# Large performance regression in nightly build

**URL:** <https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440>\
**Category:** Performance\
**Created:** [February 14, 2022, 3:39pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440 "2022-02-14T15:39:14Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![joa-quim](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/joa-quim/32/227_2.png) [@joa-quim](https://discourse.julialang.org/u/joa-quim)\
**Post date:** [February 14, 2022, 3:39pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/1 "2022-02-14T15:39:14Z")

</div>

Update the 1.8dev version today and …

```julia
  | | |_| | | | (_| | | Version 1.8.0-DEV.1519 (2022-02-13)
 _/ |\ __'_|_|_|\__'_| | Commit 1978580e72 (1 day old master)
|__/ |

julia> cd("C:/v"); @time using GMT
[Info: Precompiling GMT [5752ebe1-31b9-557e-87aa-f909b540aa54]
 28.079569 seconds (2.25 M allocations: 138.780 MiB, 0.20% gc time, 7.73% compilation time)

julia> tic(); plot(rand(5,2)); toc()
elapsed time: 23.5743263 seconds
23.5743263

```

vs 1.7

```julia
  | | |_| | | | (_| | | Version 1.7.0 (2021-11-30)
 _/ |\ __'_|_|_|\__'_| | Official https://julialang.org/ release
|__/ |

julia> cd("C:/v"); @time using GMT
[Info: Precompiling GMT [5752ebe1-31b9-557e-87aa-f909b540aa54]
 20.927768 seconds (1.92 M allocations: 120.275 MiB, 0.23% gc time, 3.40% compilation time)

julia> tic(); plot(rand(5,2)); toc()
elapsed time: 6.4301933 seconds
6.4301933

```

the 1.8dev from ~1 month ago used to run TTFP in 5.5 sec.

---

<div class="post-metadata">

**Author:** ![Oscar\_Smith](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/oscar_smith/32/25343_2.png) [@Oscar\_Smith](https://discourse.julialang.org/u/Oscar_Smith)\
**Post date:** [February 14, 2022, 3:51pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/2 "2022-02-14T15:51:17Z")

</div>

Is this the same version of `GMT`?

---

<div class="post-metadata">

**Author:** ![joa-quim](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/joa-quim/32/227_2.png) [@joa-quim](https://discourse.julialang.org/u/joa-quim)\
**Post date:** [February 14, 2022, 3:54pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/3 "2022-02-14T15:54:02Z")

</div>

Yes, It’s the one I build (and distribute on Windows)

---

<div class="post-metadata">

**Author:** ![joa-quim](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/joa-quim/32/227_2.png) [@joa-quim](https://discourse.julialang.org/u/joa-quim)\
**Post date:** [February 15, 2022, 5:35pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/4 "2022-02-15T17:35:54Z")

</div>

No idea why it was so slow yesterday (and I tried several times). Trying again today sows more reasonable numbers, but the runtime is still slower (~0.75 sec) than the ~1 month ago version.

```julia
  | | |_| | | | (_| | | Version 1.8.0-DEV.1519 (2022-02-13)
 _/ |\ __'_|_|_|\__'_| | Commit 1978580e72 (2 days old master)
|__/ |

julia> cd("C:/v"); @time using GMT
[Info: Precompiling GMT [5752ebe1-31b9-557e-87aa-f909b540aa54]
 23.074192 seconds (2.25 M allocations: 138.718 MiB, 0.10% gc time, 2.79% compilation time)

julia> tic(); plot(rand(5,2)); toc()
elapsed time: 6.2270409 seconds
6.2270409

```

---

<div class="post-metadata">

**Author:** ![Krastanov](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/krastanov/32/6817_2.png) [@Krastanov](https://discourse.julialang.org/u/Krastanov)\
**Post date:** [February 15, 2022, 5:39pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/5 "2022-02-15T17:39:41Z")

</div>

Few weeks ago I did some `git bisect` between 1.7 and nightly. There were multiple speed-ups and slow-downs of TTFX. Compilation/inference/etc performance definitely has many significant ups and downs over the weeks. I have been thinking of setting up a daily tracker with the TTFX for various libraries on nightly.

---

<div class="post-metadata">

**Author:** ![longemen3000](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/longemen3000/32/7298_2.png) [@longemen3000](https://discourse.julialang.org/u/longemen3000)\
**Post date:** [February 15, 2022, 10:09pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/6 "2022-02-15T22:09:01Z")

</div>

oh, good, i was experimenting the same thing with nightly CI

---

<div class="post-metadata">

**Author:** ![joa-quim](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/joa-quim/32/227_2.png) [@joa-quim](https://discourse.julialang.org/u/joa-quim)\
**Post date:** [February 25, 2022, 11:05pm UTC](https://discourse.julialang.org/t/large-performance-regression-in-nightly-build/76440/7 "2022-02-25T23:05:13Z")

</div>

And seems to be happening again

```julia
  | | |_| | | | (_| | | Version 1.9.0-DEV.84 (2022-02-24)
 _/ |\ __'_|_|_|\__'_| | Commit 6b29ebda35 (1 day old master)
|__/ |

julia> cd("C:/v"); @time using GMT
  1.394454 seconds (1.03 M allocations: 76.999 MiB, 1.85% gc time, 44.77% compilation time)

julia> tic(); plot(rand(5,2)); toc()
elapsed time: 9.3089013 seconds
9.3089013

```