Skip to content

Intervals.NET.Benchmarks.Benchmarks.VariableStepDomainBenchmarks-report-github.md

Latest commit

 

History

History
181 lines (147 loc) · 10.6 KB

Intervals.NET.Benchmarks.Benchmarks.VariableStepDomainBenchmarks-report-github.md

File metadata and controls

181 lines (147 loc) · 10.6 KB

BenchmarkDotNet v0.13.12, Windows 10 (10.0.19045.6456/22H2/2022Update)
Intel Core i7-1065G7 CPU 1.30GHz, 1 CPU, 8 logical and 4 physical cores
.NET SDK 8.0.403
  [Host]     : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  DefaultJob : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
Method Mean Error StdDev Median Ratio RatioSD Completed Work Items Lock Contentions Allocated Alloc Ratio
DateTimeDomain_Add_5BusinessDays 39.4229 ns 0.4022 ns 0.3140 ns 39.4494 ns 1.00 0.00 - - - NA
DateTimeDomain_Add_20BusinessDays 186.6182 ns 6.0837 ns 17.9380 ns 188.0122 ns 4.08 0.16 - - - NA
DateTimeDomain_Add_100BusinessDays 1,081.4680 ns 17.0909 ns 15.1506 ns 1,077.4330 ns 27.35 0.38 - - - NA
DateOnlyDomain_Add_5BusinessDays 19.1567 ns 0.2813 ns 0.2631 ns 19.1349 ns 0.49 0.01 - - - NA
DateOnlyDomain_Add_20BusinessDays 75.4422 ns 1.0876 ns 1.5598 ns 75.0753 ns 1.92 0.05 - - - NA
DateTimeDomain_Distance_5Days 52.1208 ns 1.0981 ns 1.5394 ns 51.7783 ns 1.31 0.03 - - - NA
DateTimeDomain_Distance_30Days 308.2804 ns 5.7216 ns 5.3520 ns 308.0863 ns 7.80 0.15 - - - NA
DateTimeDomain_Distance_365Days 3,617.8666 ns 68.9153 ns 84.6342 ns 3,607.1239 ns 91.20 2.96 - - - NA
DateOnlyDomain_Distance_30Days 137.0923 ns 5.9434 ns 16.6659 ns 132.0591 ns 3.39 0.26 - - - NA
DateTimeDomain_Floor_Monday 4.1847 ns 0.1358 ns 0.1204 ns 4.1585 ns 0.11 0.00 - - - NA
DateTimeDomain_Floor_Saturday 5.2575 ns 0.0937 ns 0.0876 ns 5.2646 ns 0.13 0.00 - - - NA
DateOnlyDomain_Floor_Monday 0.7161 ns 0.0429 ns 0.0380 ns 0.7161 ns 0.02 0.00 - - - NA
DateTimeDomain_Ceiling_Monday 4.9287 ns 0.1305 ns 0.1019 ns 4.8977 ns 0.13 0.00 - - - NA
DateTimeDomain_Ceiling_FridayWithTime 9.6422 ns 0.2095 ns 0.1749 ns 9.6177 ns 0.24 0.00 - - - NA
DateTimeDomain_Ceiling_Sunday 9.0433 ns 0.1274 ns 0.1129 ns 9.0213 ns 0.23 0.00 - - - NA
DateTimeDomain_Span_5Days 67.9617 ns 1.4306 ns 2.1412 ns 67.3845 ns 1.72 0.07 - - - NA
DateTimeDomain_Span_30Days 315.4466 ns 4.8310 ns 4.0341 ns 315.2319 ns 8.01 0.09 - - - NA
DateTimeDomain_Span_365Days 3,479.8399 ns 56.1281 ns 52.5022 ns 3,456.6425 ns 88.58 1.55 - - - NA
DateOnlyDomain_Span_30Days 131.1173 ns 2.6968 ns 6.4092 ns 130.3988 ns 3.32 0.14 - - - NA
DateTimeDomain_ExpandByRatio_30Days 440.8805 ns 4.7828 ns 3.9939 ns 440.8098 ns 11.19 0.09 - - - NA
DateTimeDomain_ExpandByRatio_365Days 4,825.0244 ns 75.1212 ns 66.5930 ns 4,821.7590 ns 122.30 2.09 - - - NA
DateOnlyDomain_ExpandByRatio_30Days 195.9311 ns 3.9509 ns 7.0228 ns 194.8128 ns 4.96 0.12 - - - NA

Summary

What This Measures

Variable-step domain performance—business day calculations that must iterate through ranges to skip weekends. Tests DateTime and DateOnly business day domains across operations that require day-by-day checking (O(N) complexity).

Key Performance Insights

⏱️ Add Operation: O(N) Iteration Required

  • 5 business days: 39.4 ns (DateTime), 19.2 ns (DateOnly)
  • 20 business days: 186.6 ns (DateTime), 75.4 ns (DateOnly)
  • 100 business days: 1,081 ns (DateTime)
  • Scaling: ~9.4 ns per business day (DateTime), ~3.8 ns per day (DateOnly)

📏 Distance: O(N) Day-by-Day Counting

  • 5 days (weekdays): 52.1 ns
  • 30 days (~21 business days): 308.3 ns
  • 365 days (~260 business days): 3,617.9 ns (3.6 μs)
  • Scaling: ~9.9 ns per calendar day checked

📐 Span: Distance + Floor/Ceiling Alignment

  • 5 days: 67.96 ns (combines Distance with alignment)
  • 30 days: 315.4 ns
  • 365 days: 3,479.8 ns (3.5 μs)
  • Overhead: ~15 ns more than Distance (for boundary alignment)

🔢 ExpandByRatio: Span + Add Operations

  • 30 days: 440.9 ns (Span: 315ns + two Add operations)
  • 365 days: 4,825 ns (4.8 μs)
  • Composition: Combines multiple O(N) operations

⚡ Floor/Ceiling: O(1) Boundary Alignment

  • DateTime Floor: 4.2-5.3 ns (check day of week)
  • DateTime Ceiling: 4.9-9.6 ns (may need forward scan to Monday)
  • DateOnly Floor: 0.72 ns (simpler, no time component)

Performance Scaling Analysis

Operation        5 days    30 days    365 days    Per-Day Cost
──────────────────────────────────────────────────────────────
Add (DateTime)   39 ns     187 ns     1,081 ns    ~9.4 ns
Distance         52 ns     308 ns     3,618 ns    ~9.9 ns
Span             68 ns     315 ns     3,480 ns    ~9.5 ns
ExpandByRatio    N/A       441 ns     4,825 ns    ~13.2 ns

DateTime vs DateOnly Performance

Operation           DateTime (ns)    DateOnly (ns)    Speedup
──────────────────────────────────────────────────────────────
Add (5 days)        39.4             19.2             2.1×
Add (20 days)       186.6            75.4             2.5×
Distance (30 days)  308.3            137.1            2.2×
Span (30 days)      315.4            131.1            2.4×

Why DateOnly is faster:

  • No time component to preserve
  • Simpler day arithmetic
  • Fewer allocations/checks

Design Trade-offs (Aligned with README Philosophy)

Why operations are O(N):

  • ❌ Cannot use arithmetic: weekends are non-uniform gaps
  • ✅ Must iterate day-by-day to check IsBusinessDay()
  • ✅ No way around iteration for variable-step domains
  • ⚠️ Performance scales linearly with range size

When O(N) is acceptable:

  • Typical business scenarios: < 100 business days (< 1 μs)
  • SLA calculations: "5 business days from now" (40 ns)
  • Deadline tracking: "15 business days remaining" (150 ns)

When O(N) becomes expensive:

  • Year-long calculations: 365 days = 3.6 μs
  • Multi-year ranges: 1,000 days = ~10 μs
  • High-frequency: millions of calls may accumulate cost

Practical Recommendations

✅ Use business day domains for:

  • Contract deadlines: "5 business days to respond" (40 ns)
  • SLA tracking: "2 business days turnaround" (20 ns)
  • Workweek calculations: "business days this month" (~300 ns for 30-day range)

⚠️ Be aware of scaling:

  • Short ranges (< 30 days): Sub-microsecond performance
  • Medium ranges (30-90 days): 300-900 ns (acceptable)
  • Long ranges (365+ days): 3-10 μs (still fast, but watch loop counts)

✅ Optimization strategies:

  • Cache business day counts if calculating repeatedly
  • Use DateOnly instead of DateTime when possible (2× faster)
  • Consider fixed-step approximations for very long ranges

Real-World Impact

SLA Deadline Calculator (10,000 requests/second):

var deadline = domain.Add(DateTime.Now, 5); // 5 business days
Cost: 39 ns × 10,000 = 390 microseconds/second
Impact: Negligible (0.039% CPU)

Business Days Remaining Widget (updated 60×/second):

var remaining = range.Span(businessDayDomain); // 30-day range
Cost: 315 ns × 60 = 18.9 microseconds/second
Impact: Negligible

Annual Business Day Report (1000 year-long calculations):

var businessDays = yearRange.Span(domain); // 365-day range
Cost: 3,618 ns × 1000 = 3.6 milliseconds
Impact: Acceptable for batch reporting

Memory Behavior

All operations:  0 bytes allocated
Iteration:       Stack-only, no heap allocations
Domain instance: Stateless, reusable

Comparison: Variable vs Fixed-Step

Characteristic Variable-Step (Business Days) Fixed-Step (Calendar Days)
Complexity O(N) - iteration required O(1) - pure arithmetic
Add (5 steps) 39 ns 0.74 ns (53× faster)
Distance (30) 308 ns 7.58 ns (41× faster)
Span (365) 3,480 ns ~20 ns (174× faster)
Use case Business logic, SLAs Simple date math

Why This Matters

These benchmarks show that variable-step domains have O(N) performance but remain practical for business scenarios:

  • ✅ Short ranges (< 30 days): Sub-microsecond
  • ✅ Medium ranges (30-90 days): Low microseconds
  • ✅ Long ranges (365 days): Still only 3.6 μs
  • ⚠️ Be mindful in tight loops with long ranges

Choose variable-step domains when business logic requires it (weekends, holidays), and fixed-step domains when O(1) performance is critical.