Loading pkgs/development/libraries/science/math/cudnn/extension.nix +69 −238 Original line number Diff line number Diff line # Support matrix can be found at # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-880/support-matrix/index.html # Type aliases # Release = { # version: String, # minCudaVersion: String, # maxCudaVersion: String, # url: String, # hash: String, # } final: prev: let inherit (final) callPackage; inherit (prev) cudatoolkit cudaVersion lib pkgs; inherit (prev.lib.versions) major; inherit (prev) cudaVersion; inherit (prev.lib) attrsets lists versions strings trivial; ### CuDNN # Utilities # majorMinorPatch :: String -> String majorMinorPatch = (trivial.flip trivial.pipe) [ (versions.splitVersion) (lists.take 3) (strings.concatStringsSep ".") ]; buildCuDnnPackage = args: let useCudatoolkitRunfile = lib.versionOlder cudaVersion "11.3.999"; in callPackage ./generic.nix { inherit useCudatoolkitRunfile; } args; # Compute versioned attribute name to be used in this package set # computeName :: String -> String computeName = version: "cudnn_${strings.replaceStrings ["."] ["_"] (majorMinorPatch version)}"; toUnderscore = str: lib.replaceStrings ["."] ["_"] str; # Check whether a CUDNN release supports our CUDA version # Thankfully we're able to do lexicographic comparison on the version strings # isSupported :: Release -> Bool isSupported = release: strings.versionAtLeast cudaVersion release.minCudaVersion && strings.versionAtLeast release.maxCudaVersion cudaVersion; majorMinorPatch = str: lib.concatStringsSep "." (lib.take 3 (lib.splitVersion str)); # useCudatoolkitRunfile :: Bool useCudatoolkitRunfile = strings.versionOlder cudaVersion "11.3.999"; cuDnnPackages = with lib; let # Check whether a file is supported for our cuda version isSupported = fileData: elem cudaVersion fileData.supportedCudaVersions; # Return the first file that is supported. In practice there should only ever be one anyway. supportedFile = files: findFirst isSupported null files; # Supported versions with versions as keys and file as value supportedVersions = filterAttrs (version: file: file !=null ) (mapAttrs (version: files: supportedFile files) cuDnnVersions); # Compute versioned attribute name to be used in this package set computeName = version: "cudnn_${toUnderscore version}"; # Add all supported builds as attributes allBuilds = mapAttrs' (version: file: nameValuePair (computeName version) (buildCuDnnPackage (removeAttrs file ["fileVersion"]))) supportedVersions; # Set the default attributes, e.g. cudnn = cudnn_8_3_1; defaultBuild = { "cudnn" = if allBuilds ? ${computeName cuDnnDefaultVersion} then allBuilds.${computeName cuDnnDefaultVersion} else throw "cudnn-${cuDnnDefaultVersion} does not support your cuda version ${cudaVersion}"; }; in allBuilds // defaultBuild; # buildCuDnnPackage :: Release -> Derivation buildCuDnnPackage = callPackage ./generic.nix {inherit useCudatoolkitRunfile;}; cuDnnVersions = let urlPrefix = "https://developer.download.nvidia.com/compute/redist/cudnn"; in { "7.4.2" = [ rec { fileVersion = "10.0"; fullVersion = "7.4.2.24"; hash = "sha256-Lt/IagK1DRfojEeJVaMy5qHoF05+U6NFi06lH68C2qM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-10.0-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.0" ]; } ]; "7.6.5" = [ rec { fileVersion = "10.0"; fullVersion = "7.6.5.32"; hash = "sha256-KDVeOV8LK5OsLIO2E2CzW6bNA3fkTni+GXtrYbS0kro="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.0" ]; } rec { fileVersion = "10.1"; fullVersion = "7.6.5.32"; hash = "sha256-fq7IA5osMKsLx1jTA1iHZ2k972v0myJIWiwAvy4TbLM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.1" ]; } rec { fileVersion = "10.2"; fullVersion = "7.6.5.32"; hash = "sha256-fq7IA5osMKsLx1jTA1iHZ2k972v0myJIWiwAvy4TbLN="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.2" ]; } ]; "8.0.5" = [ rec { fileVersion = "10.1"; fullVersion = "8.0.5.39"; hash = "sha256-kJCElSmIlrM6qVBjo0cfk8NmJ9esAcF9w211xl7qSgA="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "10.1" ]; } rec { fileVersion = "10.2"; fullVersion = "8.0.5.39"; hash = "sha256-IfhMBcZ78eyFnnfDjM1b8VSWT6HDCPRJlZvkw1bjgvM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.0"; fullVersion = "8.0.5.39"; hash = "sha256-ThbueJXetKixwZS4ErpJWG730mkCBRQB03F1EYmKm3M="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "11.0" ]; } rec { fileVersion = "11.1"; fullVersion = "8.0.5.39"; hash = "sha256-HQRr+nk5navMb2yxUHkYdUQ5RC6gyp4Pvs3URvmwDM4="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "11.1" ]; } ]; "8.1.1" = [ rec { fileVersion = "10.2"; fullVersion = "8.1.1.33"; hash = "sha256-Kkp7mabpv6aQ6xm7QeSVU/KnpJGls6v8rpAOFmxbbr0="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-811/support-matrix/index.html#cudnn-versions-810-811 supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.2"; fullVersion = "8.1.1.33"; hash = "sha256-mKh4TpKGLyABjSDCgbMNSgzZUfk2lPZDPM9K6cUCumo="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-811/support-matrix/index.html#cudnn-versions-810-811 supportedCudaVersions = [ "11.0" "11.1" "11.2" ]; } ]; "8.2.4" = [ rec { fileVersion = "10.2"; fullVersion = "8.2.4.15"; hash = "sha256-0jyUoxFaHHcRamwSfZF1+/WfcjNkN08mo0aZB18yIvE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-824/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.4"; fullVersion = "8.2.4.15"; hash = "sha256-Dl0t+JC5ln76ZhnaQhMQ2XMjVlp58FoajLm3Fluq0Nc="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-824/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" ]; } ]; "8.3.3" = [ rec { fileVersion = "10.2"; fullVersion = "8.3.3.40"; hash = "sha256-2FVPKzLmKV1fyPOsJeaPlAWLAYyAHaucFD42gS+JJqs="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-833/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.5"; fullVersion = "8.3.3.40"; hash = "sha256-6r6Wx1zwPqT1N5iU2RTx+K4UzqsSGYnoSwg22Sf7dzE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-833/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" ]; } ]; "8.4.1" = [ rec { fileVersion = "10.2"; fullVersion = "8.4.1.50"; hash = "sha256-I88qMmU6lIiLVmaPuX7TTbisgTav839mssxUo3lQNjg="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-841/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.6"; fullVersion = "8.4.1.50"; hash = "sha256-7JbSN22B/KQr3T1MPXBambKaBlurV/kgVhx2PinGfQE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-841/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" ]; } ]; "8.5.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.5.0.96"; hash = "sha256-1mzhbbzR40WKkHnQLtJHhg0vYgf7G8a0OBcCwIOkJjM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-850/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.7"; fullVersion = "8.5.0.96"; hash = "sha256-VFSm/ZTwCHKMqumtrZk8ToXvNjAuJrzkO+p9RYpee20="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-850/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" ]; } ]; "8.6.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.6.0.163"; hash = "sha256-t4sr/GrFqqdxu2VhaJQk5K1Xm/0lU4chXG8hVL09R9k="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-860/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.8"; fullVersion = "8.6.0.163"; hash = "sha256-u8OW30cpTGV+3AnGAGdNYIyxv8gLgtz0VHBgwhcRFZ4="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-860/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" "11.8" ]; } ]; "8.7.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.7.0.84"; hash = "sha256-bZhaqc8+GbPV2FQvvbbufd8VnEJgvfkICc2N3/gitRg="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-870/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.8"; fullVersion = "8.7.0.84"; hash = "sha256-l2xMunIzyXrnQAavq1Fyl2MAukD1slCiH4z3H1nJ920="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-870/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" "11.8" ]; } ]; # Reverse the list to have the latest release first # cudnnReleases :: List Release cudnnReleases = lists.reverseList (builtins.import ./releases.nix); # Check whether a CUDNN release supports our CUDA version # supportedReleases :: NonEmptyList Release supportedReleases = let filtered = builtins.filter isSupported cudnnReleases; nonEmptyFiltered = trivial.throwIf (filtered == []) '' CUDNN does not support your cuda version ${cudaVersion} '' filtered; in nonEmptyFiltered; # The latest release is the first element of the list and will be our default choice # latestReleaseName :: String latestReleaseName = computeName (builtins.head supportedReleases).version; # Function to transform our releases into build attributes # toBuildAttrs :: Release -> { name: String, value: Derivation } toBuildAttrs = release: { name = computeName release.version; value = buildCuDnnPackage release; }; # Default attributes cuDnnDefaultVersion = { "10.0" = "7.4.2"; "10.1" = "8.0.5"; "10.2" = "8.7.0"; "11.0" = "8.7.0"; "11.1" = "8.7.0"; "11.2" = "8.7.0"; "11.3" = "8.7.0"; "11.4" = "8.7.0"; "11.5" = "8.7.0"; "11.6" = "8.7.0"; "11.7" = "8.7.0"; "11.8" = "8.7.0"; }.${cudaVersion} or "8.7.0"; # Add all supported builds as attributes # allBuilds :: AttrSet String Derivation allBuilds = builtins.listToAttrs (builtins.map toBuildAttrs supportedReleases); in cuDnnPackages # The latest release will be our default build # defaultBuild :: AttrSet String Derivation defaultBuild.cudnn = allBuilds.${latestReleaseName}; # builds :: AttrSet String Derivation builds = allBuilds // defaultBuild; in builds pkgs/development/libraries/science/math/cudnn/generic.nix +103 −88 Original line number Diff line number Diff line { stdenv , lib , zlib , useCudatoolkitRunfile ? false , cudaVersion , cudaMajorVersion , cudatoolkit # if cuda>=11: only used for .cc , libcublas ? null # cuda <11 doesn't ship redist packages , autoPatchelfHook , autoAddOpenGLRunpathHook , fetchurl , # The distributed version of CUDNN includes both dynamically liked .so files, { stdenv, lib, zlib, useCudatoolkitRunfile ? false, cudaVersion, cudaMajorVersion, cudatoolkit, # if cuda>=11: only used for .cc libcublas ? null, # cuda <11 doesn't ship redist packages autoPatchelfHook, autoAddOpenGLRunpathHook, fetchurl, # The distributed version of CUDNN includes both dynamically liked .so files, # as well as statically linked .a files. However, CUDNN is quite large # (multiple gigabytes), so you can save some space in your nix store by # removing the statically linked libraries if you are not using them. # # Setting this to true removes the statically linked .a files. # Setting this to false keeps these statically linked .a files. removeStatic ? false removeStatic ? false, }: { version, url, hash, minCudaVersion, maxCudaVersion, }: { fullVersion , url , hash , supportedCudaVersions ? [ ] }: assert useCudatoolkitRunfile || (libcublas != null); let assert useCudatoolkitRunfile || (libcublas != null); let inherit (cudatoolkit) cc; inherit (lib) lists strings trivial versions; # majorMinorPatch :: String -> String majorMinorPatch = (trivial.flip trivial.pipe) [ (versions.splitVersion) (lists.take 3) (strings.concatStringsSep ".") ]; majorMinorPatch = version: lib.concatStringsSep "." (lib.take 3 (lib.splitVersion version)); version = majorMinorPatch fullVersion; # versionTriple :: String # Version with three components: major.minor.patch versionTriple = majorMinorPatch version; # cudatoolkit_root :: Derivation cudatoolkit_root = if useCudatoolkitRunfile then cudatoolkit Loading @@ -40,7 +48,7 @@ let in stdenv.mkDerivation { pname = "cudatoolkit-${cudaMajorVersion}-cudnn"; inherit version; version = versionTriple; src = fetchurl { inherit url hash; Loading @@ -64,35 +72,40 @@ stdenv.mkDerivation { # # Note also that version <=8.3.0 contained a subdirectory "lib64/" but in # version 8.3.2 it seems to have been renamed to simply "lib/". installPhase = '' installPhase = '' runHook preInstall mkdir -p $out cp -a include $out/include [ -d "lib/" ] && cp -a lib $out/lib [ -d "lib64/" ] && cp -a lib64 $out/lib64 '' + lib.optionalString removeStatic '' '' + strings.optionalString removeStatic '' rm -f $out/lib/*.a rm -f $out/lib64/*.a '' + '' '' + '' runHook postInstall ''; # Without --add-needed autoPatchelf forgets $ORIGIN on cuda>=8.0.5. postFixup = lib.optionalString (lib.versionAtLeast fullVersion "8.0.5") '' postFixup = strings.optionalString (strings.versionAtLeast versionTriple "8.0.5") '' patchelf $out/lib/libcudnn.so --add-needed libcudnn_cnn_infer.so ''; passthru = { inherit useCudatoolkitRunfile; cudatoolkit = lib.warn '' cudatoolkit = trivial.warn '' cudnn.cudatoolkit passthru attribute is deprecated; if your derivation uses cudnn directly, it should probably consume cudaPackages instead '' cudatoolkit; majorVersion = lib.versions.major version; majorVersion = versions.major versionTriple; }; meta = with lib; { Loading @@ -101,7 +114,9 @@ stdenv.mkDerivation { # official version constraints (as recorded in default.nix). In some cases # you _may_ be able to smudge version constraints, just know that you're # embarking into unknown and unsupported territory when doing so. broken = !(elem cudaVersion supportedCudaVersions); broken = strings.versionOlder cudaVersion minCudaVersion || strings.versionOlder maxCudaVersion cudaVersion; description = "NVIDIA CUDA Deep Neural Network library (cuDNN)"; homepage = "https://developer.nvidia.com/cudnn"; sourceProvenance = with sourceTypes; [binaryNativeCode]; Loading pkgs/development/libraries/science/math/cudnn/releases.nix 0 → 100644 +170 −0 File added.Preview size limit exceeded, changes collapsed. Show changes Loading
pkgs/development/libraries/science/math/cudnn/extension.nix +69 −238 Original line number Diff line number Diff line # Support matrix can be found at # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-880/support-matrix/index.html # Type aliases # Release = { # version: String, # minCudaVersion: String, # maxCudaVersion: String, # url: String, # hash: String, # } final: prev: let inherit (final) callPackage; inherit (prev) cudatoolkit cudaVersion lib pkgs; inherit (prev.lib.versions) major; inherit (prev) cudaVersion; inherit (prev.lib) attrsets lists versions strings trivial; ### CuDNN # Utilities # majorMinorPatch :: String -> String majorMinorPatch = (trivial.flip trivial.pipe) [ (versions.splitVersion) (lists.take 3) (strings.concatStringsSep ".") ]; buildCuDnnPackage = args: let useCudatoolkitRunfile = lib.versionOlder cudaVersion "11.3.999"; in callPackage ./generic.nix { inherit useCudatoolkitRunfile; } args; # Compute versioned attribute name to be used in this package set # computeName :: String -> String computeName = version: "cudnn_${strings.replaceStrings ["."] ["_"] (majorMinorPatch version)}"; toUnderscore = str: lib.replaceStrings ["."] ["_"] str; # Check whether a CUDNN release supports our CUDA version # Thankfully we're able to do lexicographic comparison on the version strings # isSupported :: Release -> Bool isSupported = release: strings.versionAtLeast cudaVersion release.minCudaVersion && strings.versionAtLeast release.maxCudaVersion cudaVersion; majorMinorPatch = str: lib.concatStringsSep "." (lib.take 3 (lib.splitVersion str)); # useCudatoolkitRunfile :: Bool useCudatoolkitRunfile = strings.versionOlder cudaVersion "11.3.999"; cuDnnPackages = with lib; let # Check whether a file is supported for our cuda version isSupported = fileData: elem cudaVersion fileData.supportedCudaVersions; # Return the first file that is supported. In practice there should only ever be one anyway. supportedFile = files: findFirst isSupported null files; # Supported versions with versions as keys and file as value supportedVersions = filterAttrs (version: file: file !=null ) (mapAttrs (version: files: supportedFile files) cuDnnVersions); # Compute versioned attribute name to be used in this package set computeName = version: "cudnn_${toUnderscore version}"; # Add all supported builds as attributes allBuilds = mapAttrs' (version: file: nameValuePair (computeName version) (buildCuDnnPackage (removeAttrs file ["fileVersion"]))) supportedVersions; # Set the default attributes, e.g. cudnn = cudnn_8_3_1; defaultBuild = { "cudnn" = if allBuilds ? ${computeName cuDnnDefaultVersion} then allBuilds.${computeName cuDnnDefaultVersion} else throw "cudnn-${cuDnnDefaultVersion} does not support your cuda version ${cudaVersion}"; }; in allBuilds // defaultBuild; # buildCuDnnPackage :: Release -> Derivation buildCuDnnPackage = callPackage ./generic.nix {inherit useCudatoolkitRunfile;}; cuDnnVersions = let urlPrefix = "https://developer.download.nvidia.com/compute/redist/cudnn"; in { "7.4.2" = [ rec { fileVersion = "10.0"; fullVersion = "7.4.2.24"; hash = "sha256-Lt/IagK1DRfojEeJVaMy5qHoF05+U6NFi06lH68C2qM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-10.0-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.0" ]; } ]; "7.6.5" = [ rec { fileVersion = "10.0"; fullVersion = "7.6.5.32"; hash = "sha256-KDVeOV8LK5OsLIO2E2CzW6bNA3fkTni+GXtrYbS0kro="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.0" ]; } rec { fileVersion = "10.1"; fullVersion = "7.6.5.32"; hash = "sha256-fq7IA5osMKsLx1jTA1iHZ2k972v0myJIWiwAvy4TbLM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.1" ]; } rec { fileVersion = "10.2"; fullVersion = "7.6.5.32"; hash = "sha256-fq7IA5osMKsLx1jTA1iHZ2k972v0myJIWiwAvy4TbLN="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${cudatoolkit.majorVersion}-linux-x64-v${fullVersion}.tgz"; supportedCudaVersions = [ "10.2" ]; } ]; "8.0.5" = [ rec { fileVersion = "10.1"; fullVersion = "8.0.5.39"; hash = "sha256-kJCElSmIlrM6qVBjo0cfk8NmJ9esAcF9w211xl7qSgA="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "10.1" ]; } rec { fileVersion = "10.2"; fullVersion = "8.0.5.39"; hash = "sha256-IfhMBcZ78eyFnnfDjM1b8VSWT6HDCPRJlZvkw1bjgvM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.0"; fullVersion = "8.0.5.39"; hash = "sha256-ThbueJXetKixwZS4ErpJWG730mkCBRQB03F1EYmKm3M="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "11.0" ]; } rec { fileVersion = "11.1"; fullVersion = "8.0.5.39"; hash = "sha256-HQRr+nk5navMb2yxUHkYdUQ5RC6gyp4Pvs3URvmwDM4="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-805/support-matrix/index.html supportedCudaVersions = [ "11.1" ]; } ]; "8.1.1" = [ rec { fileVersion = "10.2"; fullVersion = "8.1.1.33"; hash = "sha256-Kkp7mabpv6aQ6xm7QeSVU/KnpJGls6v8rpAOFmxbbr0="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-811/support-matrix/index.html#cudnn-versions-810-811 supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.2"; fullVersion = "8.1.1.33"; hash = "sha256-mKh4TpKGLyABjSDCgbMNSgzZUfk2lPZDPM9K6cUCumo="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-811/support-matrix/index.html#cudnn-versions-810-811 supportedCudaVersions = [ "11.0" "11.1" "11.2" ]; } ]; "8.2.4" = [ rec { fileVersion = "10.2"; fullVersion = "8.2.4.15"; hash = "sha256-0jyUoxFaHHcRamwSfZF1+/WfcjNkN08mo0aZB18yIvE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-824/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.4"; fullVersion = "8.2.4.15"; hash = "sha256-Dl0t+JC5ln76ZhnaQhMQ2XMjVlp58FoajLm3Fluq0Nc="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/cudnn-${fileVersion}-linux-x64-v${fullVersion}.tgz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-824/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" ]; } ]; "8.3.3" = [ rec { fileVersion = "10.2"; fullVersion = "8.3.3.40"; hash = "sha256-2FVPKzLmKV1fyPOsJeaPlAWLAYyAHaucFD42gS+JJqs="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-833/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.5"; fullVersion = "8.3.3.40"; hash = "sha256-6r6Wx1zwPqT1N5iU2RTx+K4UzqsSGYnoSwg22Sf7dzE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-833/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" ]; } ]; "8.4.1" = [ rec { fileVersion = "10.2"; fullVersion = "8.4.1.50"; hash = "sha256-I88qMmU6lIiLVmaPuX7TTbisgTav839mssxUo3lQNjg="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-841/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.6"; fullVersion = "8.4.1.50"; hash = "sha256-7JbSN22B/KQr3T1MPXBambKaBlurV/kgVhx2PinGfQE="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-841/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" ]; } ]; "8.5.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.5.0.96"; hash = "sha256-1mzhbbzR40WKkHnQLtJHhg0vYgf7G8a0OBcCwIOkJjM="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-850/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.7"; fullVersion = "8.5.0.96"; hash = "sha256-VFSm/ZTwCHKMqumtrZk8ToXvNjAuJrzkO+p9RYpee20="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-850/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" ]; } ]; "8.6.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.6.0.163"; hash = "sha256-t4sr/GrFqqdxu2VhaJQk5K1Xm/0lU4chXG8hVL09R9k="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-860/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.8"; fullVersion = "8.6.0.163"; hash = "sha256-u8OW30cpTGV+3AnGAGdNYIyxv8gLgtz0VHBgwhcRFZ4="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-860/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" "11.8" ]; } ]; "8.7.0" = [ rec { fileVersion = "10.2"; fullVersion = "8.7.0.84"; hash = "sha256-bZhaqc8+GbPV2FQvvbbufd8VnEJgvfkICc2N3/gitRg="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-870/support-matrix/index.html supportedCudaVersions = [ "10.2" ]; } rec { fileVersion = "11.8"; fullVersion = "8.7.0.84"; hash = "sha256-l2xMunIzyXrnQAavq1Fyl2MAukD1slCiH4z3H1nJ920="; url = "${urlPrefix}/v${majorMinorPatch fullVersion}/local_installers/${fileVersion}/cudnn-linux-x86_64-${fullVersion}_cuda${major fileVersion}-archive.tar.xz"; # https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-870/support-matrix/index.html supportedCudaVersions = [ "11.0" "11.1" "11.2" "11.3" "11.4" "11.5" "11.6" "11.7" "11.8" ]; } ]; # Reverse the list to have the latest release first # cudnnReleases :: List Release cudnnReleases = lists.reverseList (builtins.import ./releases.nix); # Check whether a CUDNN release supports our CUDA version # supportedReleases :: NonEmptyList Release supportedReleases = let filtered = builtins.filter isSupported cudnnReleases; nonEmptyFiltered = trivial.throwIf (filtered == []) '' CUDNN does not support your cuda version ${cudaVersion} '' filtered; in nonEmptyFiltered; # The latest release is the first element of the list and will be our default choice # latestReleaseName :: String latestReleaseName = computeName (builtins.head supportedReleases).version; # Function to transform our releases into build attributes # toBuildAttrs :: Release -> { name: String, value: Derivation } toBuildAttrs = release: { name = computeName release.version; value = buildCuDnnPackage release; }; # Default attributes cuDnnDefaultVersion = { "10.0" = "7.4.2"; "10.1" = "8.0.5"; "10.2" = "8.7.0"; "11.0" = "8.7.0"; "11.1" = "8.7.0"; "11.2" = "8.7.0"; "11.3" = "8.7.0"; "11.4" = "8.7.0"; "11.5" = "8.7.0"; "11.6" = "8.7.0"; "11.7" = "8.7.0"; "11.8" = "8.7.0"; }.${cudaVersion} or "8.7.0"; # Add all supported builds as attributes # allBuilds :: AttrSet String Derivation allBuilds = builtins.listToAttrs (builtins.map toBuildAttrs supportedReleases); in cuDnnPackages # The latest release will be our default build # defaultBuild :: AttrSet String Derivation defaultBuild.cudnn = allBuilds.${latestReleaseName}; # builds :: AttrSet String Derivation builds = allBuilds // defaultBuild; in builds
pkgs/development/libraries/science/math/cudnn/generic.nix +103 −88 Original line number Diff line number Diff line { stdenv , lib , zlib , useCudatoolkitRunfile ? false , cudaVersion , cudaMajorVersion , cudatoolkit # if cuda>=11: only used for .cc , libcublas ? null # cuda <11 doesn't ship redist packages , autoPatchelfHook , autoAddOpenGLRunpathHook , fetchurl , # The distributed version of CUDNN includes both dynamically liked .so files, { stdenv, lib, zlib, useCudatoolkitRunfile ? false, cudaVersion, cudaMajorVersion, cudatoolkit, # if cuda>=11: only used for .cc libcublas ? null, # cuda <11 doesn't ship redist packages autoPatchelfHook, autoAddOpenGLRunpathHook, fetchurl, # The distributed version of CUDNN includes both dynamically liked .so files, # as well as statically linked .a files. However, CUDNN is quite large # (multiple gigabytes), so you can save some space in your nix store by # removing the statically linked libraries if you are not using them. # # Setting this to true removes the statically linked .a files. # Setting this to false keeps these statically linked .a files. removeStatic ? false removeStatic ? false, }: { version, url, hash, minCudaVersion, maxCudaVersion, }: { fullVersion , url , hash , supportedCudaVersions ? [ ] }: assert useCudatoolkitRunfile || (libcublas != null); let assert useCudatoolkitRunfile || (libcublas != null); let inherit (cudatoolkit) cc; inherit (lib) lists strings trivial versions; # majorMinorPatch :: String -> String majorMinorPatch = (trivial.flip trivial.pipe) [ (versions.splitVersion) (lists.take 3) (strings.concatStringsSep ".") ]; majorMinorPatch = version: lib.concatStringsSep "." (lib.take 3 (lib.splitVersion version)); version = majorMinorPatch fullVersion; # versionTriple :: String # Version with three components: major.minor.patch versionTriple = majorMinorPatch version; # cudatoolkit_root :: Derivation cudatoolkit_root = if useCudatoolkitRunfile then cudatoolkit Loading @@ -40,7 +48,7 @@ let in stdenv.mkDerivation { pname = "cudatoolkit-${cudaMajorVersion}-cudnn"; inherit version; version = versionTriple; src = fetchurl { inherit url hash; Loading @@ -64,35 +72,40 @@ stdenv.mkDerivation { # # Note also that version <=8.3.0 contained a subdirectory "lib64/" but in # version 8.3.2 it seems to have been renamed to simply "lib/". installPhase = '' installPhase = '' runHook preInstall mkdir -p $out cp -a include $out/include [ -d "lib/" ] && cp -a lib $out/lib [ -d "lib64/" ] && cp -a lib64 $out/lib64 '' + lib.optionalString removeStatic '' '' + strings.optionalString removeStatic '' rm -f $out/lib/*.a rm -f $out/lib64/*.a '' + '' '' + '' runHook postInstall ''; # Without --add-needed autoPatchelf forgets $ORIGIN on cuda>=8.0.5. postFixup = lib.optionalString (lib.versionAtLeast fullVersion "8.0.5") '' postFixup = strings.optionalString (strings.versionAtLeast versionTriple "8.0.5") '' patchelf $out/lib/libcudnn.so --add-needed libcudnn_cnn_infer.so ''; passthru = { inherit useCudatoolkitRunfile; cudatoolkit = lib.warn '' cudatoolkit = trivial.warn '' cudnn.cudatoolkit passthru attribute is deprecated; if your derivation uses cudnn directly, it should probably consume cudaPackages instead '' cudatoolkit; majorVersion = lib.versions.major version; majorVersion = versions.major versionTriple; }; meta = with lib; { Loading @@ -101,7 +114,9 @@ stdenv.mkDerivation { # official version constraints (as recorded in default.nix). In some cases # you _may_ be able to smudge version constraints, just know that you're # embarking into unknown and unsupported territory when doing so. broken = !(elem cudaVersion supportedCudaVersions); broken = strings.versionOlder cudaVersion minCudaVersion || strings.versionOlder maxCudaVersion cudaVersion; description = "NVIDIA CUDA Deep Neural Network library (cuDNN)"; homepage = "https://developer.nvidia.com/cudnn"; sourceProvenance = with sourceTypes; [binaryNativeCode]; Loading
pkgs/development/libraries/science/math/cudnn/releases.nix 0 → 100644 +170 −0 File added.Preview size limit exceeded, changes collapsed. Show changes