CUTLASS 2.6.1 - functional and performance enhancements to strided DGRAD, fixes, and tuning

* cutlass 2.6 update

* remove debug prints

* cutlass 2.6.1 (minor update)

* Updated CHANGELOG.

* Minor edit to readme to indicate patch version.

* Minor edit to readme.

Co-authored-by:  Haicheng Wu <haichengw@nvidia.com>, Andrew Kerr <akerr@nvidia.com>

tools/profiler/src/gemm_operation_profiler.cu

@@ -791,7 +791,7 @@ bool GemmOperationProfiler::verify_with_reference_(
     handle.set_provider(provider);
 
     Status status = handle.gemm_universal(
-      library::GemmUniversalMode::kGemm,
+      problem_.mode,
       gemm_workspace_.configuration.problem_size.m(),
       gemm_workspace_.configuration.problem_size.n(),
       gemm_workspace_.configuration.problem_size.k(),

tools/profiler/src/gpu_timer.h

@@ -29,6 +29,7 @@
 #pragma once
 
 #include <cuda_runtime.h>
+#include "cutlass/cutlass.h"
 
 namespace cutlass {
 namespace profiler {

tools/profiler/src/options.cu

@@ -425,7 +425,9 @@ void Options::Profiling::print_usage(std::ostream &out) const {
     << "    Number of ms to sleep between profiling periods (ms).\n\n"
 
     << "  --profiling-enabled=<bool>                   "
-    << "    If true, profiling is actually conducted.\n\n";
+    << "    If true, profiling is actually conducted.\n\n"
+
+  ;
 }
 
 void Options::Profiling::print_options(std::ostream &out, int indent) const {