Various pointer-related improvements

[christiangnrd]

Each commit should be a very quick review. Best viewed with whitespace hidden.

Some code coverage improvements, bug fixes, and simplification.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.02%. Comparing base (1af3aef) to head (a5acbf5).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #765      +/-   ##
==========================================
+ Coverage   82.66%   83.02%   +0.36%     
==========================================
  Files          62       62              
  Lines        2872     2852      -20     
==========================================
- Hits         2374     2368       -6     
+ Misses        498      484      -14     

☔ 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: a5acbf5	Previous: 1af3aef	Ratio
array/accumulate/Float32/1d	1157708 ns	1113770.5 ns	1.04
array/accumulate/Float32/dims=1	1574625.5 ns	1551084 ns	1.02
array/accumulate/Float32/dims=1L	9826688 ns	9823458.5 ns	1.00
array/accumulate/Float32/dims=2	1918875 ns	1869125 ns	1.03
array/accumulate/Float32/dims=2L	7253791.5 ns	7234500 ns	1.00
array/accumulate/Int64/1d	1270791 ns	1259292 ns	1.01
array/accumulate/Int64/dims=1	1819479.5 ns	1833292 ns	0.99
array/accumulate/Int64/dims=1L	11682958 ns	11548646 ns	1.01
array/accumulate/Int64/dims=2	2200208.5 ns	2156000 ns	1.02
array/accumulate/Int64/dims=2L	9796271 ns	9765500 ns	1.00
array/broadcast	610125 ns	608792 ns	1.00
array/construct	6709 ns	6000 ns	1.12
array/permutedims/2d	1191167 ns	1167020.5 ns	1.02
array/permutedims/3d	1699625 ns	1679667 ns	1.01
array/permutedims/4d	2419334 ns	2380542 ns	1.02
array/private/copy	585583 ns	596666 ns	0.98
array/private/copyto!/cpu_to_gpu	806583 ns	786083.5 ns	1.03
array/private/copyto!/gpu_to_cpu	812000 ns	796625 ns	1.02
array/private/copyto!/gpu_to_gpu	647958 ns	631500 ns	1.03
array/private/iteration/findall/bool	1408584 ns	1411938 ns	1.00
array/private/iteration/findall/int	1578875 ns	1563375 ns	1.01
array/private/iteration/findfirst/bool	2068437.5 ns	2023833 ns	1.02
array/private/iteration/findfirst/int	2133458.5 ns	2045020.5 ns	1.04
array/private/iteration/findmin/1d	2530542 ns	2470417 ns	1.02
array/private/iteration/findmin/2d	1828667 ns	1774854 ns	1.03
array/private/iteration/logical	2701334 ns	2618500 ns	1.03
array/private/iteration/scalar	4795375 ns	4934917 ns	0.97
array/random/rand/Float32	584250 ns	634500 ns	0.92
array/random/rand/Int64	794375 ns	771417 ns	1.03
array/random/rand!/Float32	584833 ns	578709 ns	1.01
array/random/rand!/Int64	553916 ns	547250 ns	1.01
array/random/randn/Float32	976292 ns	1008833.5 ns	0.97
array/random/randn!/Float32	755750 ns	714625 ns	1.06
array/reductions/mapreduce/Float32/1d	1073333 ns	1041062.5 ns	1.03
array/reductions/mapreduce/Float32/dims=1	846583.5 ns	838292 ns	1.01
array/reductions/mapreduce/Float32/dims=1L	1346709 ns	1563812.5 ns	0.86
array/reductions/mapreduce/Float32/dims=2	861917 ns	864854 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	1828000 ns	1798229.5 ns	1.02
array/reductions/mapreduce/Int64/1d	1546000 ns	1534250 ns	1.01
array/reductions/mapreduce/Int64/dims=1	1113250 ns	1102709 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	2035687 ns	1996770.5 ns	1.02
array/reductions/mapreduce/Int64/dims=2	1175167 ns	1183062.5 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	3639562.5 ns	3617646 ns	1.01
array/reductions/reduce/Float32/1d	1051062 ns	1029479.5 ns	1.02
array/reductions/reduce/Float32/dims=1	843166 ns	830333 ns	1.02
array/reductions/reduce/Float32/dims=1L	1333499.5 ns	1317584 ns	1.01
array/reductions/reduce/Float32/dims=2	867958.5 ns	856145.5 ns	1.01
array/reductions/reduce/Float32/dims=2L	1816875 ns	1795250 ns	1.01
array/reductions/reduce/Int64/1d	1399437.5 ns	1500500 ns	0.93
array/reductions/reduce/Int64/dims=1	1112792 ns	1106458 ns	1.01
array/reductions/reduce/Int64/dims=1L	2017437.5 ns	2016083 ns	1.00
array/reductions/reduce/Int64/dims=2	1173833 ns	1154354.5 ns	1.02
array/reductions/reduce/Int64/dims=2L	4246374.5 ns	4213709 ns	1.01
array/shared/copy	243000 ns	276625 ns	0.88
array/shared/copyto!/cpu_to_gpu	82042 ns	97833 ns	0.84
array/shared/copyto!/gpu_to_cpu	83958 ns	98208 ns	0.85
array/shared/copyto!/gpu_to_gpu	83666 ns	122000 ns	0.69
array/shared/iteration/findall/bool	1432541 ns	1433249.5 ns	1.00
array/shared/iteration/findall/int	1577875 ns	1571812.5 ns	1.00
array/shared/iteration/findfirst/bool	1653354.5 ns	1630083 ns	1.01
array/shared/iteration/findfirst/int	1676834 ns	1635084 ns	1.03
array/shared/iteration/findmin/1d	2134292 ns	2098979 ns	1.02
array/shared/iteration/findmin/2d	1819958.5 ns	1779458 ns	1.02
array/shared/iteration/logical	2265937.5 ns	2292083 ns	0.99
array/shared/iteration/scalar	213375 ns	202833 ns	1.05
integration/byval/reference	1589042 ns	1549208 ns	1.03
integration/byval/slices=1	1604833 ns	1564292 ns	1.03
integration/byval/slices=2	2669875 ns	2597625 ns	1.03
integration/byval/slices=3	10157500 ns	8449250 ns	1.20
integration/metaldevrt	881833 ns	863917 ns	1.02
kernel/indexing	608500 ns	594792 ns	1.02
kernel/indexing_checked	621708 ns	606541.5 ns	1.03
kernel/launch	12667 ns	11833 ns	1.07
kernel/rand	570459 ns	569167 ns	1.00
latency/import	1424736708.5 ns	1420952187.5 ns	1.00
latency/precompile	25499036000 ns	25612710875 ns	1.00
latency/ttfp	2349867458.5 ns	2339829875 ns	1.00
metal/synchronization/context	20541 ns	20000 ns	1.03
metal/synchronization/stream	19667 ns	19000 ns	1.04

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

[christiangnrd]

Copy-related performance improvements and regressions seem to be noise.