Module profiling 

GPU timestamp profiling using wgpu::Features::TIMESTAMP_QUERY.

Provides an opt-in GpuTimestampProfiler that records GPU-side timestamps around compute passes via a wgpu::QuerySet and reports per-pass timings in microseconds after resolution.

The profiler is graceful: GpuTimestampProfiler::try_new returns None whenever the adapter does not advertise the required features (wgpu::Features::TIMESTAMP_QUERY and wgpu::Features::TIMESTAMP_QUERY_INSIDE_ENCODERS — the latter is required to call wgpu::CommandEncoder::write_timestamp outside of a render/compute pass). Callers that do not enable these features when constructing the GpuContext will simply receive None and can fall back to wall-clock timing.

Quick sketch

use oxigdal_gpu::{GpuContext, GpuTimestampProfiler};
let mut prof = match GpuTimestampProfiler::try_new(ctx, 16) {
    Some(p) => p,
    None => return Ok(()), // adapter lacks TIMESTAMP_QUERY support
};
let mut encoder = ctx.device().create_command_encoder(&Default::default());
if let Some(slot) = prof.begin_pass(&mut encoder, "blur") {
    // ... record compute work into `encoder` ...
    prof.end_pass(&mut encoder, slot);
}
ctx.queue().submit([encoder.finish()]);
for t in prof.resolve(ctx)? {
    println!("{} took {:.3} us", t.label, t.duration_us);
}

Notes on wgpu features

• wgpu::Features::TIMESTAMP_QUERY is the WebGPU baseline feature that enables creating wgpu::QueryType::Timestamp query sets and using them within render/compute pass descriptors.
• wgpu::Features::TIMESTAMP_QUERY_INSIDE_ENCODERS is the native-only extension that allows calling wgpu::CommandEncoder::write_timestamp between passes. This is the API used by GpuTimestampProfiler::begin_pass / GpuTimestampProfiler::end_pass.

Both features must be requested via GpuContextConfig::with_features before creating the GpuContext.

Structs

GpuTimestampProfiler

Opt-in GPU timestamp profiler backed by a single wgpu::QuerySet.

PassTiming

Profiled timing for a single GPU pass.