Remove short name option

tensilelite/Tensile/BenchmarkProblems.py

@@ -37,8 +37,9 @@
 from Tensile.Toolchain.Component import Assembler
 from Tensile.SolutionStructs.Problem import ProblemType, ProblemSizes
 from Tensile.SolutionStructs.Solution import Solution
-from Tensile.SolutionStructs.Validators.MatrixInstruction import matrixInstructionToMIParameters, validateMIParameters
-from Tensile.SolutionStructs.Naming import getMinNaming, getNameMin, getSerialNaming, getNameFull, getKeyNoInternalArgs
+from Tensile.SolutionStructs.Validators.MatrixInstruction import matrixInstructionToMIParameters, \
+                                                                 validateMIParameters
+from Tensile.SolutionStructs.Naming import getKeyNoInternalArgs, getSolutionNameMin, getKernelNameMin
 
 from .BenchmarkStructs import BenchmarkProcess, constructForkPermutations
 from .Contractions import ProblemType as ContractionsProblemType
@@ -55,7 +56,6 @@
 from Tensile.Common.GlobalParameters import globalParameters, startTime
 
 
-
 def _generateForkedSolutions(problemType, constantParams, forkPermutations, assembler: Assembler, \
                             debugConfig: DebugConfig, depthUConfig: DepthUConfig, isaInfoMap: Dict[IsaVersion, IsaInfo]):
     """Creates a list with a Solution object for each parameter combination in forkPermutations"""
@@ -202,7 +202,6 @@ def writeBenchmarkFiles(
         asmToolchain: AssemblyToolchain,
         srcToolchain: SourceToolchain,
         sourcePath: Path,
-        useShortNames: bool,
         debugConfig: DebugConfig,
         depthUConfig: DepthUConfig,
         deviceId: int,
@@ -235,14 +234,7 @@ def writeBenchmarkFiles(
                 kernelHelperObjs.append(ko)
                 kernelHelperNames.add(kname)
 
-    kernelSerialNaming = getSerialNaming(kernels)
-    kernelMinNaming = getMinNaming(kernels)
-    kernelWriterAssembly = KernelWriterAssembly(
-                               kernelMinNaming,
-                               kernelSerialNaming,
-                               asmToolchain.assembler,
-                               debugConfig,
-                           )
+    kernelWriterAssembly = KernelWriterAssembly(asmToolchain.assembler, debugConfig)
 
     cmdLineArchs = [var for isa in isaInfoMap.keys() for var in gfxToVariants(isaToGfx(isa))]
     # cmdLineArchs = [variant isaToGfx(isa) for isa in isaInfoMap.keys() for gfxToVariants()]
@@ -258,14 +250,14 @@ def writeBenchmarkFiles(
                             kernelWriterAssembly,
                             debugConfig.splitGSU,
                             cmdLineArchs,
-                            kernelSerialNaming,
-                            kernelMinNaming,
                             errorTolerant=True,
                             generateSourcesAndExit=globalParameters["GenerateSourcesAndExit"], # put in debug config
                             compress=False,
-                            useShortNames=useShortNames
                         )
     # ^ this is where solutions is mutated
+    for s in solutions:
+        s["SolutionNameMin"] = getSolutionNameMin(solution, debugConfig.splitGSU)
+        s["KernelNameMin"]   = getKernelNameMin(solution, debugConfig.splitGSU)
 
     newLibraryDir = ensurePath(sourcePath / 'library')
     newLibraryFile = os.path.join(newLibraryDir, "TensileLibrary")
@@ -278,7 +270,7 @@ def writeBenchmarkFiles(
                      depthUConfig,
                      isaInfoMap,
                  )
-    newLibrary.applyNaming(debugConfig.splitGSU, kernelMinNaming)
+    newLibrary.applyNaming(debugConfig.splitGSU)
     LibraryIO.write(newLibraryFile, state(newLibrary), globalParameters["LibraryFormat"])
 
     codeObjectFiles = [os.path.relpath(f, sourcePath) \
@@ -322,7 +314,7 @@ def writeBenchmarkFiles(
 
 def _benchmarkProblemType(problemTypeConfig, problemSizeGroupConfig, problemSizeGroupIdx, useCache,
                          asmToolchain: AssemblyToolchain, srcToolchain: SourceToolchain, cCompiler: str,
-                         buildTmpPath: Path, benchmarkProblemsPath: Path, useShortNames: bool,
+                         buildTmpPath: Path, benchmarkProblemsPath: Path,
                          debugConfig: DebugConfig, depthUConfig: DepthUConfig, deviceId: int,
                          gfxName: str, isaInfoMap: Dict[str, IsaInfo]
     ):
@@ -430,7 +422,7 @@ def _benchmarkProblemType(problemTypeConfig, problemSizeGroupConfig, problemSize
                 printExit(msg)
 
             for solution in solutions:
-                print2("#    ({}:{}) {}".format(0, 0, getNameFull(solution, debugConfig.splitGSU)))
+                print2("#    ({}:{}) {}".format(0, 0, getSolutionNameMin(solution, debugConfig.splitGSU)))
             print2(HR)
 
             # write benchmarkFiles
@@ -439,7 +431,7 @@ def _benchmarkProblemType(problemTypeConfig, problemSizeGroupConfig, problemSize
                     benchmarkStep.problemSizes, benchmarkStep.biasTypeArgs, \
                     benchmarkStep.factorDimArgs, benchmarkStep.activationArgs, \
                     benchmarkStep.icacheFlushArgs, shortName, [], asmToolchain, srcToolchain, \
-                    sourcePath, useShortNames, debugConfig, depthUConfig, deviceId, gfxName, isaInfoMap)
+                    sourcePath, debugConfig, depthUConfig, deviceId, gfxName, isaInfoMap)
             # ^ this mutates solutions
 
             # write cache data
@@ -457,12 +449,11 @@ def _benchmarkProblemType(problemTypeConfig, problemSizeGroupConfig, problemSize
                     .format(len(solutions), prevCount ))
 
             # add SolutionIndex and SolutionNameMin into benchmark yaml
-            solutionMinNaming = getMinNaming(solutions)
             for i in range(0, len(solutions)):
                 solution = solutions[i]
                 solution["SolutionIndex"] = i
-                solution["SolutionNameMin"] = getNameMin(solution, solutionMinNaming, debugConfig.splitGSU)
-                solution["KernelNameMin"]   = getNameMin(solution, solutionMinNaming, debugConfig.splitGSU, True)
+                solution["SolutionNameMin"] = getSolutionNameMin(solution, debugConfig.splitGSU)
+                solution["KernelNameMin"]   = getKernelNameMin(solution, debugConfig.splitGSU)
         else:
             solutions = None
             print1("# Using cached solution data")
@@ -512,7 +503,6 @@ def main(
     cCompiler: str,
     outputPath: Path,
     buildTmpPath: Path,
-    useShortNames: bool,
     debugConfig: DebugConfig,
     depthUConfig: DepthUConfig,
     deviceId: int,
@@ -567,7 +557,6 @@ def main(
                             cCompiler,
                             buildTmpPath,
                             benchmarkProblemsPath,
-                            useShortNames,
                             debugConfig,
                             depthUConfig,
                             deviceId,

tensilelite/Tensile/ClientWriter.py

@@ -88,7 +88,7 @@ class ClientLogLevel(Enum):
 ################################################################################
 # Main
 ################################################################################
-def main(config, assembler: Assembler, cCompiler: str, isaInfoMap, outputPath: Path, deviceId: int, gfxName: str, useShortNames: bool=False):
+def main(config, assembler: Assembler, cCompiler: str, isaInfoMap, outputPath: Path, deviceId: int, gfxName: str):
 
   libraryLogicPath = ensurePath(outputPath / LIBRARY_LOGIC_DIR)
   clientLibraryPath = ensurePath(outputPath / LIBRARY_CLIENT_DIR)
@@ -115,7 +115,7 @@ def main(config, assembler: Assembler, cCompiler: str, isaInfoMap, outputPath: P
   else:
     env["PYTHONPATH"] = module_path
 
-  createLibraryScript = getBuildClientLibraryScript(clientLibraryPath, libraryLogicPath, str(assembler.path), isaToGfx(list(isaInfoMap.keys())[0]), useShortNames)
+  createLibraryScript = getBuildClientLibraryScript(clientLibraryPath, libraryLogicPath, str(assembler.path), isaToGfx(list(isaInfoMap.keys())[0]))
   subprocess.run(shlex.split(createLibraryScript), env=env, cwd=clientLibraryPath)
   coList = glob(os.path.join(clientLibraryPath, "library/*.co"))
   yamlList = glob(os.path.join(clientLibraryPath, "library/*.yaml"))
@@ -230,7 +230,7 @@ def runClient(libraryLogicPath, forBenchmark, enableTileSelection, cxxCompiler:
 
   return process.returncode
 
-def getBuildClientLibraryScript(buildPath, libraryLogicPath, cxxCompiler, targetGfx, useShortNames: bool=False):
+def getBuildClientLibraryScript(buildPath, libraryLogicPath, cxxCompiler, targetGfx):
   import io
   runScriptFile = io.StringIO()
 
@@ -239,9 +239,6 @@ def getBuildClientLibraryScript(buildPath, libraryLogicPath, cxxCompiler, target
   if not globalParameters["LazyLibraryLoading"]:
     callCreateLibraryCmd += " --no-lazy-library-loading"
 
-  if useShortNames:
-    callCreateLibraryCmd += " --short-file-names"
-
   if globalParameters.get("AsmDebug", False):
     callCreateLibraryCmd += " --asm-debug"
 

tensilelite/Tensile/Common/GlobalParameters.py

@@ -596,7 +596,6 @@ def assignGlobalParameters(config, isaInfoMap: Dict[IsaVersion, IsaInfo]):
     # The following keys may be present in the config, but are not (or no longer) global parameters.
     ignoreKeys = [
         "Architecture",
-        "ShortNames",
         "PrintLevel",
         "Device",
         "UseCompression",

tensilelite/Tensile/Common/RequiredParameters.py

@@ -0,0 +1,98 @@
+################################################################################
+#
+# Copyright (C) 2025 Advanced Micro Devices, Inc. All rights reserved.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell cop-
+# ies of the
+# Software, and to permit persons to whom the Software is furnished
+# to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IM-
+# PLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+# FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
+# COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNE-
+# CTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+################################################################################
+
+from functools import lru_cache
+from .ValidParameters import validParameters
+
+
+@lru_cache
+def getRequiredParametersFull() -> set:
+    return frozenset(validParameters.keys())
+
+
+@lru_cache
+def getRequiredParametersMin() -> set:
+    return frozenset({
+        '1LDSBuffer',
+        'ActivationFuncCall',
+        'AssertFree0ElementMultiple',
+        'AssertFree1ElementMultiple',
+        'AssertSummationElementMultiple',
+        'ClusterLocalRead',
+        'ConvertAfterDS',
+        'DirectToVgprA',
+        'DirectToVgprB',
+        'ExpandPointerSwap',
+        'ForceDisableShadowInit',
+        'GlobalReadPerMfma',
+        'GlobalReadVectorWidthA',
+        'GlobalReadVectorWidthB',
+        'GlobalSplitUAlgorithm',
+        'GroupLoadStore',
+        'ISA',
+        'InnerUnroll',
+        'Kernel',
+        'LdsBlockSizePerPadA',
+        'LdsBlockSizePerPadB',
+        'LdsBlockSizePerPadMetadata',
+        'LdsPadA',
+        'LdsPadB',
+        'LdsPadMetadata',
+        'LocalReadVectorWidth',
+        'LocalWritePerMfma',
+        'MIArchVgpr',
+        'MaxOccupancy',
+        'NonTemporal',
+        'NonTemporalA',
+        'NonTemporalB',
+        'NonTemporalC',
+        'NonTemporalD',
+        'NonTemporalMetadata',
+        'NumElementsPerBatchStore',
+        'NumLoadsCoalescedA',
+        'NumLoadsCoalescedB',
+        'OptNoLoadLoop',
+        'PrefetchGlobalRead',
+        'PrefetchLocalRead',
+        'PreloadKernArgs',
+        'ScheduleIterAlg',
+        'SourceSwap',
+        'StorePriorityOpt',
+        'StoreRemapVectorWidth',
+        'StoreSyncOpt',
+        'StoreVectorWidth',
+        'StreamK',
+        'StreamKXCCMapping',
+        'TransposeLDS',
+        'UnrollLoopSwapGlobalReadOrder',
+        'Use64bShadowLimit',
+        'UseInstOffsetForGRO',
+        'UseSgprForGRO',
+        'VectorStore',
+        'VectorWidthA',
+        'VectorWidthB',
+        'WaveSeparateGlobalReadA',
+        'WaveSeparateGlobalReadB',
+        'WavefrontSize',
+        'WorkGroup',
+    })

tensilelite/Tensile/KernelWriter.py

@@ -42,7 +42,7 @@
 from .Activation import ActivationModule
 from .Common import printWarning, roundUp, print2, DebugConfig, DataDirection, \
   INDEX_CHARS, IsaVersion
-from Tensile.SolutionStructs.Naming import getKernelName
+from Tensile.SolutionStructs.Naming import getKernelNameMin
 from Tensile.Toolchain.Component import Assembler
 
 import abc
@@ -369,13 +369,9 @@ class KernelWriter(metaclass=abc.ABCMeta):
   ##############################################################################
   def __init__(
       self,
-      kernelMinNaming,
-      kernelSerialNaming,
       assembler: Assembler,
       debugConfig: DebugConfig,
     ):
-    self.kernelMinNaming = kernelMinNaming
-    self.kernelSerialNaming = kernelSerialNaming
     self.assembler = assembler
     self.debugConfig = debugConfig
 
@@ -3273,7 +3269,7 @@ def _initKernel(self, kernel, tensorParametersA, tensorParametersB):
     ti.setKernel(version, kernel["WavefrontSize"])
 
     self.consts = ConstValues()
-    self.states = StateValues(version=version, kernel=kernel, kernelName=getKernelName(self.kernelMinNaming, self.debugConfig.splitGSU, kernel))
+    self.states = StateValues(version=version, kernel=kernel, kernelName=getKernelNameMin(kernel, self.debugConfig.splitGSU))
     self.vgprs  = StateVgprs()
     self.sgprs  = collections.OrderedDict()
     self.codes  = CodeModules()
@@ -5439,7 +5435,7 @@ def getSourceFileString(self, kernel) -> Tuple[int, str]:
     pass
 
   def getHeaderFileString(self, kernel):
-    kernelName = getKernelName(self.kernelMinNaming, self.debugConfig.splitGSU, kernel)
+    kernelName = getKernelNameMin(kernel, self.debugConfig.splitGSU)
     fileString = "" # CHeader
     fileString += "extern const unsigned char %s_coba[]; // code object byte array\n" % kernelName
 

tensilelite/Tensile/KernelWriterAssembly.py

@@ -93,16 +93,14 @@ class KernelWriterAssembly(KernelWriter):
   ##############################################################################
   def __init__(
       self,
-      kernelMinNaming,
-      kernelSerialNaming,
       assembler: Assembler,
       debugConfig: DebugConfig,
     ):
-    super(KernelWriterAssembly, self).__init__(kernelMinNaming, kernelSerialNaming, assembler, debugConfig)
+    super(KernelWriterAssembly, self).__init__(assembler, debugConfig)
 
 
-  def _getCustomKernelSource(self, useShortNames, kernel, CustomKernelDirectory):
-    kernelName = getKernelFileBase(useShortNames, self.debugConfig.splitGSU, self.kernelMinNaming, self.kernelSerialNaming, kernel)
+  def _getCustomKernelSource(self, kernel, CustomKernelDirectory):
+    kernelName = getKernelFileBase(self.debugConfig.splitGSU, kernel)
     with open(os.path.join(CustomKernelDirectory, (kernelName + ".s"))) as f:
       rocmVersion = self.assembler.rocm_version
       if not (rocmVersion.major >= 6 and rocmVersion.patch >= 32650):
@@ -123,17 +121,15 @@ def _getCustomKernelSource(self, useShortNames, kernel, CustomKernelDirectory):
     return code
 
 
-  def getSourceFileString(self,
-                          kernel,
-                          useShortNames: bool=False) -> Tuple[int, str]:
+  def getSourceFileString(self, kernel) -> Tuple[int, str]:
     assert kernel["KernelLanguage"] == "Assembly"
     # Skip if .o files will have already been built for this file
     if kernel.duplicate:
       self.language = "ASM"
       return (0, "") # should this be an non zero number
 
     try:
-      code = self._getCustomKernelSource(useShortNames, kernel, CUSTOM_KERNEL_PATH) if isCustomKernelConfig(kernel) else self._getKernelSource(kernel)
+      code = self._getCustomKernelSource(kernel, CUSTOM_KERNEL_PATH) if isCustomKernelConfig(kernel) else self._getKernelSource(kernel)
       errcode = 0
     except RuntimeError as e:
       printWarning(f"Failed to generate assembly source code for {kernel}: {e}")