DerivaSharp.Gpu.Linux 1.1.0

DerivaSharp

Financial derivatives pricing in modern C# for quantitative research, model validation, and desk analytics.

DerivaSharp provides typed instruments and interchangeable pricing engines under a consistent Black–Scholes–Merton interface. It covers closed-form benchmarks, finite-difference and tree methods, numerical integration, and TorchSharp Monte Carlo on CPU or CUDA.

The library is deliberately focused: single-underlying contracts, flat volatility and continuously compounded risk-free/dividend rates, explicit contractual schedules, and transparent numerical methods. It is a good fit for pricing research, engine cross-checks, and focused analytics services—not a complete market-data, curve-building, or portfolio-risk platform.

Pricing coverage

Every Black–Scholes–Merton engine exposes value, spot Greeks, time Greeks, volatility Greeks, rho, scenario grids, and implied volatility through the same base API. Autocallable engines also support implied coupon-rate solving.

Install

DerivaSharp targets .NET 10. Choose one package path:

Install only one TorchSharp runtime path in an application. The GPU packages bring in DerivaSharp.MonteCarlo and the platform-specific CUDA runtime.

Quick start

Price a one-year at-the-money European call and calculate its risk measures:

using DerivaSharp.Instruments;
using DerivaSharp.Models;
using DerivaSharp.PricingEngines;
using DerivaSharp.Time;

DateOnly valuationDate = new(2025, 1, 6);
EuropeanOption option = new(
    OptionType.Call,
    strikePrice: 100.0,
    effectiveDate: valuationDate,
    expirationDate: valuationDate.AddDays(365));

BsmModelParameters model = new(
    volatility: 0.30,
    riskFreeRate: 0.04,
    dividendYield: 0.01);

PricingContext<BsmModelParameters> context = new(
    model,
    AssetPrice: 100.0,
    valuationDate,
    NullCalendar.Shared);

AnalyticEuropeanEngine engine = new();
PricingResult result = engine.ValueAndGreeks(option, context);
double impliedVolatility = engine.ImpliedVolatility(option, context, optionPrice: 13.151137);

Console.WriteLine($"Value: {result.Value:F6}");             // 13.151137
Console.WriteLine($"Delta: {result.Delta:F6}");             // 0.592749
Console.WriteLine($"Implied vol: {impliedVolatility:P2}");   // 30.00%

Swap the engine without changing the instrument or market context:

BsmPricingEngine<EuropeanOption> finiteDifference =
    new FdEuropeanEngine(FiniteDifferenceScheme.CrankNicolson, 500, 500);

BsmPricingEngine<EuropeanOption> monteCarlo =
    new McEuropeanEngine(pathCount: 500_000, stepCount: 2, useCuda: true, seed: 42);

The Monte Carlo example requires DerivaSharp.MonteCarlo and one compatible TorchSharp runtime package.

Conventions

• Volatility and rates use decimal units: 0.20 means 20%.
• Time to expiry uses Actual/365 Fixed.
• Theta, Charm, and Color are reported per calendar day.
• Vega, Vanna, Zomma, and Rho are reported per one percentage-point move.
• PricingContext<T> carries valuation date, spot, model parameters, and the trading calendar used for schedules and path-dependent grids.
• Built-in calendars include an all-days NullCalendar and an SSE trading calendar; custom calendars implement ICalendar.

Structured-product schedules

Contractual observation dates are explicit and validated against the pricing calendar. Monthly schedules can be generated with following adjustment:

DateOnly effectiveDate = new(2025, 1, 6);
DateOnly expirationDate = new(2026, 1, 6);

IReadOnlyList<DateOnly> observationDates = Schedule.CreateMonthly(
    effectiveDate,
    expirationDate,
    lockUpMonths: 3,
    SseCalendar.Shared);

See the research notebooks for fuller workflows:

• European option value and Greeks
• Snowball finite-difference and Monte Carlo profiles
• Accumulator analytics

Development

The .NET 10 SDK is required.

dotnet restore DerivaSharp.slnx
dotnet build DerivaSharp.slnx --no-restore
dotnet test --project tests/DerivaSharp.Tests/DerivaSharp.Tests.csproj --no-build
dotnet test --project tests/DerivaSharp.MonteCarlo.Tests/DerivaSharp.MonteCarlo.Tests.csproj --no-build

Bug reports and focused contributions are welcome through GitHub Issues and pull requests.

License

DerivaSharp is available under the MIT License.