bitnet-benchmark-categories.mdc

---
description: "Define categories of benchmarks for the BitNet project to ensure focused and comparable measurements."
globs: pkg/bitnet/**/*.go
alwaysApply: false
---

# Benchmark Categories

**Purpose:** Classify benchmarks by their semantic focus so teams can compare like with like.

## 1. Creation Benchmarks

Measure cost of allocating or initializing a component.

```go
func BenchmarkTensor_Create(b *testing.B) {
  for i := 0; i < b.N; i++ {
    NewTensor(100)
  }
}
```

## 2. Operation Benchmarks

Measure runtime of core operations on an existing instance.

```go
func BenchmarkTensor_Get(b *testing.B) {
  tensor := NewTensor(1000)
  b.ResetTimer()
  for i := 0; i < b.N; i++ {
    tensor.Get(i % 1000)
  }
}
```

## 3. Composite / Sub-operation Benchmarks

Combine multiple operations or simulate realistic sequences.

```go
func BenchmarkTensor_Sequential(b *testing.B) {
  tensor := NewTensor(1000)
  b.Run("GetSet", func(b *testing.B) {
    for i := 0; i < b.N; i++ {
      tensor.Set(1.23, i%1000)
      tensor.Get(i%1000)
    }
  })
}
```

## 4. Memory & Allocation Benchmarks

Measure allocations and memory footprint per operation.

```go
func BenchmarkAlloc_1024(b *testing.B) {
  b.ReportAllocs()
  for i := 0; i < b.N; i++ {
    _ = make([]byte, 1024)
  }
}
```

## Best Practices

* Single semantic focus per benchmark.
* Use realistic sizes and patterns.
* Report allocations with `b.ReportAllocs()`.
* Reset timers after setup (`b.ResetTimer()`).