Skip to content
STIMSMITH

Data Cache Miss Rate

Concept

Data Cache Miss Rate is a workload-characterization metric that quantifies the fraction of memory accesses to data that fail to be served by a CPU's data cache (L1, L2, etc.), forcing fetches from a lower level of the memory hierarchy. In workload-generation frameworks such as Genesys it is exposed as a user-controllable knob (split into separate L1 and L2 values) that, together with spatial- and temporal-locality parameters, determines the stride pattern of generated memory accesses.

First seen 6/13/2026
Last seen 6/13/2026
Evidence 2 chunks
Wiki v1

WIKI

Overview

Data Cache Miss Rate is a performance metric that measures the proportion of data memory references that cannot be satisfied by a given level of the CPU data cache and therefore must be resolved by a higher (slower) level of the memory hierarchy. It is the complement of the data cache hit rate for that level: miss_rate = 1 − hit_rate. The metric is commonly tracked separately per cache level (e.g., L1 and L2 data-cache miss rates) because the latency penalty and the access pattern that produces misses differ substantially between levels.

In workload-synthesis and characterization research, data cache miss rate is treated as a first-order knob because memory hierarchy behavior is widely recognized as a major determinant of application performance. [Citation: Genesys paper, Methodology section]

READ FULL ARTICLE →

NEIGHBORHOOD

2 nodes · 1 edges
graph · Data Cache Miss Rate · depth=1

RELATIONSHIPS

1 connections
Genesys ← uses 100% 2e
Genesys models L1/L2 data cache miss rates as core metrics.

CITATIONS

5 sources
5 citations — click to expand
[1] Genesys enumerates 'L1/L2 Data cache miss rates' as metric 12 (2 values) under Memory-access Characteristics, with 32 spatial-locality stride bins and 8 temporal-locality bins as related knobs. Genesys: Automatically Generating Representative Workloads (Samos 2016)
[2] Temporal locality controls the number of unique memory accesses between accesses to the same memory location and affects the achieved cache miss rates. Genesys: Automatically Generating Representative Workloads (Samos 2016)
[3] Genesys can automatically estimate strides (offsets) of load/store instructions from target data cache miss rate statistics by matching L1 hit rate first and then L2 hit rate, using a pre-computed correlation table. Genesys: Automatically Generating Representative Workloads (Samos 2016)
[4] Spatial locality is controlled by stride bins from -1K to +1K in multiples of 64 B; temporal locality is expressed as powers-of-two from 0 to 128; together they generate the sequence of memory addresses. Genesys: Automatically Generating Representative Workloads (Samos 2016)
[5] Genesys treats data locality principles as major determinants of application performance and includes data cache miss rate metrics among its user-controllable workload knobs. Genesys: Automatically Generating Representative Workloads (Samos 2016)