The Delegate Is Optional. The Display Class Is Not
A lambda that captures an int allocated 24 bytes on .NET 10.0.12; the same lambda reading a constant allocated none. The delegate was optional, the display class was not.
11 articles
Articles
Long-form writing on C#, the .NET runtime, and the frameworks built on them.
November
2026 · 1 article
October
2026 · 4 articles
A Shared lock Cost the Same as a Private One at One Thread, and Not at Two
On .NET 10.0.12 a shared lock matched a per-thread lock at one thread and cost more from two up, where Monitor.LockContentionCount left zero.
The Query EF Core Writes When You Ask for a Count
On EF Core 10.0.12, AsEnumerable() before Where().Count() sent a full-table SELECT instead of COUNT(*) and allocated 1056.36 KB against 49.53 KB at 1,000 rows.
clrstack Shows the Waiter and syncblk Names the Lock Holder
In a dotnet-dump of a contested lock, clrstack -all marked no thread as the owner. syncblk named it, and a suspended await appeared only in dumpasync.
A Container Given 1.2 CPUs Tells the Runtime It Has 2
Environment.ProcessorCount is the CPU quota rounded up, never below 1. The thread pool minimum follows it, and Server GC switches itself off at 1.
September
2026 · 3 articles
The Box You Pay For Is the One That Escapes
Assigning a struct to an interface allocated 0 bytes. Passing it across a call boundary allocated 24. A constrained generic allocated none, until its body asked for an object.
The Middleware That Looked Expensive Was Not the One Spending the Bytes
Removing UseRouting dropped per-request allocation by 2.29 KB in isolation — separated from authorization's share, routing itself costs about 24 bytes.
dotnet-counters Shows the Queue's Delta, Not Its Depth
A thread pool queue held at two thousand work items read 1986 through one provider and -13 through the other, in the same second, in the same process.
August
2026 · 2 articles
Server GC, Workstation GC, and the Numbers Between Them
Server GC finished 2.7 times faster and collected 17 times less often — and made the slowest request in the run 5 to 13 times slower. Both numbers are real.
Where Thread Pool Starvation Actually Comes From
Blocking on a task is the case the thread pool compensates for. Blocking on a sleep or an event is the case it cannot see, and it finished 2.5 times slower.
