Improvements to large buffer tests

[christiangnrd]

Depends on #728 and a way to select runners with at least 16 GB memory

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic main) to apply these changes.

Click here to view the suggested changes.

diff --git a/test/largebroadcast.jl b/test/largebroadcast.jl
index be0bf22b..b037458b 100644
--- a/test/largebroadcast.jl
+++ b/test/largebroadcast.jl
@@ -1,13 +1,15 @@
 const N = Int(typemax(UInt32)) + 1
 const T = Int8
 
-@testset "len = $n" for n in ((N÷2) - 4, (N÷2), (N÷2) + 4, N - 1024, N - 3, N - 1, N, N + 4)
+@testset "len = $n" for n in ((N ÷ 2) - 4, (N ÷ 2), (N ÷ 2) + 4, N - 1024, N - 3, N - 1, N, N + 4)
     A = MtlArray{T}(undef, n)
     # Known working method to zero out array
     Metal.unsafe_fill!(device(A), pointer(A), T(0), n * sizeof(T); async = false)
 
-    _dims = [(n,), (n, 1), (1, n), (n, 1, 1), (1, n, 1), (1, 1, n),
-             (n, 1, 1, 1), (1, n, 1, 1), (1, 1, n, 1), (1, 1, 1, n),]
+    _dims = [
+        (n,), (n, 1), (1, n), (n, 1, 1), (1, n, 1), (1, 1, n),
+        (n, 1, 1, 1), (1, n, 1, 1), (1, 1, n, 1), (1, 1, 1, n),
+    ]
     if n == 2^32
         push!(_dims, (2^16, 2^16))
         push!(_dims, (2^16, 2^8, 2^8))
@@ -17,7 +19,7 @@ const T = Int8
         # These must be run first to ensure we test
         # the unspecialized broadcast kernels
         Metal._broadcast_shapes[CartesianIndices(dims)] = Metal.BROADCAST_SPECIALIZATION_THRESHOLD - 1
-        unspec_val = T((i-1) * 2 + 1)
+        unspec_val = T((i - 1) * 2 + 1)
         arr = reshape(A, dims)
         arr .= unspec_val
         @test all(==(unspec_val), arr)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.28%. Comparing base (f7b0829) to head (f2a7fa1).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #735      +/-   ##
==========================================
- Coverage   82.59%   82.28%   -0.32%     
==========================================
  Files          62       62              
  Lines        2862     2873      +11     
==========================================
  Hits         2364     2364              
- Misses        498      509      +11     

☔ 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: f2a7fa1	Previous: f7b0829	Ratio
latency/precompile	24994097291 ns	25035410125 ns	1.00
latency/ttfp	2272859917 ns	2272092167 ns	1.00
latency/import	1445333042 ns	1444131708 ns	1.00
integration/metaldevrt	864354.5 ns	870500 ns	0.99
integration/byval/slices=1	1591292 ns	1578834 ns	1.01
integration/byval/slices=3	11237083 ns	10274687 ns	1.09
integration/byval/reference	1567667 ns	1566167 ns	1.00
integration/byval/slices=2	2643125 ns	2622959 ns	1.01
kernel/indexing	602292 ns	625542 ns	0.96
kernel/indexing_checked	627458 ns	615520.5 ns	1.02
kernel/launch	11625 ns	11417 ns	1.02
kernel/rand	568041 ns	566500 ns	1.00
array/construct	6500 ns	6084 ns	1.07
array/broadcast	598167 ns	602917 ns	0.99
array/random/randn/Float32	968666.5 ns	1001708 ns	0.97
array/random/randn!/Float32	743125 ns	750583 ns	0.99
array/random/rand!/Int64	549833.5 ns	551375 ns	1.00
array/random/rand!/Float32	586208 ns	587041 ns	1.00
array/random/rand/Int64	787917 ns	777958 ns	1.01
array/random/rand/Float32	642333 ns	590000 ns	1.09
array/accumulate/Int64/1d	1280229 ns	1254458 ns	1.02
array/accumulate/Int64/dims=1	1841125 ns	1827292 ns	1.01
array/accumulate/Int64/dims=2	2189750 ns	2165896 ns	1.01
array/accumulate/Int64/dims=1L	11558895.5 ns	11587604 ns	1.00
array/accumulate/Int64/dims=2L	9795208 ns	9813792 ns	1.00
array/accumulate/Float32/1d	1125500 ns	1121542 ns	1.00
array/accumulate/Float32/dims=1	1566375 ns	1398896.5 ns	1.12
array/accumulate/Float32/dims=2	1890042 ns	1890312.5 ns	1.00
array/accumulate/Float32/dims=1L	9798959 ns	9780770.5 ns	1.00
array/accumulate/Float32/dims=2L	7258792 ns	7265249.5 ns	1.00
array/reductions/reduce/Int64/1d	1519437.5 ns	1357812.5 ns	1.12
array/reductions/reduce/Int64/dims=1	1103250 ns	1116917 ns	0.99
array/reductions/reduce/Int64/dims=2	1153458.5 ns	1188313 ns	0.97
array/reductions/reduce/Int64/dims=1L	1984917 ns	1986958.5 ns	1.00
array/reductions/reduce/Int64/dims=2L	4718354 ns	4191792 ns	1.13
array/reductions/reduce/Float32/1d	929542 ns	1028770.5 ns	0.90
array/reductions/reduce/Float32/dims=1	831875 ns	830375 ns	1.00
array/reductions/reduce/Float32/dims=2	925833.5 ns	861042 ns	1.08
array/reductions/reduce/Float32/dims=1L	1558208.5 ns	1318208 ns	1.18
array/reductions/reduce/Float32/dims=2L	1820875 ns	1794083 ns	1.01
array/reductions/mapreduce/Int64/1d	1519479.5 ns	1551042 ns	0.98
array/reductions/mapreduce/Int64/dims=1	1105292 ns	1125167 ns	0.98
array/reductions/mapreduce/Int64/dims=2	1187312.5 ns	1206208 ns	0.98
array/reductions/mapreduce/Int64/dims=1L	2022833 ns	1995875 ns	1.01
array/reductions/mapreduce/Int64/dims=2L	3579875 ns	3621563 ns	0.99
array/reductions/mapreduce/Float32/1d	943917 ns	1027916 ns	0.92
array/reductions/mapreduce/Float32/dims=1	832833 ns	830833 ns	1.00
array/reductions/mapreduce/Float32/dims=2	862708 ns	862209 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	1327354.5 ns	1332666 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	1803459 ns	1842000 ns	0.98
array/private/copyto!/gpu_to_gpu	629709 ns	632500 ns	1.00
array/private/copyto!/cpu_to_gpu	789708 ns	780667 ns	1.01
array/private/copyto!/gpu_to_cpu	804625 ns	787750 ns	1.02
array/private/iteration/findall/int	1556291 ns	1564000 ns	1.00
array/private/iteration/findall/bool	1412146 ns	1393667 ns	1.01
array/private/iteration/findfirst/int	2062333 ns	2101875 ns	0.98
array/private/iteration/findfirst/bool	2041229.5 ns	2049749.5 ns	1.00
array/private/iteration/scalar	4025000 ns	4317166 ns	0.93
array/private/iteration/logical	2632459 ns	2622584 ns	1.00
array/private/iteration/findmin/1d	2504666 ns	2507959 ns	1.00
array/private/iteration/findmin/2d	1792750 ns	1788250 ns	1.00
array/private/copy	568792 ns	556437.5 ns	1.02
array/shared/copyto!/gpu_to_gpu	82916 ns	83375 ns	0.99
array/shared/copyto!/cpu_to_gpu	82375 ns	87084 ns	0.95
array/shared/copyto!/gpu_to_cpu	81792 ns	82625 ns	0.99
array/shared/iteration/findall/int	1559042 ns	1552250 ns	1.00
array/shared/iteration/findall/bool	1432083 ns	1424958 ns	1.01
array/shared/iteration/findfirst/int	1665458 ns	1688500 ns	0.99
array/shared/iteration/findfirst/bool	1641209 ns	1642438 ns	1.00
array/shared/iteration/scalar	207167 ns	205792 ns	1.01
array/shared/iteration/logical	2246584 ns	2261000 ns	0.99
array/shared/iteration/findmin/1d	2123958 ns	2133500 ns	1.00
array/shared/iteration/findmin/2d	1792083 ns	1797312.5 ns	1.00
array/shared/copy	235583 ns	240542 ns	0.98
array/permutedims/4d	2399833.5 ns	2390688 ns	1.00
array/permutedims/2d	1179208 ns	1170542 ns	1.01
array/permutedims/3d	1686792 ns	1675145.5 ns	1.01
metal/synchronization/stream	19416 ns	17291 ns	1.12
metal/synchronization/context	20291 ns	17667 ns	1.15

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