feat: Expose opt_level compiler config option

[haakon-e]

Pass the opt_level keyword argument through the compiler pipeline, allowing users to control the optimization level for Metal shader compilation via _compiler_config and the @metal macro.

---

If this feature is of interest, I'm happy to add tests, docs, etc., as needed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.67%. Comparing base (65fac52) to head (01439d5).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #769      +/-   ##
==========================================
+ Coverage   80.45%   80.67%   +0.21%     
==========================================
  Files          61       61              
  Lines        2855     2846       -9     
==========================================
- Hits         2297     2296       -1     
+ Misses        558      550       -8     

☔ View full report in Codecov by Sentry.
📢 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.

[github-actions]

Metal Benchmarks

Details

Benchmark suite	Current: 01439d5	Previous: acbbd34	Ratio
array/accumulate/Float32/1d	1122834 ns	1132709 ns	0.99
array/accumulate/Float32/dims=1	1538875 ns	1560896 ns	0.99
array/accumulate/Float32/dims=1L	9839917 ns	9836000 ns	1.00
array/accumulate/Float32/dims=2	1885583 ns	1881562.5 ns	1.00
array/accumulate/Float32/dims=2L	7231041.5 ns	7213791.5 ns	1.00
array/accumulate/Int64/1d	1255833 ns	1251709 ns	1.00
array/accumulate/Int64/dims=1	1842146 ns	1843416.5 ns	1.00
array/accumulate/Int64/dims=1L	11667812.5 ns	11749875 ns	0.99
array/accumulate/Int64/dims=2	2160791 ns	2128375 ns	1.02
array/accumulate/Int64/dims=2L	9731500 ns	9792208 ns	0.99
array/broadcast	609250 ns	600250 ns	1.01
array/construct	6000 ns	6375 ns	0.94
array/permutedims/2d	1170208 ns	1169958 ns	1.00
array/permutedims/3d	1670500 ns	1647292 ns	1.01
array/permutedims/4d	2387667 ns	2390042 ns	1.00
array/private/copy	568792 ns	547895.5 ns	1.04
array/private/copyto!/cpu_to_gpu	803958 ns	789667 ns	1.02
array/private/copyto!/gpu_to_cpu	807584 ns	802833.5 ns	1.01
array/private/copyto!/gpu_to_gpu	648479.5 ns	645417 ns	1.00
array/private/iteration/findall/bool	1421937 ns	1401271 ns	1.01
array/private/iteration/findall/int	1564396 ns	1552853.5 ns	1.01
array/private/iteration/findfirst/bool	2033625 ns	2046729.5 ns	0.99
array/private/iteration/findfirst/int	2069500 ns	2093500 ns	0.99
array/private/iteration/findmin/1d	2510000 ns	2506709 ns	1.00
array/private/iteration/findmin/2d	1792875 ns	1792000 ns	1.00
array/private/iteration/logical	2596709 ns	2638750 ns	0.98
array/private/iteration/scalar	5593562.5 ns	4889812.5 ns	1.14
array/random/rand/Float32	1170625 ns	1157334 ns	1.01
array/random/rand/Int64	1333667 ns	1281416 ns	1.04
array/random/rand!/Float32	909667 ns	933708 ns	0.97
array/random/rand!/Int64	868750 ns	877500 ns	0.99
array/random/randn/Float32	1072333.5 ns	1075625 ns	1.00
array/random/randn!/Float32	815041 ns	821083.5 ns	0.99
array/reductions/mapreduce/Float32/1d	1041812.5 ns	1026395.5 ns	1.02
array/reductions/mapreduce/Float32/dims=1	831041.5 ns	847750 ns	0.98
array/reductions/mapreduce/Float32/dims=1L	1333709 ns	1320958 ns	1.01
array/reductions/mapreduce/Float32/dims=2	851291 ns	856062.5 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	1819833 ns	1802542 ns	1.01
array/reductions/mapreduce/Int64/1d	1523666.5 ns	1356854.5 ns	1.12
array/reductions/mapreduce/Int64/dims=1	1097000 ns	1113917 ns	0.98
array/reductions/mapreduce/Int64/dims=1L	2010833 ns	2048021 ns	0.98
array/reductions/mapreduce/Int64/dims=2	1140104 ns	1157312.5 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	3618167 ns	3584708 ns	1.01
array/reductions/reduce/Float32/1d	1045417 ns	1030312.5 ns	1.01
array/reductions/reduce/Float32/dims=1	823437.5 ns	849459 ns	0.97
array/reductions/reduce/Float32/dims=1L	1317459 ns	1282250 ns	1.03
array/reductions/reduce/Float32/dims=2	852459 ns	779458.5 ns	1.09
array/reductions/reduce/Float32/dims=2L	1807771 ns	1811125 ns	1.00
array/reductions/reduce/Int64/1d	1534833 ns	1352583.5 ns	1.13
array/reductions/reduce/Int64/dims=1	1093833 ns	1105625.5 ns	0.99
array/reductions/reduce/Int64/dims=1L	2018458 ns	2038209 ns	0.99
array/reductions/reduce/Int64/dims=2	1149583.5 ns	1165084 ns	0.99
array/reductions/reduce/Int64/dims=2L	4224500.5 ns	4246833.5 ns	0.99
array/shared/copy	251875 ns	230667 ns	1.09
array/shared/copyto!/cpu_to_gpu	80791 ns	83667 ns	0.97
array/shared/copyto!/gpu_to_cpu	81250 ns	82334 ns	0.99
array/shared/copyto!/gpu_to_gpu	81583 ns	83209 ns	0.98
array/shared/iteration/findall/bool	1428500 ns	1414771 ns	1.01
array/shared/iteration/findall/int	1559895.5 ns	1566833 ns	1.00
array/shared/iteration/findfirst/bool	1631874.5 ns	1634542 ns	1.00
array/shared/iteration/findfirst/int	1650417 ns	1650729 ns	1.00
array/shared/iteration/findmin/1d	2092875 ns	2110166.5 ns	0.99
array/shared/iteration/findmin/2d	1795021 ns	1792229 ns	1.00
array/shared/iteration/logical	2452125 ns	2237270.5 ns	1.10
array/shared/iteration/scalar	199917 ns	207375 ns	0.96
integration/byval/reference	1563375 ns	1550250 ns	1.01
integration/byval/slices=1	1579687 ns	1587604.5 ns	1.00
integration/byval/slices=2	2612854.5 ns	2607167 ns	1.00
integration/byval/slices=3	7805520.5 ns	7747541 ns	1.01
integration/metaldevrt	864395.5 ns	880084 ns	0.98
kernel/indexing	621750 ns	597895.5 ns	1.04
kernel/indexing_checked	629124.5 ns	586250 ns	1.07
kernel/launch	11083 ns	11917 ns	0.93
kernel/rand	571167 ns	576375 ns	0.99
latency/import	1422002520.5 ns	1422194646 ns	1.00
latency/precompile	25434593458 ns	25428193000 ns	1.00
latency/ttfp	2340542917 ns	2340052813 ns	1.00
metal/synchronization/context	19667 ns	20166.5 ns	0.98
metal/synchronization/stream	18584 ns	19250 ns	0.97

This comment was automatically generated by workflow using github-action-benchmark.