Fix compatibility for CUDA 13.

[maleadt]

#2834 missed the compat bits and device detection.

Closes #2896

[github-actions]

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

Click here to view the suggested changes.

diff --git a/lib/cudadrv/state.jl b/lib/cudadrv/state.jl
index 5fc70d29d..fb0722f14 100644
--- a/lib/cudadrv/state.jl
+++ b/lib/cudadrv/state.jl
@@ -218,12 +218,14 @@ function context(dev::CuDevice)
     # check if the device isn't too old
     if capability(dev) < v"5.0"
         @error("""Your $(name(dev)) GPU (compute capability $(capability(dev).major).$(capability(dev).minor)) is not supported by CUDA.jl.
-                  Please use a device with at least capability 5.0, or downgrade CUDA.jl (see the README for compatibility details).""",
+                              Please use a device with at least capability 5.0, or downgrade CUDA.jl (see the README for compatibility details).""",
                maxlog=1, _id=devidx)
     elseif runtime_version() >= v"13" && capability(dev) <= v"7.5"
-        @error("""Your $(name(dev)) GPU (compute capability $(capability(dev).major).$(capability(dev).minor)) is not supported on CUDA 13+.
-                  Please use a device with at least capability 7.5, or downgrade your NVIDIA driver to below v580.""",
-               maxlog=1, _id=devidx)
+        @error(
+            """Your $(name(dev)) GPU (compute capability $(capability(dev).major).$(capability(dev).minor)) is not supported on CUDA 13+.
+            Please use a device with at least capability 7.5, or downgrade your NVIDIA driver to below v580.""",
+            maxlog = 1, _id = devidx
+        )
     end
     # ... or too new
     if !in(capability(dev), cuda_compat().cap)
diff --git a/src/compatibility.jl b/src/compatibility.jl
index 272a51ebf..8a06bdaad 100644
--- a/src/compatibility.jl
+++ b/src/compatibility.jl
@@ -41,15 +41,15 @@ const cuda_cap_db = Dict(
     v"3.0"   => between(v"4.2", v"10.2"),
     v"3.2"   => between(v"6.0", v"10.2"),
     v"3.5"   => between(v"5.0", v"11.8"),
-    v"3.7"   => between(v"6.5", v"11.8"),
-    v"5.0"   => between(v"6.0", v"12.9"),
-    v"5.2"   => between(v"7.0", v"12.9"),
-    v"5.3"   => between(v"7.5", v"12.9"),
-    v"6.0"   => between(v"8.0", v"12.9"),
-    v"6.1"   => between(v"8.0", v"12.9"),
-    v"6.2"   => between(v"8.0", v"12.9"),
-    v"7.0"   => between(v"9.0", v"12.9"),
-    v"7.2"   => between(v"9.2", v"12.9"),
+    v"3.7" => between(v"6.5", v"11.8"),
+    v"5.0" => between(v"6.0", v"12.9"),
+    v"5.2" => between(v"7.0", v"12.9"),
+    v"5.3" => between(v"7.5", v"12.9"),
+    v"6.0" => between(v"8.0", v"12.9"),
+    v"6.1" => between(v"8.0", v"12.9"),
+    v"6.2" => between(v"8.0", v"12.9"),
+    v"7.0" => between(v"9.0", v"12.9"),
+    v"7.2" => between(v"9.2", v"12.9"),
     v"7.5"   => between(v"10.0", highest),
     v"8.0"   => between(v"11.0", highest),
     v"8.6"   => between(v"11.1", highest),

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 00141af	Previous: 6c6977f	Ratio
latency/precompile	43747535057.5 ns	43775122238.5 ns	1.00
latency/ttfp	7306977522 ns	7276778491 ns	1.00
latency/import	3865063206 ns	3836255124 ns	1.01
integration/volumerhs	9625850 ns	9623790.5 ns	1.00
integration/byval/slices=1	146811 ns	146826 ns	1.00
integration/byval/slices=3	425646 ns	426011 ns	1.00
integration/byval/reference	144934 ns	145073 ns	1.00
integration/byval/slices=2	286202 ns	286240 ns	1.00
integration/cudadevrt	103644 ns	103460 ns	1.00
kernel/indexing	14328 ns	14196 ns	1.01
kernel/indexing_checked	15116 ns	15033 ns	1.01
kernel/occupancy	663.5375 ns	670.506329113924 ns	0.99
kernel/launch	2199.6666666666665 ns	2162.5555555555557 ns	1.02
kernel/rand	18468 ns	16555 ns	1.12
array/reverse/1d	19699 ns	19646 ns	1.00
array/reverse/2dL_inplace	66763 ns	66804 ns	1.00
array/reverse/1dL	69936 ns	69923 ns	1.00
array/reverse/2d	21653 ns	21677 ns	1.00
array/reverse/1d_inplace	9649 ns	9658 ns	1.00
array/reverse/2d_inplace	11005 ns	13323 ns	0.83
array/reverse/2dL	73860.5 ns	73803 ns	1.00
array/reverse/1dL_inplace	66808 ns	66812 ns	1.00
array/copy	20723 ns	20306 ns	1.02
array/iteration/findall/int	157751.5 ns	157164 ns	1.00
array/iteration/findall/bool	139359 ns	139633 ns	1.00
array/iteration/findfirst/int	160407 ns	160554.5 ns	1.00
array/iteration/findfirst/bool	161907.5 ns	160957 ns	1.01
array/iteration/scalar	74241.5 ns	72124 ns	1.03
array/iteration/logical	216961.5 ns	215036.5 ns	1.01
array/iteration/findmin/1d	49809.5 ns	49445 ns	1.01
array/iteration/findmin/2d	96400 ns	96493.5 ns	1.00
array/reductions/reduce/Int64/1d	42944 ns	42960 ns	1.00
array/reductions/reduce/Int64/dims=1	44783 ns	44742.5 ns	1.00
array/reductions/reduce/Int64/dims=2	61378.5 ns	61453 ns	1.00
array/reductions/reduce/Int64/dims=1L	88824 ns	88951 ns	1.00
array/reductions/reduce/Int64/dims=2L	88093 ns	88014.5 ns	1.00
array/reductions/reduce/Float32/1d	36277 ns	35769 ns	1.01
array/reductions/reduce/Float32/dims=1	45881 ns	51586 ns	0.89
array/reductions/reduce/Float32/dims=2	59535 ns	59511 ns	1.00
array/reductions/reduce/Float32/dims=1L	52305 ns	52474 ns	1.00
array/reductions/reduce/Float32/dims=2L	72046 ns	71419 ns	1.01
array/reductions/mapreduce/Int64/1d	43535 ns	43189 ns	1.01
array/reductions/mapreduce/Int64/dims=1	53216.5 ns	46540.5 ns	1.14
array/reductions/mapreduce/Int64/dims=2	61398 ns	61279.5 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	88811 ns	88854 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	87956 ns	88014 ns	1.00
array/reductions/mapreduce/Float32/1d	36670.5 ns	36287 ns	1.01
array/reductions/mapreduce/Float32/dims=1	41606 ns	41466 ns	1.00
array/reductions/mapreduce/Float32/dims=2	59638 ns	59744 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52436 ns	52550 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	71915 ns	71985 ns	1.00
array/broadcast	20004 ns	20047 ns	1.00
array/copyto!/gpu_to_gpu	13004 ns	11191 ns	1.16
array/copyto!/cpu_to_gpu	215322 ns	213964 ns	1.01
array/copyto!/gpu_to_cpu	284829 ns	284661.5 ns	1.00
array/accumulate/Int64/1d	124632 ns	124888 ns	1.00
array/accumulate/Int64/dims=1	83493 ns	83130 ns	1.00
array/accumulate/Int64/dims=2	157849 ns	157680 ns	1.00
array/accumulate/Int64/dims=1L	1709343 ns	1709578 ns	1.00
array/accumulate/Int64/dims=2L	966272.5 ns	966045 ns	1.00
array/accumulate/Float32/1d	109267 ns	108910 ns	1.00
array/accumulate/Float32/dims=1	80442 ns	80564 ns	1.00
array/accumulate/Float32/dims=2	147384 ns	147715 ns	1.00
array/accumulate/Float32/dims=1L	1618007.5 ns	1618612 ns	1.00
array/accumulate/Float32/dims=2L	697763 ns	698318 ns	1.00
array/construct	1280.7 ns	1287.5 ns	0.99
array/random/randn/Float32	45216.5 ns	43976 ns	1.03
array/random/randn!/Float32	24668 ns	24816 ns	0.99
array/random/rand!/Int64	27287 ns	27267 ns	1.00
array/random/rand!/Float32	8946 ns	8653.333333333334 ns	1.03
array/random/rand/Int64	38233 ns	38285 ns	1.00
array/random/rand/Float32	13128 ns	13026 ns	1.01
array/permutedims/4d	60540 ns	60152.5 ns	1.01
array/permutedims/2d	54289.5 ns	53934 ns	1.01
array/permutedims/3d	55099.5 ns	54649.5 ns	1.01
array/sorting/1d	2757149.5 ns	2757180 ns	1.00
array/sorting/by	3344173.5 ns	3343619 ns	1.00
array/sorting/2d	1080103 ns	1080730 ns	1.00
cuda/synchronization/stream/auto	1034.3 ns	1029.6 ns	1.00
cuda/synchronization/stream/nonblocking	7569.5 ns	7270.700000000001 ns	1.04
cuda/synchronization/stream/blocking	850.5135135135135 ns	850.7415730337078 ns	1.00
cuda/synchronization/context/auto	1162.8 ns	1157 ns	1.01
cuda/synchronization/context/nonblocking	8437.6 ns	6964 ns	1.21
cuda/synchronization/context/blocking	901.4444444444445 ns	888.5208333333334 ns	1.01

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

[maleadt]

Timeout happens on master too.