Merge branch 'main' into java

.github/workflows/main.yaml

@@ -3,6 +3,7 @@ on: [push, pull_request]
 
 
 jobs:
+
   test-java:
     runs-on: ubuntu-18.04
     defaults:
@@ -15,6 +16,36 @@ jobs:
       - name: Test run 
         if: ${{ ! cancelled() }}
         run: java -jar target/java-stream.jar --arraysize 2048
+
+  test-julia:
+    runs-on: ubuntu-18.04
+    defaults:
+      run:
+        working-directory: ./JuliaStream.jl
+    steps:
+      - uses: actions/checkout@v2
+      - name: Setup project
+        run: julia             --project -e 'import Pkg; Pkg.instantiate()'
+      - name: Test run PlainStream.jl
+        if: ${{ ! cancelled() }}
+        run: julia             --project src/PlainStream.jl       --arraysize 2048
+      - name: Test run ThreadedStream.jl
+        if: ${{ ! cancelled() }}
+        run: julia --threads 2 --project src/ThreadedStream.jl    --arraysize 2048
+      - name: Test run DistributedStream.jl (no flag)
+        if: ${{ ! cancelled() }}
+        run: julia             --project src/DistributedStream.jl --arraysize 2048
+      - name: Test run DistributedStream.jl (-p 2)
+        if: ${{ ! cancelled() }}
+        run: julia -p 2        --project src/DistributedStream.jl --arraysize 2048
+      - name: Test run CUDAStream.jl
+        if: ${{ ! cancelled() }}
+        run: julia             --project src/CUDAStream.jl        --list
+      - name: Test run AMDGPUStream.jl
+        if: ${{ ! cancelled() }}
+        run: julia             --project src/AMDGPUStream.jl      --list
+
+
   test:
     runs-on: ubuntu-18.04
     steps:

CITATION.cff

@@ -0,0 +1,14 @@
+cff-version: 1.1.0
+message: If you use this software, please cite it as below.
+authors:
+  - family-names: Deakin
+    given-names: Tom
+    affiliation: University of Bristol
+    website: https://hpc.tomdeakin.com
+  - family-names: McIntosh-Smith
+    given-names: Simon
+    affiliation: University of Bristol
+    website: https://uob-hpc.github.io
+title: BabelStream
+version: 3.4
+date-released: 2019-04-10

JuliaStream.jl/.JuliaFormatter.toml

@@ -0,0 +1,2 @@
+indent = 2
+margin = 100

JuliaStream.jl/.gitignore

@@ -0,0 +1,5 @@
+*.jl.cov
+*.jl.*.cov
+*.jl.mem
+/docs/build/
+/docs/Manifest.toml

JuliaStream.jl/AMDGPU/Manifest.toml

@@ -0,0 +1,415 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[AMDGPU]]
+deps = ["AbstractFFTs", "Adapt", "BinaryProvider", "CEnum", "GPUArrays", "GPUCompiler", "HIP_jll", "LLVM", "Libdl", "LinearAlgebra", "MacroTools", "Pkg", "Printf", "ROCmDeviceLibs_jll", "Random", "Requires", "Setfield", "hsa_rocr_jll"]
+git-tree-sha1 = "d64c97447a753cfbf0158d6c7be513f34526d559"
+uuid = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+version = "0.2.12"
+
+[[AbstractFFTs]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "485ee0867925449198280d4af84bdb46a2a404d0"
+uuid = "621f4979-c628-5d54-868e-fcf4e3e8185c"
+version = "1.0.1"
+
+[[Adapt]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "84918055d15b3114ede17ac6a7182f68870c16f7"
+uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
+version = "3.3.1"
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[ArgTools]]
+uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+
+[[Artifacts]]
+uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
+
+[[Base64]]
+uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
+
+[[BinaryProvider]]
+deps = ["Libdl", "Logging", "SHA"]
+git-tree-sha1 = "ecdec412a9abc8db54c0efc5548c64dfce072058"
+uuid = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
+version = "0.5.10"
+
+[[Bzip2_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "19a35467a82e236ff51bc17a3a44b69ef35185a2"
+uuid = "6e34b625-4abd-537c-b88f-471c36dfa7a0"
+version = "1.0.8+0"
+
+[[CEnum]]
+git-tree-sha1 = "215a9aa4a1f23fbd05b92769fdd62559488d70e9"
+uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+version = "0.4.1"
+
+[[ConstructionBase]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "f74e9d5388b8620b4cee35d4c5a618dd4dc547f4"
+uuid = "187b0558-2788-49d3-abe0-74a17ed4e7c9"
+version = "1.3.0"
+
+[[Dates]]
+deps = ["Printf"]
+uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
+
+[[Downloads]]
+deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+
+[[Elfutils_jll]]
+deps = ["Artifacts", "Bzip2_jll", "JLLWrappers", "Libdl", "Pkg", "XZ_jll", "Zlib_jll", "argp_standalone_jll", "fts_jll", "obstack_jll"]
+git-tree-sha1 = "8f9fcde6d89b0a3ca51cb2028beab462705c5436"
+uuid = "ab5a07f8-06af-567f-a878-e8bb879eba5a"
+version = "0.182.0+0"
+
+[[ExprTools]]
+git-tree-sha1 = "b7e3d17636b348f005f11040025ae8c6f645fe92"
+uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+version = "0.1.6"
+
+[[Future]]
+deps = ["Random"]
+uuid = "9fa8497b-333b-5362-9e8d-4d0656e87820"
+
+[[GPUArrays]]
+deps = ["AbstractFFTs", "Adapt", "LinearAlgebra", "Printf", "Random", "Serialization", "Statistics"]
+git-tree-sha1 = "ececbf05f8904c92814bdbd0aafd5540b0bf2e9a"
+uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
+version = "7.0.1"
+
+[[GPUCompiler]]
+deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "TimerOutputs", "UUIDs"]
+git-tree-sha1 = "4ed2616d5e656c8716736b64da86755467f26cf5"
+uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
+version = "0.12.9"
+
+[[HIP_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "ROCmCompilerSupport_jll", "ROCmDeviceLibs_jll", "ROCmOpenCLRuntime_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "5097d8f7b6842156ab0928371b3d03fefd8decab"
+uuid = "2696aab5-0948-5276-aa9a-2a86a37016b8"
+version = "4.0.0+1"
+
+[[InteractiveUtils]]
+deps = ["Markdown"]
+uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
+
+[[JLLWrappers]]
+deps = ["Preferences"]
+git-tree-sha1 = "642a199af8b68253517b80bd3bfd17eb4e84df6e"
+uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
+version = "1.3.0"
+
+[[LLVM]]
+deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
+git-tree-sha1 = "23a47d417a3cd9c2e73c854bac7dd4731c105ef7"
+uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
+version = "4.4.0"
+
+[[LLVMExtra_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "9c360e5ce980b88bb31a7b086dbb19469008154b"
+uuid = "dad2f222-ce93-54a1-a47d-0025e8a3acab"
+version = "0.0.10+0"
+
+[[LibCURL]]
+deps = ["LibCURL_jll", "MozillaCACerts_jll"]
+uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+
+[[LibCURL_jll]]
+deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
+uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+
+[[LibGit2]]
+deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
+uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
+
+[[LibSSH2_jll]]
+deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
+uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+
+[[Libdl]]
+uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+
+[[Libgcrypt_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libgpg_error_jll", "Pkg"]
+git-tree-sha1 = "64613c82a59c120435c067c2b809fc61cf5166ae"
+uuid = "d4300ac3-e22c-5743-9152-c294e39db1e4"
+version = "1.8.7+0"
+
+[[Libglvnd_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libX11_jll", "Xorg_libXext_jll"]
+git-tree-sha1 = "7739f837d6447403596a75d19ed01fd08d6f56bf"
+uuid = "7e76a0d4-f3c7-5321-8279-8d96eeed0f29"
+version = "1.3.0+3"
+
+[[Libgpg_error_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "c333716e46366857753e273ce6a69ee0945a6db9"
+uuid = "7add5ba3-2f88-524e-9cd5-f83b8a55f7b8"
+version = "1.42.0+0"
+
+[[Libiconv_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "42b62845d70a619f063a7da093d995ec8e15e778"
+uuid = "94ce4f54-9a6c-5748-9c1c-f9c7231a4531"
+version = "1.16.1+1"
+
+[[LinearAlgebra]]
+deps = ["Libdl"]
+uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[MacroTools]]
+deps = ["Markdown", "Random"]
+git-tree-sha1 = "0fb723cd8c45858c22169b2e42269e53271a6df7"
+uuid = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
+version = "0.5.7"
+
+[[Markdown]]
+deps = ["Base64"]
+uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
+
+[[MbedTLS_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+
+[[MozillaCACerts_jll]]
+uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+
+[[NUMA_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "778f9bd14400cff2c32ed357e12766ac0e3d766e"
+uuid = "7f51dc2b-bb24-59f8-b771-bb1490e4195d"
+version = "2.0.13+1"
+
+[[NetworkOptions]]
+uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[Pkg]]
+deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+
+[[Preferences]]
+deps = ["TOML"]
+git-tree-sha1 = "00cfd92944ca9c760982747e9a1d0d5d86ab1e5a"
+uuid = "21216c6a-2e73-6563-6e65-726566657250"
+version = "1.2.2"
+
+[[Printf]]
+deps = ["Unicode"]
+uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+
+[[REPL]]
+deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
+uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
+
+[[ROCmCompilerSupport_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "ROCmDeviceLibs_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "56ddcfb5d8b60c9f8c1bc619886f8d363fd1926d"
+uuid = "8fbdd1d2-db62-5cd0-981e-905da1486e17"
+version = "4.0.0+1"
+
+[[ROCmDeviceLibs_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Zlib_jll"]
+git-tree-sha1 = "d764f0f28b5af89aa004871a6a38e5d061f77257"
+uuid = "873c0968-716b-5aa7-bb8d-d1e2e2aeff2d"
+version = "4.0.0+0"
+
+[[ROCmOpenCLRuntime_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libglvnd_jll", "Pkg", "ROCmCompilerSupport_jll", "ROCmDeviceLibs_jll", "Xorg_libX11_jll", "Xorg_xorgproto_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "f9e3e2cb40a7990535efa7da9b9dd0e0b458a973"
+uuid = "10ae2a08-2eea-53f8-8c20-eec175020e9f"
+version = "4.0.0+1"
+
+[[Random]]
+deps = ["Serialization"]
+uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+
+[[Requires]]
+deps = ["UUIDs"]
+git-tree-sha1 = "4036a3bd08ac7e968e27c203d45f5fff15020621"
+uuid = "ae029012-a4dd-5104-9daa-d747884805df"
+version = "1.1.3"
+
+[[SHA]]
+uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+
+[[Serialization]]
+uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+
+[[Setfield]]
+deps = ["ConstructionBase", "Future", "MacroTools", "Requires"]
+git-tree-sha1 = "fca29e68c5062722b5b4435594c3d1ba557072a3"
+uuid = "efcf1570-3423-57d1-acb7-fd33fddbac46"
+version = "0.7.1"
+
+[[Sockets]]
+uuid = "6462fe0b-24de-5631-8697-dd941f90decc"
+
+[[SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
+[[TOML]]
+deps = ["Dates"]
+uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+
+[[Tar]]
+deps = ["ArgTools", "SHA"]
+uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[TimerOutputs]]
+deps = ["ExprTools", "Printf"]
+git-tree-sha1 = "209a8326c4f955e2442c07b56029e88bb48299c7"
+uuid = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
+version = "0.5.12"
+
+[[UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"
+
+[[Unicode]]
+uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
+
+[[XML2_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libiconv_jll", "Pkg", "Zlib_jll"]
+git-tree-sha1 = "1acf5bdf07aa0907e0a37d3718bb88d4b687b74a"
+uuid = "02c8fc9c-b97f-50b9-bbe4-9be30ff0a78a"
+version = "2.9.12+0"
+
+[[XSLT_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libgcrypt_jll", "Libgpg_error_jll", "Libiconv_jll", "Pkg", "XML2_jll", "Zlib_jll"]
+git-tree-sha1 = "91844873c4085240b95e795f692c4cec4d805f8a"
+uuid = "aed1982a-8fda-507f-9586-7b0439959a61"
+version = "1.1.34+0"
+
+[[XZ_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "a921669cd9a45c23031fd4eb904f5cc3d20de415"
+uuid = "ffd25f8a-64ca-5728-b0f7-c24cf3aae800"
+version = "5.2.5+2"
+
+[[Xorg_libX11_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libxcb_jll", "Xorg_xtrans_jll"]
+git-tree-sha1 = "5be649d550f3f4b95308bf0183b82e2582876527"
+uuid = "4f6342f7-b3d2-589e-9d20-edeb45f2b2bc"
+version = "1.6.9+4"
+
+[[Xorg_libXau_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "4e490d5c960c314f33885790ed410ff3a94ce67e"
+uuid = "0c0b7dd1-d40b-584c-a123-a41640f87eec"
+version = "1.0.9+4"
+
+[[Xorg_libXdmcp_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "4fe47bd2247248125c428978740e18a681372dd4"
+uuid = "a3789734-cfe1-5b06-b2d0-1dd0d9d62d05"
+version = "1.1.3+4"
+
+[[Xorg_libXext_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libX11_jll"]
+git-tree-sha1 = "b7c0aa8c376b31e4852b360222848637f481f8c3"
+uuid = "1082639a-0dae-5f34-9b06-72781eeb8cb3"
+version = "1.3.4+4"
+
+[[Xorg_libpthread_stubs_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "6783737e45d3c59a4a4c4091f5f88cdcf0908cbb"
+uuid = "14d82f49-176c-5ed1-bb49-ad3f5cbd8c74"
+version = "0.1.0+3"
+
+[[Xorg_libxcb_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "XSLT_jll", "Xorg_libXau_jll", "Xorg_libXdmcp_jll", "Xorg_libpthread_stubs_jll"]
+git-tree-sha1 = "daf17f441228e7a3833846cd048892861cff16d6"
+uuid = "c7cfdc94-dc32-55de-ac96-5a1b8d977c5b"
+version = "1.13.0+3"
+
+[[Xorg_xorgproto_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "9a9eb8ce756fe0bca01b4be16da770e18d264972"
+uuid = "c4d99508-4286-5418-9131-c86396af500b"
+version = "2019.2.0+2"
+
+[[Xorg_xtrans_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "79c31e7844f6ecf779705fbc12146eb190b7d845"
+uuid = "c5fb5394-a638-5e4d-96e5-b29de1b5cf10"
+version = "1.4.0+3"
+
+[[Zlib_jll]]
+deps = ["Libdl"]
+uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+
+[[argp_standalone_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "feaf9f6293003c2bf53056fd6930d677ed340b34"
+uuid = "c53206cc-00f7-50bf-ad1e-3ae1f6e49bc3"
+version = "1.3.1+0"
+
+[[fts_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "78732b942383d2cb521df8a1a0814911144e663d"
+uuid = "d65627f6-89bd-53e8-8ab5-8b75ff535eee"
+version = "1.2.7+1"
+
+[[hsa_rocr_jll]]
+deps = ["Artifacts", "Elfutils_jll", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg", "Zlib_jll", "hsakmt_roct_jll"]
+git-tree-sha1 = "df8d73efec8b1e53ad527d208f5343c0368f0fcd"
+uuid = "dd59ff1a-a01a-568d-8b29-0669330f116a"
+version = "4.0.0+0"
+
+[[hsakmt_roct_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg"]
+git-tree-sha1 = "80e0c9940e15cfd6f1f1e9d9f3953ec4d48d3d4a"
+uuid = "1cecccd7-a9b6-5045-9cdc-a44c19b16d76"
+version = "4.0.0+0"
+
+[[nghttp2_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+
+[[obstack_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "1c4a6b66e934fc6db4649cb2910c72f53bbfea7e"
+uuid = "c88a4935-d25e-5644-aacc-5db6f1b8ef79"
+version = "1.2.2+0"
+
+[[p7zip_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"

JuliaStream.jl/AMDGPU/Project.toml

@@ -0,0 +1,7 @@
+[deps]
+AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/CUDA/Manifest.toml

@@ -0,0 +1,316 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[AbstractFFTs]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "485ee0867925449198280d4af84bdb46a2a404d0"
+uuid = "621f4979-c628-5d54-868e-fcf4e3e8185c"
+version = "1.0.1"
+
+[[Adapt]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "84918055d15b3114ede17ac6a7182f68870c16f7"
+uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
+version = "3.3.1"
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[ArgTools]]
+uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+
+[[Artifacts]]
+uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
+
+[[BFloat16s]]
+deps = ["LinearAlgebra", "Test"]
+git-tree-sha1 = "4af69e205efc343068dc8722b8dfec1ade89254a"
+uuid = "ab4f0b2a-ad5b-11e8-123f-65d77653426b"
+version = "0.1.0"
+
+[[Base64]]
+uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
+
+[[CEnum]]
+git-tree-sha1 = "215a9aa4a1f23fbd05b92769fdd62559488d70e9"
+uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+version = "0.4.1"
+
+[[CUDA]]
+deps = ["AbstractFFTs", "Adapt", "BFloat16s", "CEnum", "CompilerSupportLibraries_jll", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LazyArtifacts", "Libdl", "LinearAlgebra", "Logging", "Printf", "Random", "Random123", "RandomNumbers", "Reexport", "Requires", "SparseArrays", "SpecialFunctions", "TimerOutputs"]
+git-tree-sha1 = "c583f3ccdce071b8a8bce9bf3d5d5409eaf36d2b"
+uuid = "052768ef-5323-5732-b1bb-66c8b64840ba"
+version = "3.4.1"
+
+[[ChainRulesCore]]
+deps = ["Compat", "LinearAlgebra", "SparseArrays"]
+git-tree-sha1 = "bdc0937269321858ab2a4f288486cb258b9a0af7"
+uuid = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+version = "1.3.0"
+
+[[Compat]]
+deps = ["Base64", "Dates", "DelimitedFiles", "Distributed", "InteractiveUtils", "LibGit2", "Libdl", "LinearAlgebra", "Markdown", "Mmap", "Pkg", "Printf", "REPL", "Random", "SHA", "Serialization", "SharedArrays", "Sockets", "SparseArrays", "Statistics", "Test", "UUIDs", "Unicode"]
+git-tree-sha1 = "727e463cfebd0c7b999bbf3e9e7e16f254b94193"
+uuid = "34da2185-b29b-5c13-b0c7-acf172513d20"
+version = "3.34.0"
+
+[[CompilerSupportLibraries_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+
+[[Dates]]
+deps = ["Printf"]
+uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
+
+[[DelimitedFiles]]
+deps = ["Mmap"]
+uuid = "8bb1440f-4735-579b-a4ab-409b98df4dab"
+
+[[Distributed]]
+deps = ["Random", "Serialization", "Sockets"]
+uuid = "8ba89e20-285c-5b6f-9357-94700520ee1b"
+
+[[DocStringExtensions]]
+deps = ["LibGit2"]
+git-tree-sha1 = "a32185f5428d3986f47c2ab78b1f216d5e6cc96f"
+uuid = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
+version = "0.8.5"
+
+[[Downloads]]
+deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+
+[[ExprTools]]
+git-tree-sha1 = "b7e3d17636b348f005f11040025ae8c6f645fe92"
+uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+version = "0.1.6"
+
+[[GPUArrays]]
+deps = ["Adapt", "LinearAlgebra", "Printf", "Random", "Serialization", "Statistics"]
+git-tree-sha1 = "8fac1cf7d6ce0f2249c7acaf25d22e1e85c4a07f"
+uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
+version = "8.0.2"
+
+[[GPUCompiler]]
+deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "TimerOutputs", "UUIDs"]
+git-tree-sha1 = "4ed2616d5e656c8716736b64da86755467f26cf5"
+uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
+version = "0.12.9"
+
+[[InteractiveUtils]]
+deps = ["Markdown"]
+uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
+
+[[IrrationalConstants]]
+git-tree-sha1 = "f76424439413893a832026ca355fe273e93bce94"
+uuid = "92d709cd-6900-40b7-9082-c6be49f344b6"
+version = "0.1.0"
+
+[[JLLWrappers]]
+deps = ["Preferences"]
+git-tree-sha1 = "642a199af8b68253517b80bd3bfd17eb4e84df6e"
+uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
+version = "1.3.0"
+
+[[LLVM]]
+deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
+git-tree-sha1 = "23a47d417a3cd9c2e73c854bac7dd4731c105ef7"
+uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
+version = "4.4.0"
+
+[[LLVMExtra_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "9c360e5ce980b88bb31a7b086dbb19469008154b"
+uuid = "dad2f222-ce93-54a1-a47d-0025e8a3acab"
+version = "0.0.10+0"
+
+[[LazyArtifacts]]
+deps = ["Artifacts", "Pkg"]
+uuid = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
+
+[[LibCURL]]
+deps = ["LibCURL_jll", "MozillaCACerts_jll"]
+uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+
+[[LibCURL_jll]]
+deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
+uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+
+[[LibGit2]]
+deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
+uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
+
+[[LibSSH2_jll]]
+deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
+uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+
+[[Libdl]]
+uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+
+[[LinearAlgebra]]
+deps = ["Libdl"]
+uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+
+[[LogExpFunctions]]
+deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
+git-tree-sha1 = "3d682c07e6dd250ed082f883dc88aee7996bf2cc"
+uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
+version = "0.3.0"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[Markdown]]
+deps = ["Base64"]
+uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
+
+[[MbedTLS_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+
+[[Mmap]]
+uuid = "a63ad114-7e13-5084-954f-fe012c677804"
+
+[[MozillaCACerts_jll]]
+uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+
+[[NetworkOptions]]
+uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+
+[[OpenSpecFun_jll]]
+deps = ["Artifacts", "CompilerSupportLibraries_jll", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "13652491f6856acfd2db29360e1bbcd4565d04f1"
+uuid = "efe28fd5-8261-553b-a9e1-b2916fc3738e"
+version = "0.5.5+0"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[Pkg]]
+deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+
+[[Preferences]]
+deps = ["TOML"]
+git-tree-sha1 = "00cfd92944ca9c760982747e9a1d0d5d86ab1e5a"
+uuid = "21216c6a-2e73-6563-6e65-726566657250"
+version = "1.2.2"
+
+[[Printf]]
+deps = ["Unicode"]
+uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+
+[[REPL]]
+deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
+uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
+
+[[Random]]
+deps = ["Serialization"]
+uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+
+[[Random123]]
+deps = ["Libdl", "Random", "RandomNumbers"]
+git-tree-sha1 = "0e8b146557ad1c6deb1367655e052276690e71a3"
+uuid = "74087812-796a-5b5d-8853-05524746bad3"
+version = "1.4.2"
+
+[[RandomNumbers]]
+deps = ["Random", "Requires"]
+git-tree-sha1 = "043da614cc7e95c703498a491e2c21f58a2b8111"
+uuid = "e6cf234a-135c-5ec9-84dd-332b85af5143"
+version = "1.5.3"
+
+[[Reexport]]
+git-tree-sha1 = "45e428421666073eab6f2da5c9d310d99bb12f9b"
+uuid = "189a3867-3050-52da-a836-e630ba90ab69"
+version = "1.2.2"
+
+[[Requires]]
+deps = ["UUIDs"]
+git-tree-sha1 = "4036a3bd08ac7e968e27c203d45f5fff15020621"
+uuid = "ae029012-a4dd-5104-9daa-d747884805df"
+version = "1.1.3"
+
+[[SHA]]
+uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+
+[[Serialization]]
+uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+
+[[SharedArrays]]
+deps = ["Distributed", "Mmap", "Random", "Serialization"]
+uuid = "1a1011a3-84de-559e-8e89-a11a2f7dc383"
+
+[[Sockets]]
+uuid = "6462fe0b-24de-5631-8697-dd941f90decc"
+
+[[SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[SpecialFunctions]]
+deps = ["ChainRulesCore", "LogExpFunctions", "OpenSpecFun_jll"]
+git-tree-sha1 = "a322a9493e49c5f3a10b50df3aedaf1cdb3244b7"
+uuid = "276daf66-3868-5448-9aa4-cd146d93841b"
+version = "1.6.1"
+
+[[Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
+[[TOML]]
+deps = ["Dates"]
+uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+
+[[Tar]]
+deps = ["ArgTools", "SHA"]
+uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+
+[[Test]]
+deps = ["InteractiveUtils", "Logging", "Random", "Serialization"]
+uuid = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[TimerOutputs]]
+deps = ["ExprTools", "Printf"]
+git-tree-sha1 = "209a8326c4f955e2442c07b56029e88bb48299c7"
+uuid = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
+version = "0.5.12"
+
+[[UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"
+
+[[Unicode]]
+uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
+
+[[Zlib_jll]]
+deps = ["Libdl"]
+uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+
+[[nghttp2_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+
+[[p7zip_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"

JuliaStream.jl/CUDA/Project.toml

@@ -0,0 +1,7 @@
+[deps]
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/KernelAbstractions/Manifest.toml

@@ -0,0 +1,547 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[AMDGPU]]
+deps = ["AbstractFFTs", "Adapt", "BinaryProvider", "CEnum", "GPUArrays", "GPUCompiler", "HIP_jll", "LLVM", "Libdl", "LinearAlgebra", "MacroTools", "Pkg", "Printf", "ROCmDeviceLibs_jll", "Random", "Requires", "Setfield", "hsa_rocr_jll"]
+git-tree-sha1 = "d64c97447a753cfbf0158d6c7be513f34526d559"
+uuid = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+version = "0.2.12"
+
+[[AbstractFFTs]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "485ee0867925449198280d4af84bdb46a2a404d0"
+uuid = "621f4979-c628-5d54-868e-fcf4e3e8185c"
+version = "1.0.1"
+
+[[Adapt]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "84918055d15b3114ede17ac6a7182f68870c16f7"
+uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
+version = "3.3.1"
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[ArgTools]]
+uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+
+[[Artifacts]]
+uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
+
+[[BFloat16s]]
+deps = ["LinearAlgebra", "Test"]
+git-tree-sha1 = "4af69e205efc343068dc8722b8dfec1ade89254a"
+uuid = "ab4f0b2a-ad5b-11e8-123f-65d77653426b"
+version = "0.1.0"
+
+[[Base64]]
+uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
+
+[[BinaryProvider]]
+deps = ["Libdl", "Logging", "SHA"]
+git-tree-sha1 = "ecdec412a9abc8db54c0efc5548c64dfce072058"
+uuid = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
+version = "0.5.10"
+
+[[Bzip2_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "19a35467a82e236ff51bc17a3a44b69ef35185a2"
+uuid = "6e34b625-4abd-537c-b88f-471c36dfa7a0"
+version = "1.0.8+0"
+
+[[CEnum]]
+git-tree-sha1 = "215a9aa4a1f23fbd05b92769fdd62559488d70e9"
+uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+version = "0.4.1"
+
+[[CUDA]]
+deps = ["AbstractFFTs", "Adapt", "BFloat16s", "CEnum", "CompilerSupportLibraries_jll", "DataStructures", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LazyArtifacts", "Libdl", "LinearAlgebra", "Logging", "Printf", "Random", "Random123", "RandomNumbers", "Reexport", "Requires", "SparseArrays", "SpecialFunctions", "TimerOutputs"]
+git-tree-sha1 = "5e696e37e51b01ae07bd9f700afe6cbd55250bce"
+uuid = "052768ef-5323-5732-b1bb-66c8b64840ba"
+version = "3.3.4"
+
+[[CUDAKernels]]
+deps = ["Adapt", "CUDA", "Cassette", "KernelAbstractions", "SpecialFunctions", "StaticArrays"]
+git-tree-sha1 = "81f76297b63c67723b1d60f5e7e002ae3393974b"
+uuid = "72cfdca4-0801-4ab0-bf6a-d52aa10adc57"
+version = "0.3.0"
+
+[[Cassette]]
+git-tree-sha1 = "b4b1d61ebbae2bc69a45e3a6b8439b4e411bc131"
+uuid = "7057c7e9-c182-5462-911a-8362d720325c"
+version = "0.3.8"
+
+[[ChainRulesCore]]
+deps = ["Compat", "LinearAlgebra", "SparseArrays"]
+git-tree-sha1 = "bdc0937269321858ab2a4f288486cb258b9a0af7"
+uuid = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+version = "1.3.0"
+
+[[Compat]]
+deps = ["Base64", "Dates", "DelimitedFiles", "Distributed", "InteractiveUtils", "LibGit2", "Libdl", "LinearAlgebra", "Markdown", "Mmap", "Pkg", "Printf", "REPL", "Random", "SHA", "Serialization", "SharedArrays", "Sockets", "SparseArrays", "Statistics", "Test", "UUIDs", "Unicode"]
+git-tree-sha1 = "727e463cfebd0c7b999bbf3e9e7e16f254b94193"
+uuid = "34da2185-b29b-5c13-b0c7-acf172513d20"
+version = "3.34.0"
+
+[[CompilerSupportLibraries_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+
+[[ConstructionBase]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "f74e9d5388b8620b4cee35d4c5a618dd4dc547f4"
+uuid = "187b0558-2788-49d3-abe0-74a17ed4e7c9"
+version = "1.3.0"
+
+[[DataStructures]]
+deps = ["Compat", "InteractiveUtils", "OrderedCollections"]
+git-tree-sha1 = "7d9d316f04214f7efdbb6398d545446e246eff02"
+uuid = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
+version = "0.18.10"
+
+[[Dates]]
+deps = ["Printf"]
+uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
+
+[[DelimitedFiles]]
+deps = ["Mmap"]
+uuid = "8bb1440f-4735-579b-a4ab-409b98df4dab"
+
+[[Distributed]]
+deps = ["Random", "Serialization", "Sockets"]
+uuid = "8ba89e20-285c-5b6f-9357-94700520ee1b"
+
+[[DocStringExtensions]]
+deps = ["LibGit2"]
+git-tree-sha1 = "a32185f5428d3986f47c2ab78b1f216d5e6cc96f"
+uuid = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
+version = "0.8.5"
+
+[[Downloads]]
+deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+
+[[Elfutils_jll]]
+deps = ["Artifacts", "Bzip2_jll", "JLLWrappers", "Libdl", "Pkg", "XZ_jll", "Zlib_jll", "argp_standalone_jll", "fts_jll", "obstack_jll"]
+git-tree-sha1 = "8f9fcde6d89b0a3ca51cb2028beab462705c5436"
+uuid = "ab5a07f8-06af-567f-a878-e8bb879eba5a"
+version = "0.182.0+0"
+
+[[ExprTools]]
+git-tree-sha1 = "b7e3d17636b348f005f11040025ae8c6f645fe92"
+uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+version = "0.1.6"
+
+[[Future]]
+deps = ["Random"]
+uuid = "9fa8497b-333b-5362-9e8d-4d0656e87820"
+
+[[GPUArrays]]
+deps = ["AbstractFFTs", "Adapt", "LinearAlgebra", "Printf", "Random", "Serialization", "Statistics"]
+git-tree-sha1 = "ececbf05f8904c92814bdbd0aafd5540b0bf2e9a"
+uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
+version = "7.0.1"
+
+[[GPUCompiler]]
+deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "TimerOutputs", "UUIDs"]
+git-tree-sha1 = "4ed2616d5e656c8716736b64da86755467f26cf5"
+uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
+version = "0.12.9"
+
+[[HIP_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "ROCmCompilerSupport_jll", "ROCmDeviceLibs_jll", "ROCmOpenCLRuntime_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "5097d8f7b6842156ab0928371b3d03fefd8decab"
+uuid = "2696aab5-0948-5276-aa9a-2a86a37016b8"
+version = "4.0.0+1"
+
+[[InteractiveUtils]]
+deps = ["Markdown"]
+uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
+
+[[IrrationalConstants]]
+git-tree-sha1 = "f76424439413893a832026ca355fe273e93bce94"
+uuid = "92d709cd-6900-40b7-9082-c6be49f344b6"
+version = "0.1.0"
+
+[[JLLWrappers]]
+deps = ["Preferences"]
+git-tree-sha1 = "642a199af8b68253517b80bd3bfd17eb4e84df6e"
+uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
+version = "1.3.0"
+
+[[KernelAbstractions]]
+deps = ["Adapt", "Cassette", "InteractiveUtils", "MacroTools", "SpecialFunctions", "StaticArrays", "UUIDs"]
+git-tree-sha1 = "5e6c70389c1b1e40adb81664ca8cea6ce8127afc"
+uuid = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+version = "0.7.0"
+
+[[LLVM]]
+deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
+git-tree-sha1 = "23a47d417a3cd9c2e73c854bac7dd4731c105ef7"
+uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
+version = "4.4.0"
+
+[[LLVMExtra_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "9c360e5ce980b88bb31a7b086dbb19469008154b"
+uuid = "dad2f222-ce93-54a1-a47d-0025e8a3acab"
+version = "0.0.10+0"
+
+[[LazyArtifacts]]
+deps = ["Artifacts", "Pkg"]
+uuid = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
+
+[[LibCURL]]
+deps = ["LibCURL_jll", "MozillaCACerts_jll"]
+uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+
+[[LibCURL_jll]]
+deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
+uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+
+[[LibGit2]]
+deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
+uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
+
+[[LibSSH2_jll]]
+deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
+uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+
+[[Libdl]]
+uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+
+[[Libgcrypt_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libgpg_error_jll", "Pkg"]
+git-tree-sha1 = "64613c82a59c120435c067c2b809fc61cf5166ae"
+uuid = "d4300ac3-e22c-5743-9152-c294e39db1e4"
+version = "1.8.7+0"
+
+[[Libglvnd_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libX11_jll", "Xorg_libXext_jll"]
+git-tree-sha1 = "7739f837d6447403596a75d19ed01fd08d6f56bf"
+uuid = "7e76a0d4-f3c7-5321-8279-8d96eeed0f29"
+version = "1.3.0+3"
+
+[[Libgpg_error_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "c333716e46366857753e273ce6a69ee0945a6db9"
+uuid = "7add5ba3-2f88-524e-9cd5-f83b8a55f7b8"
+version = "1.42.0+0"
+
+[[Libiconv_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "42b62845d70a619f063a7da093d995ec8e15e778"
+uuid = "94ce4f54-9a6c-5748-9c1c-f9c7231a4531"
+version = "1.16.1+1"
+
+[[LinearAlgebra]]
+deps = ["Libdl"]
+uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+
+[[LogExpFunctions]]
+deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
+git-tree-sha1 = "3d682c07e6dd250ed082f883dc88aee7996bf2cc"
+uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
+version = "0.3.0"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[MacroTools]]
+deps = ["Markdown", "Random"]
+git-tree-sha1 = "0fb723cd8c45858c22169b2e42269e53271a6df7"
+uuid = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
+version = "0.5.7"
+
+[[Markdown]]
+deps = ["Base64"]
+uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
+
+[[MbedTLS_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+
+[[Mmap]]
+uuid = "a63ad114-7e13-5084-954f-fe012c677804"
+
+[[MozillaCACerts_jll]]
+uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+
+[[NUMA_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "778f9bd14400cff2c32ed357e12766ac0e3d766e"
+uuid = "7f51dc2b-bb24-59f8-b771-bb1490e4195d"
+version = "2.0.13+1"
+
+[[NetworkOptions]]
+uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+
+[[OpenSpecFun_jll]]
+deps = ["Artifacts", "CompilerSupportLibraries_jll", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "13652491f6856acfd2db29360e1bbcd4565d04f1"
+uuid = "efe28fd5-8261-553b-a9e1-b2916fc3738e"
+version = "0.5.5+0"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[Pkg]]
+deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+
+[[Preferences]]
+deps = ["TOML"]
+git-tree-sha1 = "00cfd92944ca9c760982747e9a1d0d5d86ab1e5a"
+uuid = "21216c6a-2e73-6563-6e65-726566657250"
+version = "1.2.2"
+
+[[Printf]]
+deps = ["Unicode"]
+uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+
+[[REPL]]
+deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
+uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
+
+[[ROCKernels]]
+deps = ["AMDGPU", "Adapt", "Cassette", "KernelAbstractions", "SpecialFunctions", "StaticArrays"]
+git-tree-sha1 = "41105b861342637dde17797bdd9aaa537aca646b"
+uuid = "7eb9e9f0-4bd3-4c4c-8bef-26bd9629d9b9"
+version = "0.2.0"
+
+[[ROCmCompilerSupport_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "ROCmDeviceLibs_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "56ddcfb5d8b60c9f8c1bc619886f8d363fd1926d"
+uuid = "8fbdd1d2-db62-5cd0-981e-905da1486e17"
+version = "4.0.0+1"
+
+[[ROCmDeviceLibs_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Zlib_jll"]
+git-tree-sha1 = "d764f0f28b5af89aa004871a6a38e5d061f77257"
+uuid = "873c0968-716b-5aa7-bb8d-d1e2e2aeff2d"
+version = "4.0.0+0"
+
+[[ROCmOpenCLRuntime_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libglvnd_jll", "Pkg", "ROCmCompilerSupport_jll", "ROCmDeviceLibs_jll", "Xorg_libX11_jll", "Xorg_xorgproto_jll", "hsa_rocr_jll"]
+git-tree-sha1 = "f9e3e2cb40a7990535efa7da9b9dd0e0b458a973"
+uuid = "10ae2a08-2eea-53f8-8c20-eec175020e9f"
+version = "4.0.0+1"
+
+[[Random]]
+deps = ["Serialization"]
+uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+
+[[Random123]]
+deps = ["Libdl", "Random", "RandomNumbers"]
+git-tree-sha1 = "0e8b146557ad1c6deb1367655e052276690e71a3"
+uuid = "74087812-796a-5b5d-8853-05524746bad3"
+version = "1.4.2"
+
+[[RandomNumbers]]
+deps = ["Random", "Requires"]
+git-tree-sha1 = "043da614cc7e95c703498a491e2c21f58a2b8111"
+uuid = "e6cf234a-135c-5ec9-84dd-332b85af5143"
+version = "1.5.3"
+
+[[Reexport]]
+git-tree-sha1 = "45e428421666073eab6f2da5c9d310d99bb12f9b"
+uuid = "189a3867-3050-52da-a836-e630ba90ab69"
+version = "1.2.2"
+
+[[Requires]]
+deps = ["UUIDs"]
+git-tree-sha1 = "4036a3bd08ac7e968e27c203d45f5fff15020621"
+uuid = "ae029012-a4dd-5104-9daa-d747884805df"
+version = "1.1.3"
+
+[[SHA]]
+uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+
+[[Serialization]]
+uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+
+[[Setfield]]
+deps = ["ConstructionBase", "Future", "MacroTools", "Requires"]
+git-tree-sha1 = "fca29e68c5062722b5b4435594c3d1ba557072a3"
+uuid = "efcf1570-3423-57d1-acb7-fd33fddbac46"
+version = "0.7.1"
+
+[[SharedArrays]]
+deps = ["Distributed", "Mmap", "Random", "Serialization"]
+uuid = "1a1011a3-84de-559e-8e89-a11a2f7dc383"
+
+[[Sockets]]
+uuid = "6462fe0b-24de-5631-8697-dd941f90decc"
+
+[[SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[SpecialFunctions]]
+deps = ["ChainRulesCore", "LogExpFunctions", "OpenSpecFun_jll"]
+git-tree-sha1 = "a322a9493e49c5f3a10b50df3aedaf1cdb3244b7"
+uuid = "276daf66-3868-5448-9aa4-cd146d93841b"
+version = "1.6.1"
+
+[[StaticArrays]]
+deps = ["LinearAlgebra", "Random", "Statistics"]
+git-tree-sha1 = "3240808c6d463ac46f1c1cd7638375cd22abbccb"
+uuid = "90137ffa-7385-5640-81b9-e52037218182"
+version = "1.2.12"
+
+[[Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
+[[TOML]]
+deps = ["Dates"]
+uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+
+[[Tar]]
+deps = ["ArgTools", "SHA"]
+uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+
+[[Test]]
+deps = ["InteractiveUtils", "Logging", "Random", "Serialization"]
+uuid = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[TimerOutputs]]
+deps = ["ExprTools", "Printf"]
+git-tree-sha1 = "209a8326c4f955e2442c07b56029e88bb48299c7"
+uuid = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
+version = "0.5.12"
+
+[[UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"
+
+[[Unicode]]
+uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
+
+[[XML2_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libiconv_jll", "Pkg", "Zlib_jll"]
+git-tree-sha1 = "1acf5bdf07aa0907e0a37d3718bb88d4b687b74a"
+uuid = "02c8fc9c-b97f-50b9-bbe4-9be30ff0a78a"
+version = "2.9.12+0"
+
+[[XSLT_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Libgcrypt_jll", "Libgpg_error_jll", "Libiconv_jll", "Pkg", "XML2_jll", "Zlib_jll"]
+git-tree-sha1 = "91844873c4085240b95e795f692c4cec4d805f8a"
+uuid = "aed1982a-8fda-507f-9586-7b0439959a61"
+version = "1.1.34+0"
+
+[[XZ_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "a921669cd9a45c23031fd4eb904f5cc3d20de415"
+uuid = "ffd25f8a-64ca-5728-b0f7-c24cf3aae800"
+version = "5.2.5+2"
+
+[[Xorg_libX11_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libxcb_jll", "Xorg_xtrans_jll"]
+git-tree-sha1 = "5be649d550f3f4b95308bf0183b82e2582876527"
+uuid = "4f6342f7-b3d2-589e-9d20-edeb45f2b2bc"
+version = "1.6.9+4"
+
+[[Xorg_libXau_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "4e490d5c960c314f33885790ed410ff3a94ce67e"
+uuid = "0c0b7dd1-d40b-584c-a123-a41640f87eec"
+version = "1.0.9+4"
+
+[[Xorg_libXdmcp_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "4fe47bd2247248125c428978740e18a681372dd4"
+uuid = "a3789734-cfe1-5b06-b2d0-1dd0d9d62d05"
+version = "1.1.3+4"
+
+[[Xorg_libXext_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "Xorg_libX11_jll"]
+git-tree-sha1 = "b7c0aa8c376b31e4852b360222848637f481f8c3"
+uuid = "1082639a-0dae-5f34-9b06-72781eeb8cb3"
+version = "1.3.4+4"
+
+[[Xorg_libpthread_stubs_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "6783737e45d3c59a4a4c4091f5f88cdcf0908cbb"
+uuid = "14d82f49-176c-5ed1-bb49-ad3f5cbd8c74"
+version = "0.1.0+3"
+
+[[Xorg_libxcb_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "XSLT_jll", "Xorg_libXau_jll", "Xorg_libXdmcp_jll", "Xorg_libpthread_stubs_jll"]
+git-tree-sha1 = "daf17f441228e7a3833846cd048892861cff16d6"
+uuid = "c7cfdc94-dc32-55de-ac96-5a1b8d977c5b"
+version = "1.13.0+3"
+
+[[Xorg_xorgproto_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "9a9eb8ce756fe0bca01b4be16da770e18d264972"
+uuid = "c4d99508-4286-5418-9131-c86396af500b"
+version = "2019.2.0+2"
+
+[[Xorg_xtrans_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "79c31e7844f6ecf779705fbc12146eb190b7d845"
+uuid = "c5fb5394-a638-5e4d-96e5-b29de1b5cf10"
+version = "1.4.0+3"
+
+[[Zlib_jll]]
+deps = ["Libdl"]
+uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+
+[[argp_standalone_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "feaf9f6293003c2bf53056fd6930d677ed340b34"
+uuid = "c53206cc-00f7-50bf-ad1e-3ae1f6e49bc3"
+version = "1.3.1+0"
+
+[[fts_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "78732b942383d2cb521df8a1a0814911144e663d"
+uuid = "d65627f6-89bd-53e8-8ab5-8b75ff535eee"
+version = "1.2.7+1"
+
+[[hsa_rocr_jll]]
+deps = ["Artifacts", "Elfutils_jll", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg", "Zlib_jll", "hsakmt_roct_jll"]
+git-tree-sha1 = "df8d73efec8b1e53ad527d208f5343c0368f0fcd"
+uuid = "dd59ff1a-a01a-568d-8b29-0669330f116a"
+version = "4.0.0+0"
+
+[[hsakmt_roct_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg"]
+git-tree-sha1 = "80e0c9940e15cfd6f1f1e9d9f3953ec4d48d3d4a"
+uuid = "1cecccd7-a9b6-5045-9cdc-a44c19b16d76"
+version = "4.0.0+0"
+
+[[nghttp2_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+
+[[obstack_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "1c4a6b66e934fc6db4649cb2910c72f53bbfea7e"
+uuid = "c88a4935-d25e-5644-aacc-5db6f1b8ef79"
+version = "1.2.2+0"
+
+[[p7zip_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"

JuliaStream.jl/KernelAbstractions/Project.toml

@@ -0,0 +1,11 @@
+[deps]
+AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
+CUDAKernels = "72cfdca4-0801-4ab0-bf6a-d52aa10adc57"
+KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+ROCKernels = "7eb9e9f0-4bd3-4c4c-8bef-26bd9629d9b9"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/Manifest.toml

@@ -0,0 +1,493 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[AMDGPU]]
+deps = ["AbstractFFTs", "Adapt", "BinaryProvider", "CEnum", "GPUArrays", "GPUCompiler", "LLVM", "Libdl", "LinearAlgebra", "MacroTools", "Printf", "Random", "Requires", "Setfield", "hsa_rocr_jll", "hsakmt_roct_jll"]
+git-tree-sha1 = "04fdb3923ac6f55fa7347dce0f0f6f10e321e2e9"
+uuid = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+version = "0.2.7"
+
+[[AbstractFFTs]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "485ee0867925449198280d4af84bdb46a2a404d0"
+uuid = "621f4979-c628-5d54-868e-fcf4e3e8185c"
+version = "1.0.1"
+
+[[Adapt]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "84918055d15b3114ede17ac6a7182f68870c16f7"
+uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
+version = "3.3.1"
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[ArgTools]]
+uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+
+[[Artifacts]]
+uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
+
+[[BFloat16s]]
+deps = ["LinearAlgebra", "Test"]
+git-tree-sha1 = "4af69e205efc343068dc8722b8dfec1ade89254a"
+uuid = "ab4f0b2a-ad5b-11e8-123f-65d77653426b"
+version = "0.1.0"
+
+[[Base64]]
+uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
+
+[[BinaryProvider]]
+deps = ["Libdl", "Logging", "SHA"]
+git-tree-sha1 = "ecdec412a9abc8db54c0efc5548c64dfce072058"
+uuid = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
+version = "0.5.10"
+
+[[Bzip2_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "19a35467a82e236ff51bc17a3a44b69ef35185a2"
+uuid = "6e34b625-4abd-537c-b88f-471c36dfa7a0"
+version = "1.0.8+0"
+
+[[CEnum]]
+git-tree-sha1 = "215a9aa4a1f23fbd05b92769fdd62559488d70e9"
+uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+version = "0.4.1"
+
+[[CUDA]]
+deps = ["AbstractFFTs", "Adapt", "BFloat16s", "CEnum", "CompilerSupportLibraries_jll", "DataStructures", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LazyArtifacts", "Libdl", "LinearAlgebra", "Logging", "MacroTools", "Memoize", "Printf", "Random", "Random123", "RandomNumbers", "Reexport", "Requires", "SparseArrays", "SpecialFunctions", "TimerOutputs"]
+git-tree-sha1 = "364179416eabc34c9ca32126a6bdb431680c3bad"
+uuid = "052768ef-5323-5732-b1bb-66c8b64840ba"
+version = "3.2.1"
+
+[[CUDAKernels]]
+deps = ["Adapt", "CUDA", "Cassette", "KernelAbstractions", "SpecialFunctions", "StaticArrays"]
+git-tree-sha1 = "81f76297b63c67723b1d60f5e7e002ae3393974b"
+uuid = "72cfdca4-0801-4ab0-bf6a-d52aa10adc57"
+version = "0.3.0"
+
+[[Cassette]]
+git-tree-sha1 = "b4b1d61ebbae2bc69a45e3a6b8439b4e411bc131"
+uuid = "7057c7e9-c182-5462-911a-8362d720325c"
+version = "0.3.8"
+
+[[ChainRulesCore]]
+deps = ["Compat", "LinearAlgebra", "SparseArrays"]
+git-tree-sha1 = "bdc0937269321858ab2a4f288486cb258b9a0af7"
+uuid = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+version = "1.3.0"
+
+[[Compat]]
+deps = ["Base64", "Dates", "DelimitedFiles", "Distributed", "InteractiveUtils", "LibGit2", "Libdl", "LinearAlgebra", "Markdown", "Mmap", "Pkg", "Printf", "REPL", "Random", "SHA", "Serialization", "SharedArrays", "Sockets", "SparseArrays", "Statistics", "Test", "UUIDs", "Unicode"]
+git-tree-sha1 = "727e463cfebd0c7b999bbf3e9e7e16f254b94193"
+uuid = "34da2185-b29b-5c13-b0c7-acf172513d20"
+version = "3.34.0"
+
+[[CompilerSupportLibraries_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+
+[[ConstructionBase]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "f74e9d5388b8620b4cee35d4c5a618dd4dc547f4"
+uuid = "187b0558-2788-49d3-abe0-74a17ed4e7c9"
+version = "1.3.0"
+
+[[DataStructures]]
+deps = ["Compat", "InteractiveUtils", "OrderedCollections"]
+git-tree-sha1 = "7d9d316f04214f7efdbb6398d545446e246eff02"
+uuid = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
+version = "0.18.10"
+
+[[Dates]]
+deps = ["Printf"]
+uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
+
+[[DelimitedFiles]]
+deps = ["Mmap"]
+uuid = "8bb1440f-4735-579b-a4ab-409b98df4dab"
+
+[[Distributed]]
+deps = ["Random", "Serialization", "Sockets"]
+uuid = "8ba89e20-285c-5b6f-9357-94700520ee1b"
+
+[[DocStringExtensions]]
+deps = ["LibGit2"]
+git-tree-sha1 = "a32185f5428d3986f47c2ab78b1f216d5e6cc96f"
+uuid = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
+version = "0.8.5"
+
+[[Downloads]]
+deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+
+[[Elfutils_jll]]
+deps = ["Artifacts", "Bzip2_jll", "JLLWrappers", "Libdl", "Pkg", "XZ_jll", "Zlib_jll", "argp_standalone_jll", "fts_jll", "obstack_jll"]
+git-tree-sha1 = "8f9fcde6d89b0a3ca51cb2028beab462705c5436"
+uuid = "ab5a07f8-06af-567f-a878-e8bb879eba5a"
+version = "0.182.0+0"
+
+[[ExprTools]]
+git-tree-sha1 = "b7e3d17636b348f005f11040025ae8c6f645fe92"
+uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+version = "0.1.6"
+
+[[Future]]
+deps = ["Random"]
+uuid = "9fa8497b-333b-5362-9e8d-4d0656e87820"
+
+[[GPUArrays]]
+deps = ["AbstractFFTs", "Adapt", "LinearAlgebra", "Printf", "Random", "Serialization", "Statistics"]
+git-tree-sha1 = "df5b8569904c5c10e84c640984cfff054b18c086"
+uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
+version = "6.4.1"
+
+[[GPUCompiler]]
+deps = ["DataStructures", "ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "Scratch", "Serialization", "TimerOutputs", "UUIDs"]
+git-tree-sha1 = "42d635f6d87af125b86288df3819f805fb4d851a"
+uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
+version = "0.11.5"
+
+[[InteractiveUtils]]
+deps = ["Markdown"]
+uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
+
+[[IrrationalConstants]]
+git-tree-sha1 = "f76424439413893a832026ca355fe273e93bce94"
+uuid = "92d709cd-6900-40b7-9082-c6be49f344b6"
+version = "0.1.0"
+
+[[JLLWrappers]]
+deps = ["Preferences"]
+git-tree-sha1 = "642a199af8b68253517b80bd3bfd17eb4e84df6e"
+uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
+version = "1.3.0"
+
+[[KernelAbstractions]]
+deps = ["Adapt", "Cassette", "InteractiveUtils", "MacroTools", "SpecialFunctions", "StaticArrays", "UUIDs"]
+git-tree-sha1 = "5e6c70389c1b1e40adb81664ca8cea6ce8127afc"
+uuid = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+version = "0.7.0"
+
+[[LLVM]]
+deps = ["CEnum", "Libdl", "Printf", "Unicode"]
+git-tree-sha1 = "f57ac3fd2045b50d3db081663837ac5b4096947e"
+uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
+version = "3.9.0"
+
+[[LazyArtifacts]]
+deps = ["Artifacts", "Pkg"]
+uuid = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
+
+[[LibCURL]]
+deps = ["LibCURL_jll", "MozillaCACerts_jll"]
+uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+
+[[LibCURL_jll]]
+deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
+uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+
+[[LibGit2]]
+deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
+uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
+
+[[LibSSH2_jll]]
+deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
+uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+
+[[Libdl]]
+uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+
+[[LinearAlgebra]]
+deps = ["Libdl"]
+uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+
+[[LogExpFunctions]]
+deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
+git-tree-sha1 = "3d682c07e6dd250ed082f883dc88aee7996bf2cc"
+uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
+version = "0.3.0"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[MacroTools]]
+deps = ["Markdown", "Random"]
+git-tree-sha1 = "0fb723cd8c45858c22169b2e42269e53271a6df7"
+uuid = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
+version = "0.5.7"
+
+[[Markdown]]
+deps = ["Base64"]
+uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
+
+[[MbedTLS_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+
+[[Memoize]]
+deps = ["MacroTools"]
+git-tree-sha1 = "2b1dfcba103de714d31c033b5dacc2e4a12c7caa"
+uuid = "c03570c3-d221-55d1-a50c-7939bbd78826"
+version = "0.4.4"
+
+[[Mmap]]
+uuid = "a63ad114-7e13-5084-954f-fe012c677804"
+
+[[MozillaCACerts_jll]]
+uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+
+[[NEO_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "gmmlib_jll", "libigc_jll", "oneAPI_Level_Zero_Headers_jll"]
+git-tree-sha1 = "c753dd029eb0837658bf8eaee041c19e4ce5bb8c"
+uuid = "700fe977-ac61-5f37-bbc8-c6c4b2b6a9fd"
+version = "21.12.19358+0"
+
+[[NUMA_jll]]
+deps = ["Libdl", "Pkg"]
+git-tree-sha1 = "778f9bd14400cff2c32ed357e12766ac0e3d766e"
+uuid = "7f51dc2b-bb24-59f8-b771-bb1490e4195d"
+version = "2.0.13+1"
+
+[[NetworkOptions]]
+uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+
+[[OpenSpecFun_jll]]
+deps = ["Artifacts", "CompilerSupportLibraries_jll", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "13652491f6856acfd2db29360e1bbcd4565d04f1"
+uuid = "efe28fd5-8261-553b-a9e1-b2916fc3738e"
+version = "0.5.5+0"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[Pkg]]
+deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+
+[[Preferences]]
+deps = ["TOML"]
+git-tree-sha1 = "00cfd92944ca9c760982747e9a1d0d5d86ab1e5a"
+uuid = "21216c6a-2e73-6563-6e65-726566657250"
+version = "1.2.2"
+
+[[Printf]]
+deps = ["Unicode"]
+uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+
+[[REPL]]
+deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
+uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
+
+[[ROCKernels]]
+deps = ["AMDGPU", "Adapt", "Cassette", "KernelAbstractions", "SpecialFunctions", "StaticArrays"]
+git-tree-sha1 = "41105b861342637dde17797bdd9aaa537aca646b"
+uuid = "7eb9e9f0-4bd3-4c4c-8bef-26bd9629d9b9"
+version = "0.2.0"
+
+[[Random]]
+deps = ["Serialization"]
+uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+
+[[Random123]]
+deps = ["Libdl", "Random", "RandomNumbers"]
+git-tree-sha1 = "0e8b146557ad1c6deb1367655e052276690e71a3"
+uuid = "74087812-796a-5b5d-8853-05524746bad3"
+version = "1.4.2"
+
+[[RandomNumbers]]
+deps = ["Random", "Requires"]
+git-tree-sha1 = "043da614cc7e95c703498a491e2c21f58a2b8111"
+uuid = "e6cf234a-135c-5ec9-84dd-332b85af5143"
+version = "1.5.3"
+
+[[Reexport]]
+git-tree-sha1 = "45e428421666073eab6f2da5c9d310d99bb12f9b"
+uuid = "189a3867-3050-52da-a836-e630ba90ab69"
+version = "1.2.2"
+
+[[Requires]]
+deps = ["UUIDs"]
+git-tree-sha1 = "4036a3bd08ac7e968e27c203d45f5fff15020621"
+uuid = "ae029012-a4dd-5104-9daa-d747884805df"
+version = "1.1.3"
+
+[[SHA]]
+uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+
+[[SPIRV_LLVM_Translator_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "8cca87d57f6ddf19373cc9791fddc741406c8fbf"
+uuid = "4a5d46fc-d8cf-5151-a261-86b458210efb"
+version = "11.0.0+2"
+
+[[SPIRV_Tools_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "c0324b7e07bc4649f755bfe7e00f7c6ed6aa353f"
+uuid = "6ac6d60f-d740-5983-97d7-a4482c0689f4"
+version = "2021.2.0+0"
+
+[[Scratch]]
+deps = ["Dates"]
+git-tree-sha1 = "0b4b7f1393cff97c33891da2a0bf69c6ed241fda"
+uuid = "6c6a2e73-6563-6170-7368-637461726353"
+version = "1.1.0"
+
+[[Serialization]]
+uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+
+[[Setfield]]
+deps = ["ConstructionBase", "Future", "MacroTools", "Requires"]
+git-tree-sha1 = "fca29e68c5062722b5b4435594c3d1ba557072a3"
+uuid = "efcf1570-3423-57d1-acb7-fd33fddbac46"
+version = "0.7.1"
+
+[[SharedArrays]]
+deps = ["Distributed", "Mmap", "Random", "Serialization"]
+uuid = "1a1011a3-84de-559e-8e89-a11a2f7dc383"
+
+[[Sockets]]
+uuid = "6462fe0b-24de-5631-8697-dd941f90decc"
+
+[[SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[SpecialFunctions]]
+deps = ["ChainRulesCore", "LogExpFunctions", "OpenSpecFun_jll"]
+git-tree-sha1 = "a322a9493e49c5f3a10b50df3aedaf1cdb3244b7"
+uuid = "276daf66-3868-5448-9aa4-cd146d93841b"
+version = "1.6.1"
+
+[[StaticArrays]]
+deps = ["LinearAlgebra", "Random", "Statistics"]
+git-tree-sha1 = "3240808c6d463ac46f1c1cd7638375cd22abbccb"
+uuid = "90137ffa-7385-5640-81b9-e52037218182"
+version = "1.2.12"
+
+[[Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
+[[TOML]]
+deps = ["Dates"]
+uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+
+[[Tar]]
+deps = ["ArgTools", "SHA"]
+uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+
+[[Test]]
+deps = ["InteractiveUtils", "Logging", "Random", "Serialization"]
+uuid = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[TimerOutputs]]
+deps = ["ExprTools", "Printf"]
+git-tree-sha1 = "209a8326c4f955e2442c07b56029e88bb48299c7"
+uuid = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
+version = "0.5.12"
+
+[[UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"
+
+[[Unicode]]
+uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
+
+[[XZ_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "a921669cd9a45c23031fd4eb904f5cc3d20de415"
+uuid = "ffd25f8a-64ca-5728-b0f7-c24cf3aae800"
+version = "5.2.5+2"
+
+[[Zlib_jll]]
+deps = ["Libdl"]
+uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+
+[[argp_standalone_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "feaf9f6293003c2bf53056fd6930d677ed340b34"
+uuid = "c53206cc-00f7-50bf-ad1e-3ae1f6e49bc3"
+version = "1.3.1+0"
+
+[[fts_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "78732b942383d2cb521df8a1a0814911144e663d"
+uuid = "d65627f6-89bd-53e8-8ab5-8b75ff535eee"
+version = "1.2.7+1"
+
+[[gmmlib_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "4067ef455d4fa67febe26efc3f9565a9bb7ba911"
+uuid = "09858cae-167c-5acb-9302-fddc6874d481"
+version = "20.3.2+0"
+
+[[hsa_rocr_jll]]
+deps = ["Artifacts", "Elfutils_jll", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg", "Zlib_jll", "hsakmt_roct_jll"]
+git-tree-sha1 = "42189f176d6ae4f37c0c0e652fec339bb0bfab5d"
+uuid = "dd59ff1a-a01a-568d-8b29-0669330f116a"
+version = "3.7.0+1"
+
+[[hsakmt_roct_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "NUMA_jll", "Pkg"]
+git-tree-sha1 = "8a9ee6c091e952e4ea6585d15131d43f789ae041"
+uuid = "1cecccd7-a9b6-5045-9cdc-a44c19b16d76"
+version = "3.8.0+0"
+
+[[libigc_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "6140dbf267f7ab57fb791b49f2114374218b5c20"
+uuid = "94295238-5935-5bd7-bb0f-b00942e9bdd5"
+version = "1.0.6712+0"
+
+[[nghttp2_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+
+[[obstack_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "1c4a6b66e934fc6db4649cb2910c72f53bbfea7e"
+uuid = "c88a4935-d25e-5644-aacc-5db6f1b8ef79"
+version = "1.2.2+0"
+
+[[oneAPI]]
+deps = ["Adapt", "CEnum", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LinearAlgebra", "NEO_jll", "Printf", "Random", "SPIRV_LLVM_Translator_jll", "SPIRV_Tools_jll", "SpecialFunctions", "oneAPI_Level_Zero_Loader_jll"]
+git-tree-sha1 = "b4a4b84c864e75fe885a1643525f0c97ce310dd9"
+uuid = "8f75cd03-7ff8-4ecb-9b8f-daf728133b1b"
+version = "0.1.3"
+
+[[oneAPI_Level_Zero_Headers_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "48982fbfd2f3d0a30d644563dcf96892d252b395"
+uuid = "f4bc562b-d309-54f8-9efb-476e56f0410d"
+version = "1.1.2+1"
+
+[[oneAPI_Level_Zero_Loader_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "oneAPI_Level_Zero_Headers_jll"]
+git-tree-sha1 = "1fa53dfdd32a732f09c254c86403e1abab653fb2"
+uuid = "13eca655-d68d-5b81-8367-6d99d727ab01"
+version = "1.3.6+0"
+
+[[p7zip_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"

JuliaStream.jl/Project.toml

@@ -0,0 +1,19 @@
+name = "JuliaStream"
+uuid = "1bdcc9b7-f5ed-4705-bc7b-be1b748ec681"
+authors = ["Wei-Chen Lin <wl14928@bristol.ac.uk>"]
+version = "3.4.0"
+
+[deps]
+AMDGPU = "21141c5a-9bdb-4563-92ae-f87d6854732e"
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
+CUDAKernels = "72cfdca4-0801-4ab0-bf6a-d52aa10adc57"
+Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"
+ExprTools = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+ROCKernels = "7eb9e9f0-4bd3-4c4c-8bef-26bd9629d9b9"
+oneAPI = "8f75cd03-7ff8-4ecb-9b8f-daf728133b1b"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/README.md

@@ -0,0 +1,67 @@
+JuliaStream.jl
+==============
+
+This is an implementation of BabelStream in Julia which contains the following variants:
+
+ * `PlainStream.jl` - Single threaded `for`
+ * `ThreadedStream.jl` - Threaded implementation with `Threads.@threads` macros
+ * `DistributedStream.jl` - Process based parallelism with `@distributed` macros
+ * `CUDAStream.jl` - Direct port of BabelStream's native CUDA implementation using [CUDA.jl](https://github.com/JuliaGPU/CUDA.jl)
+ * `AMDGPUStream.jl` - Direct port of BabelStream's native HIP implementation using [AMDGPU.jl](https://github.com/JuliaGPU/AMDGPU.jl)
+ * `oneAPIStream.jl` - Direct port of BabelStream's native SYCL implementation using [oneAPI.jl](https://github.com/JuliaGPU/oneAPI.jl)
+ * `KernelAbstractions.jl` - Direct port of miniBUDE's native CUDA implementation using [KernelAbstractions.jl](https://github.com/JuliaGPU/KernelAbstractions.jl)
+
+### Build & Run
+
+Prerequisites
+
+ * Julia >= 1.6+
+
+A set of reduced dependency projects are available for the following backend and implementations:
+
+ * `AMDGPU` supports:
+   - `AMDGPUStream.jl`
+ * `CUDA` supports:
+   - `CUDAStream.jl`
+ * `oneAPI` supports:
+   - `oneAPIStream.jl`
+ * `KernelAbstractions` supports:
+   - `KernelAbstractionsStream.jl`
+ * `Threaded` supports:
+   - `PlainStream.jl`
+   - `ThreadedStream.jl`
+   - `DistributedStream.jl`
+
+With Julia on path, run your selected benchmark with:
+
+```shell
+> cd JuliaStream.jl
+> julia --project=<BACKEND> -e 'import Pkg; Pkg.instantiate()' # only required on first run
+> julia --project=<BACKEND> src/<IMPL>Stream.jl
+```
+
+For example. to run the CUDA implementation:
+
+```shell
+> cd JuliaStream.jl
+> julia --project=CUDA -e 'import Pkg; Pkg.instantiate()' 
+> julia --project=CUDA src/CUDAStream.jl
+```
+
+**Important:**
+ * Julia is 1-indexed, so N >= 1 in `--device N`.
+ * Thread count for `ThreadedStream` must be set via the `JULIA_NUM_THREADS` environment variable (e.g `export JULIA_NUM_THREADS=$(nproc)`) otherwise it defaults to 1.
+ * Worker count for `DistributedStream` is set with `-p <N>` as per the [documentation](https://docs.julialang.org/en/v1/manual/distributed-computing).
+ * Certain implementations such as CUDA and AMDGPU will do hardware detection at runtime and may download and/or compile further software packages for the platform.
+
+***
+
+Alternatively, the top-level project `Project.toml` contains all dependencies needed to run all implementations in `src`.
+There may be instances where some packages are locked to an older version because of transitive dependency requirements.
+
+To run the benchmark using the top-level project, run the benchmark with:
+```shell
+> cd JuliaStream.jl
+> julia --project -e 'import Pkg; Pkg.instantiate()'  
+> julia --project src/<IMPL>Stream.jl
+```

JuliaStream.jl/Threaded/Manifest.toml

@@ -0,0 +1,31 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"

JuliaStream.jl/Threaded/Project.toml

@@ -0,0 +1,6 @@
+[deps]
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/oneAPI/Manifest.toml

@@ -0,0 +1,319 @@
+# This file is machine-generated - editing it directly is not advised
+
+[[Adapt]]
+deps = ["LinearAlgebra"]
+git-tree-sha1 = "84918055d15b3114ede17ac6a7182f68870c16f7"
+uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
+version = "3.3.1"
+
+[[ArgParse]]
+deps = ["Logging", "TextWrap"]
+git-tree-sha1 = "3102bce13da501c9104df33549f511cd25264d7d"
+uuid = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+version = "1.1.4"
+
+[[ArgTools]]
+uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+
+[[Artifacts]]
+uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
+
+[[Base64]]
+uuid = "2a0f44e3-6c83-55bd-87e4-b1978d98bd5f"
+
+[[CEnum]]
+git-tree-sha1 = "215a9aa4a1f23fbd05b92769fdd62559488d70e9"
+uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+version = "0.4.1"
+
+[[ChainRulesCore]]
+deps = ["Compat", "LinearAlgebra", "SparseArrays"]
+git-tree-sha1 = "bdc0937269321858ab2a4f288486cb258b9a0af7"
+uuid = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+version = "1.3.0"
+
+[[Compat]]
+deps = ["Base64", "Dates", "DelimitedFiles", "Distributed", "InteractiveUtils", "LibGit2", "Libdl", "LinearAlgebra", "Markdown", "Mmap", "Pkg", "Printf", "REPL", "Random", "SHA", "Serialization", "SharedArrays", "Sockets", "SparseArrays", "Statistics", "Test", "UUIDs", "Unicode"]
+git-tree-sha1 = "727e463cfebd0c7b999bbf3e9e7e16f254b94193"
+uuid = "34da2185-b29b-5c13-b0c7-acf172513d20"
+version = "3.34.0"
+
+[[CompilerSupportLibraries_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+
+[[Dates]]
+deps = ["Printf"]
+uuid = "ade2ca70-3891-5945-98fb-dc099432e06a"
+
+[[DelimitedFiles]]
+deps = ["Mmap"]
+uuid = "8bb1440f-4735-579b-a4ab-409b98df4dab"
+
+[[Distributed]]
+deps = ["Random", "Serialization", "Sockets"]
+uuid = "8ba89e20-285c-5b6f-9357-94700520ee1b"
+
+[[DocStringExtensions]]
+deps = ["LibGit2"]
+git-tree-sha1 = "a32185f5428d3986f47c2ab78b1f216d5e6cc96f"
+uuid = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
+version = "0.8.5"
+
+[[Downloads]]
+deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+
+[[ExprTools]]
+git-tree-sha1 = "b7e3d17636b348f005f11040025ae8c6f645fe92"
+uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+version = "0.1.6"
+
+[[GPUArrays]]
+deps = ["Adapt", "LinearAlgebra", "Printf", "Random", "Serialization", "Statistics"]
+git-tree-sha1 = "8fac1cf7d6ce0f2249c7acaf25d22e1e85c4a07f"
+uuid = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
+version = "8.0.2"
+
+[[GPUCompiler]]
+deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "TimerOutputs", "UUIDs"]
+git-tree-sha1 = "4ed2616d5e656c8716736b64da86755467f26cf5"
+uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
+version = "0.12.9"
+
+[[InteractiveUtils]]
+deps = ["Markdown"]
+uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
+
+[[IrrationalConstants]]
+git-tree-sha1 = "f76424439413893a832026ca355fe273e93bce94"
+uuid = "92d709cd-6900-40b7-9082-c6be49f344b6"
+version = "0.1.0"
+
+[[JLLWrappers]]
+deps = ["Preferences"]
+git-tree-sha1 = "642a199af8b68253517b80bd3bfd17eb4e84df6e"
+uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
+version = "1.3.0"
+
+[[LLVM]]
+deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
+git-tree-sha1 = "23a47d417a3cd9c2e73c854bac7dd4731c105ef7"
+uuid = "929cbde3-209d-540e-8aea-75f648917ca0"
+version = "4.4.0"
+
+[[LLVMExtra_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "9c360e5ce980b88bb31a7b086dbb19469008154b"
+uuid = "dad2f222-ce93-54a1-a47d-0025e8a3acab"
+version = "0.0.10+0"
+
+[[LibCURL]]
+deps = ["LibCURL_jll", "MozillaCACerts_jll"]
+uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+
+[[LibCURL_jll]]
+deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
+uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+
+[[LibGit2]]
+deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
+uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
+
+[[LibSSH2_jll]]
+deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
+uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+
+[[Libdl]]
+uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+
+[[LinearAlgebra]]
+deps = ["Libdl"]
+uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+
+[[LogExpFunctions]]
+deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
+git-tree-sha1 = "3d682c07e6dd250ed082f883dc88aee7996bf2cc"
+uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
+version = "0.3.0"
+
+[[Logging]]
+uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
+
+[[Markdown]]
+deps = ["Base64"]
+uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
+
+[[MbedTLS_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+
+[[Mmap]]
+uuid = "a63ad114-7e13-5084-954f-fe012c677804"
+
+[[MozillaCACerts_jll]]
+uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+
+[[NEO_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "gmmlib_jll", "libigc_jll", "oneAPI_Level_Zero_Headers_jll"]
+git-tree-sha1 = "2bfc354b5684821dcc88f1e477cefd0dd03c60b5"
+uuid = "700fe977-ac61-5f37-bbc8-c6c4b2b6a9fd"
+version = "21.31.20514+0"
+
+[[NetworkOptions]]
+uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+
+[[OpenSpecFun_jll]]
+deps = ["Artifacts", "CompilerSupportLibraries_jll", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "13652491f6856acfd2db29360e1bbcd4565d04f1"
+uuid = "efe28fd5-8261-553b-a9e1-b2916fc3738e"
+version = "0.5.5+0"
+
+[[OrderedCollections]]
+git-tree-sha1 = "85f8e6578bf1f9ee0d11e7bb1b1456435479d47c"
+uuid = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
+version = "1.4.1"
+
+[[Parameters]]
+deps = ["OrderedCollections", "UnPack"]
+git-tree-sha1 = "2276ac65f1e236e0a6ea70baff3f62ad4c625345"
+uuid = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+version = "0.12.2"
+
+[[Pkg]]
+deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+
+[[Preferences]]
+deps = ["TOML"]
+git-tree-sha1 = "00cfd92944ca9c760982747e9a1d0d5d86ab1e5a"
+uuid = "21216c6a-2e73-6563-6e65-726566657250"
+version = "1.2.2"
+
+[[Printf]]
+deps = ["Unicode"]
+uuid = "de0858da-6303-5e67-8744-51eddeeeb8d7"
+
+[[REPL]]
+deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
+uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
+
+[[Random]]
+deps = ["Serialization"]
+uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+
+[[SHA]]
+uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+
+[[SPIRV_LLVM_Translator_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "8cca87d57f6ddf19373cc9791fddc741406c8fbf"
+uuid = "4a5d46fc-d8cf-5151-a261-86b458210efb"
+version = "11.0.0+2"
+
+[[SPIRV_Tools_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "c0324b7e07bc4649f755bfe7e00f7c6ed6aa353f"
+uuid = "6ac6d60f-d740-5983-97d7-a4482c0689f4"
+version = "2021.2.0+0"
+
+[[Serialization]]
+uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
+
+[[SharedArrays]]
+deps = ["Distributed", "Mmap", "Random", "Serialization"]
+uuid = "1a1011a3-84de-559e-8e89-a11a2f7dc383"
+
+[[Sockets]]
+uuid = "6462fe0b-24de-5631-8697-dd941f90decc"
+
+[[SparseArrays]]
+deps = ["LinearAlgebra", "Random"]
+uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
+
+[[SpecialFunctions]]
+deps = ["ChainRulesCore", "LogExpFunctions", "OpenSpecFun_jll"]
+git-tree-sha1 = "a322a9493e49c5f3a10b50df3aedaf1cdb3244b7"
+uuid = "276daf66-3868-5448-9aa4-cd146d93841b"
+version = "1.6.1"
+
+[[Statistics]]
+deps = ["LinearAlgebra", "SparseArrays"]
+uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+
+[[TOML]]
+deps = ["Dates"]
+uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+
+[[Tar]]
+deps = ["ArgTools", "SHA"]
+uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+
+[[Test]]
+deps = ["InteractiveUtils", "Logging", "Random", "Serialization"]
+uuid = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[[TextWrap]]
+git-tree-sha1 = "9250ef9b01b66667380cf3275b3f7488d0e25faf"
+uuid = "b718987f-49a8-5099-9789-dcd902bef87d"
+version = "1.0.1"
+
+[[TimerOutputs]]
+deps = ["ExprTools", "Printf"]
+git-tree-sha1 = "209a8326c4f955e2442c07b56029e88bb48299c7"
+uuid = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
+version = "0.5.12"
+
+[[UUIDs]]
+deps = ["Random", "SHA"]
+uuid = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[[UnPack]]
+git-tree-sha1 = "387c1f73762231e86e0c9c5443ce3b4a0a9a0c2b"
+uuid = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
+version = "1.0.2"
+
+[[Unicode]]
+uuid = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
+
+[[Zlib_jll]]
+deps = ["Libdl"]
+uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+
+[[gmmlib_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "0d5e5461d21b14853b4c332045c57d2601c403bd"
+uuid = "09858cae-167c-5acb-9302-fddc6874d481"
+version = "21.2.1+0"
+
+[[libigc_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "b30a895e7ea52991a3f984ab0302c42858d766c0"
+uuid = "94295238-5935-5bd7-bb0f-b00942e9bdd5"
+version = "1.0.8173+0"
+
+[[nghttp2_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+
+[[oneAPI]]
+deps = ["Adapt", "CEnum", "ExprTools", "GPUArrays", "GPUCompiler", "LLVM", "LinearAlgebra", "NEO_jll", "Printf", "Random", "SPIRV_LLVM_Translator_jll", "SPIRV_Tools_jll", "SpecialFunctions", "oneAPI_Level_Zero_Headers_jll", "oneAPI_Level_Zero_Loader_jll"]
+git-tree-sha1 = "92e8eefdd4694597994590230ab329545804bdb3"
+uuid = "8f75cd03-7ff8-4ecb-9b8f-daf728133b1b"
+version = "0.2.0"
+
+[[oneAPI_Level_Zero_Headers_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg"]
+git-tree-sha1 = "e1d123ff9ada6c469a1eaf57e33a74c3cb26a5a4"
+uuid = "f4bc562b-d309-54f8-9efb-476e56f0410d"
+version = "1.2.13+0"
+
+[[oneAPI_Level_Zero_Loader_jll]]
+deps = ["Artifacts", "JLLWrappers", "Libdl", "Pkg", "oneAPI_Level_Zero_Headers_jll"]
+git-tree-sha1 = "50124857f7e87420655929a9c8ca86749826af11"
+uuid = "13eca655-d68d-5b81-8367-6d99d727ab01"
+version = "1.4.1+0"
+
+[[p7zip_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"

JuliaStream.jl/oneAPI/Project.toml

@@ -0,0 +1,7 @@
+[deps]
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
+Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
+oneAPI = "8f75cd03-7ff8-4ecb-9b8f-daf728133b1b"
+
+[compat]
+julia = "1.6"

JuliaStream.jl/src/AMDGPUStream.jl

@@ -0,0 +1,167 @@
+# AMDGPU.jl doesn't support CPU agents, so this isn't a feature-complete ROCmStream, only AMD GPUs
+include("Stream.jl")
+using AMDGPU
+
+const ROCData = StreamData{T,ROCArray{T}} where {T}
+const TBSize = 1024::Int
+const DotBlocks = 256::Int
+
+function devices()::Vector{DeviceWithRepr}
+  try
+    # AMDGPU.agents()'s internal iteration order isn't stable
+    sorted = sort(AMDGPU.get_agents(:gpu), by = repr)
+    map(x -> (x, repr(x), "AMDGPU.jl"), sorted)
+  catch
+    # probably unsupported
+    String[]
+  end
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  device::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{ROCData{T},Nothing} where {T}
+
+  if arraysize % TBSize != 0
+    error("arraysize ($(arraysize)) must be divisible by $(TBSize)!")
+  end
+
+  # XXX AMDGPU doesn't expose an API for setting the default like CUDA.device!()
+  # but AMDGPU.get_default_agent returns DEFAULT_AGENT so we can do it by hand
+  AMDGPU.DEFAULT_AGENT[] = device[1]
+  selected = AMDGPU.get_default_agent()
+  if !silent
+    println("Using GPU HSA device: $(AMDGPU.get_name(selected)) ($(repr(selected)))")
+    println("Kernel parameters   : <<<$(arraysize),$(TBSize)>>>")
+  end
+  return (
+    ROCData{T}(
+      ROCArray{T}(undef, arraysize),
+      ROCArray{T}(undef, arraysize),
+      ROCArray{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    ),
+    nothing,
+  )
+end
+
+function init_arrays!(data::ROCData{T}, _, init::Tuple{T,T,T}) where {T}
+  AMDGPU.fill!(data.a, init[1])
+  AMDGPU.fill!(data.b, init[2])
+  AMDGPU.fill!(data.c, init[3])
+end
+
+function copy!(data::ROCData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, c::AbstractArray{T})
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    @inbounds c[i] = a[i]
+    return
+  end
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = data.size kernel(data.a, data.c)
+  )
+end
+
+function mul!(data::ROCData{T}, _) where {T}
+  function kernel(b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    @inbounds b[i] = scalar * c[i]
+    return
+  end
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = data.size kernel(data.b, data.c, data.scalar)
+  )
+end
+
+function add!(data::ROCData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T})
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    @inbounds c[i] = a[i] + b[i]
+    return
+  end
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = data.size kernel(data.a, data.b, data.c)
+  )
+end
+
+function triad!(data::ROCData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    @inbounds a[i] = b[i] + (scalar * c[i])
+    return
+  end
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = data.size kernel(
+      data.a,
+      data.b,
+      data.c,
+      data.scalar,
+    )
+  )
+end
+
+function nstream!(data::ROCData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    @inbounds a[i] += b[i] + scalar * c[i]
+    return
+  end
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = data.size kernel(
+      data.a,
+      data.b,
+      data.c,
+      data.scalar,
+    )
+  )
+end
+
+function dot(data::ROCData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, size::Int, partial::AbstractArray{T})
+    tb_sum = ROCDeviceArray((TBSize,), alloc_local(:reduce, T, TBSize))
+    local_i = workitemIdx().x
+    @inbounds tb_sum[local_i] = 0.0
+
+    # do dot first
+    i = (workgroupIdx().x - 1) * workgroupDim().x + workitemIdx().x  # only workgroupIdx starts at 1
+    while i <= size
+      @inbounds tb_sum[local_i] += a[i] * b[i]
+      i += TBSize * DotBlocks # XXX don't use (workgroupDim().x * gridDimWG().x) here
+    end
+
+    # then tree reduction
+    offset = workgroupDim().x ÷ 2
+    while offset > 0
+      sync_workgroup()
+      if (local_i - 1) < offset
+        @inbounds tb_sum[local_i] += tb_sum[local_i+offset]
+      end
+      offset ÷= 2
+    end
+
+    if (local_i == 1)
+      @inbounds partial[workgroupIdx().x] = tb_sum[local_i]
+    end
+
+    return
+  end
+  partial_sum = ROCArray{T}(undef, DotBlocks)
+  AMDGPU.wait(
+    @roc groupsize = TBSize gridsize = TBSize * DotBlocks kernel(
+      data.a,
+      data.b,
+      data.size,
+      partial_sum,
+    )
+  )
+  return sum(partial_sum)
+end
+
+function read_data(data::ROCData{T}, _)::VectorData{T} where {T}
+  return VectorData{T}(data.a, data.b, data.c, data.scalar, data.size)
+end
+
+main()

JuliaStream.jl/src/CUDAStream.jl

@@ -0,0 +1,152 @@
+include("Stream.jl")
+using CUDA
+
+const CuData = StreamData{T,CuArray{T}} where {T}
+const TBSize = 1024::Int
+const DotBlocks = 256::Int
+
+function devices()::Vector{DeviceWithRepr}
+  return !CUDA.functional(false) ? String[] :
+         map(d -> (d, "$(CUDA.name(d)) ($(repr(d)))", "CUDA.jl"), CUDA.devices())
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  device::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{CuData{T},Nothing} where {T}
+
+  if arraysize % TBSize != 0
+    error("arraysize ($(arraysize)) must be divisible by $(TBSize)!")
+  end
+
+  CUDA.device!(device[1])
+  selected = CUDA.device()
+  # show_reason is set to true here so it dumps CUDA info 
+  # for us regardless of whether it's functional
+  if !CUDA.functional(true)
+    error("Non-functional CUDA configuration")
+  end
+  if !silent
+    println("Using CUDA device: $(CUDA.name(selected)) ($(repr(selected)))")
+    println("Kernel parameters: <<<$(arraysize ÷ TBSize),$(TBSize)>>>")
+  end
+  return (
+    CuData{T}(
+      CuArray{T}(undef, arraysize),
+      CuArray{T}(undef, arraysize),
+      CuArray{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    ),
+    nothing,
+  )
+end
+
+function init_arrays!(data::CuData{T}, _, init::Tuple{T,T,T}) where {T}
+  fill!(data.a, init[1])
+  fill!(data.b, init[2])
+  fill!(data.c, init[3])
+end
+
+function copy!(data::CuData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, c::AbstractArray{T})
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    @inbounds c[i] = a[i]
+    return
+  end
+  @cuda blocks = data.size ÷ TBSize threads = TBSize kernel(data.a, data.c)
+  CUDA.synchronize()
+end
+
+function mul!(data::CuData{T}, _) where {T}
+  function kernel(b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    @inbounds b[i] = scalar * c[i]
+    return
+  end
+  @cuda blocks = data.size ÷ TBSize threads = TBSize kernel(data.b, data.c, data.scalar)
+  CUDA.synchronize()
+end
+
+function add!(data::CuData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T})
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    @inbounds c[i] = a[i] + b[i]
+    return
+  end
+  @cuda blocks = data.size ÷ TBSize threads = TBSize kernel(data.a, data.b, data.c)
+  CUDA.synchronize()
+end
+
+function triad!(data::CuData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    @inbounds a[i] = b[i] + (scalar * c[i])
+    return
+  end
+  @cuda blocks = data.size ÷ TBSize threads = TBSize kernel(
+    data.a,
+    data.b,
+    data.c,
+    data.scalar,
+  )
+  CUDA.synchronize()
+end
+
+function nstream!(data::CuData{T}, _) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    @inbounds a[i] += b[i] + scalar * c[i]
+    return
+  end
+  @cuda blocks = data.size ÷ TBSize threads = TBSize kernel(
+    data.a,
+    data.b,
+    data.c,
+    data.scalar,
+  )
+  CUDA.synchronize()
+end
+
+function dot(data::CuData{T}, _) where {T}
+  # direct port of the reduction in CUDAStream.cu 
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, size::Int, partial::AbstractArray{T})
+    tb_sum = @cuStaticSharedMem(T, TBSize)
+    local_i = threadIdx().x
+    @inbounds tb_sum[local_i] = 0.0
+
+    # do dot first
+    i = (blockIdx().x - 1) * blockDim().x + threadIdx().x
+    while i <= size
+      @inbounds tb_sum[local_i] += a[i] * b[i]
+      i += blockDim().x * gridDim().x
+    end
+
+    # then tree reduction
+    offset = blockDim().x ÷ 2
+    while offset > 0
+      sync_threads()
+      if (local_i - 1) < offset
+        @inbounds tb_sum[local_i] += tb_sum[local_i+offset]
+      end
+      offset ÷= 2
+    end
+
+    if (local_i == 1)
+      @inbounds partial[blockIdx().x] = tb_sum[local_i]
+    end
+
+    return
+  end
+  partial_sum = CuArray{T}(undef, DotBlocks)
+  @cuda blocks = DotBlocks threads = TBSize kernel(data.a, data.b, data.size, partial_sum)
+  return sum(partial_sum)
+end
+
+function read_data(data::CuData{T}, _)::VectorData{T} where {T}
+  return VectorData{T}(data.a, data.b, data.c, data.scalar, data.size)
+end
+
+main()

JuliaStream.jl/src/DistributedStream.jl

@@ -0,0 +1,85 @@
+using Distributed
+
+@everywhere using Pkg
+@everywhere Pkg.activate("."; io = devnull) # don't spam `Activating environment at...`
+@everywhere include("StreamData.jl")
+@everywhere include("Stream.jl")
+@everywhere using SharedArrays
+@everywhere const SharedArrayData = StreamData{T,SharedArray{T}} where {T}
+
+function devices()::Vector{DeviceWithRepr}
+  return [(undef, "CPU (localhost) $(nworkers())P", "Distributed.jl")]
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  _::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{SharedArrayData{T},Nothing} where {T}
+
+  if !silent
+    println("Using max $(nworkers()) process(es) + 1 master")
+  end
+  return (
+    SharedArrayData{T}(
+      SharedArray{T}(arraysize),
+      SharedArray{T}(arraysize),
+      SharedArray{T}(arraysize),
+      scalar,
+      arraysize,
+    ),
+    nothing,
+  )
+end
+
+function init_arrays!(data::SharedArrayData{T}, _, init::Tuple{T,T,T}) where {T}
+
+  @sync @distributed for i = 1:data.size
+    @inbounds data.a[i] = init[1]
+    @inbounds data.b[i] = init[2]
+    @inbounds data.c[i] = init[3]
+  end
+end
+
+function copy!(data::SharedArrayData{T}, _) where {T}
+  @sync @distributed for i = 1:data.size
+    @inbounds data.c[i] = data.a[i]
+  end
+end
+
+function mul!(data::SharedArrayData{T}, _) where {T}
+  @sync @distributed for i = 1:data.size
+    @inbounds data.b[i] = data.scalar * data.c[i]
+  end
+end
+
+function add!(data::SharedArrayData{T}, _) where {T}
+  @sync @distributed for i = 1:data.size
+    @inbounds data.c[i] = data.a[i] + data.b[i]
+  end
+end
+
+function triad!(data::SharedArrayData{T}, _) where {T}
+  @sync @distributed for i = 1:data.size
+    @inbounds data.a[i] = data.b[i] + (data.scalar * data.c[i])
+  end
+end
+
+function nstream!(data::SharedArrayData{T}, _) where {T}
+  @sync @distributed for i = 1:data.size
+    @inbounds data.a[i] += data.b[i] + data.scalar * data.c[i]
+  end
+end
+
+function dot(data::SharedArrayData{T}, _) where {T}
+  return @distributed (+) for i = 1:data.size
+    @inbounds data.a[i] * data.b[i]
+  end
+end
+
+function read_data(data::SharedArrayData{T}, _)::VectorData{T} where {T}
+  return VectorData{T}(data.a, data.b, data.c, data.scalar, data.size)
+end
+
+main()

JuliaStream.jl/src/JuliaStream.jl

@@ -0,0 +1,4 @@
+module JuliaStream
+end
+
+println("Please run benchmarks directly via `julia --project src/<IMPL>Stream.jl`")

JuliaStream.jl/src/KernelAbstractionsStream.jl

@@ -0,0 +1,255 @@
+using ROCKernels, CUDAKernels, KernelAbstractions, CUDA, AMDGPU
+include("Stream.jl")
+
+const CuData = StreamData{T,CUDA.CuArray{T}} where {T}
+const ROCData = StreamData{T,AMDGPU.ROCArray{T}} where {T}
+
+const TBSize = 1024::Int
+const DotBlocks = 256::Int
+
+@enum Backend cuda rocm cpu
+
+struct Context
+  backend::Backend
+  device::Device
+end
+
+function list_rocm_devices()::Vector{DeviceWithRepr}
+  try
+    # AMDGPU.agents()'s internal iteration order isn't stable
+    sorted = sort(AMDGPU.get_agents(:gpu), by = repr)
+    map(x -> (x, repr(x), rocm), sorted)
+  catch
+    # probably unsupported
+    []
+  end
+end
+
+function list_cuda_devices()::Vector{DeviceWithRepr}
+  return !CUDA.functional(false) ? String[] :
+         map(d -> (d, "$(CUDA.name(d)) ($(repr(d)))", cuda), CUDA.devices())
+end
+
+function devices()::Vector{DeviceWithRepr}
+  cudas = list_cuda_devices()
+  rocms = list_rocm_devices()
+  cpus = [(undef, "$(Sys.cpu_info()[1].model) ($(Threads.nthreads())T)", cpu)]
+  vcat(cpus, cudas, rocms)
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  device::DeviceWithRepr,
+  silent::Bool,
+) where {T}
+
+  if arraysize % TBSize != 0
+    error("arraysize ($(arraysize)) must be divisible by $(TBSize)!")
+  end
+
+  (selected, _, backend) = device
+  if backend == cpu
+    if !silent
+      println("Using CPU with max $(Threads.nthreads()) threads")
+    end
+    partialsum = Vector{T}(undef, DotBlocks)
+    data = VectorData{T}(
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    )
+    backenddevice = CPU()
+  elseif backend == cuda
+    CUDA.device!(selected)
+    if CUDA.device() != selected
+      error("Cannot select CUDA device, expecting $selected, but got $(CUDA.device())")
+    end
+    if !CUDA.functional(true)
+      error("Non-functional CUDA configuration")
+    end
+    if !silent
+      println("Using CUDA device: $(CUDA.name(selected)) ($(repr(selected)))")
+    end
+    partialsum = CuArray{T}(undef, DotBlocks)
+    data = CuData{T}(
+      CuArray{T}(undef, arraysize),
+      CuArray{T}(undef, arraysize),
+      CuArray{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    )
+    backenddevice = CUDADevice()
+  elseif backend == rocm
+    AMDGPU.DEFAULT_AGENT[] = selected
+    if AMDGPU.get_default_agent() != selected
+      error(
+        "Cannot select HSA device, expecting $selected, but got $(AMDGPU.get_default_agent())",
+      )
+    end
+    if !silent
+      println("Using GPU HSA device: $(AMDGPU.get_name(selected)) ($(repr(selected)))")
+    end
+    partialsum = ROCArray{T}(undef, DotBlocks)
+    data = ROCData{T}(
+      ROCArray{T}(undef, arraysize),
+      ROCArray{T}(undef, arraysize),
+      ROCArray{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    )
+    backenddevice = ROCDevice()
+  else
+    error("unsupported backend $(backend)")
+  end
+
+  if !silent
+    println("Kernel parameters   : <<<$(data.size),$(TBSize)>>>")
+  end
+  return (data, Context(backend, backenddevice))
+end
+
+function init_arrays!(
+  data::StreamData{T,C},
+  context::Context,
+  init::Tuple{T,T,T},
+) where {T,C}
+  if context.backend == cpu
+    Threads.@threads for i = 1:data.size
+      @inbounds data.a[i] = init[1]
+      @inbounds data.b[i] = init[2]
+      @inbounds data.c[i] = init[3]
+    end
+  elseif context.backend == cuda
+    CUDA.fill!(data.a, init[1])
+    CUDA.fill!(data.b, init[2])
+    CUDA.fill!(data.c, init[3])
+  elseif context.backend == rocm
+    AMDGPU.fill!(data.a, init[1])
+    AMDGPU.fill!(data.b, init[2])
+    AMDGPU.fill!(data.c, init[3])
+  else
+    error("unsupported backend $(backend)")
+  end
+end
+
+function copy!(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(@Const(a::AbstractArray{T}), c)
+    i = @index(Global)
+    @inbounds c[i] = a[i]
+  end
+  wait(kernel(context.device, TBSize)(data.a, data.c, ndrange = data.size))
+end
+
+function mul!(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(b::AbstractArray{T}, @Const(c::AbstractArray{T}), scalar::T)
+    i = @index(Global)
+    @inbounds b[i] = scalar * c[i]
+  end
+  wait(kernel(context.device, TBSize)(data.b, data.c, data.scalar, ndrange = data.size))
+end
+
+function add!(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(@Const(a::AbstractArray{T}), @Const(b::AbstractArray{T}), c)
+    i = @index(Global)
+    @inbounds c[i] = a[i] + b[i]
+  end
+  wait(kernel(context.device, TBSize)(data.a, data.b, data.c, ndrange = data.size))
+end
+
+function triad!(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(a::AbstractArray{T}, @Const(b::AbstractArray{T}), @Const(c), scalar::T)
+    i = @index(Global)
+    @inbounds a[i] = b[i] + (scalar * c[i])
+  end
+  wait(
+    kernel(context.device, TBSize)(
+      data.a,
+      data.b,
+      data.c,
+      data.scalar,
+      ndrange = data.size,
+    ),
+  )
+end
+
+function nstream!(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(a::AbstractArray{T}, @Const(b::AbstractArray{T}), @Const(c), scalar::T)
+    i = @index(Global)
+    @inbounds a[i] += b[i] + scalar * c[i]
+  end
+  wait(
+    kernel(context.device, TBSize)(
+      data.a,
+      data.b,
+      data.c,
+      data.scalar,
+      ndrange = data.size,
+    ),
+  )
+end
+
+function dot(data::StreamData{T,C}, context::Context) where {T,C}
+  @kernel function kernel(@Const(a::AbstractArray{T}), @Const(b::AbstractArray{T}), size::Int, partial::AbstractArray{T})
+    local_i = @index(Local)
+    group_i = @index(Group)
+    tb_sum = @localmem T TBSize
+    @inbounds tb_sum[local_i] = 0.0
+
+    # do dot first
+    i = @index(Global)
+    while i <= size
+      @inbounds tb_sum[local_i] += a[i] * b[i]
+      i += TBSize * DotBlocks
+    end
+
+    # then tree reduction
+    # FIXME this does not compile when targeting CPUs:
+    # see https://github.com/JuliaGPU/KernelAbstractions.jl/issues/262
+    offset = @private Int64 (1,)
+    @inbounds begin
+      offset[1] = @groupsize()[1] ÷ 2
+      while offset[1] > 0
+        @synchronize
+        if (local_i - 1) < offset[1]
+          tb_sum[local_i] += tb_sum[local_i+offset[1]]
+        end
+        offset[1] ÷= 2
+      end
+    end
+
+    if (local_i == 1)
+      @inbounds partial[group_i] = tb_sum[local_i]
+    end
+  end
+
+  if context.backend == cpu
+    partial_sum = Vector{T}(undef, DotBlocks)
+  elseif context.backend == cuda
+    partial_sum = CuArray{T}(undef, DotBlocks)
+  elseif context.backend == rocm
+    partial_sum = ROCArray{T}(undef, DotBlocks)
+  else
+    error("unsupported backend $(backend)")
+  end
+
+  wait(
+    kernel(context.device, TBSize)(
+      data.a,
+      data.b,
+      data.size,
+      partial_sum,
+      ndrange = TBSize * DotBlocks,
+    ),
+  )
+
+  return sum(partial_sum)
+end
+
+function read_data(data::StreamData{T,C}, _::Context)::VectorData{T} where {T,C}
+  return VectorData{T}(data.a, data.b, data.c, data.scalar, data.size)
+end
+
+main()

JuliaStream.jl/src/PlainStream.jl

@@ -0,0 +1,75 @@
+include("Stream.jl")
+
+function devices()::Vector{DeviceWithRepr}
+  return [(undef, "CPU", "Palin")]
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  _::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{VectorData{T},Nothing} where {T}
+  return (
+    VectorData{T}(
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    ),
+    nothing
+  )
+end
+
+function init_arrays!(data::VectorData{T}, _, init::Tuple{T,T,T}) where {T}
+  for i = 1:data.size
+    @inbounds data.a[i] = init[1]
+    @inbounds data.b[i] = init[2]
+    @inbounds data.c[i] = init[3]
+  end
+end
+
+function copy!(data::VectorData{T}, _) where {T}
+  for i = 1:data.size
+    @inbounds data.c[i] = data.a[i]
+  end
+end
+
+function mul!(data::VectorData{T}, _) where {T}
+  for i = 1:data.size
+    @inbounds data.b[i] = data.scalar * data.c[i]
+  end
+end
+
+function add!(data::VectorData{T}, _) where {T}
+  for i = 1:data.size
+    @inbounds data.c[i] = data.a[i] + data.b[i]
+  end
+end
+
+function triad!(data::VectorData{T}, _) where {T}
+  for i = 1:data.size
+    @inbounds data.a[i] = data.b[i] + (data.scalar * data.c[i])
+  end
+end
+
+function nstream!(data::VectorData{T}, _) where {T}
+  for i = 1:data.size
+    @inbounds data.a[i] += data.b[i] + data.scalar * data.c[i]
+  end
+end
+
+function dot(data::VectorData{T}, _) where {T}
+  sum = zero(T)
+  for i = 1:data.size
+    @inbounds sum += data.a[i] * data.b[i]
+  end
+  return sum
+end
+
+function read_data(data::VectorData{T}, _)::VectorData{T} where {T}
+  return data
+end
+
+main()

JuliaStream.jl/src/Stream.jl

@@ -0,0 +1,300 @@
+using ArgParse
+using Parameters
+using Printf
+using Base: Float64, Int
+
+include("StreamData.jl")
+
+const VectorData = StreamData{T,Vector{T}} where {T}
+
+const DeviceWithRepr = Tuple{Any,String,Any}
+
+struct Timings
+  copy::Vector{Float64}
+  mul::Vector{Float64}
+  add::Vector{Float64}
+  triad::Vector{Float64}
+  dot::Vector{Float64}
+  Timings(n) = new(zeros(n), zeros(n), zeros(n), zeros(n), zeros(n))
+end
+
+@enum Benchmark All Triad Nstream
+
+function run_all!(data::StreamData{T,C}, context, times::Int)::Tuple{Timings,T} where {T,C}
+  timings = Timings(times)
+  lastSum::T = 0
+  for i = 1:times
+    @inbounds timings.copy[i] = @elapsed copy!(data, context)
+    @inbounds timings.mul[i] = @elapsed mul!(data, context)
+    @inbounds timings.add[i] = @elapsed add!(data, context)
+    @inbounds timings.triad[i] = @elapsed triad!(data, context)
+    @inbounds timings.dot[i] = @elapsed lastSum = dot(data, context)
+  end
+  return (timings, lastSum)
+end
+
+function run_triad!(data::StreamData{T,C}, context, times::Int)::Float64 where {T,C}
+  return @elapsed for _ = 1:times
+    triad!(data, context)
+  end
+end
+
+function run_nstream!(
+  data::StreamData{T,C},
+  context,
+  times::Int,
+)::Vector{Float64} where {T,C}
+  timings::Vector{Float64} = zeros(times)
+  for i = 1:times
+    @inbounds timings[i] = @elapsed nstream!(data, context)
+  end
+  return timings
+end
+
+function check_solutions(
+  data::StreamData{T,C},
+  times::Int,
+  init::Tuple{T,T,T},
+  benchmark::Benchmark,
+  dot::Union{T,Nothing},
+) where {T,C}
+  (gold_a, gold_b, gold_c) = init
+  for _ = 1:times
+    if benchmark == All
+      gold_c = gold_a
+      gold_b = data.scalar * gold_c
+      gold_c = gold_a + gold_b
+      gold_a = gold_b + data.scalar * gold_c
+    elseif benchmark == Triad
+      gold_a = gold_b + data.scalar * gold_c
+    elseif benchmark == Nstream
+      gold_a += gold_b + data.scalar * gold_c
+    else
+      error("Unknown benchmark", benchmark)
+    end
+  end
+
+  tolerance = eps(T) * 100
+  function validate_xs(name::String, xs::AbstractArray{T}, from::T)
+    error = (map(x -> abs(x - from), xs) |> sum) / length(xs)
+    failed = error > tolerance
+    if failed
+      println("Validation failed on $name. Average error $error")
+    end
+    !failed
+  end
+  a_valid = validate_xs("a", data.a, gold_a)
+  b_valid = validate_xs("b", data.b, gold_b)
+  c_valid = validate_xs("c", data.c, gold_c)
+  dot_valid =
+    dot !== nothing ?
+    begin
+      gold_sum = gold_a * gold_b * data.size
+      error = abs((dot - gold_sum) / gold_sum)
+      failed = error > 1.0e-8
+      if failed
+        println(
+          "Validation failed on sum. Error $error \nSum was $dot but should be $gold_sum",
+        )
+      end
+      !failed
+    end : true
+
+  a_valid && b_valid && c_valid && dot_valid
+end
+
+@with_kw mutable struct Config
+  list::Bool = false
+  device::Int = 1
+  numtimes::Int = 100
+  arraysize::Int = 33554432
+  float::Bool = false
+  triad_only::Bool = false
+  nstream_only::Bool = false
+  csv::Bool = false
+  mibibytes::Bool = false
+end
+
+function parse_options(given::Config)
+  s = ArgParseSettings()
+  @add_arg_table s begin
+    "--list"
+    help = "List available devices"
+    action = :store_true
+    "--device", "-d"
+    help = "Select device at DEVICE, NOTE: Julia is 1-indexed"
+    arg_type = Int
+    default = given.device
+    "--numtimes", "-n"
+    help = "Run the test NUMTIMES times (NUM >= 2)"
+    arg_type = Int
+    default = given.numtimes
+    "--arraysize", "-s"
+    help = "Use ARRAYSIZE elements in the array"
+    arg_type = Int
+    default = given.arraysize
+    "--float"
+    help = "Use floats (rather than doubles)"
+    action = :store_true
+    "--triad_only"
+    help = "Only run triad"
+    action = :store_true
+    "--nstream_only"
+    help = "Only run nstream"
+    action = :store_true
+    "--csv"
+    help = "Output as csv table"
+    action = :store_true
+    "--mibibytes"
+    help = "Use MiB=2^20 for bandwidth calculation (default MB=10^6)"
+    action = :store_true
+  end
+  args = parse_args(s)
+  # surely there's a better way than doing this:
+  for (arg, val) in args
+    setproperty!(given, Symbol(arg), val)
+  end
+end
+
+const DefaultInit = (0.1, 0.2, 0.0)
+const DefaultScalar = 0.4
+const Version = "3.4.0"
+
+function main()
+
+  config::Config = Config()
+  parse_options(config)
+
+  if config.list
+    for (i, (_,repr, impl)) in enumerate(devices())
+      println("[$i] ($impl) $repr")
+    end
+    exit(0)
+  end
+
+  ds = devices()
+  # TODO implement substring device match
+  if config.device < 1 || config.device > length(ds)
+    error(
+      "Device $(config.device) out of range (1..$(length(ds))), NOTE: Julia is 1-indexed",
+    )
+  else
+    device = ds[config.device]
+  end
+
+  type = config.float ? Float32 : Float64
+
+  if config.nstream_only && !config.triad_only
+    benchmark = Nstream
+  elseif !config.nstream_only && config.triad_only
+    benchmark = Triad
+  elseif !config.nstream_only && !config.triad_only
+    benchmark = All
+  elseif config.nstream_only && config.triad_only
+    error("Both triad and nstream are enabled, pick one or omit both to run all benchmarks")
+  else
+    error("Invalid config: $(repr(config))")
+  end
+
+  array_bytes = config.arraysize * sizeof(type)
+  total_bytes = array_bytes * 3
+  (mega_scale, mega_suffix, giga_scale, giga_suffix) =
+    !config.mibibytes ? (1.0e-6, "MB", 1.0e-9, "GB") : (2^-20, "MiB", 2^-30, "GiB")
+
+  if !config.csv
+    println("""BabelStream
+               Version: $Version
+               Implementation: Julia; $(PROGRAM_FILE)""")
+    println("Running kernels $(config.numtimes) times")
+    if benchmark == Triad
+      println("Number of elements: $(config.arraysize)")
+    end
+    println("Precision: $(config.float ?  "float" :   "double")")
+    r1 = n -> round(n; digits = 1)
+    println(
+      "Array size: $(r1(mega_scale * array_bytes)) $mega_suffix(=$(r1(giga_scale * array_bytes)) $giga_suffix)",
+    )
+    println(
+      "Total size: $(r1(mega_scale * total_bytes)) $mega_suffix(=$(r1(giga_scale * total_bytes)) $giga_suffix)",
+    )
+  end
+
+  function mk_row(xs::Vector{Float64}, name::String, total_bytes::Int)
+    tail = Base.rest(xs)
+    min = Iterators.minimum(tail)
+    max = Iterators.maximum(tail)
+    avg = Iterators.sum(tail) / Iterators.length(tail)
+    mbps = mega_scale * total_bytes / min
+    if config.csv
+      return [
+        ("function", name),
+        ("num_times", config.numtimes),
+        ("n_elements", config.arraysize),
+        ("sizeof", total_bytes),
+        ("max_m$(  config.mibibytes ? "i" : "")bytes_per_sec", mbps),
+        ("min_runtime", min),
+        ("max_runtime", max),
+        ("avg_runtime", avg),
+      ]
+    else
+      return [
+        ("Function", name),
+        ("M$(config.mibibytes ? "i" : "")Bytes/sec", round(mbps; digits = 3)),
+        ("Min (sec)", round(min; digits = 5)),
+        ("Max", round(max; digits = 5)),
+        ("Average", round(avg; digits = 5)),
+      ]
+    end
+  end
+
+  function tabulate(rows::Vector{Tuple{String,Any}}...)
+    header = Base.first(rows)
+    padding = config.csv ? 0 : 12
+    sep = config.csv ? "," : ""
+    map(x -> rpad(x[1], padding), header) |> x -> join(x, sep) |> println
+    for row in rows
+      map(x -> rpad(x[2], padding), row) |> x -> join(x, sep) |> println
+    end
+  end
+
+  init::Tuple{type,type,type} = DefaultInit
+  scalar::type = DefaultScalar
+
+  GC.enable(false)
+
+  (data, context) = make_stream(config.arraysize, scalar, device, config.csv)
+  init_arrays!(data, context, init)
+  if benchmark == All
+    (timings, sum) = run_all!(data, context, config.numtimes)
+    valid = check_solutions(read_data(data, context), config.numtimes, init, benchmark, sum)
+    tabulate(
+      mk_row(timings.copy, "Copy", 2 * array_bytes),
+      mk_row(timings.mul, "Mul", 2 * array_bytes),
+      mk_row(timings.add, "Add", 3 * array_bytes),
+      mk_row(timings.triad, "Triad", 3 * array_bytes),
+      mk_row(timings.dot, "Dot", 2 * array_bytes),
+    )
+  elseif benchmark == Nstream
+    timings = run_nstream!(data, context, config.numtimes)
+    valid =
+      check_solutions(read_data(data, context), config.numtimes, init, benchmark, nothing)
+    tabulate(mk_row(timings, "Nstream", 4 * array_bytes))
+  elseif benchmark == Triad
+    elapsed = run_triad!(data, context, config.numtimes)
+    valid =
+      check_solutions(read_data(data, context), config.numtimes, init, benchmark, nothing)
+    total_bytes = 3 * array_bytes * config.numtimes
+    bandwidth = mega_scale * (total_bytes / elapsed)
+    println("Runtime (seconds): $(round(elapsed; digits=5))")
+    println("Bandwidth ($giga_suffix/s): $(round(bandwidth; digits=3)) ")
+  else
+    error("Bad benchmark $(benchmark)")
+  end
+  
+  GC.enable(true)
+
+  if !valid
+    exit(1)
+  end
+
+end

JuliaStream.jl/src/StreamData.jl

@@ -0,0 +1,7 @@
+struct StreamData{T,C<:AbstractArray{T}}
+  a::C
+  b::C
+  c::C
+  scalar::T
+  size::Int
+end

JuliaStream.jl/src/ThreadedStream.jl

@@ -0,0 +1,112 @@
+include("Stream.jl")
+
+function devices()::Vector{DeviceWithRepr}
+  return [(undef, "$(Sys.cpu_info()[1].model) ($(Threads.nthreads())T)", "Threaded")]
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  _::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{VectorData{T},Nothing} where {T}
+  if !silent
+    println("Using max $(Threads.nthreads()) threads")
+  end
+  return (
+    VectorData{T}(
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      Vector{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    ),
+    nothing
+  )
+end
+
+function init_arrays!(data::VectorData{T}, _, init::Tuple{T,T,T}) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.a[i] = init[1]
+    @inbounds data.b[i] = init[2]
+    @inbounds data.c[i] = init[3]
+  end
+end
+
+function copy!(data::VectorData{T}, _) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.c[i] = data.a[i]
+  end
+end
+
+function mul!(data::VectorData{T}, _) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.b[i] = data.scalar * data.c[i]
+  end
+end
+
+function add!(data::VectorData{T}, _) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.c[i] = data.a[i] + data.b[i]
+  end
+end
+
+function triad!(data::VectorData{T}, _) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.a[i] = data.b[i] + (data.scalar * data.c[i])
+  end
+end
+
+function nstream!(data::VectorData{T}, _) where {T}
+  Threads.@threads for i = 1:data.size
+    @inbounds data.a[i] += data.b[i] + data.scalar * data.c[i]
+  end
+end
+
+# Threads.@threads/Threads.@spawn doesn't support OpenMP's firstprivate, etc
+function static_par_ranged(f::Function, range::Int, n::Int)
+  stride = range ÷ n
+  rem = range % n
+  strides = map(0:n) do i
+    width = stride + (i < rem ? 1 : 0)
+    offset = i < rem ? (stride + 1) * i : ((stride + 1) * rem) + (stride * (i - rem))
+    (offset, width)
+  end
+  ccall(:jl_enter_threaded_region, Cvoid, ())
+  try
+    foreach(wait, map(1:n) do group
+      (offset, size) = strides[group]
+      task = Task(() -> f(group, offset+1, offset+size))
+      task.sticky = true
+      ccall(:jl_set_task_tid, Cvoid, (Any, Cint), task, group-1) # ccall, so 0-based for group
+      schedule(task)
+    end)
+  finally
+    ccall(:jl_exit_threaded_region, Cvoid, ())
+  end
+end
+
+function dot(data::VectorData{T}, _) where {T}
+  partial = Vector{T}(undef,  Threads.nthreads())
+  static_par_ranged(data.size, Threads.nthreads()) do group, startidx, endidx
+    acc = zero(T)
+    @simd for i = startidx:endidx
+      @inbounds acc += data.a[i] * data.b[i]
+    end
+    @inbounds partial[group] = acc
+  end
+  return sum(partial)
+  # This doesn't do well on aarch64 because of the excessive Threads.threadid() ccall 
+  # and inhibited vectorisation from the lack of @simd 
+  # partial = zeros(T, Threads.nthreads())
+  # Threads.@threads for i = 1:data.size
+  #   @inbounds partial[Threads.threadid()] += (data.a[i] * data.b[i])
+  # end
+  # return sum(partial)
+end
+
+function read_data(data::VectorData{T}, _)::VectorData{T} where {T}
+  return data
+end
+
+main()

JuliaStream.jl/src/oneAPIStream.jl

@@ -0,0 +1,170 @@
+using Base.Iterators: println
+using Base.Iterators: println
+using Printf: Iterators
+
+include("Stream.jl")
+using oneAPI
+
+const oneData = StreamData{T,oneArray{T}} where {T}
+const DotWGSize = 256::Int
+
+function devices()::Vector{DeviceWithRepr}
+  all = map(oneL0.devices, oneL0.drivers()) |> Iterators.flatten |> Iterators.collect
+  map(dev -> (dev, repr("text/plain", dev), "oneAPi.jl"), all)
+end
+
+function make_stream(
+  arraysize::Int,
+  scalar::T,
+  device::DeviceWithRepr,
+  silent::Bool,
+)::Tuple{oneData{T},Int} where {T}
+
+  oneAPI.allowscalar(false)
+  oneAPI.device!(device[1])
+
+  props = oneL0.compute_properties(oneAPI.device())
+  groupsize = min(props.maxTotalGroupSize, arraysize)
+
+  if arraysize % groupsize != 0
+    error("arraysize ($(arraysize)) must be divisible by $(groupsize)!")
+  end
+
+  if !silent
+    println("Using L0 device: $(repr("text/plain",device[1]))")
+    println("Kernel parameters   : <<<$(arraysize),$(groupsize)>>>")
+  end
+  return (
+    oneData{T}(
+      oneArray{T}(undef, arraysize),
+      oneArray{T}(undef, arraysize),
+      oneArray{T}(undef, arraysize),
+      scalar,
+      arraysize,
+    ),
+    groupsize,
+  )
+end
+
+function init_arrays!(data::oneData{T}, _, init::Tuple{T,T,T}) where {T}
+  oneAPI.fill!(data.a, init[1])
+  oneAPI.fill!(data.b, init[2])
+  oneAPI.fill!(data.c, init[3])
+end
+
+function copy!(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(a::AbstractArray{T}, c::AbstractArray{T})
+    i = get_global_id()
+    @inbounds c[i] = a[i]
+    return
+  end
+  @oneapi items = groupsize groups = data.size ÷ groupsize kernel( # 
+    data.a,
+    data.c,
+  )
+  oneAPI.synchronize()
+end
+
+function mul!(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = get_global_id()
+    @inbounds b[i] = scalar * c[i]
+    return
+  end
+  @oneapi items = groupsize groups = data.size ÷ groupsize kernel( #
+    data.b,
+    data.c,
+    data.scalar,
+  )
+  oneAPI.synchronize()
+end
+
+function add!(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T})
+    i = get_global_id()
+    @inbounds c[i] = a[i] + b[i]
+    return
+  end
+  @oneapi items = groupsize groups = data.size ÷ groupsize kernel( #
+    data.a,
+    data.b,
+    data.c,
+  )
+  oneAPI.synchronize()
+end
+
+function triad!(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = get_global_id()
+    @inbounds a[i] = b[i] + (scalar * c[i])
+    return
+  end
+  @oneapi items = groupsize groups = data.size ÷ groupsize kernel( #
+    data.a,
+    data.b,
+    data.c,
+    data.scalar,
+  )
+  oneAPI.synchronize()
+end
+
+function nstream!(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, c::AbstractArray{T}, scalar::T)
+    i = get_global_id()
+    @inbounds a[i] += b[i] + scalar * c[i]
+    return
+  end
+  @oneapi items = groupsize groups = data.size ÷ groupsize kernel( #
+    data.a,
+    data.b,
+    data.c,
+    data.scalar,
+  )
+  oneAPI.synchronize()
+end
+
+function dot(data::oneData{T}, groupsize::Int) where {T}
+  function kernel(a::AbstractArray{T}, b::AbstractArray{T}, size::Int, partial::AbstractArray{T})
+    wg_sum = @LocalMemory(T, (DotWGSize,))
+    li = get_local_id()
+    @inbounds wg_sum[li] = 0.0
+
+    # do dot first
+    i = get_global_id()
+    while i <= size
+      @inbounds wg_sum[li] += a[i] * b[i]
+      i += get_global_size()
+    end
+
+    # then tree reduction
+    offset = get_local_size() ÷ 2
+    while offset > 0
+      barrier()
+      if li <= offset
+        @inbounds wg_sum[li] += wg_sum[li+offset]
+      end
+      offset ÷= 2
+    end
+
+    if li == 1
+      @inbounds partial[get_group_id()] =  wg_sum[li]
+    end
+
+    return
+  end
+  partial_sum = oneArray{T}(undef, groupsize)
+  @oneapi items = groupsize groups = DotWGSize kernel(
+    data.a,
+    data.b,
+    data.size,
+    partial_sum,
+  )
+  oneAPI.synchronize()
+  return sum(partial_sum)
+end
+
+function read_data(data::oneData{T}, _)::VectorData{T} where {T}
+  return VectorData{T}(data.a, data.b, data.c, data.scalar, data.size)
+end
+
+main()

JuliaStream.jl/update_all.sh

@@ -0,0 +1,7 @@
+#!/bin/bash
+# shellcheck disable=SC2034 disable=SC2153
+
+for BACKEND in  "." "AMDGPU" "CUDA" "oneAPI" "Threaded" "KernelAbstractions" 
+do
+  julia --project="$BACKEND" -e 'import Pkg; Pkg.resolve(); Pkg.instantiate(); Pkg.update(); Pkg.gc();' 
+done

README.md

@@ -23,6 +23,8 @@ Currently implemented are:
 
 This code was previously called GPU-STREAM.
 
+This project also contains implementations in alternative languages with different build systems:
+* Scala - [scala-stream](./scala-stream)
 
 How is this different to STREAM?
 --------------------------------
@@ -159,10 +161,12 @@ Citing
 
 Please cite BabelStream via this reference:
 
-> Deakin T, Price J, Martineau M, McIntosh-Smith S. GPU-STREAM v2.0: Benchmarking the achievable memory bandwidth of many-core processors across diverse parallel programming models. 2016. Paper presented at P^3MA Workshop at ISC High Performance, Frankfurt, Germany.
+> Deakin T, Price J, Martineau M, McIntosh-Smith S. GPU-STREAM v2.0: Benchmarking the achievable memory bandwidth of many-core processors across diverse parallel programming models. 2016. Paper presented at P^3MA Workshop at ISC High Performance, Frankfurt, Germany. DOI: 10.1007/978- 3-319-46079-6_34
 
 **Other BabelStream publications:**
 
+> Deakin T, Price J, Martineau M, McIntosh-Smith S. Evaluating attainable memory bandwidth of parallel programming models via BabelStream. International Journal of Computational Science and Engineering. Special issue. Vol. 17, No. 3, pp. 247–262. 2018.DOI: 10.1504/IJCSE.2018.095847
+
 > Deakin T, McIntosh-Smith S. GPU-STREAM: Benchmarking the achievable memory bandwidth of Graphics Processing Units. 2015. Poster session presented at IEEE/ACM SuperComputing, Austin, United States.
 You can view the [Poster and Extended Abstract](http://sc15.supercomputing.org/sites/all/themes/SC15images/tech_poster/tech_poster_pages/post150.html).
 
@@ -171,8 +175,6 @@ You can view the [Poster and Extended Abstract](http://sc16.supercomputing.org/s
 
 > Raman K, Deakin T, Price J, McIntosh-Smith S. Improving achieved memory bandwidth from C++ codes on Intel Xeon Phi Processor (Knights Landing). IXPUG Spring Meeting, Cambridge, UK, 2017.
 
-> Deakin T, Price J, Martineau M, McIntosh-Smith S. Evaluating attainable memory bandwidth of parallel programming models via BabelStream. International Journal of Computational Science and Engineering. Special issue (in press). 2017.
-
 > Deakin T, Price J, McIntosh-Smith S. Portable methods for measuring cache hierarchy performance. 2017. Poster sessions presented at IEEE/ACM SuperComputing, Denver, United States.
 You can view the [Poster and Extended Abstract](http://sc17.supercomputing.org/SC17%20Archive/tech_poster/tech_poster_pages/post155.html)
 

scala-stream/.bsp/sbt.json

@@ -0,0 +1 @@
+{"name":"sbt","version":"1.5.2","bspVersion":"2.0.0-M5","languages":["scala"],"argv":["/usr/lib/jvm/java-11-openjdk-11.0.11.0.9-2.fc33.x86_64/bin/java","-Xms100m","-Xmx100m","-classpath","/home/tom/.local/share/JetBrains/Toolbox/apps/IDEA-U/ch-0/211.7142.45.plugins/Scala/launcher/sbt-launch.jar","xsbt.boot.Boot","-bsp","--sbt-launch-jar=/home/tom/.local/share/JetBrains/Toolbox/apps/IDEA-U/ch-0/211.7142.45.plugins/Scala/launcher/sbt-launch.jar"]}

scala-stream/.gitignore

@@ -0,0 +1 @@
+target/

scala-stream/.jvmopts

@@ -0,0 +1,2 @@
+-Xmx4096m
+-Xss4m

scala-stream/.scalafmt.conf

@@ -0,0 +1,34 @@
+version = "3.0.0-RC2"
+runner.dialect = scala3
+
+style = defaultWithAlign
+
+maxColumn = 100
+
+align.preset = more
+
+rewrite.rules = [
+  AvoidInfix
+  RedundantBraces
+  RedundantParens
+  AsciiSortImports
+  PreferCurlyFors
+]
+
+rewrite.neverInfix.excludeFilters = [until
+  to
+  by
+  eq
+  ne
+  "should.*"
+  "contain.*"
+  "must.*"
+  in
+  be
+  taggedAs
+  thrownBy
+  synchronized
+  have
+  when
+  size
+  theSameElementsAs]

scala-stream/README.md

@@ -0,0 +1,102 @@
+ScalaStream
+===========
+
+This is an implementation of BabelStream
+in [Scala 3](https://docs.scala-lang.org/scala3/new-in-scala3.html) on the JVM. In theory, this
+implementation also covers Java. Scala and Java, like any other programming language, has its own
+ecosystem of library supported parallel programming frameworks, we currently implement the
+following:
+
+* Parallel streams (introduced in Java 8) - `src/main/scala/scalastream/J8SStream.scala`
+* [Scala Parallel Collections](https://github.com/scala/scala-parallel-collections)
+    - `src/main/scala/scalastream/ParStream.scala`
+
+As the benchmark is relatively simple, we also implement some baselines:
+
+* Single threaded Scala `for` (i.e `foreach` sugar) - `src/main/scala/scalastream/PlainStream.scala`
+* Manually parallelism with Java executors - `src/main/scala/scalastream/ThreadedStream.scala`
+
+### Performance considerations
+
+As Scala 3 defaults to Scala 2.13's standard library, we roll our own `Fractional` typeclass with
+liberal use of inlining and specialisation. This is motivated by 2.13 stdlib's lack of
+specialisation for primitives types on the default `Fractional` and `Numeric` typeclasses.
+
+The use of [Spire](https://github.com/typelevel/spire) to mitigate this was attempted, however, due
+to its use of Scala 2 macros, it currently doesn't compile with Scala 3.
+
+### Build & Run
+
+Prerequisites
+
+* JDK >= 8 on any of its supported platform; known working implementations:
+    - OpenJDK
+      distributions ([Amazon Corretto](https://docs.aws.amazon.com/corretto/latest/corretto-11-ug/downloads-list.html)
+      , [Azul](https://www.azul.com/downloads/?version=java-11-lts&package=jdk)
+      , [AdoptOpenJDK](https://adoptopenjdk.net/), etc)
+    - Oracle Graal CE/EE 8+
+
+To run the benchmark, first create a binary:
+
+```shell
+> ./sbt assembly
+```
+
+The binary will be located at `./target/scala-3.0.0/scala-stream.jar`. Run it with:
+
+```shell
+> java -version 
+openjdk version "11.0.11" 2021-04-20
+OpenJDK Runtime Environment 18.9 (build 11.0.11+9)
+OpenJDK 64-Bit Server VM 18.9 (build 11.0.11+9, mixed mode, sharing)
+> java  -jar target/scala-3.0.0/scala-stream.jar --help
+
+```
+
+For best results, benchmark with the following JVM flags:
+
+```
+-XX:-UseOnStackReplacement     # disable OSR, not useful for this benchmark as we are measuring peak performance  
+-XX:-TieredCompilation         # disable C1, go straight to C2 
+-XX:ReservedCodeCacheSize=512m # don't flush compiled code out of cache at any point 
+```
+
+Worked example:
+
+```shell
+> java -XX:-UseOnStackReplacement -XX:-TieredCompilation -XX:ReservedCodeCacheSize=512m -jar target/scala-3.0.0/scala-stream.jar
+
+BabelStream
+Version: 3.4.0
+Implementation: Scala Parallel Collections; Scala (Java 11.0.11; Red Hat, Inc.; home=/usr/lib/jvm/java-11-openjdk-11.0.11.0.9-2.fc33.x86_64)
+Running kernels 100 times
+Precision: double
+Array size: 268.4 MB (=0.3 GB)
+Total size: 805.3 MB (=0.8 GB)
+Function    MBytes/sec  Min (sec)   Max         Average     
+Copy        4087.077    0.13136     0.24896     0.15480     
+Mul         2934.709    0.18294     0.28706     0.21627     
+Add         3016.342    0.26698     0.39835     0.31119     
+Triad       3016.496    0.26697     0.37612     0.31040     
+Dot         2216.096    0.24226     0.41235     0.28264
+
+```
+
+### Graal Native Image
+
+The port has partial support for Graal Native Image, to generate one, run:
+
+```shell
+> ./sbt nativeImage
+```
+
+The ELF binary will be located at `./target/native-image/scala-stream`, relocation should work on
+the same architecture the binary is built on.
+
+There's an ongoing bug with Scala 3 's use of `lazy val`s where the program crashes at declaration
+site. Currently, Scala Parallel Collections uses this feature internally, so selecting this device
+will crash at runtime.
+
+The bug originates from the use of  `Unsafe` in `lazy val` for thready safety guarantees. It seems
+that Graal only supports limited uses of this JVM implementation detail and Scala 3 happens to be on
+the unsupported side.  

scala-stream/build.sbt

@@ -0,0 +1,29 @@
+lazy val mainCls = Some("scalastream.App")
+
+lazy val root = (project in file("."))
+  .enablePlugins(NativeImagePlugin)
+  .settings(
+    scalaVersion := "3.0.0",
+    version := "3.4.0",
+    organization := "uk.ac.bristol.uob-hpc",
+    organizationName := "University of Bristol",
+    Compile / mainClass := mainCls,
+    assembly / mainClass := mainCls,
+    scalacOptions ~= filterConsoleScalacOptions,
+    assembly / assemblyJarName := "scala-stream.jar",
+    nativeImageOptions := Seq(
+      "--no-fallback",
+      "-H:ReflectionConfigurationFiles=../../reflect-config.json"
+    ),
+    nativeImageVersion := "21.1.0",
+    (Global / excludeLintKeys) += nativeImageVersion,
+    name := "scala-stream",
+    libraryDependencies ++= Seq(
+      // Lazy val implementation in Scala 3 triggers an exception in nativeImage, use 2_13 for arg parsing for now otherwise we can't get to the benchmarking part
+      ("com.github.scopt" %% "scopt" % "4.0.1").cross(CrossVersion.for3Use2_13),
+      // par also uses lazy val at some point, so it doesn't work in nativeImage
+      "org.scala-lang.modules" %% "scala-parallel-collections" % "1.0.3",
+      "net.openhft"             % "affinity"                   % "3.21ea1",
+      "org.slf4j"               % "slf4j-simple"               % "1.7.30" // for affinity
+    )
+  )

scala-stream/project/build.properties

@@ -0,0 +1 @@
+sbt.version=1.5.2

scala-stream/project/plugins.sbt

@@ -0,0 +1,6 @@
+addSbtPlugin("com.timushev.sbt"          % "sbt-updates"      % "0.5.3")
+addSbtPlugin("io.github.davidgregory084" % "sbt-tpolecat"     % "0.1.17")
+addSbtPlugin("org.scalameta"             % "sbt-native-image" % "0.3.0")
+addSbtPlugin("com.eed3si9n"              % "sbt-assembly"     % "0.15.0")
+addSbtPlugin("ch.epfl.scala"             % "sbt-scalafix"     % "0.9.27")
+addSbtPlugin("org.scalameta"             % "sbt-scalafmt"     % "2.4.2")

scala-stream/reflect-config.json

@@ -0,0 +1,11 @@
+[
+  {
+    "name": "sun.misc.Unsafe",
+    "fields": [
+      {
+        "name": "theUnsafe",
+        "allowUnsafeAccess": true
+      }
+    ]
+  }
+]

scala-stream/sbt

@@ -0,0 +1,3 @@
+#!/usr/bin/env bash
+
+./sbt-dist/bin/sbt "$@"

scala-stream/sbt-dist/bin/sbt

@@ -0,0 +1,177 @@
+#!/usr/bin/env bash
+
+
+###  ------------------------------- ###
+###  Helper methods for BASH scripts ###
+###  ------------------------------- ###
+
+realpath () {
+(
+  TARGET_FILE="$1"
+  FIX_CYGPATH="$2"
+
+  cd "$(dirname "$TARGET_FILE")"
+  TARGET_FILE=$(basename "$TARGET_FILE")
+
+  COUNT=0
+  while [ -L "$TARGET_FILE" -a $COUNT -lt 100 ]
+  do
+      TARGET_FILE=$(readlink "$TARGET_FILE")
+      cd "$(dirname "$TARGET_FILE")"
+      TARGET_FILE=$(basename "$TARGET_FILE")
+      COUNT=$(($COUNT + 1))
+  done
+
+  # make sure we grab the actual windows path, instead of cygwin's path.
+  if [[ "x$FIX_CYGPATH" != "x" ]]; then
+    echo "$(cygwinpath "$(pwd -P)/$TARGET_FILE")"
+  else
+    echo "$(pwd -P)/$TARGET_FILE"
+  fi
+)
+}
+
+
+# Uses uname to detect if we're in the odd cygwin environment.
+is_cygwin() {
+  local os=$(uname -s)
+  case "$os" in
+    CYGWIN*) return 0 ;;
+    MINGW*) return 0 ;;
+    MSYS*) return 0 ;;
+    *)  return 1 ;;
+  esac
+}
+
+# TODO - Use nicer bash-isms here.
+CYGWIN_FLAG=$(if is_cygwin; then echo true; else echo false; fi)
+
+
+# This can fix cygwin style /cygdrive paths so we get the
+# windows style paths.
+cygwinpath() {
+  local file="$1"
+  if [[ "$CYGWIN_FLAG" == "true" ]]; then
+    echo $(cygpath -w $file)
+  else
+    echo $file
+  fi
+}
+
+. "$(dirname "$(realpath "$0")")/sbt-launch-lib.bash"
+
+
+declare -r noshare_opts="-Dsbt.global.base=project/.sbtboot -Dsbt.boot.directory=project/.boot -Dsbt.ivy.home=project/.ivy"
+declare -r sbt_opts_file=".sbtopts"
+declare -r etc_sbt_opts_file="/etc/sbt/sbtopts"
+declare -r dist_sbt_opts_file="${sbt_home}/conf/sbtopts"
+declare -r win_sbt_opts_file="${sbt_home}/conf/sbtconfig.txt"
+
+usage() {
+ cat <<EOM
+Usage: `basename "$0"` [options]
+
+  -h | -help         print this message
+  -v | -verbose      this runner is chattier
+  -d | -debug        set sbt log level to debug
+  -no-colors         disable ANSI color codes
+  -sbt-create        start sbt even if current directory contains no sbt project
+  -sbt-dir   <path>  path to global settings/plugins directory (default: ~/.sbt)
+  -sbt-boot  <path>  path to shared boot directory (default: ~/.sbt/boot in 0.11 series)
+  -ivy       <path>  path to local Ivy repository (default: ~/.ivy2)
+  -mem    <integer>  set memory options (default: $sbt_default_mem, which is $(get_mem_opts))
+  -no-share          use all local caches; no sharing
+  -no-global         uses global caches, but does not use global ~/.sbt directory.
+  -jvm-debug <port>  Turn on JVM debugging, open at the given port.
+  -batch             Disable interactive mode
+
+  # sbt version (default: from project/build.properties if present, else latest release)
+  -sbt-version  <version>   use the specified version of sbt
+  -sbt-jar      <path>      use the specified jar as the sbt launcher
+  -sbt-rc                   use an RC version of sbt
+  -sbt-snapshot             use a snapshot version of sbt
+
+  # java version (default: java from PATH, currently $(java -version 2>&1 | grep version))
+  -java-home <path>         alternate JAVA_HOME
+
+  # jvm options and output control
+  JAVA_OPTS          environment variable, if unset uses "$java_opts"
+  .jvmopts           if this file exists in the current directory, its contents
+                     are appended to JAVA_OPTS
+  SBT_OPTS           environment variable, if unset uses "$default_sbt_opts"
+  .sbtopts           if this file exists in the current directory, its contents
+                     are prepended to the runner args
+  /etc/sbt/sbtopts   if this file exists, it is prepended to the runner args
+  -Dkey=val          pass -Dkey=val directly to the java runtime
+  -J-X               pass option -X directly to the java runtime 
+                     (-J is stripped)
+  -S-X               add -X to sbt's scalacOptions (-S is stripped)
+
+In the case of duplicated or conflicting options, the order above
+shows precedence: JAVA_OPTS lowest, command line options highest.
+EOM
+}
+
+
+
+process_my_args () {
+  while [[ $# -gt 0 ]]; do
+    case "$1" in
+     -no-colors) addJava "-Dsbt.log.noformat=true" && shift ;;
+      -no-share) addJava "$noshare_opts" && shift ;;
+     -no-global) addJava "-Dsbt.global.base=$(pwd)/project/.sbtboot" && shift ;;
+      -sbt-boot) require_arg path "$1" "$2" && addJava "-Dsbt.boot.directory=$2" && shift 2 ;;
+       -sbt-dir) require_arg path "$1" "$2" && addJava "-Dsbt.global.base=$2" && shift 2 ;;
+     -debug-inc) addJava "-Dxsbt.inc.debug=true" && shift ;;
+         -batch) exec </dev/null && shift ;;
+
+    -sbt-create) sbt_create=true && shift ;;
+
+            new) sbt_new=true && addResidual "$1" && shift ;;
+
+              *) addResidual "$1" && shift ;;
+    esac
+  done
+  
+  # Now, ensure sbt version is used.
+  [[ "${sbt_version}XXX" != "XXX" ]] && addJava "-Dsbt.version=$sbt_version" 
+
+  # Confirm a user's intent if the current directory does not look like an sbt
+  # top-level directory and neither the -sbt-create option nor the "new"
+  # command was given.
+  [[ -f ./build.sbt || -d ./project || -n "$sbt_create" || -n "$sbt_new" ]] || {
+    echo "[warn] Neither build.sbt nor a 'project' directory in the current directory: $(pwd)"
+    while true; do
+      echo 'c) continue'
+      echo 'q) quit'
+
+      read -p '? ' || exit 1
+      case "$REPLY" in
+        c|C) break ;;
+        q|Q) exit 1 ;;
+      esac
+    done
+  }
+}
+
+loadConfigFile() {
+  # Make sure the last line is read even if it doesn't have a terminating \n
+  cat "$1" | sed $'/^\#/d;s/\r$//' | while read -r line || [[ -n "$line" ]]; do
+    eval echo $line
+  done
+}
+
+# Here we pull in the default settings configuration.
+[[ -f "$dist_sbt_opts_file" ]] && set -- $(loadConfigFile "$dist_sbt_opts_file") "$@"
+
+# Here we pull in the global settings configuration.
+[[ -f "$etc_sbt_opts_file" ]] && set -- $(loadConfigFile "$etc_sbt_opts_file") "$@"
+
+#  Pull in the project-level config file, if it exists.
+[[ -f "$sbt_opts_file" ]] && set -- $(loadConfigFile "$sbt_opts_file") "$@"
+
+#  Pull in the project-level java config, if it exists.
+[[ -f ".jvmopts" ]] && export JAVA_OPTS="$JAVA_OPTS $(loadConfigFile .jvmopts)"
+
+run "$@"
+

scala-stream/sbt-dist/bin/sbt-launch-lib.bash

@@ -0,0 +1,363 @@
+#!/usr/bin/env bash
+#
+
+# A library to simplify using the SBT launcher from other packages.
+# Note: This should be used by tools like giter8/conscript etc.
+
+# TODO - Should we merge the main SBT script with this library?
+
+declare -a residual_args
+declare -a java_args
+declare -a scalac_args
+declare -a sbt_commands
+declare java_cmd=java
+declare java_version
+declare init_sbt_version="1.2.8"
+declare sbt_default_mem=1024
+
+declare SCRIPT=$0
+while [ -h "$SCRIPT" ] ; do
+  ls=$(ls -ld "$SCRIPT")
+  # Drop everything prior to ->
+  link=$(expr "$ls" : '.*-> \(.*\)$')
+  if expr "$link" : '/.*' > /dev/null; then
+    SCRIPT="$link"
+  else
+    SCRIPT=$(dirname "$SCRIPT")/"$link"
+  fi
+done
+declare -r sbt_bin_dir="$(dirname "$SCRIPT")"
+declare -r sbt_home="$(dirname "$sbt_bin_dir")"
+
+echoerr () {
+  echo 1>&2 "$@"
+}
+vlog () {
+  [[ $verbose || $debug ]] && echoerr "$@"
+}
+dlog () {
+  [[ $debug ]] && echoerr "$@"
+}
+
+jar_file () {
+  echo "$(cygwinpath "${sbt_home}/bin/sbt-launch.jar")"
+}
+
+acquire_sbt_jar () {
+  sbt_jar="$(jar_file)"
+
+  if [[ ! -f "$sbt_jar" ]]; then
+    echoerr "Could not find launcher jar: $sbt_jar"
+    exit 2
+  fi
+}
+
+rt_export_file () {
+  echo "${sbt_bin_dir}/java9-rt-export.jar"
+}
+
+execRunner () {
+  # print the arguments one to a line, quoting any containing spaces
+  [[ $verbose || $debug ]] && echo "# Executing command line:" && {
+    for arg; do
+      if printf "%s\n" "$arg" | grep -q ' '; then
+        printf "\"%s\"\n" "$arg"
+      else
+        printf "%s\n" "$arg"
+      fi
+    done
+    echo ""
+  }
+
+  # THis used to be exec, but we loose the ability to re-hook stty then
+  # for cygwin...  Maybe we should flag the feature here...
+  "$@"
+}
+
+addJava () {
+  dlog "[addJava] arg = '$1'"
+  java_args=( "${java_args[@]}" "$1" )
+}
+addSbt () {
+  dlog "[addSbt] arg = '$1'"
+  sbt_commands=( "${sbt_commands[@]}" "$1" )
+}
+addResidual () {
+  dlog "[residual] arg = '$1'"
+  residual_args=( "${residual_args[@]}" "$1" )
+}
+addDebugger () {
+  addJava "-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=$1"
+}
+
+get_mem_opts () {
+  # if we detect any of these settings in ${JAVA_OPTS} or ${JAVA_TOOL_OPTIONS} we need to NOT output our settings.
+  # The reason is the Xms/Xmx, if they don't line up, cause errors.
+  if [[ "${JAVA_OPTS}" == *-Xmx* ]] || [[ "${JAVA_OPTS}" == *-Xms* ]] || [[ "${JAVA_OPTS}" == *-XX:MaxPermSize* ]] || [[ "${JAVA_OPTS}" == *-XX:MaxMetaspaceSize* ]] || [[ "${JAVA_OPTS}" == *-XX:ReservedCodeCacheSize* ]]; then
+    echo ""
+  elif [[ "${JAVA_TOOL_OPTIONS}" == *-Xmx* ]] || [[ "${JAVA_TOOL_OPTIONS}" == *-Xms* ]] || [[ "${JAVA_TOOL_OPTIONS}" == *-XX:MaxPermSize* ]] || [[ "${JAVA_TOOL_OPTIONS}" == *-XX:MaxMetaspaceSize* ]] || [[ "${JAVA_TOOL_OPTIONS}" == *-XX:ReservedCodeCacheSize* ]]; then
+    echo ""
+  elif [[ "${SBT_OPTS}" == *-Xmx* ]] || [[ "${SBT_OPTS}" == *-Xms* ]] || [[ "${SBT_OPTS}" == *-XX:MaxPermSize* ]] || [[ "${SBT_OPTS}" == *-XX:MaxMetaspaceSize* ]] || [[ "${SBT_OPTS}" == *-XX:ReservedCodeCacheSize* ]]; then
+    echo ""
+  else
+    # a ham-fisted attempt to move some memory settings in concert
+    # so they need not be messed around with individually.
+    local mem=${1:-$sbt_default_mem}
+    local codecache=$(( $mem / 8 ))
+    (( $codecache > 128 )) || codecache=128
+    (( $codecache < 512 )) || codecache=512
+    local class_metadata_size=$(( $codecache * 2 ))
+    if [[ -z $java_version ]]; then
+        java_version=$(jdk_version)
+    fi
+    local class_metadata_opt=$((( $java_version < 8 )) && echo "MaxPermSize" || echo "MaxMetaspaceSize")
+
+    local arg_xms=$([[ "${java_args[@]}" == *-Xms* ]] && echo "" || echo "-Xms${mem}m")
+    local arg_xmx=$([[ "${java_args[@]}" == *-Xmx* ]] && echo "" || echo "-Xmx${mem}m")
+    local arg_rccs=$([[ "${java_args[@]}" == *-XX:ReservedCodeCacheSize* ]] && echo "" || echo "-XX:ReservedCodeCacheSize=${codecache}m")
+    local arg_meta=$([[ "${java_args[@]}" == *-XX:${class_metadata_opt}* && ! (( $java_version < 8 )) ]] && echo "" || echo "-XX:${class_metadata_opt}=${class_metadata_size}m")
+
+    echo "${arg_xms} ${arg_xmx} ${arg_rccs} ${arg_meta}"
+  fi
+}
+
+get_gc_opts () {
+  local older_than_9=$(( $java_version < 9 ))
+
+  if [[ "$older_than_9" == "1" ]]; then
+    # don't need to worry about gc
+    echo ""
+  elif [[ "${JAVA_OPTS}" =~ Use.*GC ]] || [[ "${JAVA_TOOL_OPTIONS}" =~ Use.*GC ]] || [[ "${SBT_OPTS}" =~ Use.*GC ]] ; then
+    # GC arg has been passed in - don't change
+    echo ""
+  else
+    # Java 9+ so revert to old
+    echo "-XX:+UseParallelGC"
+  fi
+}
+
+require_arg () {
+  local type="$1"
+  local opt="$2"
+  local arg="$3"
+  if [[ -z "$arg" ]] || [[ "${arg:0:1}" == "-" ]]; then
+    echo "$opt requires <$type> argument"
+    exit 1
+  fi
+}
+
+is_function_defined() {
+  declare -f "$1" > /dev/null
+}
+
+# parses JDK version from the -version output line.
+# 8 for 1.8.0_nn, 9 for 9-ea etc, and "no_java" for undetected
+jdk_version() {
+  local result
+  local lines=$("$java_cmd" -Xms32M -Xmx32M -version 2>&1 | tr '\r' '\n')
+  local IFS=$'\n'
+  for line in $lines; do
+    if [[ (-z $result) && ($line = *"version \""*) ]]
+    then
+      local ver=$(echo $line | sed -e 's/.*version "\(.*\)"\(.*\)/\1/; 1q')
+      # on macOS sed doesn't support '?'
+      if [[ $ver = "1."* ]]
+      then
+        result=$(echo $ver | sed -e 's/1\.\([0-9]*\)\(.*\)/\1/; 1q')
+      else
+        result=$(echo $ver | sed -e 's/\([0-9]*\)\(.*\)/\1/; 1q')
+      fi
+    fi
+  done
+  if [[ -z $result ]]
+  then
+    result=no_java
+  fi
+  echo "$result"
+}
+
+process_args () {
+  while [[ $# -gt 0 ]]; do
+    case "$1" in
+       -h|-help) usage; exit 1 ;;
+    -v|-verbose) verbose=1 && shift ;;
+      -d|-debug) debug=1 && addSbt "-debug" && shift ;;
+
+           -ivy) require_arg path "$1" "$2" && addJava "-Dsbt.ivy.home=$2" && shift 2 ;;
+           -mem) require_arg integer "$1" "$2" && sbt_mem="$2" && shift 2 ;;
+     -jvm-debug) require_arg port "$1" "$2" && addDebugger $2 && shift 2 ;;
+         -batch) exec </dev/null && shift ;;
+
+       -sbt-jar) require_arg path "$1" "$2" && sbt_jar="$2" && shift 2 ;;
+   -sbt-version) require_arg version "$1" "$2" && sbt_version="$2" && shift 2 ;;
+     -java-home) require_arg path "$1" "$2" &&
+                 java_cmd="$2/bin/java" &&
+                 export JAVA_HOME="$2" &&
+                 export JDK_HOME="$2" &&
+                 export PATH="$2/bin:$PATH" &&
+                 shift 2 ;;
+
+          "-D*") addJava "$1" && shift ;;
+            -J*) addJava "${1:2}" && shift ;;
+              *) addResidual "$1" && shift ;;
+    esac
+  done
+
+  is_function_defined process_my_args && {
+    myargs=("${residual_args[@]}")
+    residual_args=()
+    process_my_args "${myargs[@]}"
+  }
+
+  java_version="$(jdk_version)"
+  vlog "[process_args] java_version = '$java_version'"
+}
+
+# Extracts the preloaded directory from either -Dsbt.preloaded or -Dsbt.global.base
+# properties by looking at:
+#   - _JAVA_OPTIONS environment variable,
+#   - SBT_OPTS environment variable,
+#   - JAVA_OPTS environment variable and
+#   - properties set by command-line options
+# in that order. The last one will be chosen such that `sbt.preloaded` is
+# always preferred over `sbt.global.base`.
+getPreloaded() {
+  local -a _java_options_array
+  local -a sbt_opts_array
+  local -a java_opts_array
+  read -a _java_options_array <<< "$_JAVA_OPTIONS"
+  read -a sbt_opts_array <<< "$SBT_OPTS"
+  read -a java_opts_array <<< "$JAVA_OPTS"
+
+  local args_to_check=(
+    "${_java_options_array[@]}"
+    "${sbt_opts_array[@]}"
+    "${java_opts_array[@]}"
+    "${java_args[@]}")
+  local via_global_base="$HOME/.sbt/preloaded"
+  local via_explicit=""
+
+  for opt in "${args_to_check[@]}"; do
+    if [[ "$opt" == -Dsbt.preloaded=* ]]; then
+      via_explicit="${opt#-Dsbt.preloaded=}"
+    elif [[ "$opt" == -Dsbt.global.base=* ]]; then
+      via_global_base="${opt#-Dsbt.global.base=}/preloaded"
+    fi
+  done
+
+  echo "${via_explicit:-${via_global_base}}"
+}
+
+syncPreloaded() {
+  local source_preloaded="$sbt_home/lib/local-preloaded/"
+  local target_preloaded="$(getPreloaded)"
+  if [[ "$init_sbt_version" == "" ]]; then
+    # FIXME: better $init_sbt_version detection
+    init_sbt_version="$(ls -1 "$source_preloaded/org.scala-sbt/sbt/")"
+  fi
+  [[ -f "$target_preloaded/org.scala-sbt/sbt/$init_sbt_version/jars/sbt.jar" ]] || {
+    # lib/local-preloaded exists (This is optional)
+    [[ -d "$source_preloaded" ]] && {
+      command -v rsync >/dev/null 2>&1 && {
+        mkdir -p "$target_preloaded"
+        rsync -a --ignore-existing "$source_preloaded" "$target_preloaded"
+      }
+    }
+  }
+}
+
+# Detect that we have java installed.
+checkJava() {
+  local required_version="$1"
+  # Now check to see if it's a good enough version
+  local good_enough="$(expr $java_version ">=" $required_version)"
+  if [[ "$java_version" == "" ]]; then
+    echo
+    echo "No Java Development Kit (JDK) installation was detected."
+    echo Please go to http://www.oracle.com/technetwork/java/javase/downloads/ and download.
+    echo
+    exit 1
+  elif [[ "$good_enough" != "1" ]]; then
+    echo
+    echo "The Java Development Kit (JDK) installation you have is not up to date."
+    echo $script_name requires at least version $required_version+, you have
+    echo version $java_version
+    echo
+    echo Please go to http://www.oracle.com/technetwork/java/javase/downloads/ and download
+    echo a valid JDK and install before running $script_name.
+    echo
+    exit 1
+  fi
+}
+
+copyRt() {
+  local at_least_9="$(expr $java_version ">=" 9)"
+  if [[ "$at_least_9" == "1" ]]; then
+    rtexport=$(rt_export_file)
+    # The grep for java9-rt-ext- matches the filename prefix printed in Export.java
+    java9_ext=$("$java_cmd" ${JAVA_OPTS} ${SBT_OPTS:-$default_sbt_opts} ${java_args[@]} \
+      -jar "$rtexport" --rt-ext-dir | grep java9-rt-ext-)
+    java9_rt=$(echo "$java9_ext/rt.jar")
+    vlog "[copyRt] java9_rt = '$java9_rt'"
+    if [[ ! -f "$java9_rt" ]]; then
+      echo Copying runtime jar.
+      mkdir -p "$java9_ext"
+      execRunner "$java_cmd" \
+        ${JAVA_OPTS} \
+        ${SBT_OPTS:-$default_sbt_opts} \
+        ${java_args[@]} \
+        -jar "$rtexport" \
+        "${java9_rt}"
+    fi
+    addJava "-Dscala.ext.dirs=${java9_ext}"
+  fi
+}
+
+run() {
+  # process the combined args, then reset "$@" to the residuals
+  process_args "$@"
+  set -- "${residual_args[@]}"
+  argumentCount=$#
+
+  # Copy preloaded repo to user's preloaded directory
+  syncPreloaded
+
+  # no jar? download it.
+  [[ -f "$sbt_jar" ]] || acquire_sbt_jar "$sbt_version" || {
+    # still no jar? uh-oh.
+    echo "Download failed. Obtain the sbt-launch.jar manually and place it at $sbt_jar"
+    exit 1
+  }
+
+  # TODO - java check should be configurable...
+  checkJava "6"
+
+  # Java 9 support
+  copyRt
+
+  #If we're in cygwin, we should use the windows config, and terminal hacks
+  if [[ "$CYGWIN_FLAG" == "true" ]]; then
+    stty -icanon min 1 -echo > /dev/null 2>&1
+    addJava "-Djline.terminal=jline.UnixTerminal"
+    addJava "-Dsbt.cygwin=true"
+  fi
+
+  # run sbt
+  execRunner "$java_cmd" \
+    $(get_mem_opts $sbt_mem) \
+    $(get_gc_opts) \
+    ${JAVA_OPTS} \
+    ${SBT_OPTS:-$default_sbt_opts} \
+    ${java_args[@]} \
+    -jar "$sbt_jar" \
+    "${sbt_commands[@]}" \
+    "${residual_args[@]}"
+
+  exit_code=$?
+
+  # Clean up the terminal from cygwin hacks.
+  if [[ "$CYGWIN_FLAG" == "true" ]]; then
+    stty icanon echo > /dev/null 2>&1
+  fi
+  exit $exit_code
+}

scala-stream/sbt-dist/bin/sbt.bat

@@ -0,0 +1,212 @@
+@REM SBT launcher script
+@REM
+@REM Environment:
+@REM JAVA_HOME - location of a JDK home dir (mandatory)
+@REM SBT_OPTS  - JVM options (optional)
+@REM Configuration:
+@REM sbtconfig.txt found in the SBT_HOME.
+
+@REM   ZOMG! We need delayed expansion to build up CFG_OPTS later
+@setlocal enabledelayedexpansion
+
+@echo off
+set SBT_HOME=%~dp0
+set SBT_ARGS=
+
+rem FIRST we load the config file of extra options.
+set FN=%SBT_HOME%\..\conf\sbtconfig.txt
+set CFG_OPTS=
+FOR /F "tokens=* eol=# usebackq delims=" %%i IN ("%FN%") DO (
+  set DO_NOT_REUSE_ME=%%i
+  rem ZOMG (Part #2) WE use !! here to delay the expansion of
+  rem CFG_OPTS, otherwise it remains "" for this loop.
+  set CFG_OPTS=!CFG_OPTS! !DO_NOT_REUSE_ME!
+)
+
+rem poor man's jenv (which is not available on Windows)
+IF DEFINED JAVA_HOMES (
+  IF EXIST .java-version FOR /F %%A IN (.java-version) DO (
+    SET JAVA_HOME=%JAVA_HOMES%\%%A
+    SET JDK_HOME=%JAVA_HOMES%\%%A
+  )
+)
+rem must set PATH or wrong javac is used for java projects
+IF DEFINED JAVA_HOME SET "PATH=%JAVA_HOME%\bin;%PATH%"
+
+rem users can set JAVA_OPTS via .jvmopts (sbt-extras style)
+IF EXIST .jvmopts FOR /F %%A IN (.jvmopts) DO (
+  SET _jvmopts_line=%%A
+  IF NOT "!_jvmopts_line:~0,1!"=="#" (
+    SET JAVA_OPTS=%%A !JAVA_OPTS!
+  )
+)
+rem We use the value of the JAVACMD environment variable if defined
+set _JAVACMD=%JAVACMD%
+
+if "%_JAVACMD%"=="" (
+  if not "%JAVA_HOME%"=="" (
+    if exist "%JAVA_HOME%\bin\java.exe" set "_JAVACMD=%JAVA_HOME%\bin\java.exe"
+  )
+)
+
+if "%_JAVACMD%"=="" set _JAVACMD=java
+
+rem We use the value of the JAVA_OPTS environment variable if defined, rather than the config.
+set _JAVA_OPTS=%JAVA_OPTS%
+if "%_JAVA_OPTS%"=="" set _JAVA_OPTS=%CFG_OPTS%
+
+set INIT_SBT_VERSION=1.2.8
+
+:args_loop
+if "%~1" == "" goto args_end
+
+if "%~1" == "-jvm-debug" (
+  set JVM_DEBUG=true
+  set /a JVM_DEBUG_PORT=5005 2>nul >nul
+) else if "!JVM_DEBUG!" == "true" (
+  set /a JVM_DEBUG_PORT=%1 2>nul >nul
+  if not "%~1" == "!JVM_DEBUG_PORT!" (
+    set SBT_ARGS=!SBT_ARGS! %1
+  )
+) else if /I "%~1" == "new" (
+  set sbt_new=true
+  set SBT_ARGS=!SBT_ARGS! %1
+) else (
+  set SBT_ARGS=!SBT_ARGS! %1
+)
+
+shift
+goto args_loop
+:args_end
+
+rem Confirm a user's intent if the current directory does not look like an sbt
+rem top-level directory and the "new" command was not given.
+if not exist build.sbt (
+  if not exist project\ (
+    if not defined sbt_new (
+      echo [warn] Neither build.sbt nor a 'project' directory in the current directory: %CD%
+      setlocal
+:confirm
+      echo c^) continue
+      echo q^) quit
+
+      set /P reply=?^ 
+      if /I "!reply!" == "c" (
+        goto confirm_end
+      ) else if /I "!reply!" == "q" (
+        exit /B 1
+      )
+
+      goto confirm
+:confirm_end
+      endlocal
+    )
+  )
+)
+
+call :process
+
+call :checkjava
+
+call :copyrt
+
+if defined JVM_DEBUG_PORT (
+  set _JAVA_OPTS=!_JAVA_OPTS! -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=!JVM_DEBUG_PORT!
+)
+
+call :sync_preloaded
+
+call :run %SBT_ARGS%
+
+if ERRORLEVEL 1 goto error
+goto end
+
+:run
+
+"%_JAVACMD%" %_JAVA_OPTS% %SBT_OPTS% -cp "%SBT_HOME%sbt-launch.jar" xsbt.boot.Boot %*
+goto :eof
+
+:process
+rem Parses x out of 1.x; for example 8 out of java version 1.8.0_xx
+rem Otherwise, parses the major version; 9 out of java version 9-ea
+set JAVA_VERSION=0
+for /f "tokens=3" %%g in ('"%_JAVACMD%" -Xms32M -Xmx32M -version 2^>^&1 ^| findstr /i version') do (
+  set JAVA_VERSION=%%g
+)
+set JAVA_VERSION=%JAVA_VERSION:"=%
+for /f "delims=.-_ tokens=1-2" %%v in ("%JAVA_VERSION%") do (
+  if /I "%%v" EQU "1" (
+    set JAVA_VERSION=%%w
+  ) else (
+    set JAVA_VERSION=%%v
+  )
+)
+exit /B 0
+
+:checkjava
+set required_version=6
+if /I %JAVA_VERSION% GEQ %required_version% (
+  exit /B 0
+)
+echo.
+echo The Java Development Kit (JDK) installation you have is not up to date.
+echo sbt requires at least version %required_version%+, you have
+echo version %JAVA_VERSION%
+echo.
+echo Please go to http://www.oracle.com/technetwork/java/javase/downloads/ and download
+echo a valid JDK and install before running sbt.
+echo.
+exit /B 1
+
+:copyrt
+if /I %JAVA_VERSION% GEQ 9 (
+  set rtexport=!SBT_HOME!java9-rt-export.jar
+
+  "%_JAVACMD%" %_JAVA_OPTS% %SBT_OPTS% -jar "!rtexport!" --rt-ext-dir > "%TEMP%.\rtext.txt"
+  set /p java9_ext= < "%TEMP%.\rtext.txt"
+  set java9_rt=!java9_ext!\rt.jar
+
+  if not exist "!java9_rt!" (
+    mkdir "!java9_ext!"
+    "%_JAVACMD%" %_JAVA_OPTS% %SBT_OPTS% -jar "!rtexport!" "!java9_rt!"
+  )
+  set _JAVA_OPTS=!_JAVA_OPTS! -Dscala.ext.dirs="!java9_ext!"
+
+  rem check to see if a GC has been set in the opts
+  echo !_JAVA_OPTS! | findstr /r "Use.*GC" >nul
+  if ERRORLEVEL 1 (
+    rem don't have a GC set - revert to old GC
+    set _JAVA_OPTS=!_JAVA_OPTS! -XX:+UseParallelGC
+  )
+)
+exit /B 0
+
+:sync_preloaded
+if "%INIT_SBT_VERSION%"=="" (
+  rem FIXME: better %INIT_SBT_VERSION% detection
+  FOR /F "tokens=* USEBACKQ" %%F IN (`dir /b "%SBT_HOME%\..\lib\local-preloaded\org.scala-sbt\sbt" /B`) DO (
+    SET INIT_SBT_VERSION=%%F
+  )
+)
+set PRELOAD_SBT_JAR="%UserProfile%\.sbt\preloaded\org.scala-sbt\sbt\%INIT_SBT_VERSION%\jars\sbt.jar"
+if /I %JAVA_VERSION% GEQ 8 (
+  where robocopy >nul 2>nul
+  if %ERRORLEVEL% equ 0 (
+    REM echo %PRELOAD_SBT_JAR%
+    if not exist %PRELOAD_SBT_JAR% (
+      if exist "%SBT_HOME%\..\lib\local-preloaded\" (
+        echo "about to robocopy"
+        robocopy "%SBT_HOME%\..\lib\local-preloaded" "%UserProfile%\.sbt\preloaded" /E
+      )
+    )
+  )
+)
+exit /B 0
+
+:error
+@endlocal
+exit /B 1
+
+:end
+@endlocal
+exit /B 0

scala-stream/sbt-dist/conf/sbtconfig.txt

@@ -0,0 +1,14 @@
+# Set the java args to high
+
+-Xmx512M
+
+-XX:MaxPermSize=256m
+
+-XX:ReservedCodeCacheSize=128m
+
+
+
+# Set the extra SBT options
+
+-Dsbt.log.format=true
+

scala-stream/sbt-dist/conf/sbtopts

@@ -0,0 +1,49 @@
+# ------------------------------------------------ #
+#  The SBT Configuration file.                     #
+# ------------------------------------------------ #
+
+
+# Disable ANSI color codes
+#
+#-no-colors
+
+# Starts sbt even if the current directory contains no sbt project.
+#
+-sbt-create
+
+# Path to global settings/plugins directory (default: ~/.sbt)
+#
+#-sbt-dir  /etc/sbt
+
+# Path to shared boot directory (default: ~/.sbt/boot in 0.11 series)
+#
+#-sbt-boot ~/.sbt/boot  
+
+# Path to local Ivy repository (default: ~/.ivy2)
+#
+#-ivy ~/.ivy2
+
+# set memory options
+#
+#-mem   <integer>  
+
+# Use local caches for projects, no sharing.
+#
+#-no-share
+
+# Put SBT in offline mode.
+#
+#-offline
+
+# Sets the SBT version to use.
+#-sbt-version  0.11.3
+
+# Scala version (default: latest release)
+#
+#-scala-home <path>        
+#-scala-version <version>
+
+# java version (default: java from PATH, currently $(java -version |& grep version))
+#
+#-java-home <path>
+

scala-stream/src/main/scala/scalastream/J8SStream.scala

@@ -0,0 +1,44 @@
+package scalastream
+
+import scalastream.App.{Config, Data}
+
+import scala.collection.immutable.ArraySeq
+import scala.reflect.{ClassTag, classTag}
+
+class J8SStream[@specialized(Float, Double) A: Fractional: ClassTag](val config: Config[A])
+    extends ScalaStream[A]:
+
+  private var a: Array[A] = _
+  private var b: Array[A] = _
+  private var c: Array[A] = _
+  private val scalar: A   = config.scalar
+
+  inline private def stream =
+    java.util.stream.IntStream.range(0, config.options.arraysize).parallel()
+
+  override inline def initArrays(): Unit =
+    a = Array.ofDim(config.options.arraysize)
+    b = Array.ofDim(config.options.arraysize)
+    c = Array.ofDim(config.options.arraysize)
+    stream.forEach { i =>
+      a(i) = config.init._1
+      b(i) = config.init._2
+      c(i) = config.init._3
+    }
+
+  override inline def copy(): Unit    = stream.forEach(i => c(i) = a(i))
+  override inline def mul(): Unit     = stream.forEach(i => b(i) = scalar * c(i))
+  override inline def add(): Unit     = stream.forEach(i => c(i) = a(i) + b(i))
+  override inline def triad(): Unit   = stream.forEach(i => a(i) = b(i) + scalar * c(i))
+  override inline def nstream(): Unit = stream.forEach(i => a(i) = b(i) * scalar * c(i))
+  override inline def dot(): A =
+    // horrible special-case for double, there isn't a mapToFloat so we give up on that
+    val cls = classTag[A].runtimeClass
+    if java.lang.Double.TYPE == cls then
+      stream
+        .mapToDouble(i => (a(i) * b(i)).asInstanceOf[Double])
+        .reduce(0, (l: Double, r: Double) => l + r)
+        .asInstanceOf[A]
+    else stream.mapToObj[A](i => a(i) * b(i)).reduce(0.fractional, (l: A, r: A) => l + r)
+
+  override inline def data(): Data[A] = Data(a.to(ArraySeq), b.to(ArraySeq), c.to(ArraySeq))

scala-stream/src/main/scala/scalastream/ParStream.scala

@@ -0,0 +1,36 @@
+package scalastream
+
+import scalastream.App.{Config, Data}
+
+import scala.collection.immutable.ArraySeq
+import scala.collection.parallel.CollectionConverters._
+import scala.reflect.ClassTag
+class ParStream[@specialized(Float, Double) A: Fractional: ClassTag](val config: Config[A])
+    extends ScalaStream[A]:
+
+  private var a: Array[A] = _
+  private var b: Array[A] = _
+  private var c: Array[A] = _
+  private val scalar: A   = config.scalar
+
+  inline private def indices = (0 until config.options.arraysize).par
+
+  override inline def initArrays(): Unit =
+    a = Array.ofDim(config.options.arraysize)
+    b = Array.ofDim(config.options.arraysize)
+    c = Array.ofDim(config.options.arraysize)
+
+    for i <- indices do
+      a(i) = config.init._1
+      b(i) = config.init._2
+      c(i) = config.init._3
+
+  override inline def copy(): Unit    = for i <- indices do c(i) = a(i)
+  override inline def mul(): Unit     = for i <- indices do b(i) = scalar * c(i)
+  override inline def add(): Unit     = for i <- indices do c(i) = a(i) + b(i)
+  override inline def triad(): Unit   = for i <- indices do a(i) = b(i) + scalar * c(i)
+  override inline def nstream(): Unit = for i <- indices do a(i) = b(i) * scalar * c(i)
+  override inline def dot(): A =
+    indices.aggregate[A](0.fractional)((acc, i) => acc + (a(i) * b(i)), _ + _)
+
+  override inline def data(): Data[A] = Data(a.to(ArraySeq), b.to(ArraySeq), c.to(ArraySeq))

scala-stream/src/main/scala/scalastream/PlainStream.scala

@@ -0,0 +1,31 @@
+package scalastream
+
+import scalastream.App.{Config, Data}
+
+import scala.collection.immutable.ArraySeq
+import scala.reflect.ClassTag
+class PlainStream[@specialized(Float, Double) A: Fractional: ClassTag](val config: Config[A])
+    extends ScalaStream[A]:
+
+  private var a: Array[A] = _
+  private var b: Array[A] = _
+  private var c: Array[A] = _
+  private val scalar: A   = config.scalar
+
+  override inline def initArrays(): Unit =
+    a = Array.fill(config.options.arraysize)(config.init._1)
+    b = Array.fill(config.options.arraysize)(config.init._2)
+    c = Array.fill(config.options.arraysize)(config.init._3)
+
+  private inline def indices = 0 until config.options.arraysize
+
+  override inline def copy(): Unit    = for i <- indices do c(i) = a(i)
+  override inline def mul(): Unit     = for i <- indices do b(i) = scalar * c(i)
+  override inline def add(): Unit     = for i <- indices do c(i) = a(i) + b(i)
+  override inline def triad(): Unit   = for i <- indices do a(i) = b(i) + (scalar * c(i))
+  override inline def nstream(): Unit = for i <- indices do a(i) = b(i) * scalar * c(i)
+  override inline def dot(): A =
+    var acc: A = 0.fractional
+    for i <- indices do acc = acc + (a(i) * b(i))
+    acc
+  override inline def data(): Data[A] = Data(a.to(ArraySeq), b.to(ArraySeq), c.to(ArraySeq))

scala-stream/src/main/scala/scalastream/ScalaStream.scala

@@ -0,0 +1,369 @@
+package scalastream
+import scalastream.App.{Config, Data, Timings}
+
+import java.util.concurrent.TimeUnit
+import scala.collection.immutable.ArraySeq
+import scala.collection.mutable.ArrayBuffer
+import scala.concurrent.duration.{Duration, FiniteDuration, SECONDS}
+import scala.math.{Pi, pow}
+import scala.reflect.ClassTag
+import scopt.OParser
+
+transparent trait ScalaStream[@specialized(Float, Double) A]:
+
+  def config: Config[A]
+
+  def initArrays(): Unit
+  def copy(): Unit
+  def mul(): Unit
+  def add(): Unit
+  def triad(): Unit
+  def nstream(): Unit
+  def dot(): A
+
+  transparent inline def timed[R](f: => R): (FiniteDuration, R) =
+    val start = System.nanoTime()
+    val r     = f
+    val end   = System.nanoTime()
+    FiniteDuration(end - start, TimeUnit.NANOSECONDS) -> r
+
+  inline def runAll(times: Int)(using Fractional[A]): (Timings[Vector[FiniteDuration]], A) =
+    val copy  = ArrayBuffer.fill[FiniteDuration](times)(Duration.Zero)
+    val mul   = ArrayBuffer.fill[FiniteDuration](times)(Duration.Zero)
+    val add   = ArrayBuffer.fill[FiniteDuration](times)(Duration.Zero)
+    val triad = ArrayBuffer.fill[FiniteDuration](times)(Duration.Zero)
+    val dot   = ArrayBuffer.fill[FiniteDuration](times)(Duration.Zero)
+
+    var lastSum: A = 0.fractional
+
+    for i <- 0 until times do
+      copy(i) = timed(this.copy())._1
+      mul(i) = timed(this.mul())._1
+      add(i) = timed(this.add())._1
+      triad(i) = timed(this.triad())._1
+      val (dot_, sum) = timed(this.dot())
+      dot(i) = dot_
+      lastSum = sum
+    val s = lastSum
+
+    (
+      Timings(
+        copy = copy.toVector,
+        mul = mul.toVector,
+        add = add.toVector,
+        triad = triad.toVector,
+        dot = dot.toVector
+      ),
+      s
+    )
+
+  def runTriad(times: Int): FiniteDuration           = timed(for _ <- 0 until times do triad())._1
+  def runNStream(times: Int): Vector[FiniteDuration] = Vector.fill(times)(timed(nstream())._1)
+
+  def data(): Data[A]
+
+
+trait Fractional[@specialized(Double, Float) A]:
+  def toFractional(f: Float): A
+  def toFractional(f: Double): A
+  def compare(x: A, y: A): Int
+  def add(x: A, y: A): A
+  def sub(x: A, y: A): A
+  def mul(x: A, y: A): A
+  def div(x: A, y: A): A
+  def abs(x: A): A
+  extension (x: Float) inline def fractional  = toFractional(x)
+  extension (x: Double) inline def fractional = toFractional(x)
+  extension (x: Int) inline def fractional    = toFractional(x.toFloat)
+  extension (x: Long) inline def fractional   = toFractional(x.toDouble)
+  extension (x: A)
+    inline def +(y: A)  = add(x, y)
+    inline def -(y: A)  = sub(x, y)
+    inline def *(y: A)  = mul(x, y)
+    inline def /(y: A)  = div(x, y)
+    inline def >(y: A)  = compare(x, y) > 0
+    inline def <(y: A)  = compare(x, y) < 0
+    inline def abs_     = abs(x)
+end Fractional
+
+given FloatFractional: Fractional[Float] with
+  inline def toFractional(f: Float): Float    = f
+  inline def toFractional(f: Double): Float   = f.toFloat
+  inline def compare(x: Float, y: Float): Int = x.compare(y)
+  inline def add(x: Float, y: Float): Float   = x + y
+  inline def sub(x: Float, y: Float): Float   = x - y
+  inline def mul(x: Float, y: Float): Float   = x * y
+  inline def div(x: Float, y: Float): Float   = x / y
+  inline def abs(x: Float): Float             = math.abs(x)
+
+given DoubleFractional: Fractional[Double] with
+  inline def toFractional(f: Float): Double     = f.toDouble
+  inline def toFractional(f: Double): Double    = f
+  inline def compare(x: Double, y: Double): Int = x.compare(y)
+  inline def add(x: Double, y: Double): Double  = x + y
+  inline def sub(x: Double, y: Double): Double  = x - y
+  inline def mul(x: Double, y: Double): Double  = x * y
+  inline def div(x: Double, y: Double): Double  = x / y
+  inline def abs(x: Double): Double             = math.abs(x)
+
+object App:
+
+  final val Version: String = "3.4.0"
+
+  case class Config[@specialized(Double, Float) A](
+      options: Options,
+      benchmark: Benchmark,
+      typeSize: Int,
+      ulp: A,
+      scalar: A,
+      init: (A, A, A)
+  )
+
+  case class Timings[A](copy: A, mul: A, add: A, triad: A, dot: A)
+  case class Data[A](@specialized(Double, Float) a: ArraySeq[A], b: ArraySeq[A], c: ArraySeq[A])
+
+  case class Options(
+      list: Boolean = false,
+      device: Int = 0,
+      numtimes: Int = 100,
+      arraysize: Int = 33554432,
+      float: Boolean = false,
+      triad_only: Boolean = false,
+      nstream_only: Boolean = false,
+      csv: Boolean = false,
+      mibibytes: Boolean = false
+  )
+
+  object Options:
+    val Default = Options()
+    val builder = OParser.builder[Options]
+    val parser1 =
+      import builder._
+      OParser.sequence(
+        programName("scala-stream"),
+        head("ScalaStream", s"$Version"),
+        opt[Unit]('l', "list").text("List available devices").action((_, x) => x.copy(list = true)),
+        opt[Int]('d', "device")
+          .text(s"Select device at <device>, defaults to ${Default.device}")
+          .action((v, x) => x.copy(device = v)),
+        opt[Int]('n', "numtimes")
+          .text(s"Run the test <numtimes> times (NUM >= 2), defaults to ${Default.numtimes}")
+          .validate {
+            case n if n >= 2 => success
+            case n           => failure(s"$n <= 2")
+          }
+          .action((n, x) => x.copy(numtimes = n)),
+        opt[Int]('a', "arraysize")
+          .text(s"Use <arraysize> elements in the array, defaults to ${Default.arraysize}")
+          .action((v, x) => x.copy(arraysize = v)),
+        opt[Unit]('f', "float")
+          .text("Use floats (rather than doubles)")
+          .action((_, x) => x.copy(float = true)),
+        opt[Unit]('t', "triad_only")
+          .text("Only run triad")
+          .action((_, x) => x.copy(triad_only = true)),
+        opt[Unit]('n', "nstream_only")
+          .text("Only run nstream")
+          .action((_, x) => x.copy(nstream_only = true)),
+        opt[Unit]('c', "csv").text("Output as csv table").action((_, x) => x.copy(csv = true)),
+        opt[Unit]('m', "mibibytes")
+          .text("Use MiB=2^20 for bandwidth calculation (default MB=10^6)")
+          .action((_, x) => x.copy(mibibytes = true)),
+        help('h', "help").text("prints this usage text")
+      )
+
+  enum Benchmark:
+    case All, NStream, Triad
+
+  implicit class RichDuration(private val d: Duration) extends AnyVal:
+    def seconds: Double = d.toUnit(SECONDS)
+
+  def validate[A: Fractional](vec: Data[A], config: Config[A], dotSum: Option[A] = None): Unit =
+
+    var (goldA, goldB, goldC) = config.init
+    for _ <- 0 until config.options.numtimes do
+      config.benchmark match
+        case Benchmark.All =>
+          goldC = goldA
+          goldB = config.scalar * goldC
+          goldC = goldA + goldB
+          goldA = goldB + config.scalar * goldC
+        case Benchmark.Triad =>
+          goldA = goldB + config.scalar * goldC
+        case Benchmark.NStream =>
+          goldA += goldB + config.scalar * goldC
+
+    val tolerance = config.ulp * (100.fractional)
+    def validateXs(name: String, xs: Seq[A], from: A): Unit =
+      val error = xs.map(x => (x - from).abs_).fold(0.fractional)(_ + _) / xs.size.fractional
+      if error > tolerance then
+        Console.err.println(s"Validation failed on $name. Average error $error ")
+
+    validateXs("a", vec.a, goldA)
+    validateXs("b", vec.b, goldB)
+    validateXs("c", vec.c, goldC)
+
+    dotSum.foreach { sum =>
+      val goldSum = (goldA * goldB) * (config.options.arraysize).fractional
+      val error   = ((sum - goldSum) / goldSum).abs_
+      if error > 1.fractional / 100000000.fractional then
+        Console.err.println(
+          s"Validation failed on sum. Error $error \nSum was $sum but should be $goldSum"
+        )
+    }
+
+  inline def run[A: Fractional: ClassTag](
+      name: String,
+      config: Config[A],
+      mkStream: Config[A] => ScalaStream[A]
+  ): Unit =
+
+    val opt = config.options
+
+    val arrayBytes = opt.arraysize * config.typeSize
+    val totalBytes = arrayBytes * 3
+    val (megaScale, megaSuffix, gigaScale, gigaSuffix) =
+      if !opt.mibibytes then (1.0e-6, "MB", 1.0e-9, "GB")
+      else (pow(2.0, -20), "MiB", pow(2.0, -30), "GiB")
+
+    if !opt.csv then
+
+      val vendor = System.getProperty("java.vendor")
+      val ver    = System.getProperty("java.version")
+      val home   = System.getProperty("java.home")
+      println(
+        s"""BabelStream
+           |Version: $Version
+           |Implementation: $name; Scala (Java $ver; $vendor; home=$home)""".stripMargin
+      )
+
+      println(s"Running ${config.benchmark match {
+        case Benchmark.All     => "kernels"
+        case Benchmark.Triad   => "triad"
+        case Benchmark.NStream => "nstream"
+      }} ${opt.numtimes} times")
+
+      if config.benchmark == Benchmark.Triad then println(s"Number of elements: ${opt.arraysize}")
+
+      println(s"Precision: ${if opt.float then "float" else "double"}")
+      println(
+        f"Array size: ${megaScale * arrayBytes}%.1f $megaSuffix (=${gigaScale * arrayBytes}%.1f $gigaSuffix)"
+      )
+      println(
+        f"Total size: ${megaScale * totalBytes}%.1f $megaSuffix (=${gigaScale * totalBytes}%.1f $gigaSuffix)"
+      )
+
+    def mkRow(xs: Vector[FiniteDuration], name: String, totalBytes: Int) =
+      val tail = xs.tail
+      (tail.minOption.map(_.seconds), tail.maxOption.map(_.seconds)) match
+        case (Some(min), Some(max)) =>
+          val avg  = (tail.foldLeft(Duration.Zero)(_ + _) / tail.size.toDouble).seconds
+          val mbps = megaScale * totalBytes.toDouble / min
+          if opt.csv then
+            Vector(
+              "function"                                                -> name,
+              "num_times"                                               -> opt.numtimes.toString,
+              "n_elements"                                              -> opt.arraysize.toString,
+              "sizeof"                                                  -> totalBytes.toString,
+              s"max_m${if opt.mibibytes then "i" else ""}bytes_per_sec" -> mbps.toString,
+              "min_runtime"                                             -> min.toString,
+              "max_runtime"                                             -> max.toString,
+              "avg_runtime"                                             -> avg.toString
+            )
+          else
+            Vector(
+              "Function"                                        -> name,
+              s"M${if opt.mibibytes then "i" else ""}Bytes/sec" -> f"$mbps%.3f",
+              "Min (sec)"                                       -> f"$min%.5f",
+              "Max"                                             -> f"$max%.5f",
+              "Average"                                         -> f"$avg%.5f"
+            )
+        case (_, _) => sys.error(s"No min/max element for $name(size=$totalBytes)")
+
+    def tabulate(rows: Vector[(String, String)]*): Unit = rows.toList match
+      case Nil => sys.error(s"Empty tabulation")
+      case header :: _ =>
+        val padding = if opt.csv then 0 else 12
+        val sep     = if opt.csv then "," else ""
+        println(header.map(_._1.padTo(padding, ' ')).mkString(sep))
+        println(rows.map(_.map(_._2.padTo(padding, ' ')).mkString(sep)).mkString("\n"))
+
+    val stream = mkStream(config)
+    stream.initArrays()
+    config.benchmark match
+      case Benchmark.All =>
+        val (results, sum) = stream.runAll(opt.numtimes)
+        validate(stream.data(), config, Some(sum))
+        tabulate(
+          mkRow(results.copy, "Copy", 2 * arrayBytes),
+          mkRow(results.mul, "Mul", 2 * arrayBytes),
+          mkRow(results.add, "Add", 3 * arrayBytes),
+          mkRow(results.triad, "Triad", 3 * arrayBytes),
+          mkRow(results.dot, "Dot", 2 * arrayBytes)
+        )
+      case Benchmark.NStream =>
+        val result = stream.runNStream(opt.numtimes)
+        validate(stream.data(), config)
+        tabulate(mkRow(result, "Nstream", 4 * arrayBytes))
+      case Benchmark.Triad =>
+        val results    = stream.runTriad(opt.numtimes)
+        val totalBytes = 3 * arrayBytes * opt.numtimes
+        val bandwidth  = megaScale * (totalBytes / results.seconds)
+        println(f"Runtime (seconds): ${results.seconds}%.5f")
+        println(f"Bandwidth ($gigaSuffix/s): $bandwidth%.3f ")
+
+  inline def devices[A: Fractional: ClassTag]: Vector[(String, Config[A] => ScalaStream[A])] =
+    Vector(
+      "Scala Parallel Collections" -> (ParStream(_)),
+      "Java 8 Stream"              -> (J8SStream(_)),
+      "Threaded"                   -> (ThreadStream(_)),
+      "Serial"                     -> (PlainStream(_))
+    )
+
+  inline def runWith[A: Fractional: ClassTag](i: Int, config: Config[A]): Unit =
+    devices[A].lift(i) match
+      case None                   => println(s"Device index out of bounds: $i")
+      case Some((name, mkStream)) => run(name, config, mkStream)
+
+  def main(args: Array[String]): Unit =
+
+    def handleOpt(opt: Options) =
+      val benchmark = (opt.nstream_only, opt.triad_only) match
+        case (true, false)  => Benchmark.NStream
+        case (false, true)  => Benchmark.Triad
+        case (false, false) => Benchmark.All
+        case (true, true) =>
+          throw new RuntimeException(
+            "Both triad and nstream are enabled, pick one or omit both to run all benchmarks"
+          )
+
+      if opt.list then
+        devices[Float].zipWithIndex.foreach { case ((name, _), i) => println(s"$i: $name") }
+      else if opt.float then
+        runWith(
+          opt.device,
+          Config(
+            options = opt,
+            benchmark = benchmark,
+            typeSize = 4, // 32bit
+            ulp = math.ulp(Float.MaxValue),
+            scalar = 0.4f,
+            init = (0.1f, 0.2f, 0.0f)
+          )
+        )
+      else
+        runWith(
+          opt.device,
+          Config(
+            options = opt,
+            benchmark = benchmark,
+            typeSize = 8,
+            ulp = math.ulp(Double.MaxValue),
+            scalar = 0.4, // 64bit
+            init = (0.1, 0.2, 0.0)
+          )
+        )
+
+    OParser.parse(Options.parser1, args, Options.Default) match
+      case Some(config) => handleOpt(config)
+      case _            => sys.exit(1)

scala-stream/src/main/scala/scalastream/ThreadStream.scala

@@ -0,0 +1,68 @@
+package scalastream
+
+import net.openhft.affinity.{AffinityStrategies, AffinityThreadFactory}
+import scalastream.App.{Config, Data}
+
+import java.util.concurrent.{Callable, Executors}
+import scala.collection.immutable.ArraySeq
+import scala.reflect.ClassTag
+object ThreadStream {}
+class ThreadStream[@specialized(Float, Double) A: Fractional: ClassTag](val config: Config[A])
+    extends ScalaStream[A]:
+
+  private var a: Array[A] = _
+  private var b: Array[A] = _
+  private var c: Array[A] = _
+  private val scalar: A   = config.scalar
+
+  private val chunks: Int = sys.runtime.availableProcessors()
+
+  private val pool = Executors.newFixedThreadPool(
+    chunks,
+    new AffinityThreadFactory("scala-stream", true, AffinityStrategies.DIFFERENT_CORE)
+  )
+
+  private val indices = (0 until config.options.arraysize)
+    .grouped(config.options.arraysize / chunks)
+    .toSeq
+
+  private inline def forEachAll[C](c: => C)(f: (C, Int) => Unit): Seq[C] =
+    import scala.jdk.CollectionConverters._
+    val xs = pool
+      .invokeAll(
+        indices.map { r =>
+          { () =>
+            val ctx = c
+            r.foreach(f(ctx, _))
+            ctx
+          }: Callable[C]
+        }.asJavaCollection
+      )
+      .asScala
+      .map(_.get())
+      .toSeq
+    xs
+
+  override inline def initArrays(): Unit =
+    a = Array.ofDim(config.options.arraysize)
+    b = Array.ofDim(config.options.arraysize)
+    c = Array.ofDim(config.options.arraysize)
+    forEachAll(()) { (_, i) =>
+      a(i) = config.init._1
+      b(i) = config.init._2
+      c(i) = config.init._3
+    }
+    ()
+
+  class Box(var value: A)
+  override inline def copy(): Unit = { forEachAll(())((_, i) => c(i) = a(i)); () }
+  override inline def mul(): Unit = { forEachAll(())((_, i) => b(i) = scalar * c(i)); () }
+  override inline def add(): Unit = { forEachAll(())((_, i) => c(i) = a(i) + b(i)); () }
+  override inline def triad(): Unit = { forEachAll(())((_, i) => a(i) = b(i) + scalar * c(i)); () }
+  override inline def nstream(): Unit = { forEachAll(())((_, i) => a(i) = b(i) * scalar * c(i)); () }
+
+  override inline def dot(): A =
+    forEachAll(Box(0.fractional))((acc, i) => acc.value = acc.value + (a(i) * b(i)))
+      .map(_.value)
+      .fold(0.fractional)(_ + _)
+  override inline def data(): Data[A] = Data(a.to(ArraySeq), b.to(ArraySeq), c.to(ArraySeq))