compressed-source exact detector-bin-8 load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
cold original source; no prepared index |
file selection through exact complete uint16 working output |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run controlled arbitrary-source cold trials with full-volume parity, stage timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
! Blocked |
I/O and load |
prepared-index exact integer detector-bin-8 hardware load |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source-page state declared by run |
exact uint16 working volume |
3 |
|
Integrate and qualify exact-integer accumulation and residency, then run full-volume parity, repeated prepared timing, adapter/browser memory, and device-loss checks. |
WebGPU |
MacBook Pro (M5, 24 GB) |
○ Pending |
Selective loading |
64 by 64 prepared-shard rectangle |
64 × 64 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes and frame-span manifest |
exact selected resident output; harness setup excluded |
2 |
|
Run the physical browser selector with full selected-tensor parity, source bytes, browser/device memory, and timing; retain the intra-shard range-read gap explicitly. |
CPU reference |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact full-native CPU input |
complete reference product suite |
4 |
|
Run the independent full-native CPU reference product suite only when host CPU, storage, and memory ownership are uncontended; retain all hashes, timing, and peak RSS. |
CPU reference |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
○ Pending |
I/O and load |
independent full-native exact load reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source-page state declared by run; CPU exact-reference validation |
public CPU reference load; full-volume hash excluded |
4 |
|
Run only when host CPU, storage, and memory ownership are uncontended; retain exact full-volume parity, p50/p95/max, RSS, cache state, and source bytes. |
CPU reference |
MacBook Air (M2, 8 GB) |
! Blocked |
Detector products |
mean DP, total, BF, ABF, ADF, and DF reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admissible |
blocked reference product suite |
4 |
|
Implement and prove an exact bounded-streaming CPU reference product path that never requires an 18 GiB resident input before attempting this physical row. |
CPU reference |
MacBook Air (M2, 8 GB) |
! Blocked |
I/O and load |
full-native exact CPU-reference admission |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
not admitted as an in-memory CPU result on the 8 GB computer |
admission decision before allocation |
4 |
|
Keep full resident admission fail-closed; qualify only an exact bounded mmap or streaming reference that preserves all 18 GiB logical bytes without memory-pressure termination. |
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
exact full-native CPU-resident input |
complete reference product suite |
4 |
|
Add ABF, retain all product hashes, and repeat the exact physical CPU reference traversal for p50/p95/maximum and complete process-memory telemetry. |
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
independent full-native exact load reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source pages unspecified; CPU exact-reference validation only |
public CPU reference load; full-volume hash excluded |
4 |
|
Keep the canonical CPU reference current after loader or metadata changes; a repeated CPU speed distribution is not a release gate. |
CPU reference |
MacBook Pro (M5, 24 GB) |
○ Pending |
Detector products |
mean DP, total, BF, ABF, ADF, and DF reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident exact full-native CPU input |
complete reference product suite |
4 |
|
Run only after an uncontended memory preflight; retain one exact full-native CPU reference smoke, complete product hashes, pressure, swap, and peak RSS before repetition. |
CPU reference |
MacBook Pro (M5, 24 GB) |
○ Pending |
I/O and load |
independent full-native exact load reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared indexes; source-page state declared by run; CPU exact-reference validation |
public CPU reference load; full-volume hash excluded |
4 |
|
Run one bounded exact CPU-reference smoke with mmap or streaming, full-volume parity, process physical footprint, pressure, and swap; repeat only if the 18 GiB logical output is safely admitted. |
CPU reference |
Portable CI runner |
◐ Partial |
CoM, DPC, and iDPC |
CoM row/column, rotation, and iDPC reference |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident reference inputs |
complete reference publication |
4 |
|
Maintain frozen full-precision references and signal-range checks. |
CPU reference |
Portable CI runner |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
resident reference input |
complete reference product suite |
4 |
|
Keep exact reference tests current; no CPU speed claim is required. |
CPU reference |
Portable CI runner |
◐ Partial |
Display kernels |
transform, histogram, and color reference |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident reference maps |
complete reference display outputs |
4 |
|
Maintain the shared transform, histogram, and color goldens and frozen display conventions. |
CPU reference |
Portable CI runner |
◐ Partial |
FFT |
Fourier transform reference |
512 × 512 |
n/a |
n/a |
n/a |
float32 maps |
n/a |
float32 |
n/a |
resident reference maps |
complete reference Fourier output |
4 |
|
Maintain the frozen Fourier-transform reference and numerical conventions. |
CPU reference |
Portable CI runner |
Not supported |
I/O and load |
audited uint8 staging to exact uint16 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
Audited uint8 staging with uint16 resident reconstruction is not implemented for CPU reference. |
CPU reference |
Portable CI runner |
◐ Partial |
I/O and load |
native uint16 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
Promote only after the portable public API fixture retains complete-volume uint16 hash, shape, order, mask, and metadata parity. |
CPU reference |
Portable CI runner |
◐ Partial |
I/O and load |
native uint32 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint32 |
uint32 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
Add a public-API native uint32 fixture that disables advisory auto narrowing and proves complete count, dtype, shape, and metadata parity. |
CPU reference |
Portable CI runner |
◐ Partial |
I/O and load |
native uint8 exact residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint8 |
uint8 |
uint8 |
exact |
source and prepared state declared by the run |
complete exact resident output |
4 |
|
Add a repository fixture that proves public io.load preserves every native uint8 count, shape, order, and provenance field at detector bin 1. |
CPU reference |
Portable CI runner |
Not supported |
I/O and load |
exact uint16 to uint32 widening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint32 |
exact |
source and prepared state declared by the run |
complete exact widened resident output |
4 |
|
CPU reference io.load has no uint16-to-uint32 widening. |
CPU reference |
Portable CI runner |
Not supported |
I/O and load |
audited lossless uint16 to uint8 residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint8 |
exact |
source identity and complete value-range audit declared by the run |
complete audit-bound exact resident output |
4 |
|
CPU reference io.load has no uint8 output dtype. |
CPU reference |
Portable CI runner |
Not supported |
I/O and load |
explicit saturating uint16 to uint8 browse residency |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint8 |
browse-only |
source and prepared state declared by the run |
complete saturating browse resident output |
4 |
|
CPU reference io.load has no saturating uint8 browse output. |
CPU reference |
Portable CI runner |
◐ Partial |
I/O and load |
independent exact load/bin reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
portable reference; performance not promoted |
exact reference output |
4 |
|
Retain a standardized full-size CPU reference timing only when a practical physical memory envelope is available; parity remains the primary role. |
CPU reference |
Portable CI runner |
◐ Partial |
Screening |
complete screening reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
reference source |
complete reference screening result |
4 |
|
Maintain the frozen reference arrays; physical timing is not a CPU release gate. |
CPU reference |
Portable CI runner |
◐ Partial |
Single-sideband ptychography |
object, phase, and loss reference |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared reference BF evidence |
complete reference object, phase, and loss |
4 |
|
Maintain the frozen full-precision reference fixture and objective. |
CPU reference |
Portable CI runner |
◐ Partial |
Single-sideband ptychography |
200 trials plus Nelder-Mead reference |
512 × 512 |
n/a |
n/a |
n/a |
full active BF evidence |
n/a |
complex64 |
n/a |
prepared reference BF evidence |
complete 200-trial plus Nelder-Mead fit |
5 |
|
Maintain the frozen deterministic reference objective and fitted parameters. |
Retained atomic measurements#
These rows are imported from immutable evidence or explicitly registered follow-up evidence. A partial row is retained but does not satisfy a complete timing-and-parity gate.
Platform |
Computer |
State |
Module |
Operation |
Selected scan |
Source detector |
Detector bin |
Output detector |
Source dtype |
Staging dtype |
Resident dtype |
Scientific gate |
Cache/process state |
Wall boundary |
Samples |
p50 |
p95 |
Maximum |
Logical resident |
Driver allocated after load |
Driver allocated after release |
Accelerator peak |
Total-device peak |
Process/tree peak |
Process physical-footprint peak |
Swap delta |
Parity |
Device tested |
Date tested |
Revision |
Fixture ID |
Master SHA-256 |
Source identity SHA-256 |
Measurement ID |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-streamed-summaries |
n/a |
warm operating-system source pages; unspecified; no prepared source index |
quantem.gpu.screening.prepare package wall through exact complete public screening arrays |
3 |
1.208146 s |
1.241817 s |
1.245558 s |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-streamed-summaries |
n/a |
warm operating-system source pages; unspecified; no prepared source index |
quantem.gpu.screening.prepare package wall through exact complete public screening arrays |
3 |
1.663968 s |
1.718105 s |
1.724120 s |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-streamed-summaries |
n/a |
warm operating-system source pages; unspecified; no prepared source index; unrelated CPU contention retained |
quantem.gpu.screening.prepare package wall through exact complete public screening arrays |
3 |
1.546580 s |
1.552670 s |
1.553347 s |
n/a |
n/a |
n/a |
n/a |
6.084 GiB |
2.130 GiB |
n/a |
n/a |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-streamed-summaries |
n/a |
warm operating-system source pages; unspecified; empty screening-result cache |
quantem.gpu.screening.prepare package wall through exact complete six-array screening result in balanced A-B-B-A-B-A-A-B comparison |
6 |
1.356516 s |
1.642386 s |
1.730328 s |
n/a |
n/a |
n/a |
2.084 GiB |
6.084 GiB |
2.130 GiB |
n/a |
0 B |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-streamed-summaries |
n/a |
warm operating-system source pages; unspecified; empty screening-result cache |
quantem.gpu.screening.prepare package wall through exact complete six-array screening result in balanced A-B-B-A-B-A-A-B comparison |
6 |
1.204713 s |
1.325760 s |
1.329731 s |
n/a |
n/a |
n/a |
2.084 GiB |
6.084 GiB |
1.958 GiB |
n/a |
0 B |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0; driver 580.159.03 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
CUDA |
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
✓ Measured |
Screening |
screening-cache-reopen |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
not-applicable-cached-results |
n/a |
prepared immutable screening result cache; fresh-process reopen; source build excluded |
fresh-process exact-spelling prepared-result reopen through strong source validation, lazy raw-I/O import boundary, six-array materialization, and immutable ScreeningResult publication |
3 |
5.599 ms |
5.771 ms |
5.790 ms |
n/a |
n/a |
n/a |
n/a |
n/a |
0.874 GiB |
n/a |
0 B |
Pass |
NVIDIA RTX PRO 6000 Blackwell Workstation Edition; GPU0 |
2026-08-22 |
|
real-512-native-detector |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
× Refuted |
CoM, DPC, and iDPC |
CoM row/column, rotation, aligned fields, and iDPC parity smoke |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
float32 products |
full-precision |
exact resident detector-bin-1 MPS input; source pages uncontrolled or warm; no prepared product result; five pageouts; zero swapout growth |
single first-execution diagnostic only; synchronized public CoM through rotation and iDPC pipeline was 0.159022 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
n/a |
27.510 GiB |
n/a |
20.436 GiB |
28.001 GiB |
0 B |
Failed |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Detector products |
mean DP, total, BF, ABF, ADF, and DF exact parity smoke |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 source; uint64 and float32 products |
full-precision |
exact resident detector-bin-1 MPS input; source pages uncontrolled or warm; no prepared product result; five pageouts; zero swapout growth |
single first-execution diagnostic only; synchronized detector-product publication was 0.499909 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
n/a |
27.510 GiB |
n/a |
20.436 GiB |
28.001 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
× Refuted |
FFT |
first-execution Torch-MPS FFT parity smoke |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
float32 |
full-precision |
resident full-precision iDPC map; first MPS FFT execution; five pageouts; zero swapout growth |
single first-execution diagnostic only; upload, synchronized FFT, and readback publication was 0.291350 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
n/a |
27.510 GiB |
n/a |
20.436 GiB |
28.001 GiB |
0 B |
Failed |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
post-warmup full-native exact load with fresh returned destination |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
one same-process warmup; operating-system source pages uncontrolled; no eviction performed; fresh returned destination released after each trial |
public io.load return after MPS synchronization; full-volume hash and release excluded |
7 |
0.406624 s |
0.428164 s |
0.428164 s |
18.000 GiB |
18.442 GiB |
0.442 GiB |
18.442 GiB |
n/a |
18.692 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
↺ Superseded |
I/O and load |
post-warmup exact package load with fresh returned destination |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
one same-process warmup; operating-system source pages uncontrolled; no eviction performed; fresh returned destination released after each trial |
public io.load return after backend synchronization |
7 |
0.414824 s |
0.457261 s |
0.457261 s |
18.000 GiB |
18.442 GiB |
0.442 GiB |
n/a |
n/a |
0.691 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; fresh destination |
exact package load into a fresh destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.425533 s |
0.436353 s |
0.437419 s |
18.000 GiB |
n/a |
n/a |
18.442 GiB |
n/a |
0.687 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; explicit resident destination recycled |
exact package load into caller-selected resident destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.259189 s |
0.263118 s |
0.263375 s |
18.000 GiB |
n/a |
n/a |
18.442 GiB |
n/a |
0.688 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; fresh destination |
exact package load into a fresh caller-selected resident destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.462541 s |
0.479014 s |
0.483058 s |
4.500 GiB |
n/a |
n/a |
5.688 GiB |
n/a |
0.571 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; explicit resident destination recycled |
exact package load into an explicitly recycled caller-selected resident destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.359606 s |
0.361384 s |
0.361995 s |
4.500 GiB |
n/a |
n/a |
5.688 GiB |
n/a |
0.571 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
loading |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
warm operating-system source pages; uncontrolled |
fresh-process package load only; not import, cold HDF5, prepared reopen, first app product, or app E2E |
14 |
0.498708 s |
0.503702 s |
0.505150 s |
4.500 GiB |
n/a |
n/a |
5.688 GiB |
n/a |
0.575 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; fresh destination |
exact package load into a fresh caller-selected resident destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.384264 s |
0.385355 s |
0.385638 s |
1.125 GiB |
n/a |
n/a |
2.313 GiB |
n/a |
0.571 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
loading |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
warm or uncontrolled operating-system source pages after one same-lifecycle warmup; explicit resident destination recycled |
exact package load into an explicitly recycled caller-selected resident destination in balanced A-B-B-A-B-A-A-B lifecycle comparison |
8 |
0.352990 s |
0.355048 s |
0.355062 s |
1.125 GiB |
n/a |
n/a |
2.313 GiB |
n/a |
0.572 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
loading |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared plans and warm or uncontrolled operating-system source pages |
fresh-process exact package load after one excluded warm-up |
6 |
0.383772 s |
0.390006 s |
0.391839 s |
1.125 GiB |
n/a |
n/a |
2.313 GiB |
n/a |
n/a |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
loading |
512 × 512 |
192 × 192 |
8 |
24 × 24 |
uint16 |
n/a |
uint16 |
n/a |
warm operating-system source pages; uncontrolled |
immutable-source fresh-process package A-B-B-A; contended stage run excluded |
6 |
0.356969 s |
0.359302 s |
0.359820 s |
0.281 GiB |
n/a |
n/a |
1.470 GiB |
n/a |
0.572 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Python MPS |
MacBook Pro (M5, 24 GB) |
◐ Partial |
CoM, DPC, and iDPC |
CoM row/column, rotation, aligned fields, and iDPC parity smoke |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
float32 products |
n/a |
exact resident detector-bin-2 MPS input; source pages uncontrolled or warm; no prepared product result; Screen Sharing active; 73 pageouts; zero swap growth |
single contaminated diagnostic only; public CoM through rotation and iDPC pipeline was 0.080079 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
4.500 GiB |
n/a |
n/a |
5.736 GiB |
n/a |
9.657 GiB |
10.508 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Python MPS |
MacBook Pro (M5, 24 GB) |
× Refuted |
FFT |
first-execution Torch-MPS FFT parity smoke |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
float32 |
n/a |
resident full-precision iDPC map; first MPS FFT execution; Screen Sharing active; 73 pageouts; zero swap growth |
single contaminated diagnostic only; upload, synchronized FFT, and readback publication was 0.212618 seconds; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
4.500 GiB |
n/a |
n/a |
5.736 GiB |
n/a |
9.657 GiB |
10.508 GiB |
0 B |
Failed |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-2 load with destination reuse |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.225984 s |
0.231192 s |
0.231192 s |
4.500 GiB |
n/a |
n/a |
5.087 GiB |
n/a |
0.657 GiB |
5.336 GiB |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-4 load with destination reuse |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, audit creation, first allocation, pipeline compilation, application presentation, and post-boundary hashes excluded |
6 |
0.200036 s |
0.204694 s |
0.204694 s |
1.125 GiB |
n/a |
n/a |
1.712 GiB |
n/a |
1.039 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
↺ Superseded |
I/O and load |
loading-and-products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
fresh QH5 index root and destination; F_NOCACHE on source hashing and indexed descriptors; already-audited immutable source |
catalog plus pipeline compilation, plan, complete exact private-resident volume, and seven products |
7 |
0.524590 s |
1.572781 s |
1.572781 s |
18.000 GiB |
n/a |
n/a |
18.571 GiB |
n/a |
0.991 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GiB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
controlled uncached exact resident load and products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint8 |
uint16 |
exact |
fresh QH5 index root and destination; macOS F_NOCACHE on source hashing and indexed descriptors; immutable source already audited |
catalog, pipeline compilation, plan, complete exact private-resident volume, seven products, metadata, and provenance |
7 |
0.577793 s |
0.900979 s |
0.900979 s |
18.000 GiB |
n/a |
n/a |
18.571 GiB |
n/a |
0.874 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
I/O and load |
loading-and-products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable QH5 indexes; source pages unspecified |
complete exact resident volume, products, metadata, calibration, and provenance; full private readback/hash excluded |
6 |
0.313870 s |
0.318865 s |
0.318865 s |
18.000 GiB |
n/a |
n/a |
18.571 GiB |
n/a |
0.639 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
CoM, DPC, and iDPC |
dpc-idpc |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
float32 |
n/a |
already-resident 512x512 float32 CoM row and column maps; source load excluded |
explicit CoM rotation plus Fourier iDPC through synchronized publication; source load and CoM reduction excluded |
15 |
1.979 ms |
3.640 ms |
3.640 ms |
n/a |
n/a |
n/a |
0.009 GiB |
n/a |
0.034 GiB |
0.064 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
original HDF5 reread to complete exact packed resident |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
exact |
original compressed source reread every visit; existing QH5 index and source-bound packing layout; exact DPC sums reused after first load; OS pages uncontrolled |
indexed source open through synchronous complete packed-resident return; catalog and independent full-count audit separate |
7 |
1.496660 s |
2.086119 s |
2.086119 s |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
n/a |
2.531 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-09-06 |
|
original-512x512x192x192-u16-fixture-d |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
I/O and load |
prepared-index full-native exact private-resident smoke |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE descriptors; fresh exact private-resident destination; repetition stopped by pressure gate |
exact indexed full-native load and seven native-resolution products; source discovery, audit creation, application presentation, and separately timed hashes excluded |
1 |
3.424108 s |
3.424108 s |
3.424108 s |
18.000 GiB |
n/a |
n/a |
18.587 GiB |
n/a |
0.611 GiB |
18.673 GiB |
~0.706 GiB |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
◐ Partial |
I/O and load |
loading-and-products |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared QH5 indexes; F_NOCACHE source descriptors; operating-system source-page state unspecified |
package exact indexed load plus exact full-scan products; app UI, index creation, and full-volume hash excluded |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
n/a |
18.571 GiB |
n/a |
0.624 GiB |
18.658 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-2 load with destination reuse |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.678703 s |
0.697179 s |
0.697179 s |
4.500 GiB |
n/a |
n/a |
5.087 GiB |
n/a |
0.662 GiB |
5.271 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
↺ Superseded |
I/O and load |
prepared-index exact detector-bin-2 canonical smoke |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; fresh exact private-resident destination |
catalog, pipeline compilation, plan, exact indexed detector sum-bin-2 load, and seven native-resolution products; separately timed boundary hashes excluded |
1 |
0.763019 s |
0.763019 s |
0.763019 s |
4.500 GiB |
n/a |
n/a |
5.087 GiB |
n/a |
0.645 GiB |
5.200 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
↺ Superseded |
I/O and load |
prepared-index exact detector-bin-2 load with destination reuse |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, first allocation, pipeline compilation, and application excluded |
6 |
0.671309 s |
0.694479 s |
0.694479 s |
4.500 GiB |
n/a |
n/a |
5.071 GiB |
n/a |
0.646 GiB |
n/a |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
loading-and-products |
512 × 512 |
192 × 192 |
2 |
96 × 96 |
uint16 |
n/a |
uint16 |
n/a |
prepared QH5 indexes; F_NOCACHE source descriptors; operating-system source-page state unspecified; caller destination reused |
package exact indexed load plus exact full-scan products; hash and app excluded |
6 |
0.683492 s |
0.690893 s |
0.690893 s |
4.500 GiB |
n/a |
n/a |
5.071 GiB |
n/a |
0.647 GiB |
5.182 GiB |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
prepared-index exact detector-bin-4 load with destination reuse |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; controlled F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven native-resolution products; source discovery, audit creation, pipeline compilation, application presentation, and separately timed boundary hashes excluded |
6 |
0.640942 s |
0.651931 s |
0.651931 s |
1.125 GiB |
n/a |
n/a |
1.712 GiB |
n/a |
1.018 GiB |
2.319 GiB |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
↺ Superseded |
I/O and load |
prepared-index exact detector-bin-4 load with destination reuse |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable indexes; F_NOCACHE source descriptors; compiled pipeline; exact private-resident destination reused for retained trials |
exact indexed load, detector sum, and seven products; source discovery, first allocation, pipeline compilation, and application excluded |
6 |
0.630903 s |
0.638497 s |
0.638497 s |
1.125 GiB |
n/a |
n/a |
1.696 GiB |
n/a |
0.650 GiB |
n/a |
0 B |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Native Swift/Metal |
MacBook Pro (M5, 24 GB) |
✓ Measured |
I/O and load |
loading-and-products |
512 × 512 |
192 × 192 |
4 |
48 × 48 |
uint16 |
n/a |
uint16 |
n/a |
prepared QH5 indexes; F_NOCACHE source descriptors; operating-system source-page state unspecified; caller destination reused |
package load plus seven exact products; UI and full-volume hash excluded |
12 |
0.631540 s |
0.645521 s |
0.645521 s |
1.125 GiB |
n/a |
n/a |
1.696 GiB |
n/a |
0.654 GiB |
n/a |
n/a |
Pass |
Apple M5 10-core integrated GPU; 24 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
n/a |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
CoM, DPC, and iDPC |
dpc |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
float32 |
n/a |
resident exact full-native WebGPU state; source load excluded |
DPC column through awaited GPU readback; synchronized wall time, not GPU-only timestamp interval |
7 |
0.700 ms |
0.800 ms |
0.800 ms |
n/a |
n/a |
n/a |
n/a |
n/a |
6.366 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
n/a |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
CoM, DPC, and iDPC |
dpc |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
float32 |
n/a |
resident exact full-native WebGPU state; source load excluded |
DPC row through awaited GPU readback; synchronized wall time, not GPU-only timestamp interval |
7 |
0.700 ms |
0.900 ms |
0.900 ms |
n/a |
n/a |
n/a |
n/a |
n/a |
6.366 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
n/a |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
CoM, DPC, and iDPC |
idpc |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
float32 |
n/a |
resident exact full-native WebGPU state; source load excluded |
optimized-rotation iDPC through awaited GPU readback; synchronized wall time, not GPU-only timestamp interval |
7 |
1.400 ms |
1.500 ms |
1.500 ms |
n/a |
n/a |
n/a |
n/a |
n/a |
6.366 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
n/a |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
✓ Measured |
Detector products |
exact resident full detector-product plus CoM/DPC/iDPC suite |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared QH5 indexes; warm source pages; exact full-native uint16 resident input; fresh browser target per retained run |
complete internal GPU product suite from exact resident input through awaited readbacks; parity hashing, sorting, and encoding excluded |
7 |
0.482000 s |
0.483600 s |
0.483600 s |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
6.492 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
prepared-index full-native exact hardware-WebGPU distribution |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
prepared immutable block indexes; explicitly warm source pages; fresh browser target per retained run; 699 pageouts; zero swap growth; no dropped runs |
navigation through scientifically usable exact resident output and diagnostic frame checksums; exhaustive full-volume hash and application E2E excluded |
7 |
1.358000 s |
1.594000 s |
1.594000 s |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
6.500 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
↺ Superseded |
I/O and load |
prepared-index full-native exact hardware-WebGPU smoke |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable block indexes; source pages warm from the immediately preceding independent CPU reference; fresh hardware browser target |
hardware harness load wall through exact resident output; 63.153-second full-volume verification hash and application E2E excluded |
1 |
1.094000 s |
1.094000 s |
1.094000 s |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
≥6.604 GiB |
n/a |
0 B |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
↺ Superseded |
I/O and load |
loading |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared block indexes; source pages unspecified and warm after the accepted parity run |
one full-native hardware-WebGPU load preceding the accepted synchronized resident-product repetitions |
1 |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
n/a |
6.366 GiB |
n/a |
n/a |
Pass |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal-3 hardware adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
n/a |
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Selective loading |
selective-loading |
256 × 256 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable block indexes and frame-span manifest; operating-system source-page state uncontrolled/unspecified; no eviction performed; fresh browser target per repetition |
WebGPU loader wall through exact selected resident output; frame-manifest read/encoding, DevTools injection, checksum harness, products, and application E2E excluded |
5 |
0.381000 s |
0.392400 s |
0.394000 s |
4.500 GiB |
n/a |
n/a |
n/a |
n/a |
2.797 GiB |
n/a |
0 B |
Qualified probes |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal WebGPU adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Selective loading |
selective-loading |
384 × 384 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable block indexes and frame-span manifest; operating-system source-page state uncontrolled/unspecified; no eviction performed; fresh browser target per repetition |
WebGPU loader wall through exact selected resident output; frame-manifest read/encoding, DevTools injection, checksum harness, products, and application E2E excluded |
5 |
0.574000 s |
0.582000 s |
0.584000 s |
10.125 GiB |
n/a |
n/a |
n/a |
n/a |
3.629 GiB |
n/a |
0 B |
Qualified probes |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal WebGPU adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
WebGPU |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Selective loading |
selective-loading |
64 × 64 |
192 × 192 |
1 |
192 × 192 |
uint16 |
n/a |
uint16 |
n/a |
prepared immutable block indexes and frame-span manifest; operating-system source-page state uncontrolled/unspecified; no eviction performed; fresh browser target per repetition |
WebGPU loader wall through exact selected resident output; frame-manifest read/encoding, DevTools injection, checksum harness, products, and application E2E excluded |
5 |
0.147000 s |
0.154400 s |
0.156000 s |
0.281 GiB |
n/a |
n/a |
n/a |
n/a |
1.606 GiB |
n/a |
0 B |
Qualified probes |
Apple M5 Max 40-core integrated GPU; Chrome 151; Apple Metal WebGPU adapter; software=false |
2026-08-22 |
|
full-native-webgpu-512x512x192x192-u16 |
|
|
|
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
Detector products |
independent mean DP, total, BF, ADF, DF, CoM, DPC, and iDPC reference traversal |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
exact full-native CPU-resident input after a warm operating-system page-cache source load |
single independent detector-product traversal was 31.078040 seconds; source load excluded; no p50, p95, or maximum distribution |
1 |
n/a |
n/a |
n/a |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
36.450 GiB |
n/a |
n/a |
Pass |
Apple M5 Max CPU; 128 GB unified memory |
2026-08-19 |
|
real-512x512x192x192-u16-bslz4-27shard |
|
|
|
CPU reference |
MacBook Pro (M5 Max, 128 GB) |
◐ Partial |
I/O and load |
independent full-native exact load reference |
512 × 512 |
192 × 192 |
1 |
192 × 192 |
uint16 |
uint16 |
uint16 |
exact |
prepared indexes; source pages unspecified; CPU exact-reference validation only |
public CPU reference load; full-volume hash excluded |
1 |
32.788039 s |
32.788039 s |
32.788039 s |
18.000 GiB |
n/a |
n/a |
n/a |
n/a |
19.141 GiB |
n/a |
0 B |
Pass |
Apple M5 Max CPU; 128 GB unified memory |
2026-08-22 |
|
real-512x512x192x192-u16-bslz4-27shard-master-fixture-c |
|
|
|
Reproducible runbooks#
Runbook |
Owner |
Tier |
Evidence level |
Command |
Required artifacts |
|---|---|---|---|---|---|
|
Linux CUDA workstation (dual 96 GB Blackwell GPUs) |
PR smoke |
parity preflight |
|
pytest log; separate physical reconstruction/calibration timing artifact |
|
Portable CI runner plus specified backend computer |
PR smoke |
parity preflight |
|
pytest log; backend-specific physical timing from a separate runbook |
|
Physical CPU, CUDA, or Apple MPS computer named by the gate |
scheduled physical profile |
physical performance |
|
run-level synchronized timing and exact product hashes; source, full-volume, independent-reference, range-audit, mask, dtype, shape, and detector-bin identities; accelerator/driver/process peak memory plus system pressure, compressor, pageout, and swap telemetry; experiment manifest and RUNS.md row |
|
Portable CI runner plus specified backend computer |
PR smoke |
parity preflight |
|
pytest log; physical first-execution and warm timing artifact |
|
Portable CI runner plus specified backend computer |
PR smoke |
parity preflight |
|
pytest log; physical backend synchronized timing and parity artifact |
|
MacBook Pro (M5 Max, 128 GB) |
scheduled physical profile |
physical performance |
|
run-level reconstruction and fit JSONL; phase/loss parity and fitted parameters; allocator/driver/process/swap memory; experiment manifest and RUNS.md row |
|
specified physical backend computer |
scheduled physical profile |
physical performance |
|
run-level JSON with p50, p95, and maximum; exact output parity hashes; logical resident, sampled accelerator/process lower-bound peaks, lifetime process RSS, and swap telemetry; external process-tree, compressed-memory, and pageout telemetry when required by the physical gate; experiment manifest and RUNS.md row |
|
Portable CI runner |
PR smoke |
parity preflight |
|
pytest terminal log; registry validator output |
|
Linux CUDA workstation (dual 96 GB Blackwell GPUs) or MacBook Pro (M5 Max, 128 GB) |
scheduled physical profile |
physical performance |
|
run-level timing and cache lifecycle JSON; mean DP, total, BF, DF, CoM, rotation, and iDPC parity; GPU allocation/reserve/total-card and host RSS/swap telemetry; experiment manifest and RUNS.md row |
|
MacBook Pro (M5 Max, 128 GB), MacBook Air (M2, 8 GB), or MacBook Pro (M5, 24 GB) |
scheduled physical profile |
physical performance |
|
benchmark JSON, physical-storage hashes, and canonical logical-pixel hashes; source and working geometry/dtype/provenance; Metal allocation, process RSS/footprint, pressure, and swap telemetry; experiment manifest and RUNS.md row |
|
Physical Apple Silicon Mac |
scheduled physical profile |
physical performance |
|
Raw benchmark JSON lines including failures and release records; Independent full-count oracle and source/working dtype and geometry; Separate process/device memory, pressure and swap telemetry |
|
MacBook Pro (M5 Max, 128 GB) or MacBook Pro (M5, 24 GB) |
scheduled physical profile |
physical performance |
|
benchmark JSON; phase/loss/fitted-parameter parity; Metal and process memory telemetry; experiment manifest and RUNS.md row |
|
MacBook Pro (M5 Max, 128 GB), MacBook Pro (M5, 24 GB), or MacBook Air (M2, 8 GB) |
scheduled physical profile |
physical performance |
|
browser profile JSON with run-level timings; independent full-output SHA-256 plus diagnostic frame checksums; browser-tree RSS, adapter allocation when available, and swap telemetry; experiment manifest and RUNS.md row |
|
MacBook Pro (M5 Max, 128 GB) or MacBook Air (M2, 8 GB) |
scheduled physical profile |
physical performance |
|
browser timing and source-byte JSON; full selected-tensor parity or explicitly qualified probe coverage; browser-tree memory and swap telemetry; experiment manifest and RUNS.md row |
|
MacBook Pro (M5 Max, 128 GB) |
manual hardware diagnostic |
parity preflight |
|
pytest log; physical browser phase/loss parity and timing artifact |