Rewriting the Hot Path in Mojo
Translate the bottleneck with fn and SIMD.
Rewriting the Hot Path in Mojo is a free Mojo Academy lesson on CoddyKit — lesson 2 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the Mojo Academy learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
Translate the Hot Path
Now take that slow Python loop and rebuild it in Mojo. You keep the same logic but gain real performance from typed, compiled code. 🔧
Reach for fn
Write the kernel as an fn function, not def. Required types let the compiler generate tight machine code for your hot loop.
fn sum_squares(data: List[Float64]) -> Float64:
var total: Float64 = 0.0
return totalAnnotate Every Type
Give each value a concrete type. With no dynamic guessing left, Mojo can optimize the arithmetic the way C would.
var total: Float64 = 0.0
var n: Int = len(data)Port the Loop Body
Copy the inner math one step at a time. Keep the same operations so results match Python exactly before you optimize anything.
for i in range(len(data)):
total += data[i] * data[i]Verify Correctness First
Run the Mojo version on the same input and compare. Correctness comes before speed, since a fast wrong answer helps no one.
Bring In SIMD
Once it is correct, vectorize. A SIMD value packs several numbers so one instruction processes many elements at once.
alias width = 4
var acc = SIMD[DType.float64, width](0)Process in Vector Chunks
Step through the data in groups equal to the SIMD width. Each pass handles a whole chunk instead of one lonely element.
for i in range(0, n, width):
var chunk = data.load[width](i)
acc += chunk * chunkReduce the Vector
After the loop, collapse the SIMD accumulator into one number with a reduce. That folds every lane down to a single total.
var total = acc.reduce_add()Handle the Tail
If the length is not a clean multiple of the width, a few elements remain. A small tail loop adds them so nothing is dropped.
for i in range(n - (n % width), n):
total += data[i] * data[i]Re-check the Result
Compare the vectorized output to the Python baseline again. Matching results confirm your SIMD rewrite kept the math honest.
Time the New Kernel
Measure the Mojo kernel on the same data. Even single-threaded, a typed and vectorized kernel usually beats the Python loop by a lot.
Quick Check
Pick the right order for rewriting a hot path in Mojo.
Recap
You ported the loop to a typed fn, verified correctness, then vectorized with SIMD and handled the tail. The hot path is now fast Mojo. 🎯
Frequently asked questions
Is the “Rewriting the Hot Path in Mojo” lesson free?
Yes — the full text of “Rewriting the Hot Path in Mojo” is free to read here on the web, and the Mojo Academy course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the Mojo Academy course, upgrade to CoddyKit PRO.
What will I learn in “Rewriting the Hot Path in Mojo”?
Translate the bottleneck with fn and SIMD. You practise Mojo Academy with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start Mojo Academy?
No prior experience is required. Mojo Academy on CoddyKit is structured for beginners through advanced learners; this is — lesson 2 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Rewriting the Hot Path in Mojo” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this Mojo Academy lesson?
Yes. Every Mojo Academy lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- Profiling the Python Baseline
- Rewriting the Hot Path in Mojo
- Parallelizing and Tuning the Core
- Shipping the Accelerated Project