Precompile show, \, products, broadcast and concatenation of sparse Float64 arrays

[ViralBShah]

SparseArrays is not in the sysimage, and its package image holds almost no native code (a 56 KB .so), so the first call of every common operation compiles: show of a matrix takes about 0.4 s and A\b about 0.6 s.

Before (main), first calls in a fresh session, Float64 SparseMatrixCSC with Int indices:

	load	.ji + .so	show	A\b	A*x	25 common operations
main	195 ms	7.7 MB + 0.06 MB	430 ms	605 ms	115 ms	1.95 s

After:

	load	.ji + .so	show	A\b	A*x	25 common operations
this PR	220 ms	13.2 MB + 1.5 MB	2 ms	0.4 ms	0.0 ms	0.43 s

src/precompile.jl runs the common operations on small Float64 literals behind Base.generating_output(), the pattern Profile and Pkg use, so their native code goes into the package image; nothing runs at load. The solver calls are behind Base.USE_GPL_LIBS.

The cost is 25 ms of load time, a package image about 7 MB larger, and about 4 s more to build it, which Julia's build and download carry too. Left out: ComplexF64, which would take load to 229 ms and the image to 16 MB + 2.7 MB.

🤖 Generated with Claude Code

[ViralBShah]

@KristofferC Is it ok to include this in an stdlib (which is not in the system image)?

[github-actions]

TTFX: 3 robust regressions, 3 robust improvements.

SparseArrays #1103 (cfb5048) vs a49c313 · run

• SparseArrays: precompile +169.5%, run (cold) -53.9%, load+run (warm) -30.2%
• All packages (geometric mean): precompile +37.0%, load (cold) +8.5%, load+run (warm) -6.4%

Details

Julia 1.14.0-DEV.3740 · Linux X64 · 4 blocks (ABBA)

metric	base (s)	head (s)	change per block	with GC off
precompile (regression)	5.38, 5.39, 5.39, 5.4	14.5, 14.5, 14.5, 14.6	+169.9%, +168.7%, +169.5%, +169.6%	–
load (cold)	0.261, 0.26, 0.26, 0.266	0.286, 0.286, 0.286, 0.286	+9.5%, +9.7%, +10.0%, +7.7%	+9.7%, +11.2%, +10.1%, +9.8%
run (cold) (improvement)	0.693, 0.681, 0.682, 0.685	0.318, 0.316, 0.315, 0.313	-54.1%, -53.6%, -53.8%, -54.3%	-53.2%, -50.2%, -54.9%, -54.5%
load+run (warm) (improvement)	0.524, 0.524, 0.526, 0.53	0.372, 0.366, 0.367, 0.368	-29.0%, -30.1%, -30.3%, -30.5%	–

Dependents (median change over blocks)

package	precompile	load (cold)	run (cold)	load+run (warm)
Krylov	+0.8%	+9.1%	–	+9.1%
SparseMatrixColorings	-5.6%	+6.7%	–	+6.8%

all packages (geomean)	change per block	with GC off
precompile (regression)	+36.8%, +36.9%, +37.2%, +37.0%	–
load (cold) (regression)	+8.3%, +8.9%, +8.7%, +8.0%	+8.1%, +9.0%, +8.6%, +8.6%
load+run (warm) (improvement)	-6.1%, -6.2%, -6.7%, -7.5%	–

Dependents per block

Krylov

metric	base (s)	head (s)	change per block	with GC off
precompile	6.7, 6.74, 6.72, 6.74	6.72, 6.78, 6.83, 6.81	+0.4%, +0.6%, +1.7%, +1.0%	–
load (cold)	0.282, 0.282, 0.28, 0.281	0.306, 0.306, 0.307, 0.308	+8.8%, +8.8%, +9.6%, +9.4%	+8.8%, +9.2%, +9.1%, +8.4%
load+run (warm)	0.28, 0.28, 0.281, 0.286	0.305, 0.306, 0.307, 0.305	+9.0%, +9.2%, +9.4%, +6.7%	–

SparseMatrixColorings

metric	base (s)	head (s)	change per block	with GC off
precompile	7.06, 7.04, 7.05, 7.04	6.67, 6.69, 6.64, 6.65	-5.6%, -5.1%, -5.8%, -5.5%	–
load (cold)	0.334, 0.331, 0.336, 0.332	0.356, 0.357, 0.358, 0.355	+6.5%, +8.1%, +6.5%, +6.9%	+5.8%, +6.7%, +6.8%, +7.5%
load+run (warm)	0.33, 0.33, 0.331, 0.333	0.352, 0.357, 0.353, 0.355	+6.9%, +8.1%, +6.6%, +6.6%	–

_{A change is robust when every block clears the threshold (precompile 8%, load (cold) 10%, run (cold) 15%, load+run (warm) 15%), the samples of the two commits do not overlap, and the medians differ by at least the floor (precompile 0.25 s, load (cold) 0.1 s, run (cold) 0.05 s, load+run (warm) 0.05 s). Load and run are also judged with the GC off, and either counts. Over all packages whose base value is at least the floor, each metric's geometric mean must clear a threshold in every block (precompile 2%, load (cold) 2%, run (cold) 3%, load+run (warm) 3%).}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.83%. Comparing base (a49c313) to head (7963dbe).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1103   +/-   ##
=======================================
  Coverage   94.83%   94.83%           
=======================================
  Files          16       16           
  Lines       10252    10257    +5     
=======================================
+ Hits         9722     9727    +5     
  Misses        530      530           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ViralBShah]

@IanButterworth Seems like TTFX is doing its job here. Pinging for you to take a look since I thought you may find it useful to see it in action.