L2 & Last-Level Cache: Reports & Metrics

Reports and metrics

Reports & Metrics for L2 & Last-Level Cache centers on L2 hit ratio, eviction pressure, and inter-SM coherence traffic. The objective is to connect profiler evidence to root-cause mechanism and release-safe action.

Reports must turn L2 hit ratio, eviction pressure, and inter-SM coherence traffic into a release decision. Single counter improvements are insufficient without workload context and traceability.

Trend snapshot

BEFORE / AFTER — L2 & Last-Level Cache

metric quality
  ^
  |                        o target region
  |                 o post-fix validation
  |            o
  |      o baseline (failing)
  +------------------------------------------> iteration
      evidence capture  mechanism fix  closure

Use this to prove improvement is causal, not incidental.

Roofline interpretation

BANDWIDTH ROOFLINE — L2 & Last-Level Cache

performance
   ^
   |                compute ceiling
   |               /
   |              /
   |-------------/------------------ memory ceiling
   +------------------------------------------> operational intensity
      memory-bound             compute-bound

Interpretation: identify compute vs memory bound

• Track L2 hit ratio, eviction pressure, and inter-SM coherence traffic across representative workloads, not one microbenchmark.

• Include counter captures with matching compiler, driver, and firmware tags.

• Correlate scheduler stalls with memory and interconnect pressure before optimization.

• Report frame or kernel tail behavior, not only average throughput.

GPU deep dive

Bandwidth wins come from coalescing and locality discipline, not peak-memory specs alone.

Concept diagram

MEMORY HIERARCHY

register -> shared/L1 -> L2/LLC -> HBM/GDDR
access pattern quality decides latency

Metric graph

BANDWIDTH UTILIZATION

requested BW  ███████████
effective BW  ████████
wasted BW     ███

Reports and artifacts

• L1/L2 hit-rate report

• HBM efficiency counters

• coalescing transaction log

• shared-memory bank audit

Mini case study

Stencil kernel sat at 43% of peak HBM due to uncoalesced loads; layout rewrite recovered 1.6x effective bandwidth.

Debug branches

• Check transactions per request at warp granularity

• Classify cache-thrash versus true DRAM saturation

• Audit shared-memory bank conflicts before algorithm rewrites

Senior review question

Ask: which metric and benchmark pairing proves this topic is truly closed in production context?

Key takeaways

• Always pair micro-kernel metrics with end-to-end workload impact.

• Lock toolchain, driver, and launch metadata before comparing performance results.

Common pitfalls

• Optimizing occupancy without checking memory-system saturation.

• Comparing profiler captures from different driver or compiler builds.

• Declaring wins without reproducible accuracy and performance gates.

Report interpretation

A shared L2/LLC smooths off-chip traffic and inter-client sharing, but contention and policy choices can shift bottlenecks across engines. This mechanism matters because GPUs are throughput machines: a small inefficiency repeated across lanes, warps, SMs, frames, or dispatches can dominate product performance even when a unit-level diagram looks balanced.

Use L2 hit ratio, eviction pressure, and inter-SM coherence traffic as an entry point, not as the conclusion. A metric shift only becomes actionable after it is tied to a workload slice, a profiler capture, an architectural path, and a reproducible artifact such as L2 traffic breakdown, eviction reason chart, and bandwidth pressure summary.

GPU memory systems win when access regularity, cache policy, and bandwidth provisioning are co-designed. The review posture is therefore evidence-first: explain what the kernel or graphics workload asked for, how the GPU mapped it onto hardware, where useful work stopped, and which owner can change the smallest boundary safely.

For L2 & Last-Level Cache, reports should explain why L2 hit ratio, eviction pressure, and inter-SM coherence traffic changed, not merely that it changed. Ask whether the movement came from useful work, reduced waste, different scheduling, changed memory traffic, or hidden throttling.

A strong report includes counter consistency checks: the story told by occupancy should agree with issue activity; the memory story should agree with cache and transaction behavior; the silicon story should agree with clocks, voltage, thermals, and power telemetry.