update cubin export

Signed-off-by: Anthony Chang <[email protected]>

Author: rosenrodt

cpp/tensorrt_llm/kernels/trtllmGenKernels/batchedGemm/trtllmGen_bmm_export/.clang-format

@@ -0,0 +1,78 @@
+---
+AccessModifierOffset: -4
+AlignAfterOpenBracket: DontAlign
+AlignConsecutiveAssignments: None
+AlignConsecutiveDeclarations: None
+AlignOperands:   false
+AlignTrailingComments: true
+AllowAllParametersOfDeclarationOnNextLine: true
+AllowShortBlocksOnASingleLine: Empty
+AllowShortCaseLabelsOnASingleLine: true
+AllowShortFunctionsOnASingleLine: Empty
+AllowShortIfStatementsOnASingleLine: false
+AllowShortLoopsOnASingleLine: false
+AlwaysBreakAfterDefinitionReturnType: None
+AlwaysBreakAfterReturnType: None
+AlwaysBreakBeforeMultilineStrings: true
+AlwaysBreakTemplateDeclarations: Yes
+BasedOnStyle: None
+BinPackArguments: true
+BinPackParameters: true
+BreakBeforeBinaryOperators: All
+BreakBeforeBraces: Allman
+BreakBeforeTernaryOperators: true
+BreakConstructorInitializersBeforeComma: true
+ColumnLimit:     120
+CommentPragmas:  '^ IWYU pragma:'
+ConstructorInitializerAllOnOneLineOrOnePerLine: false
+ConstructorInitializerIndentWidth: 4
+ContinuationIndentWidth: 4
+Cpp11BracedListStyle: true
+DerivePointerAlignment: false
+DisableFormat:   false
+ExperimentalAutoDetectBinPacking: false
+ForEachMacros:   [ foreach, Q_FOREACH, BOOST_FOREACH ]
+IncludeBlocks: Preserve
+IncludeCategories:
+  - Regex:           '^"(llvm|llvm-c|clang|clang-c)/'
+    Priority:        2
+  - Regex:           '^(<|"(gtest|isl|json)/)'
+    Priority:        3
+  - Regex:           '.*'
+    Priority:        1
+IndentCaseLabels: false
+IndentWidth:     4
+IndentWrappedFunctionNames: false
+KeepEmptyLinesAtTheStartOfBlocks: true
+Language: Cpp
+MacroBlockBegin: ''
+MacroBlockEnd:   ''
+MaxEmptyLinesToKeep: 1
+NamespaceIndentation: None
+ObjCBlockIndentWidth: 4
+ObjCSpaceAfterProperty: true
+ObjCSpaceBeforeProtocolList: true
+PenaltyBreakBeforeFirstCallParameter: 19
+PenaltyBreakComment: 300
+PenaltyBreakFirstLessLess: 120
+PenaltyBreakString: 1000
+PenaltyExcessCharacter: 1000000
+PenaltyReturnTypeOnItsOwnLine: 60
+PointerAlignment: Left
+QualifierAlignment: Right
+ReflowComments:  true
+SeparateDefinitionBlocks: Always
+SortIncludes:    false
+SpaceAfterCStyleCast: true
+SpaceBeforeAssignmentOperators: true
+SpaceBeforeParens: ControlStatements
+SpaceInEmptyParentheses: false
+SpacesBeforeTrailingComments: 1
+SpacesInAngles:  false
+SpacesInCStyleCastParentheses: false
+SpacesInContainerLiterals: true
+SpacesInParentheses: false
+SpacesInSquareBrackets: false
+Standard:        c++14
+TabWidth:        4
+UseTab:          Never

cpp/tensorrt_llm/kernels/trtllmGenKernels/batchedGemm/trtllmGen_bmm_export/BatchedGemmEnums.h

@@ -16,8 +16,8 @@
  */
 #pragma once
 
-#include <cassert>
 #include <string>
+#include <cassert>
 
 namespace batchedGemm
 {
@@ -34,7 +34,9 @@ enum class RouteImpl
     // Use LDGSTS to do the routing
     Ldgsts = 1,
     // Use UTMALDG.GATHER4 to do the routing
-    Tma = 2
+    Tma = 2,
+    // Use LDG+STS to do the routing
+    LdgPlusSts = 3
 };
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -60,6 +62,13 @@ inline bool doesRouteImplUseTma(RouteImpl mode)
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
+inline bool doesRouteImplUseLdgPlusSts(RouteImpl mode)
+{
+    return (mode == RouteImpl::LdgPlusSts);
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
 } // namespace batchedGemm
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////