summaryrefslogtreecommitdiff
path: root/dev-util/Tensile
diff options
context:
space:
mode:
authorV3n3RiX <venerix@koprulu.sector>2022-12-19 19:49:08 +0000
committerV3n3RiX <venerix@koprulu.sector>2022-12-19 19:49:08 +0000
commitf287ecad888abdeb38e617d0485de282cd6819db (patch)
tree874d56500040734c4dbb9e437e0d5ed80a0a1886 /dev-util/Tensile
parent844ae757702c53a56ee57056873a8204d256d47e (diff)
gentoo auto-resync : 19:12:2022 - 19:49:08
Diffstat (limited to 'dev-util/Tensile')
-rw-r--r--dev-util/Tensile/Manifest7
-rw-r--r--dev-util/Tensile/Tensile-4.3.0-r1.ebuild86
-rw-r--r--dev-util/Tensile/files/Tensile-4.3.0-fix-arch-parse.patch12
-rw-r--r--dev-util/Tensile/files/Tensile-4.3.0-gentoopath.patch62
-rw-r--r--dev-util/Tensile/files/Tensile-4.3.0-gfx1031.patch119
-rw-r--r--dev-util/Tensile/files/Tensile-4.3.0-hsaco-compile-specified-arch.patch96
-rw-r--r--dev-util/Tensile/files/Tensile-4.3.0-use-ninja.patch21
7 files changed, 0 insertions, 403 deletions
diff --git a/dev-util/Tensile/Manifest b/dev-util/Tensile/Manifest
index 239cad34ae7f..6c604616e60a 100644
--- a/dev-util/Tensile/Manifest
+++ b/dev-util/Tensile/Manifest
@@ -1,9 +1,4 @@
-AUX Tensile-4.3.0-fix-arch-parse.patch 424 BLAKE2B ac89228abe3173a4cabcb90a679be3706a39cf6bebc448fb17dc9a94be5c5c75289e53fb867c17a3bc36330079905302a83736663d2f743c9805c97253b9fec0 SHA512 131e9fa7c39f12fbbbc37ace0d29162ff2805e8435985d3edab7e678d89587a2ee9cc1549eee21842c42e99e4a4fba3125fa34fec3dbc9f61cff96d8eeaf5b25
-AUX Tensile-4.3.0-gentoopath.patch 2796 BLAKE2B 10c90d69ea4c83cd36f1f61be8c716e66eba889bd10a221c4af895d30f04d851c1eba4c3d6c845d96d0e14a78f617e55a4acdd2f93db9c5fcbaeffd50ab5c6f1 SHA512 e14ca9c94f302863c1b0f9eca86ab3a7dbed9c61d1488255a5c9c0d458c800e7329a41f7b91694558a733632bee025dee5bb59a8478982a3ab1d591f4cb3a1e4
-AUX Tensile-4.3.0-gfx1031.patch 7037 BLAKE2B ea2e5dec9d5f6749a78cae4d1659a08b712e548cdca29dabde29f52a60f443eac0c10bac27667545816e5b6b89835ee5e79d766028bad79ba17e5b14a4bf5ea9 SHA512 2b1936a6e5b52bf2f6aaa70255be9f78e82329f7855a3db56d769fed13934ca2c2671f089bc5aea43c4a0e67d3ec8ad185e32f7ade03c515b3c74a683568b192
-AUX Tensile-4.3.0-hsaco-compile-specified-arch.patch 3859 BLAKE2B 4590978cff0fdb0501ba2405615775cb3f8900ce164c3d57cc59dfe82e1adc25376b6463855a045f5e0652cc985b0015566b95ff28327c59f2db875a84cb49b3 SHA512 57520f1d46f46c7cfd5b8c3ae307da07695c19479b687ce994ad9fba2e755ed7c776c3d8bee8e0642c7613a5d53d8469f329b1eb07563c13d2f7bda8b3d36687
AUX Tensile-4.3.0-output-commands.patch 1629 BLAKE2B c43395d8e8aeb956d5eb84c8a34ad4b995ed32499ceababeb22d5ca3ac5ee6e8ef353a1467a77943a441f0840b9a6ad1f53202ec8673ef3753aa2bec1395fec4 SHA512 45bdae1d87f6e31d337b22f146b9dee25f4f87357c6ac52bd843556dd742431db10efb55fb8ef815cc9941b5b9f05970d43a6fec0f2d9195c30c633e33d138bf
-AUX Tensile-4.3.0-use-ninja.patch 957 BLAKE2B cb919b224aaeb3b9d65a755ff0688aa823b70361d2ee940d62f33828c118a60c88cd0f108b6e34b22097d9c4c5122a4db2808332a97cd6e7685a4447570b7552 SHA512 3d206a093e870fc95f0ed42e08691144589112d4581f73ba1c4c74ed9bdad0812ec4f67a6076a8a3ac9ab89c61c70f1baaedb774e0908177473be2d206858bf4
AUX Tensile-5.0.2-adjust-parallel.patch 2936 BLAKE2B faca0234d615e4451b83b53a23147eb441e7929a3ef7542bef3ac5d0a383a6d1aa966373a012cc40bcbd1137a16a80a9c88b4d5459552b6b04c4819cb325e8d9 SHA512 d32f358133e0727b9ea66300f680451222190c0b7c9d91a9e0364fb2589b0ae12d69facb5872a81bd36a770ad649a22120649e6d3550b02bc6365706392caf57
AUX Tensile-5.0.2-fix-arch-parse.patch 424 BLAKE2B 43e716ccc9256c2f98cc1d2807e776599b384875161233d4a010a53f11a5764fd7a5b4c50ded1762737e9fbe7923b64e9353ee4590fd1b6b9b03ac621af35c9b SHA512 5913e7fc42bc5b43e1dcca531e4ef88b524325b497601171a9844013906c2ffc87469a4c55037c721ace5757c6aaf4d8bf8ca683f659f9d8988555c90425802a
AUX Tensile-5.0.2-gentoopath.patch 3410 BLAKE2B bbf2c406bbcf2c34efd04667788719f6f10aebda3f9d9096e6798c1a8924768df4fed30223e2f4675af88c41841d148e43e3cd63edc7523dab4618ae245b0339 SHA512 c7b99eab1f0c979f54fcac3b2dcb4a2094b9a536bb392fcf4fb62050b1dc6e61f5280d3d1c7e3ed21617ebe693a17eb2bb7535a93f700123fe6aa6f28f256514
@@ -11,10 +6,8 @@ AUX Tensile-5.0.2-gfx1031.patch 5561 BLAKE2B 244bb644a90ba34f500f4d0cc634acf041c
AUX Tensile-5.0.2-use-ninja.patch 897 BLAKE2B f06a3831a43a4b6c00e7eb231cc13901f7b4a3a924a50d85ba81f9def535951c450a01504506476c1ee88317cbdf0c0c00a57cb5cd198ab44e8e0b94aefda25d SHA512 15404d3cd5dbc7b74e81092d86f9c49cd45fe4df645602849ff515baafa3890fd68ca808cbebd6ae37e951b382ff4446a31089306c950f8549431dfbe75c3b53
AUX Tensile-5.1.3-gentoopath.patch 3350 BLAKE2B 890ab24fbd99aede187bde67c8d6f586f625c30cff34da39e12b199b022170de8df3087c3581f9e635d5d12f31db0aa944a2628733417322621ab51b0292b01b SHA512 d17a948146e7ef2447ae92b2822f53cd77b0992297fd2a43598434db4f7c2095d269138ce5c0eedda73a34e193fc56401f22471f9eca68ec6026577f07a7fca1
DIST Tensile-5.0.2-PR1419.patch.gz 43754 BLAKE2B 3b1d7333ed0115f77816806e7735e25d897dbae3f7a40ece61d655c8f016d9d1c804b586b4e0e75661e20c517f5937a61d5ead25a545f0ce2e4a178bcbb3e7be SHA512 0154d8e14d335f4ecd86e5db335289f8662cede7e5d4b93bf6c56d8e48c1d7d0457271c078c9abbbca1f4aa91f028f6ce6a89efb783ba0207d0f39fd5de8a42c
-DIST rocm-Tensile-4.3.0.tar.gz 14250149 BLAKE2B 030138eaca2a0aadd96801e6dbd72e510716dd90553ef3795c5e04e00a34a05ecae82b24f755e4033a4acfcdb1cf26291da1e7902bb090f89a010d403e832beb SHA512 126db0b413c716fba8d5be9bff7a44fd1badacbf32f3db8d0db649819177db37ebd56fd22dd3c809655f5d29675be115e698cd10bc3d0b4b23878ae3726fce47
DIST rocm-Tensile-5.0.2.tar.gz 14362090 BLAKE2B d44c3135496db41c4eb2545f6054c84493e89ef6357da881dae6859763338417e7a9e142cdce7e2de0536684bb641049ad0440964adbe68d03ca329f01efabc9 SHA512 f5ab7d7cbc5ba25498a02a6b7c07d5a9a548edcd44b439b92c8718072382adc18ae78274db1a7174e06436ac1396f3b4edafc7f491a2c87086c368e7270e64e3
DIST rocm-Tensile-5.1.3.tar.gz 14357517 BLAKE2B 7c5881cc2b1d25db39f54a99f9f0a08395a5f7a58deebe74d9868e8f7044db5ae2f3907196adf16dc6abeba0f1f47f5cce5c8900891276daff8e4ad614cf3b94 SHA512 4405cb6de65c1da1fd93910fc901402a24e996ff04eccadf483c25d8eb518e03ea9bbf00e372dd5fe377696e71c39355603095f3c5fc1e26d311ed15d8e3b734
-EBUILD Tensile-4.3.0-r1.ebuild 2764 BLAKE2B e24ddb516d8dfd9e5e691f567f5b97b279e518b38be49a2cc69e8eae100d4d2ce7dc1432247c0bd5c94656ff1a7faf88a8e8f2e8df3dbbbc491eb4323cc20706 SHA512 79859e7dd5647bc4c1eb766faa9b3116994ed74aeb900caaf1b524a1216dcadd8d0c608d097a0c4c1d51b3117b75ccf7a061b7624cad6390fd5ce37494d67987
EBUILD Tensile-5.0.2-r2.ebuild 2634 BLAKE2B f1360a69a7bba99f0abacc39d3a967626512317426c527d48c119c10a3da8adf49478e59bc6b3a8a37730b2c9ab073e6d7665e4ef7f25992a6741cc869306e23 SHA512 0ac3f2512c6555adc64797ed1ec26e072fe3ebe374251ed7771b5690f7039964bbf6616cb79b0cc8f0a3eec9efc9e90f96eea0ae138738fb1a7cd3c1e36ead35
EBUILD Tensile-5.1.3-r1.ebuild 2723 BLAKE2B b9fa24c223e4065c9ca2192b8a3ea9ee0549331cbf802ce99c6000374336fd12fea0ecee84fcbe44c39ce475905e4fd8ad8870c88ffbf527d4d00c02714747f0 SHA512 da3ccf7ac8e061474850bd81f83e616bf2077d318768808e133a68a0be3b55a24828a684f6582d294aa57c769c3f250338769bfa91cac733c2a5741bcac41846
MISC metadata.xml 350 BLAKE2B f621bb03aafc000672f022db9cad9a1fed65b912129d26ed7e717ba8e5cfb11f0d6b309905b6e12b07993a7aa0cbebf18aeaf92d643eff5419d765e31b3c52a7 SHA512 99b497edddaaf6d84159df959fb8b0a5d18ea0ecc879d600a2a1fc004554612e1422b258a2f7c5f73148caca93f9283481b53669f7247d62782401fdb112f69e
diff --git a/dev-util/Tensile/Tensile-4.3.0-r1.ebuild b/dev-util/Tensile/Tensile-4.3.0-r1.ebuild
deleted file mode 100644
index 7cb7e94cd256..000000000000
--- a/dev-util/Tensile/Tensile-4.3.0-r1.ebuild
+++ /dev/null
@@ -1,86 +0,0 @@
-# Copyright 1999-2022 Gentoo Authors
-# Distributed under the terms of the GNU General Public License v2
-
-EAPI=8
-
-PYTHON_COMPAT=( python3_{8..10} )
-DISTUTILS_USE_PEP517=setuptools
-inherit distutils-r1 prefix
-
-DESCRIPTION="Stretching GPU performance for GEMMs and tensor contractions"
-HOMEPAGE="https://github.com/ROCmSoftwarePlatform/Tensile"
-SRC_URI="https://github.com/ROCmSoftwarePlatform/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz"
-S="${WORKDIR}/${PN}-rocm-${PV}"
-
-LICENSE="MIT"
-KEYWORDS="~amd64"
-SLOT="0/$(ver_cut 1-2)"
-
-# Not compatible with recent versions of pytest
-RESTRICT="test"
-
-RDEPEND="${PYTHON_DEPS}
- dev-python/pyyaml[${PYTHON_USEDEP}]
- dev-python/msgpack[${PYTHON_USEDEP}]
- >=dev-util/rocm-smi-4.3.0
-"
-DEPEND="${RDEPEND}
- dev-util/hip:${SLOT}
-"
-
-PATCHES=( "${FILESDIR}"/${PN}-4.3.0-output-commands.patch
- "${FILESDIR}"/${PN}-4.3.0-hsaco-compile-specified-arch.patch
- "${FILESDIR}"/${PN}-4.3.0-gfx1031.patch
- "${FILESDIR}"/${PN}-4.3.0-fix-arch-parse.patch
- "${FILESDIR}"/${PN}-4.3.0-use-ninja.patch
- )
-
-src_prepare() {
- distutils-r1_src_prepare
- eapply $(prefixify_ro "${FILESDIR}"/${PN}-4.3.0-gentoopath.patch)
-
- pushd ${PN} || die
-
- sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \
- -i Source/cmake/FindROCmSMI.cmake || die
- sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \
- -i Source/CMakeLists.txt || die
- sed -e "/chmod 755/d" -i Source/TensileCreateLibrary.cmake || die # remove chmod 755 on
- sed -e "s,\${Tensile_ROOT}/bin/,,g" -i Source/TensileCreateLibrary.cmake cmake/TensileConfig.cmake || die # ${Tensile_ROOT}/bin does not exists; call command directly
-
- local Tensile_share_dir="\"${EPREFIX}/usr/share/${PN}\""
- sed -e "/HipClangVersion/s/0,0,0/$(hipconfig -v)/" \
- -e "/SourcePath/s,globalParameters\[\"ScriptPath\"\],${Tensile_share_dir}," \
- -i Common.py || die
-
- sed -e "/CMAKE_CXX_COMPILER/s,globalParameters\[\"ROCmBinPath\"\],\"${EPREFIX}/usr/lib/hip/bin\"," -i ClientExecutable.py || die
-
- sed -e "/scriptDir/s,os.path.dirname(os.path.realpath(__file__)),${Tensile_share_dir}," -i ReplacementKernels.py || die
-
- sed -e "s,os.path.dirname(os.path.realpath(__file__)),${Tensile_share_dir},g" -i ${PN}.py || die
-
- sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile/Source\", end='')|" -i __init__.py || die
-
- popd || die
-
- sed -e "/package_data/d" -e "/data_files/d" -i setup.py || die
-}
-
-python_install() {
- distutils-r1_python_install
-
- python_moduleinto Tensile
- pushd Tensile || die
- python_domodule Components
- python_newexe Utilities/merge.py ${PN}-merge
-}
-
-src_install() {
- distutils-r1_src_install
-
- pushd ${PN} || die
- insinto /usr/share/${PN}
- doins -r Configs Perf ReplacementKernels ReplacementKernels-cov3 Source
- insinto /usr/$(get_libdir)/cmake/${PN}
- doins cmake/*.cmake
-}
diff --git a/dev-util/Tensile/files/Tensile-4.3.0-fix-arch-parse.patch b/dev-util/Tensile/files/Tensile-4.3.0-fix-arch-parse.patch
deleted file mode 100644
index 01b94b32cb6e..000000000000
--- a/dev-util/Tensile/files/Tensile-4.3.0-fix-arch-parse.patch
+++ /dev/null
@@ -1,12 +0,0 @@
-Index: Tensile-rocm-4.3.0/Tensile/Tensile.py
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/Tensile.py
-+++ Tensile-rocm-4.3.0/Tensile/Tensile.py
-@@ -97,7 +97,6 @@ def addCommonArguments(argParser):
- """
-
- (key, value) = par.split("=")
-- value = eval(value)
- return (key, value)
-
- argParser.add_argument("-d", "--device", dest="device", type=int, \
diff --git a/dev-util/Tensile/files/Tensile-4.3.0-gentoopath.patch b/dev-util/Tensile/files/Tensile-4.3.0-gentoopath.patch
deleted file mode 100644
index da135e3b38ee..000000000000
--- a/dev-util/Tensile/files/Tensile-4.3.0-gentoopath.patch
+++ /dev/null
@@ -1,62 +0,0 @@
-Index: Tensile-rocm-4.3.0/Tensile/Source/client/CMakeLists.txt
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/Source/client/CMakeLists.txt
-+++ Tensile-rocm-4.3.0/Tensile/Source/client/CMakeLists.txt
-@@ -50,3 +50,5 @@ endif()
- foreach(arch IN LISTS TENSILE_GPU_ARCHS)
- target_link_libraries(tensile_client PRIVATE "--amdgpu-target=${arch}")
- endforeach(arch)
-+
-+target_link_libraries(tensile_client PRIVATE LLVMSupport)
-Index: Tensile-rocm-4.3.0/Tensile/Source/lib/CMakeLists.txt
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/Source/lib/CMakeLists.txt
-+++ Tensile-rocm-4.3.0/Tensile/Source/lib/CMakeLists.txt
-@@ -48,7 +48,7 @@ set(tensile_sources ${tensile_sources}
- )
-
- if(TENSILE_USE_LLVM)
-- find_package(LLVM 13.0 QUIET CONFIG)
-+ find_package(LLVM PATHS @GENTOO_PORTAGE_EPREFIX@/usr/lib/llvm/roc QUIET CONFIG)
- if(NOT LLVM_FOUND)
- find_package(LLVM 12.0 QUIET CONFIG)
- if(NOT LLVM_FOUND)
-Index: Tensile-rocm-4.3.0/Tensile/Common.py
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/Common.py
-+++ Tensile-rocm-4.3.0/Tensile/Common.py
-@@ -1667,7 +1667,7 @@ def assignGlobalParameters( config ):
- else:
- print2(" %24s: %8s (unspecified)" % (key, defaultValue))
-
-- globalParameters["ROCmPath"] = "/opt/rocm"
-+ globalParameters["ROCmPath"] = "@GENTOO_PORTAGE_EPREFIX@/usr"
- if "ROCM_PATH" in os.environ:
- globalParameters["ROCmPath"] = os.environ.get("ROCM_PATH")
- if "TENSILE_ROCM_PATH" in os.environ:
-Index: Tensile-rocm-4.3.0/Tensile/ClientWriter.py
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/ClientWriter.py
-+++ Tensile-rocm-4.3.0/Tensile/ClientWriter.py
-@@ -284,7 +284,7 @@ def getBuildNewClientLibraryScript(build
- import io
- runScriptFile = io.StringIO()
-
-- callCreateLibraryCmd = globalParameters["ScriptPath"] + "/bin/TensileCreateLibrary"
-+ callCreateLibraryCmd = "TensileCreateLibrary"
-
-
- if globalParameters["MergeFiles"]:
-Index: Tensile-rocm-4.3.0/Tensile/GenerateSummations.py
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/GenerateSummations.py
-+++ Tensile-rocm-4.3.0/Tensile/GenerateSummations.py
-@@ -57,7 +57,7 @@ def createLibraryForBenchmark(logicPath,
- Selection.
- """
-
-- pythonExePath = os.path.join(os.path.dirname(os.path.realpath(__file__)), "bin", "TensileCreateLibrary")
-+ pythonExePath = "TensileCreateLibrary"
- args = [pythonExePath, \
- "--merge-files", "--no-legacy-components", \
- "--new-client-only", "--no-short-file-names", "--no-library-print-debug", "--architecture=all", \
diff --git a/dev-util/Tensile/files/Tensile-4.3.0-gfx1031.patch b/dev-util/Tensile/files/Tensile-4.3.0-gfx1031.patch
deleted file mode 100644
index 99abceb7c44d..000000000000
--- a/dev-util/Tensile/files/Tensile-4.3.0-gfx1031.patch
+++ /dev/null
@@ -1,119 +0,0 @@
---- Tensile-rocm-4.3.0/pytest.ini 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/pytest.ini 2021-09-30 17:32:21.382806851 +0800
-@@ -88,6 +88,7 @@
- xfail-gfx1011: architecture
- xfail-gfx1012: architecture
- xfail-gfx1030: architecture
-+ xfail-gfx1031: architecture
- skip-gfx000: architecture
- skip-gfx900: architecture
- skip-gfx906: architecture
-@@ -97,3 +98,4 @@
- skip-gfx1011: architecture
- skip-gfx1012: architecture
- skip-gfx1030: architecture
-+ skip-gfx1031: architecture
---- Tensile-rocm-4.3.0/Tensile/Common.py 2021-10-05 14:41:22.490754124 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Common.py 2021-09-30 17:33:19.041724015 +0800
-@@ -200,7 +200,7 @@
- globalParameters["MergeFiles"] = True # F=store every solution and kernel in separate file; T=store all solutions in single file
-
- globalParameters["MaxFileName"] = 64 # If a file name would be longer than this, shorten it with a hash.
-+globalParameters["SupportedISA"] = [(8,0,3), (9,0,0), (9,0,6), (9,0,8), (9,0,10), (10,1,0), (10,1,1), (10,1,2), (10,3,0),(10,3,1)] # assembly kernels writer supports these architectures
--globalParameters["SupportedISA"] = [(8,0,3), (9,0,0), (9,0,6), (9,0,8), (9,0,10), (10,1,0), (10,1,1), (10,1,2), (10,3,0)] # assembly kernels writer supports these architectures
-
- globalParameters["GenerateManifestAndExit"] = False # Output manifest file with list of expected library objects and exit
- globalParameters["ClientBuildPath"] = "0_Build" # subdirectory for host code build directory
-@@ -265,7 +265,7 @@
- 'gfx906':'vega20', 'gfx906:xnack+':'vega20', 'gfx906:xnack-':'vega20',
- 'gfx908':'arcturus','gfx908:xnack+':'arcturus', 'gfx908:xnack-':'arcturus',
- 'gfx90a':'aldebaran', 'gfx90a:xnack+':'aldebaran', 'gfx90a:xnack-':'aldebaran',
-+ 'gfx1010':'navi10', 'gfx1011':'navi11', 'gfx1012':'navi12', 'gfx1030':'navi21', 'gfx1031':'navi22',
-- 'gfx1010':'navi10', 'gfx1011':'navi11', 'gfx1012':'navi12', 'gfx1030':'navi21'
- }
-
- def getArchitectureName(gfxName):
---- Tensile-rocm-4.3.0/Tensile/KernelWriterAssembly.py 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/KernelWriterAssembly.py 2021-09-30 17:27:06.696711144 +0800
-@@ -2257,7 +2257,7 @@
-
- # integer i8x4
- elif kernel["ProblemType"]["DataType"].isInt8x4():
-+ if self.version == (9,0,6) or self.version == (9,0,8) or self.version == (9,0,10) or self.version == (10,3,0) or self.version == (10,3,1):
-- if self.version == (9,0,6) or self.version == (9,0,8) or self.version == (9,0,10) or self.version == (10,3,0):
- for b in range(0, kernel["ThreadTile1"]):
- for a in range(0, kernel["ThreadTile0"]):
- for iui in range(0, innerUnroll):
---- Tensile-rocm-4.3.0/Tensile/Source/CMakeLists.txt 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Source/CMakeLists.txt 2021-09-30 17:29:12.846345541 +0800
-@@ -51,7 +51,7 @@
- endif()
-
- if(CMAKE_CXX_COMPILER STREQUAL "hipcc")
-+ set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 gfx1031 CACHE STRING "GPU architectures")
-- set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 CACHE STRING "GPU architectures")
- else()
- set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906 gfx908 gfx90a CACHE STRING "GPU architectures")
- endif()
-@@ -104,7 +104,7 @@
- set_property( CACHE Tensile_CODE_OBJECT_VERSION PROPERTY STRINGS V2 V3)
- set_property( CACHE Tensile_COMPILER PROPERTY STRINGS hipcc)
- set(Tensile_ARCHITECTURE all CACHE STRING "Which GPU architecture to use")
-+ set_property( CACHE Tensile_ARCHITECTURE PROPERTY STRINGS all gfx000 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 gfx1031)
-- set_property( CACHE Tensile_ARCHITECTURE PROPERTY STRINGS all gfx000 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030)
- set_property( CACHE Tensile_LIBRARY_FORMAT PROPERTY STRINGS yaml msgpack)
-
- message( STATUS "Tensile_CODE_OBJECT_VERSION from Tensile/Source/CMakeLists.txt: ${Tensile_CODE_OBJECT_VERSION}")
-@@ -206,7 +206,7 @@
- ${Tensile_RUNTIME_LANGUAGE} # HIP or OCL
- ${Tensile_COMPILER} # hipcc
- ${Tensile_CODE_OBJECT_VERSION} # V2 or V3
-+ ${Tensile_ARCHITECTURE} # all, gfx000, gfx803, gfx900, gfx906:xnack-, gfx908:xnack-, gfx90a:xnack-, gfx1010, gfx1011, gfx1012, gfx1030, gfx1031
-- ${Tensile_ARCHITECTURE} # all, gfx000, gfx803, gfx900, gfx906:xnack-, gfx908:xnack-, gfx90a:xnack-, gfx1010, gfx1011, gfx1012, gfx1030
- ${Tensile_LIBRARY_FORMAT} # yaml or msgpack
- ${Tensile_MERGE_FILES} # ON or OFF
- ${Tensile_SHORT_FILE_NAMES} # ON or OFF
---- Tensile-rocm-4.3.0/Tensile/Source/lib/include/Tensile/AMDGPU.hpp 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Source/lib/include/Tensile/AMDGPU.hpp 2021-09-30 17:48:11.277066515 +0800
-@@ -62,7 +62,8 @@
- gfx1010 = 1010,
- gfx1011 = 1011,
- gfx1012 = 1012,
-+ gfx1030 = 1030,
-+ gfx1031 = 1031
-- gfx1030 = 1030
- };
-
- AMDGPU();
---- Tensile-rocm-4.3.0/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp 2021-09-30 17:31:26.075845249 +0800
-@@ -215,6 +215,7 @@
- iot::enumCase(io, value, "gfx1011", AMDGPU::Processor::gfx1011);
- iot::enumCase(io, value, "gfx1012", AMDGPU::Processor::gfx1012);
- iot::enumCase(io, value, "gfx1030", AMDGPU::Processor::gfx1030);
-+ iot::enumCase(io, value, "gfx1031", AMDGPU::Processor::gfx1031);
- }
- };
- } // namespace Serialization
---- Tensile-rocm-4.3.0/Tensile/Source/lib/source/AMDGPU.cpp 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Source/lib/source/AMDGPU.cpp 2021-09-30 17:31:45.195486318 +0800
-@@ -82,6 +82,8 @@
- return stream << "gfx1012";
- case AMDGPU::Processor::gfx1030:
- return stream << "gfx1030";
-+ case AMDGPU::Processor::gfx1031:
-+ return stream << "gfx1031";
- }
- return stream;
- }
---- Tensile-rocm-4.3.0/Tensile/Source/TensileCreateLibrary.cmake 2021-04-22 22:35:33.000000000 +0800
-+++ Tensile-rocm-4.3.0/Tensile/Source/TensileCreateLibrary.cmake 2021-09-30 17:29:24.766121912 +0800
-@@ -138,7 +138,7 @@
- add_library(Tensile ${options} ${Tensile_SOURCE_FILES})
- # specify gpu targets
- if( Tensile_ARCHITECTURE MATCHES "all" )
-+ set( Tensile_HIP_ISA "gfx803" "gfx900" "gfx906" "gfx908" "gfx1010" "gfx1011" "gfx1012" "gfx1030", "gfx1031")
-- set( Tensile_HIP_ISA "gfx803" "gfx900" "gfx906" "gfx908" "gfx1010" "gfx1011" "gfx1012" "gfx1030")
- else()
- set( Tensile_HIP_ISA ${Tensile_ARCHITECTURE})
- endif()
diff --git a/dev-util/Tensile/files/Tensile-4.3.0-hsaco-compile-specified-arch.patch b/dev-util/Tensile/files/Tensile-4.3.0-hsaco-compile-specified-arch.patch
deleted file mode 100644
index 8e6753781a2a..000000000000
--- a/dev-util/Tensile/files/Tensile-4.3.0-hsaco-compile-specified-arch.patch
+++ /dev/null
@@ -1,96 +0,0 @@
-https://github.com/ROCmSoftwarePlatform/Tensile/issues/1395
-https://github.com/ROCmSoftwarePlatform/Tensile/pull/1398
-
---- a/Tensile/TensileCreateLibrary.py
-+++ b/Tensile/TensileCreateLibrary.py
-@@ -136,6 +136,35 @@ def which(p):
- return candidate
- return None
-
-+def splitArchs():
-+ # Helper for architecture
-+ def isSupported(arch):
-+ return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
-+ globalParameters["AsmCaps"][arch]["SupportedSource"]
-+
-+ if ";" in globalParameters["Architecture"]:
-+ wantedArchs = globalParameters["Architecture"].split(";")
-+ else:
-+ wantedArchs = globalParameters["Architecture"].split("_")
-+ archs = []
-+ cmdlineArchs = []
-+ if "all" in wantedArchs:
-+ for arch in globalParameters['SupportedISA']:
-+ if isSupported(arch):
-+ if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
-+ if (arch == (9,0,10)):
-+ archs += [gfxName(arch) + '-xnack+']
-+ cmdlineArchs += [gfxName(arch) + ':xnack+']
-+ archs += [gfxName(arch) + '-xnack-']
-+ cmdlineArchs += [gfxName(arch) + ':xnack-']
-+ else:
-+ archs += [gfxName(arch)]
-+ cmdlineArchs += [gfxName(arch)]
-+ else:
-+ for arch in wantedArchs:
-+ archs += [re.sub(":", "-", arch)]
-+ cmdlineArchs += [arch]
-+ return archs, cmdlineArchs
-
- def buildSourceCodeObjectFile(CxxCompiler, outputPath, kernelFile):
- buildPath = ensurePath(os.path.join(globalParameters['WorkingPath'], 'code_object_tmp'))
-@@ -149,24 +178,8 @@ def buildSourceCodeObjectFile(CxxCompiler, outputPath, kernelFile):
- objectFilename = base + '.o'
- soFilename = base + '.so'
-
-- def isSupported(arch):
-- return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
-- globalParameters["AsmCaps"][arch]["SupportedSource"]
--
- if (CxxCompiler == "hipcc"):
-- archs = []
-- cmdlineArchs = []
-- for arch in globalParameters['SupportedISA']:
-- if isSupported(arch):
-- if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
-- if (arch == (9,0,10)):
-- archs += [gfxName(arch) + '-xnack+']
-- cmdlineArchs += [gfxName(arch) + ':xnack+']
-- archs += [gfxName(arch) + '-xnack-']
-- cmdlineArchs += [gfxName(arch) + ':xnack-']
-- else:
-- archs += [gfxName(arch)]
-- cmdlineArchs += [gfxName(arch)]
-+ archs, cmdlineArchs = splitArchs()
-
- archFlags = ['--offload-arch=' + arch for arch in cmdlineArchs]
-
-@@ -1063,11 +1076,6 @@ def buildObjectFileNames(solutionWriter, kernelWriterSource, kernelWriterAssembl
- sourceKernels = list([k for k in kernels if k['KernelLanguage'] == 'Source'])
- asmKernels = list([k for k in kernels if k['KernelLanguage'] == 'Assembly'])
-
-- # Helper for architecture
-- def isSupported(arch):
-- return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
-- globalParameters["AsmCaps"][arch]["SupportedSource"]
--
- # Build a list of kernel object names.
- for kernel in sourceKernels:
- sourceKernelNames += [kernelWriterSource.getKernelFileBase(kernel)]
-@@ -1081,15 +1089,7 @@ def buildObjectFileNames(solutionWriter, kernelWriterSource, kernelWriterAssembl
-
- # Source based kernels are built for all supported architectures
- if (cxxCompiler == 'hipcc'):
-- sourceArchs = []
-- for arch in globalParameters['SupportedISA']:
-- if isSupported(arch):
-- if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
-- if (arch == (9,0,10)):
-- sourceArchs += [gfxName(arch) + '-xnack+']
-- sourceArchs += [gfxName(arch) + '-xnack-']
-- else:
-- sourceArchs += [gfxName(arch)]
-+ sourceArchs, _ = splitArchs()
- else:
- raise RuntimeError("Unknown compiler %s" % cxxCompiler)
-
diff --git a/dev-util/Tensile/files/Tensile-4.3.0-use-ninja.patch b/dev-util/Tensile/files/Tensile-4.3.0-use-ninja.patch
deleted file mode 100644
index 6645efd72fb1..000000000000
--- a/dev-util/Tensile/files/Tensile-4.3.0-use-ninja.patch
+++ /dev/null
@@ -1,21 +0,0 @@
-Index: Tensile-rocm-4.3.0/Tensile/ClientExecutable.py
-===================================================================
---- Tensile-rocm-4.3.0.orig/Tensile/ClientExecutable.py
-+++ Tensile-rocm-4.3.0/Tensile/ClientExecutable.py
-@@ -36,6 +36,7 @@ class CMakeEnvironment:
-
- args = ['cmake']
- args += itertools.chain.from_iterable([ ['-D', '{}={}'.format(key, value)] for key,value in self.options.items()])
-+ args += itertools.chain.from_iterable([ ['-G', 'Ninja'], [ '-D', 'CMAKE_EXPORT_COMPILE_COMMANDS=ON' ] ])
- args += [self.sourceDir]
-
- Common.print2(' '.join(args))
-@@ -43,7 +44,7 @@ class CMakeEnvironment:
- subprocess.check_call(args, cwd=Common.ensurePath(self.buildDir))
-
- def build(self):
-- args = ['make', '-j']
-+ args = ['ninja']
- Common.print2(' '.join(args))
- with Common.ClientExecutionLock():
- subprocess.check_call(args, cwd=self.buildDir)