Speedup
Speedup is the ratio of the execution time for a baseline computation to the execution time for an alternative implementation of the same computation.
Grokking Concurrency
Speedup is the ratio comparing the time required by a baseline execution with the time required by a parallel execution.
- How it is used
- Measures the performance benefit of parallel execution against an optimal sequential baseline.
- How deep it goes
- Explains fixed-work and fixed-time interpretations, bottlenecks, and a process-pool measurement without deriving a formal law.
- Examples
- Uses card sorting, shopping checkout lines, and Python password cracking to illustrate bottlenecks and measured gains.
- What you can do after
- Evaluate parallel designs by comparing execution times, identifying sequential limits, and interpreting throughput-oriented gains.
Grokking GPU Programming
Speedup is the ratio of a computation's sequential execution time to its parallel execution time.
- How it is used
- Measures how much faster parallel execution is than sequential execution and helps judge whether parallelizing is worthwhile.
- How deep it goes
- Defines speedup as a time ratio and compares Amdahl’s and Gustafson’s formulas, assumptions, and numerical predictions.
- Examples
- Uses 1,000-element array doubling and a 10%-sequential workload to contrast ideal parallelism with predicted speedup.
- What you can do after
- Evaluate potential gains by estimating sequential work and distinguishing fixed-size from growing workloads.
Learn Concurrent Programming with Go
Speedup is the ratio of a computation's execution time in a baseline configuration to its execution time in an improved or parallel configuration.
- How it is used
- Uses speedup to evaluate parallel performance, processor scalability, and the effect of sequential work.
- How deep it goes
- Explains speedup through Amdahl’s and Gustafson’s laws, examples, and figures, but stops short of deriving their formulas.
- Examples
- Illustrates speedup with house construction, builders, narrow-road transport, and a computer game’s frame rate and graphics.
- What you can do after
- Evaluate/understand why adding processors eventually produces diminishing returns or supports larger workloads.
Read how Learn Concurrent Programming with Go explains speedup →
Also covered in
- CUDA for LLMs
- Latency
- Build a DeepSeek Model (From Scratch)In early access (MEAP)
- Quantum Programming in Depth
What to know first
- CPUA central processing unit (CPU) is a computer processor that executes program instructions and performs general-purpose computational operations.
- parallelismParallelism is the execution of multiple independent tasks or computational operations at the same time rather than sequentially.
- workflowA workflow is an organized sequence of coordinated activities or tasks through which work progresses toward a defined outcome.
- meanThe mean is a measure of central location equal to the arithmetic average of numerical observations, or the probability-weighted average of a random variable’s possible values.
Related concepts
- Amdahl’s Lawmeasured by · Amdahl’s Law describes how the speedup from optimizing or parallelizing a computation is limited by the fraction that remains sequential or unoptimized.
- Gustafson's lawmeasured by · Gustafson's law models parallel speedup when problem size grows with processing resources, making fixed sequential work a smaller proportion of total work.
- operator coverageproduced by · Operator coverage is the proportion of a model graph's operator instances that execute using a specified numerical representation such as INT8.
- strong scalingmeasured by · Strong scaling is the behavior of a fixed-size computational problem as additional processors or GPUs are added, evaluated by the resulting speedup.
- parallelismmeasured by · Parallelism is the execution of multiple independent tasks or computational operations at the same time rather than sequentially.
Questions about speedup
What is speedup?
Speedup is the ratio of the execution time for a baseline computation to the execution time for an alternative implementation of the same computation.
Which Manning books explain speedup?
Grokking Concurrency, Grokking GPU Programming and Learn Concurrent Programming with Go.
What should I know before learning speedup?
Manning books that teach speedup usually introduce CPU, parallelism, workflow and mean first.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.


