Gentoo Archives: gentoo-commits

From: Andrew Ammerlaan <andrewammerlaan@g.o>
To: gentoo-commits@l.g.o
Subject: [gentoo-commits] proj/sci:master commit in: dev-util/Tensile/files/, dev-util/Tensile/
Date: Sat, 05 Mar 2022 20:52:56
Message-Id: 1646513446.aa2ea880a86367c5a2eb8b1baa412a179f209b56.andrewammerlaan@gentoo
1 commit: aa2ea880a86367c5a2eb8b1baa412a179f209b56
2 Author: Andrew Ammerlaan <andrewammerlaan <AT> gentoo <DOT> org>
3 AuthorDate: Sat Mar 5 20:50:46 2022 +0000
4 Commit: Andrew Ammerlaan <andrewammerlaan <AT> gentoo <DOT> org>
5 CommitDate: Sat Mar 5 20:50:46 2022 +0000
6 URL: https://gitweb.gentoo.org/proj/sci.git/commit/?id=aa2ea880
7
8 dev-util/Tensile: moved to ::gentoo
9
10 Signed-off-by: Andrew Ammerlaan <andrewammerlaan <AT> gentoo.org>
11
12 dev-util/Tensile/Tensile-4.0.0-r1.ebuild | 62 --------------
13 dev-util/Tensile/Tensile-4.0.0.ebuild | 62 --------------
14 dev-util/Tensile/Tensile-4.3.0.ebuild | 74 -----------------
15 dev-util/Tensile/files/Tensile-4.0.0-cmake.patch | 38 ---------
16 .../files/Tensile-4.0.0-locate-commands.patch | 62 --------------
17 .../files/Tensile-4.0.0-output-currentISA.patch | 22 -----
18 .../files/Tensile-4.0.0-setup.py-cmake.patch | 13 ---
19 ...ensile-4.3.0-hsaco-compile-specified-arch.patch | 96 ----------------------
20 .../files/Tensile-4.3.0-output-commands.patch | 23 ------
21 dev-util/Tensile/metadata.xml | 11 ---
22 10 files changed, 463 deletions(-)
23
24 diff --git a/dev-util/Tensile/Tensile-4.0.0-r1.ebuild b/dev-util/Tensile/Tensile-4.0.0-r1.ebuild
25 deleted file mode 100644
26 index ec2a738d9..000000000
27 --- a/dev-util/Tensile/Tensile-4.0.0-r1.ebuild
28 +++ /dev/null
29 @@ -1,62 +0,0 @@
30 -# Copyright 1999-2021 Gentoo Authors
31 -# Distributed under the terms of the GNU General Public License v2
32 -
33 -EAPI=7
34 -
35 -PYTHON_COMPAT=( python3_{7,8,9} )
36 -DISTUTILS_USE_SETUPTOOLS=rdepend
37 -
38 -inherit distutils-r1
39 -
40 -DESCRIPTION="Stretching GPU performance for GEMMs and tensor contractions"
41 -HOMEPAGE="https://github.com/ROCmSoftwarePlatform/Tensile"
42 -SRC_URI="https://github.com/ROCmSoftwarePlatform/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz"
43 -
44 -LICENSE="MIT"
45 -KEYWORDS="~amd64"
46 -SLOT="0"
47 -IUSE=""
48 -
49 -RDEPEND="${PYTHON_DEPS}
50 - dev-python/pyyaml[${PYTHON_USEDEP}]
51 - dev-python/msgpack[${PYTHON_USEDEP}]"
52 -DEPEND="${RDEPEND}
53 - dev-util/hip"
54 -
55 -PATCHES=( "${FILESDIR}"/${PN}-4.0.0-cmake.patch
56 - "${FILESDIR}"/${PN}-4.0.0-setup.py-cmake.patch
57 - "${FILESDIR}"/${PN}-4.0.0-locate-commands.patch
58 - "${FILESDIR}"/${PN}-4.0.0-output-currentISA.patch )
59 -
60 -S="${WORKDIR}/${PN}-rocm-${PV}"
61 -CMAKE_USE_DIR="${WORKDIR}/Source"
62 -
63 -src_prepare() {
64 - distutils-r1_src_prepare
65 -
66 - mv ${PN}/Source "${WORKDIR}"/ || die
67 - sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \
68 - -i "${WORKDIR}"/Source/cmake/FindROCmSMI.cmake || die
69 - sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \
70 - -i "${WORKDIR}"/Source/CMakeLists.txt || die
71 -
72 - mv ${PN}/cmake "${T}"/ || die
73 -
74 - sed -e "/HipClangVersion/s/0,0,0/$(ver_rs 1-3 ,)/" \
75 - -e "/SourcePath/s,os\.path\.join.*$,\"${EPREFIX}/usr/share/${PN}\"," \
76 - -i ${PN}/Common.py || die
77 -
78 - sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile\", end='')|" \
79 - -i ${PN}/__init__.py || die
80 -}
81 -
82 -src_install() {
83 - distutils-r1_src_install
84 -
85 - insinto /usr/$(get_libdir)/cmake/${PN}
86 - doins "${T}"/cmake/*.cmake
87 -
88 - insinto /usr/share/${PN}
89 - doins -r "${WORKDIR}"/Source/*
90 - dosym . /usr/share/${PN}/Source
91 -}
92
93 diff --git a/dev-util/Tensile/Tensile-4.0.0.ebuild b/dev-util/Tensile/Tensile-4.0.0.ebuild
94 deleted file mode 100644
95 index eda97d225..000000000
96 --- a/dev-util/Tensile/Tensile-4.0.0.ebuild
97 +++ /dev/null
98 @@ -1,62 +0,0 @@
99 -# Copyright 1999-2021 Gentoo Authors
100 -# Distributed under the terms of the GNU General Public License v2
101 -
102 -EAPI=7
103 -
104 -PYTHON_COMPAT=( python3_{7,8,9} )
105 -DISTUTILS_USE_SETUPTOOLS=rdepend
106 -
107 -inherit distutils-r1
108 -
109 -DESCRIPTION="Stretching GPU performance for GEMMs and tensor contractions"
110 -HOMEPAGE="https://github.com/ROCmSoftwarePlatform/Tensile"
111 -SRC_URI="https://github.com/ROCmSoftwarePlatform/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz"
112 -
113 -LICENSE="MIT"
114 -KEYWORDS="~amd64"
115 -SLOT="0"
116 -IUSE=""
117 -
118 -RDEPEND="${PYTHON_DEPS}
119 - dev-python/pyyaml[${PYTHON_USEDEP}]
120 - dev-python/msgpack[${PYTHON_USEDEP}]"
121 -DEPEND="${RDEPEND}
122 - dev-util/hip"
123 -
124 -PATCHES=( "${FILESDIR}"/${PN}-4.0.0-cmake.patch
125 - "${FILESDIR}"/${PN}-4.0.0-setup.py-cmake.patch
126 - "${FILESDIR}"/${PN}-4.0.0-locate-commands.patch
127 - "${FILESDIR}"/${PN}-4.0.0-output-currentISA.patch )
128 -
129 -S="${WORKDIR}/${PN}-rocm-${PVR}"
130 -CMAKE_USE_DIR="${WORKDIR}/Source"
131 -
132 -src_prepare() {
133 - distutils-r1_src_prepare
134 -
135 - mv ${PN}/Source "${WORKDIR}"/ || die
136 - sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \
137 - -i "${WORKDIR}"/Source/cmake/FindROCmSMI.cmake || die
138 - sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \
139 - -i "${WORKDIR}"/Source/CMakeLists.txt || die
140 -
141 - mv ${PN}/cmake "${T}"/ || die
142 -
143 - sed -e "/HipClangVersion/s/0,0,0/$(ver_rs 1-3 ,)/" \
144 - -e "/SourcePath/s,os\.path\.join.*$,\"${EPREFIX}/usr/share/${PN}\"," \
145 - -i ${PN}/Common.py || die
146 -
147 - sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile\", end='')|" \
148 - -i ${PN}/__init__.py || die
149 -}
150 -
151 -src_install() {
152 - distutils-r1_src_install
153 -
154 - insinto /usr/$(get_libdir)/cmake/${PN}
155 - doins "${T}"/cmake/*.cmake
156 -
157 - insinto /usr/share/${PN}
158 - doins -r "${WORKDIR}"/Source/*
159 - dosym . /usr/share/${PN}/Source
160 -}
161
162 diff --git a/dev-util/Tensile/Tensile-4.3.0.ebuild b/dev-util/Tensile/Tensile-4.3.0.ebuild
163 deleted file mode 100644
164 index cf33fb19c..000000000
165 --- a/dev-util/Tensile/Tensile-4.3.0.ebuild
166 +++ /dev/null
167 @@ -1,74 +0,0 @@
168 -# Copyright 1999-2022 Gentoo Authors
169 -# Distributed under the terms of the GNU General Public License v2
170 -
171 -EAPI=8
172 -
173 -PYTHON_COMPAT=( python3_{8..10} )
174 -DISTUTILS_USE_PEP517=setuptools
175 -inherit distutils-r1
176 -
177 -DESCRIPTION="Stretching GPU performance for GEMMs and tensor contractions"
178 -HOMEPAGE="https://github.com/ROCmSoftwarePlatform/Tensile"
179 -SRC_URI="https://github.com/ROCmSoftwarePlatform/Tensile/archive/rocm-${PV}.tar.gz -> rocm-Tensile-${PV}.tar.gz"
180 -S="${WORKDIR}/${PN}-rocm-${PV}"
181 -
182 -LICENSE="MIT"
183 -KEYWORDS="~amd64"
184 -SLOT="0/$(ver_cut 1-2)"
185 -
186 -# Not compatible with recent versions of pytest
187 -RESTRICT="test"
188 -
189 -RDEPEND="${PYTHON_DEPS}
190 - dev-python/pyyaml[${PYTHON_USEDEP}]
191 - dev-python/msgpack[${PYTHON_USEDEP}]
192 -"
193 -DEPEND="${RDEPEND}
194 - dev-util/hip:${SLOT}
195 -"
196 -BDEPEND="test? (
197 - dev-util/rocminfo:${SLOT}
198 -)"
199 -
200 -PATCHES=(
201 - "${FILESDIR}/Tensile-${PV}-hsaco-compile-specified-arch.patch" # backported from upstream, should remove after 4.3.0
202 - "${FILESDIR}/Tensile-4.3.0-output-commands.patch"
203 -)
204 -
205 -CMAKE_USE_DIR="${WORKDIR}/Source"
206 -
207 -distutils_enable_tests pytest
208 -
209 -src_prepare() {
210 - distutils-r1_src_prepare
211 -
212 - mv ${PN}/Source "${WORKDIR}"/ || die
213 - sed -e "/ROCM_SMI_ROOT/s,lib,$(get_libdir)," \
214 - -i "${WORKDIR}"/Source/cmake/FindROCmSMI.cmake || die
215 - sed -r -e "/TENSILE_USE_LLVM/s/ON/OFF/" \
216 - -i "${WORKDIR}"/Source/CMakeLists.txt || die
217 -
218 - sed -e "/HipClangVersion/s/0,0,0/$(ver_rs 1-3 ,)/" \
219 - -e "/SourcePath/s,os\.path\.join.*$,\"${EPREFIX}/usr/share/${PN}\"," \
220 - -i ${PN}/Common.py || die
221 -
222 - sed -e "s|os\.path\.dirname.*$|\"${EPREFIX}/usr/share/Tensile\", end='')|" \
223 - -i ${PN}/__init__.py || die
224 -}
225 -
226 -src_test() {
227 - ROCM_PATH="${EPREFIX}/usr/" distutils-r1_src_test
228 -}
229 -
230 -src_install() {
231 - distutils-r1_src_install
232 -
233 - # Move the cmake files to the correct directory
234 - mkdir -p "${ED}/usr/$(get_libdir)/cmake/${PN}" || die
235 - mv "${ED}/usr/cmake/"* "${ED}/usr/$(get_libdir)/cmake/${PN}" || die
236 - rm -r "${ED}/usr/cmake" || die
237 -
238 - insinto /usr/share/${PN}
239 - doins -r "${WORKDIR}"/Source/*
240 - dosym . /usr/share/${PN}/Source
241 -}
242
243 diff --git a/dev-util/Tensile/files/Tensile-4.0.0-cmake.patch b/dev-util/Tensile/files/Tensile-4.0.0-cmake.patch
244 deleted file mode 100644
245 index 0c39f747c..000000000
246 --- a/dev-util/Tensile/files/Tensile-4.0.0-cmake.patch
247 +++ /dev/null
248 @@ -1,38 +0,0 @@
249 -Index: Tensile-rocm-4.0.0/Tensile/cmake/TensileConfig.cmake
250 -===================================================================
251 ---- Tensile-rocm-4.0.0.orig/Tensile/cmake/TensileConfig.cmake
252 -+++ Tensile-rocm-4.0.0/Tensile/cmake/TensileConfig.cmake
253 -@@ -26,7 +26,7 @@ if(NOT DEFINED Tensile_ROOT)
254 - get_filename_component(Tensile_PREFIX "${CMAKE_CURRENT_LIST_FILE}" PATH)
255 - get_filename_component(Tensile_PREFIX "${Tensile_PREFIX}" PATH)
256 -
257 --execute_process(COMMAND "${Tensile_PREFIX}/bin/TensileGetPath" OUTPUT_VARIABLE Tensile_ROOT)
258 -+execute_process(COMMAND "TensileGetPath" OUTPUT_VARIABLE Tensile_ROOT)
259 - endif()
260 - list(APPEND CMAKE_MODULE_PATH "${Tensile_ROOT}/Source/cmake/")
261 - list(APPEND CMAKE_MODULE_PATH "${Tensile_ROOT}/Source/")
262 -@@ -93,7 +93,7 @@ function(TensileCreateLibraryFiles
263 - set(Tensile_MERGE_FILES OFF)
264 - endif()
265 -
266 -- set(Script "${Tensile_ROOT}/bin/TensileCreateLibrary")
267 -+ set(Script "TensileCreateLibrary")
268 - message(STATUS "Tensile script: ${Script}")
269 -
270 - set(Options "--new-client-only" "--no-legacy-components")
271 -Index: Tensile-rocm-4.0.0/Tensile/Source/TensileCreateLibrary.cmake
272 -===================================================================
273 ---- Tensile-rocm-4.0.0.orig/Tensile/Source/TensileCreateLibrary.cmake
274 -+++ Tensile-rocm-4.0.0/Tensile/Source/TensileCreateLibrary.cmake
275 -@@ -52,10 +52,7 @@ function(TensileCreateLibraryCmake
276 - message(STATUS "Tensile_ARCHITECTURE from TensileCreateLibraryCmake : ${Tensile_ARCHITECTURE}")
277 - message(STATUS "Tensile_LIBRARY_FORMAT from TensileCreateLibraryCmake : ${Tensile_LIBRARY_FORMAT}")
278 -
279 -- execute_process(COMMAND chmod 755 ${Tensile_ROOT}/bin/TensileCreateLibrary)
280 -- execute_process(COMMAND chmod 755 ${Tensile_ROOT}/bin/Tensile)
281 --
282 -- set(Tensile_CREATE_COMMAND "${Tensile_ROOT}/bin/TensileCreateLibrary")
283 -+ set(Tensile_CREATE_COMMAND "TensileCreateLibrary")
284 -
285 - set(Tensile_SOURCE_PATH "${PROJECT_BINARY_DIR}/Tensile")
286 - message(STATUS "Tensile_SOURCE_PATH=${Tensile_SOURCE_PATH}")
287
288 diff --git a/dev-util/Tensile/files/Tensile-4.0.0-locate-commands.patch b/dev-util/Tensile/files/Tensile-4.0.0-locate-commands.patch
289 deleted file mode 100644
290 index 78d467349..000000000
291 --- a/dev-util/Tensile/files/Tensile-4.0.0-locate-commands.patch
292 +++ /dev/null
293 @@ -1,62 +0,0 @@
294 -Index: Tensile-rocm-4.0.0/Tensile/Common.py
295 -===================================================================
296 ---- Tensile-rocm-4.0.0.orig/Tensile/Common.py
297 -+++ Tensile-rocm-4.0.0/Tensile/Common.py
298 -@@ -155,7 +155,7 @@ globalParameters["PrintTensorRef"] = 0
299 - globalParameters["PrintIndexAssignments"] = 0 # Print the tensor index assignment info
300 - globalParameters["PrintTensorRef"] = 0 # Print reference tensor. 0x1=after init; 0x2=after copy-back; 0x3=both
301 - globalParameters["PrintWinnersOnly"] = False # Only print the solutions which become the fastest
302 --globalParameters["PrintCodeCommands"] = False # print the commands used to generate the code objects (asm,link,hip-clang, etc)
303 -+globalParameters["PrintCodeCommands"] = True # print the commands used to generate the code objects (asm,link,hip-clang, etc)
304 -
305 - # TODO - remove this when NewClient is mainstream
306 - globalParameters["OldClientSourceTmp"] = True # Use an intermediate sourceTmp dir to detect file changes and minimize rebuilds on old client
307 -@@ -1521,14 +1521,14 @@ def assignGlobalParameters( config ):
308 - print2(" %24s: %8s (unspecified)" % (key, defaultValue))
309 -
310 - # ROCm Agent Enumerator Path
311 -- globalParameters["ROCmAgentEnumeratorPath"] = locateExe("/opt/rocm/bin", "rocm_agent_enumerator")
312 -+ globalParameters["ROCmAgentEnumeratorPath"] = locateExe("", "rocm_agent_enumerator")
313 - if "CxxCompiler" in config:
314 - globalParameters["CxxCompiler"] = config["CxxCompiler"]
315 -
316 - if "TENSILE_ROCM_ASSEMBLER_PATH" in os.environ:
317 - globalParameters["AssemblerPath"] = os.environ.get("TENSILE_ROCM_ASSEMBLER_PATH")
318 - elif globalParameters["AssemblerPath"] is None and globalParameters["CxxCompiler"] == "hipcc":
319 -- globalParameters["AssemblerPath"] = locateExe("/opt/rocm/llvm/bin", "clang++")
320 -+ globalParameters["AssemblerPath"] = locateExe("", "clang++")
321 - elif globalParameters["AssemblerPath"] is None and globalParameters["CxxCompiler"] == "hcc":
322 - globalParameters["AssemblerPath"] = locateExe("/opt/rocm/bin", "hcc")
323 -
324 -@@ -1536,8 +1536,8 @@ def assignGlobalParameters( config ):
325 - if globalParameters["CxxCompiler"] == "hcc":
326 - globalParameters["ExtractKernelPath"] = locateExe("/opt/rocm/bin", "extractkernel")
327 - else:
328 -- globalParameters["ExtractKernelPath"] = locateExe("/opt/rocm/hip/bin", "extractkernel")
329 -- globalParameters["ClangOffloadBundlerPath"] = locateExe("/opt/rocm/llvm/bin", "clang-offload-bundler")
330 -+ globalParameters["ExtractKernelPath"] = locateExe("", "extractkernel")
331 -+ globalParameters["ClangOffloadBundlerPath"] = locateExe("", "clang-offload-bundler")
332 -
333 - if "ROCmAgentEnumeratorPath" in config:
334 - globalParameters["ROCmAgentEnumeratorPath"] = config["ROCmAgentEnumeratorPath"]
335 -@@ -1579,20 +1579,6 @@ def assignGlobalParameters( config ):
336 - # Due to platform.linux_distribution() being deprecated, just try to run dpkg regardless.
337 - # The alternative would be to install the `distro` package.
338 - # See https://docs.python.org/3.7/library/platform.html#platform.linux_distribution
339 -- try:
340 -- if globalParameters["CxxCompiler"] == "hipcc":
341 -- output = subprocess.run(["dpkg", "-l", "hip-rocclr"], check=True, stdout=subprocess.PIPE).stdout.decode()
342 -- elif globalParameters["CxxCompiler"] == "hcc":
343 -- output = subprocess.run(["dpkg", "-l", "hcc"], check=True, stdout=subprocess.PIPE).stdout.decode()
344 --
345 -- for line in output.split('\n'):
346 -- if 'hipcc' in line:
347 -- globalParameters['HipClangVersion'] = line.split()[2]
348 -- elif 'hcc' in line:
349 -- globalParameters['HccVersion'] = line.split()[2]
350 --
351 -- except (subprocess.CalledProcessError, OSError) as e:
352 -- printWarning("Error: {} looking for package {}: {}".format('dpkg', 'hip-rocclr', e))
353 -
354 - for key in config:
355 - value = config[key]
356
357 diff --git a/dev-util/Tensile/files/Tensile-4.0.0-output-currentISA.patch b/dev-util/Tensile/files/Tensile-4.0.0-output-currentISA.patch
358 deleted file mode 100644
359 index a00f9fdca..000000000
360 --- a/dev-util/Tensile/files/Tensile-4.0.0-output-currentISA.patch
361 +++ /dev/null
362 @@ -1,22 +0,0 @@
363 -Index: Tensile-rocm-4.0.0/Tensile/TensileCreateLibrary.py
364 -===================================================================
365 ---- Tensile-rocm-4.0.0.orig/Tensile/TensileCreateLibrary.py
366 -+++ Tensile-rocm-4.0.0/Tensile/TensileCreateLibrary.py
367 -@@ -132,7 +132,7 @@ def buildSourceCodeObjectFile(CxxCompile
368 - return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
369 - globalParameters["AsmCaps"][arch]["SupportedSource"]
370 -
371 -- archs = ['gfx'+''.join(map(str,arch)) for arch in globalParameters['SupportedISA'] \
372 -+ archs = ['gfx'+''.join(map(str,arch)) for arch in [globalParameters['CurrentISA']] \
373 - if isSupported(arch)]
374 -
375 - archFlags = ['--amdgpu-target=' + arch for arch in archs]
376 -@@ -1004,7 +1004,7 @@ def buildObjectFileNames(solutionWriter,
377 - kernelHelperOjbNmaes = [ko.getKernelName() for ko in kernelHelperOjbs]
378 -
379 - # Source based kernels are built for all supported architectures
380 -- sourceArchs = ['gfx'+''.join(map(str,arch)) for arch in globalParameters['SupportedISA'] \
381 -+ sourceArchs = ['gfx'+''.join(map(str,arch)) for arch in [globalParameters['CurrentISA']] \
382 - if isSupported(arch)]
383 -
384 - # Asm based kernels target the configured ISA
385
386 diff --git a/dev-util/Tensile/files/Tensile-4.0.0-setup.py-cmake.patch b/dev-util/Tensile/files/Tensile-4.0.0-setup.py-cmake.patch
387 deleted file mode 100644
388 index e020842b2..000000000
389 --- a/dev-util/Tensile/files/Tensile-4.0.0-setup.py-cmake.patch
390 +++ /dev/null
391 @@ -1,13 +0,0 @@
392 -Index: Tensile-rocm-4.0.0/setup.py
393 -===================================================================
394 ---- Tensile-rocm-4.0.0.orig/setup.py
395 -+++ Tensile-rocm-4.0.0/setup.py
396 -@@ -30,8 +30,6 @@ setup(
397 - install_requires=readRequirementsFromTxt(),
398 - python_requires='>=3.5',
399 - packages=["Tensile"],
400 -- package_data={ "Tensile": ["Tensile/cmake/*"] },
401 -- data_files=[ ("cmake", ["Tensile/cmake/TensileConfig.cmake", "Tensile/cmake/TensileConfigVersion.cmake"]) ],
402 - include_package_data=True,
403 - entry_points={"console_scripts": [
404 - # user runs a benchmark
405
406 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
407 deleted file mode 100644
408 index 8e6753781..000000000
409 --- a/dev-util/Tensile/files/Tensile-4.3.0-hsaco-compile-specified-arch.patch
410 +++ /dev/null
411 @@ -1,96 +0,0 @@
412 -https://github.com/ROCmSoftwarePlatform/Tensile/issues/1395
413 -https://github.com/ROCmSoftwarePlatform/Tensile/pull/1398
414 -
415 ---- a/Tensile/TensileCreateLibrary.py
416 -+++ b/Tensile/TensileCreateLibrary.py
417 -@@ -136,6 +136,35 @@ def which(p):
418 - return candidate
419 - return None
420 -
421 -+def splitArchs():
422 -+ # Helper for architecture
423 -+ def isSupported(arch):
424 -+ return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
425 -+ globalParameters["AsmCaps"][arch]["SupportedSource"]
426 -+
427 -+ if ";" in globalParameters["Architecture"]:
428 -+ wantedArchs = globalParameters["Architecture"].split(";")
429 -+ else:
430 -+ wantedArchs = globalParameters["Architecture"].split("_")
431 -+ archs = []
432 -+ cmdlineArchs = []
433 -+ if "all" in wantedArchs:
434 -+ for arch in globalParameters['SupportedISA']:
435 -+ if isSupported(arch):
436 -+ if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
437 -+ if (arch == (9,0,10)):
438 -+ archs += [gfxName(arch) + '-xnack+']
439 -+ cmdlineArchs += [gfxName(arch) + ':xnack+']
440 -+ archs += [gfxName(arch) + '-xnack-']
441 -+ cmdlineArchs += [gfxName(arch) + ':xnack-']
442 -+ else:
443 -+ archs += [gfxName(arch)]
444 -+ cmdlineArchs += [gfxName(arch)]
445 -+ else:
446 -+ for arch in wantedArchs:
447 -+ archs += [re.sub(":", "-", arch)]
448 -+ cmdlineArchs += [arch]
449 -+ return archs, cmdlineArchs
450 -
451 - def buildSourceCodeObjectFile(CxxCompiler, outputPath, kernelFile):
452 - buildPath = ensurePath(os.path.join(globalParameters['WorkingPath'], 'code_object_tmp'))
453 -@@ -149,24 +178,8 @@ def buildSourceCodeObjectFile(CxxCompiler, outputPath, kernelFile):
454 - objectFilename = base + '.o'
455 - soFilename = base + '.so'
456 -
457 -- def isSupported(arch):
458 -- return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
459 -- globalParameters["AsmCaps"][arch]["SupportedSource"]
460 --
461 - if (CxxCompiler == "hipcc"):
462 -- archs = []
463 -- cmdlineArchs = []
464 -- for arch in globalParameters['SupportedISA']:
465 -- if isSupported(arch):
466 -- if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
467 -- if (arch == (9,0,10)):
468 -- archs += [gfxName(arch) + '-xnack+']
469 -- cmdlineArchs += [gfxName(arch) + ':xnack+']
470 -- archs += [gfxName(arch) + '-xnack-']
471 -- cmdlineArchs += [gfxName(arch) + ':xnack-']
472 -- else:
473 -- archs += [gfxName(arch)]
474 -- cmdlineArchs += [gfxName(arch)]
475 -+ archs, cmdlineArchs = splitArchs()
476 -
477 - archFlags = ['--offload-arch=' + arch for arch in cmdlineArchs]
478 -
479 -@@ -1063,11 +1076,6 @@ def buildObjectFileNames(solutionWriter, kernelWriterSource, kernelWriterAssembl
480 - sourceKernels = list([k for k in kernels if k['KernelLanguage'] == 'Source'])
481 - asmKernels = list([k for k in kernels if k['KernelLanguage'] == 'Assembly'])
482 -
483 -- # Helper for architecture
484 -- def isSupported(arch):
485 -- return globalParameters["AsmCaps"][arch]["SupportedISA"] and \
486 -- globalParameters["AsmCaps"][arch]["SupportedSource"]
487 --
488 - # Build a list of kernel object names.
489 - for kernel in sourceKernels:
490 - sourceKernelNames += [kernelWriterSource.getKernelFileBase(kernel)]
491 -@@ -1081,15 +1089,7 @@ def buildObjectFileNames(solutionWriter, kernelWriterSource, kernelWriterAssembl
492 -
493 - # Source based kernels are built for all supported architectures
494 - if (cxxCompiler == 'hipcc'):
495 -- sourceArchs = []
496 -- for arch in globalParameters['SupportedISA']:
497 -- if isSupported(arch):
498 -- if (arch == (9,0,6) or arch == (9,0,8) or arch == (9,0,10)):
499 -- if (arch == (9,0,10)):
500 -- sourceArchs += [gfxName(arch) + '-xnack+']
501 -- sourceArchs += [gfxName(arch) + '-xnack-']
502 -- else:
503 -- sourceArchs += [gfxName(arch)]
504 -+ sourceArchs, _ = splitArchs()
505 - else:
506 - raise RuntimeError("Unknown compiler %s" % cxxCompiler)
507 -
508
509 diff --git a/dev-util/Tensile/files/Tensile-4.3.0-output-commands.patch b/dev-util/Tensile/files/Tensile-4.3.0-output-commands.patch
510 deleted file mode 100644
511 index be5a4db21..000000000
512 --- a/dev-util/Tensile/files/Tensile-4.3.0-output-commands.patch
513 +++ /dev/null
514 @@ -1,23 +0,0 @@
515 -diff --color -uprN orig/Tensile/cmake/TensileConfig.cmake Tensile-rocm-4.3.0/Tensile/cmake/TensileConfig.cmake
516 ---- orig/Tensile/cmake/TensileConfig.cmake 2021-08-18 17:48:00.115478470 +0800
517 -+++ Tensile-rocm-4.3.0/Tensile/cmake/TensileConfig.cmake 2021-08-18 17:48:49.963478801 +0800
518 -@@ -234,6 +234,7 @@ function(TensileCreateLibraryFiles
519 - COMMENT "Generating Tensile Libraries"
520 - OUTPUT ${Tensile_EMBED_LIBRARY_SOURCE};${Tensile_MANIFEST_CONTENTS}
521 - COMMAND ${CommandLine}
522 -+ USES_TERMINAL
523 - )
524 -
525 - set("${Tensile_VAR_PREFIX}_ALL_FILES" ${Tensile_MANIFEST_CONTENTS} PARENT_SCOPE)
526 -diff --color -uprN orig/Tensile/Common.py Tensile-rocm-4.3.0/Tensile/Common.py
527 ---- orig/Tensile/Common.py 2021-08-18 17:48:00.075478470 +0800
528 -+++ Tensile-rocm-4.3.0/Tensile/Common.py 2021-08-18 17:48:23.287478624 +0800
529 -@@ -179,7 +179,7 @@ globalParameters["PrintTensorD"] = 0
530 - globalParameters["PrintTensorRef"] = 0 # Print reference tensor. 0x1=after init; 0x2=after copy-back; 0x3=both
531 - globalParameters["PrintIndexAssignments"] = 0 # Print the tensor index assignment info
532 - globalParameters["PrintWinnersOnly"] = False # Only print the solutions which become the fastest
533 --globalParameters["PrintCodeCommands"] = False # print the commands used to generate the code objects (asm,link,hip-clang, etc)
534 -+globalParameters["PrintCodeCommands"] = True # print the commands used to generate the code objects (asm,link,hip-clang, etc)
535 - globalParameters["DumpTensors"] = False # If True, dump tensors to binary files instead of printing them.
536 -
537 - # TODO - remove this when NewClient is mainstream
538
539 diff --git a/dev-util/Tensile/metadata.xml b/dev-util/Tensile/metadata.xml
540 deleted file mode 100644
541 index 336bdb639..000000000
542 --- a/dev-util/Tensile/metadata.xml
543 +++ /dev/null
544 @@ -1,11 +0,0 @@
545 -<?xml version="1.0" encoding="UTF-8"?>
546 -<!DOCTYPE pkgmetadata SYSTEM 'http://www.gentoo.org/dtd/metadata.dtd'>
547 -<pkgmetadata>
548 - <maintainer type="project">
549 - <email>sci@g.o</email>
550 - </maintainer>
551 - <maintainer type="person">
552 - <email>gentoo@××××××.net</email>
553 - <name>Wilfried Holzke</name>
554 - </maintainer>
555 -</pkgmetadata>