diff --git a/example/ck_tile/03_gemm/README.md b/example/ck_tile/03_gemm/README.md
index 20cc202176..59ef2640b7 100644
--- a/example/ck_tile/03_gemm/README.md
+++ b/example/ck_tile/03_gemm/README.md
@@ -18,7 +18,6 @@ This will result in an executable `build/bin/tile_example_gemm_basic` & `build/b
 ## example
 ```
 args:
-          -b    batch size (default:1)
           -m    m dimension (default:1024)
           -n    n dimension (default:2048)
           -k    k dimension (default:64)
@@ -29,9 +28,11 @@ args:
    -stride_b    Tensor B stride (default:0)
    -stride_c    Tensor C stride (default:0)
           -v    0. No validation, 1. Validation on CPU, 2. Validation on GPU (default:2)
-          -e    Absolute error tolerance (default:1e-5)
        -prec    data type. fp16/bf16/fp8/bf8/int8 (default:fp16)
      -warmup    number of iterations before benchmark the kernel (default:10)
      -repeat    number of iterations to benchmark the kernel (default:100)
       -timer    gpu:gpu timer, cpu:cpu timer (default:gpu)
+    -split_k    splitK value (default:1)
+       -init    0:random, 1:linear, 2:constant (default:1)
+ -persistent    0:non-persistent, 1:persistent (default:0)
 ```
diff --git a/example/ck_tile/03_gemm/gemm_basic.cpp b/example/ck_tile/03_gemm/gemm_basic.cpp
index 0d9c2d9957..25781a4ae8 100644
--- a/example/ck_tile/03_gemm/gemm_basic.cpp
+++ b/example/ck_tile/03_gemm/gemm_basic.cpp
@@ -1,15 +1,6 @@
 // SPDX-License-Identifier: MIT
 // Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
 
-#include <hip/hip_runtime.h>
-
-#include <cstring>
-#include <iostream>
-#include <ostream>
-#include <string>
-#include <tuple>
-
-#include "ck_tile/host.hpp"
 #include "gemm_utils.hpp"
 
 template <typename GemmConfig,
@@ -29,12 +20,6 @@ float gemm(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
 {
     if constexpr(Persistent)
         std::cout << "WARNING: Ignoring persistent kernel option for basic gemm." << std::endl;
-    // The kPadM, kPadN, kPadK & kBlockPerCu should also come from the Codegen part.
-    constexpr bool kPadM = false;
-    constexpr bool kPadN = false;
-    constexpr bool kPadK = false;
-
-    constexpr int kBlockPerCu = 1;
 
     // This part comes from the Codegen
     constexpr ck_tile::index_t M_Tile = 256;
@@ -56,8 +41,12 @@ float gemm(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
 
     using TilePartitioner = ck_tile::GemmTile1DPartitioner<CodegenGemmShape>;
 
-    using CodegenGemmTraits =
-        ck_tile::TileGemmTraits<kPadM, kPadN, kPadK, ALayout, BLayout, CLayout>;
+    using CodegenGemmTraits = ck_tile::TileGemmTraits<GemmConfig::kPadM,
+                                                      GemmConfig::kPadN,
+                                                      GemmConfig::kPadK,
+                                                      ALayout,
+                                                      BLayout,
+                                                      CLayout>;
 
     using CodegenPipelineProblem = ck_tile::
         GemmPipelineProblem<ADataType, BDataType, AccDataType, CodegenGemmShape, CodegenGemmTraits>;
@@ -111,28 +100,30 @@ float gemm(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
                       << std::endl;
         }
 
-        float ave_time = ck_tile::launch_kernel(
-            s, ck_tile::make_kernel<blocks.x, kBlockPerCu>(Kernel{}, grids, blocks, 0, kargs));
+        float ave_time =
+            ck_tile::launch_kernel(s,
+                                   ck_tile::make_kernel<blocks.x, GemmConfig::kBlockPerCu>(
+                                       Kernel{}, grids, blocks, 0, kargs));
 
         return ave_time;
     };
 
     if(args.k_batch == 1)
     {
-        return Run(ck_tile::integral_constant<ck_tile::memory_operation_enum,
-                                              ck_tile::memory_operation_enum::set>{});
+        return Run(MemoryOpSet{});
     }
     else
     {
-        return Run(ck_tile::integral_constant<ck_tile::memory_operation_enum,
-                                              ck_tile::memory_operation_enum::atomic_add>{});
+        return Run(MemoryOpAtomicAdd{});
     }
 }
 
 #include "run_gemm_example.inc"
 
 template <typename APrecType, typename BPrecType = APrecType, typename CPrecType = APrecType>
-int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int argc, char* argv[])
+int run_gemm_example_prec_type(std::string a_layout,
+                               std::string b_layout,
+                               ck_tile::ArgParser& arg_parser)
 {
     using Row = ck_tile::tensor_layout::gemm::RowMajor;
     using Col = ck_tile::tensor_layout::gemm::ColumnMajor;
@@ -142,12 +133,12 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
         if(a_layout == "R" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Col{}, Row{});
+                arg_parser, Row{}, Col{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Col{}, Row{});
+                arg_parser, Col{}, Col{}, Row{});
         }
         else
         {
@@ -160,22 +151,22 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
         if(a_layout == "R" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Col{}, Row{});
+                arg_parser, Row{}, Col{}, Row{});
         }
         else if(a_layout == "R" && b_layout == "R")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Row{}, Row{});
+                arg_parser, Row{}, Row{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "R")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Row{}, Row{});
+                arg_parser, Col{}, Row{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfigBase, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Col{}, Row{});
+                arg_parser, Col{}, Col{}, Row{});
         }
         else
         {
@@ -184,38 +175,34 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
     }
 }
 
-int run_gemm_example(int argc, char* argv[])
+int run_gemm_example(ck_tile::ArgParser& arg_parser)
 {
-    auto [result, arg_parser] = create_args(argc, argv);
-    if(!result)
-        return -1;
-
     std::string data_type = arg_parser.get_str("prec");
     std::string a_layout  = arg_parser.get_str("a_layout");
     std::string b_layout  = arg_parser.get_str("b_layout");
 
     if(data_type == "fp16")
     {
-        return run_gemm_example_prec_type<ck_tile::half_t>(a_layout, b_layout, argc, argv);
+        return run_gemm_example_prec_type<ck_tile::half_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf16")
     {
-        return run_gemm_example_prec_type<ck_tile::bf16_t>(a_layout, b_layout, argc, argv);
+        return run_gemm_example_prec_type<ck_tile::bf16_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "fp8")
     {
         return run_gemm_example_prec_type<ck_tile::fp8_t, ck_tile::fp8_t, ck_tile::half_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf8")
     {
         return run_gemm_example_prec_type<ck_tile::bf8_t, ck_tile::bf8_t, ck_tile::half_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "i8")
     {
         return run_gemm_example_prec_type<ck_tile::int8_t, ck_tile::int8_t, int32_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "pk_int4_t")
     {
@@ -223,7 +210,7 @@ int run_gemm_example(int argc, char* argv[])
         if constexpr(GemmConfigBase::Pipeline == CK_TILE_PIPELINE_COMPUTE_V3)
         {
             return run_gemm_example_prec_type<ck_tile::half_t, ck_tile::pk_int4_t, ck_tile::half_t>(
-                a_layout, b_layout, argc, argv);
+                a_layout, b_layout, arg_parser);
         }
         else
         {
@@ -238,9 +225,13 @@ int run_gemm_example(int argc, char* argv[])
 
 int main(int argc, char* argv[])
 {
+    auto [result, arg_parser] = create_args(argc, argv);
+    if(!result)
+        return -1;
+
     try
     {
-        return !run_gemm_example(argc, argv);
+        return !run_gemm_example(arg_parser);
     }
     catch(const std::runtime_error& e)
     {
diff --git a/example/ck_tile/03_gemm/gemm_utils.hpp b/example/ck_tile/03_gemm/gemm_utils.hpp
index cab110597b..5f477b3821 100644
--- a/example/ck_tile/03_gemm/gemm_utils.hpp
+++ b/example/ck_tile/03_gemm/gemm_utils.hpp
@@ -476,6 +476,12 @@ auto create_args(int argc, char* argv[])
     return std::make_tuple(result, arg_parser);
 }
 
+// Type aliases for memory operation integral constants
+using MemoryOpSet =
+    std::integral_constant<ck_tile::memory_operation_enum, ck_tile::memory_operation_enum::set>;
+using MemoryOpAtomicAdd = std::integral_constant<ck_tile::memory_operation_enum,
+                                                 ck_tile::memory_operation_enum::atomic_add>;
+
 // host API
 template <typename ADataType,
           typename BDataType,
diff --git a/example/ck_tile/03_gemm/gemm_weight_preshuffle.cpp b/example/ck_tile/03_gemm/gemm_weight_preshuffle.cpp
index 0a06787e2b..8a7560bf86 100644
--- a/example/ck_tile/03_gemm/gemm_weight_preshuffle.cpp
+++ b/example/ck_tile/03_gemm/gemm_weight_preshuffle.cpp
@@ -210,12 +210,13 @@ template <typename GemmConfig,
           typename APrecType,
           typename BPrecType = APrecType,
           typename CPrecType = APrecType>
-int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int argc, char* argv[])
+int run_gemm_example_prec_type(std::string a_layout,
+                               std::string b_layout,
+                               ck_tile::ArgParser& arg_parser)
 {
-    using Row                 = ck_tile::tensor_layout::gemm::RowMajor;
-    using Col                 = ck_tile::tensor_layout::gemm::ColumnMajor;
-    auto [result, arg_parser] = create_args(argc, argv);
-    bool preshuffle           = GemmConfig::Preshuffle;
+    using Row       = ck_tile::tensor_layout::gemm::RowMajor;
+    using Col       = ck_tile::tensor_layout::gemm::ColumnMajor;
+    bool preshuffle = GemmConfig::Preshuffle;
 
     if(preshuffle && (a_layout != "R" || b_layout != "C"))
     {
@@ -226,7 +227,7 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
     if(a_layout == "R" && b_layout == "C")
     {
         return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-            argc, argv, Row{}, Col{}, Row{});
+            arg_parser, Row{}, Col{}, Row{});
     }
     else
     {
@@ -235,12 +236,8 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
 }
 
 template <template <typename PreType> typename GemmConfig>
-int run_gemm_example(int argc, char* argv[])
+int run_gemm_example(ck_tile::ArgParser& arg_parser)
 {
-    auto [result, arg_parser] = create_args(argc, argv);
-    if(!result)
-        return -1;
-
     std::string data_type = arg_parser.get_str("prec");
     std::string a_layout  = arg_parser.get_str("a_layout");
     std::string b_layout  = arg_parser.get_str("b_layout");
@@ -248,26 +245,26 @@ int run_gemm_example(int argc, char* argv[])
     if(data_type == "fp16")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::half_t>, ck_tile::half_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf16")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::half_t>, ck_tile::bf16_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "fp8")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::fp8_t>,
                                           ck_tile::fp8_t,
                                           ck_tile::fp8_t,
-                                          ck_tile::half_t>(a_layout, b_layout, argc, argv);
+                                          ck_tile::half_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf8")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::bf8_t>,
                                           ck_tile::bf8_t,
                                           ck_tile::bf8_t,
-                                          ck_tile::half_t>(a_layout, b_layout, argc, argv);
+                                          ck_tile::half_t>(a_layout, b_layout, arg_parser);
     }
     else
     {
@@ -277,9 +274,13 @@ int run_gemm_example(int argc, char* argv[])
 
 int main(int argc, char* argv[])
 {
+    auto [result, arg_parser] = create_args(argc, argv);
+    if(!result)
+        return -1;
+
     try
     {
-        return !run_gemm_example<GemmConfigPreshuffle_2>(argc, argv);
+        return !run_gemm_example<GemmConfigPreshuffle_2>(arg_parser);
     }
     catch(const std::runtime_error& e)
     {
diff --git a/example/ck_tile/03_gemm/run_gemm_example.inc b/example/ck_tile/03_gemm/run_gemm_example.inc
index cc10394065..229771e536 100644
--- a/example/ck_tile/03_gemm/run_gemm_example.inc
+++ b/example/ck_tile/03_gemm/run_gemm_example.inc
@@ -272,6 +272,25 @@ auto shuffle_b(const ck_tile::HostTensor<T>& t)
     return ck_tile::reference_permute(t_view, {0, 2, 3, 1, 4});
 }
 
+template <typename CDataType>
+bool do_verify(const ck_tile::HostTensor<CDataType>& c_m_n_dev_result,
+               const ck_tile::HostTensor<CDataType>& c_m_n_ref,
+               const ck_tile::tuple<double, double>& rtol_atol,
+               const char* variant)
+{
+    bool pass = ck_tile::check_err(c_m_n_dev_result,
+                                   c_m_n_ref,
+                                   "Error: Incorrect results!",
+                                   rtol_atol.at(ck_tile::number<0>{}),
+                                   rtol_atol.at(ck_tile::number<1>{}));
+
+    std::cout << "Relative error threshold: " << rtol_atol.at(ck_tile::number<0>{})
+              << " Absolute error threshold: " << rtol_atol.at(ck_tile::number<1>{}) << std::endl;
+    std::cout << "The " << variant << " verification result is:" << (pass ? "correct" : "fail")
+              << std::endl;
+    return pass;
+}
+
 template <typename GemmConfig,
           typename ADataType,
           typename BDataType = ADataType,
@@ -279,16 +298,11 @@ template <typename GemmConfig,
           typename ALayout,
           typename BLayout,
           typename CLayout>
-int run_gemm_example_with_layouts(int argc,
-                                  char* argv[],
+int run_gemm_example_with_layouts(ck_tile::ArgParser& arg_parser,
                                   const ALayout a_layout                  = ALayout{},
                                   const BLayout b_layout                  = BLayout{},
                                   [[maybe_unused]] const CLayout c_layout = CLayout{})
 {
-    auto [result, arg_parser] = create_args(argc, argv);
-    if(!result)
-        return -1;
-
     using AccDataType = typename GemmTypeConfig<ADataType, BDataType, CDataType>::AccDataType;
 
     ck_tile::index_t M = arg_parser.get_int("m");
@@ -430,28 +444,20 @@ int run_gemm_example_with_layouts(int argc,
     c_m_n_dev_buf.FromDevice(c_m_n_dev_result.data());
     bool pass = true;
 
+    // memory on host to store gpu reference result
+    ck_tile::HostTensor<CDataType> c_m_n_ref(
+        ck_tile::host_tensor_descriptor(M, N, stride_C, is_row_major(CLayout{})));
+    c_m_n_ref.SetZero();
+
     if(arg_parser.get_int("v") == 1)
     {
-        ck_tile::HostTensor<CDataType> c_m_n_host_ref(
-            ck_tile::host_tensor_descriptor(M, N, stride_C, is_row_major(CLayout{})));
-        c_m_n_host_ref.SetZero();
-
         ck_tile::reference_gemm<ADataType, BDataType, AccDataType, CDataType>(
-            a_m_k, b_k_n, c_m_n_host_ref);
+            a_m_k, b_k_n, c_m_n_ref);
         const float max_accumulated_value =
-            *std::max_element(c_m_n_host_ref.mData.begin(), c_m_n_host_ref.mData.end());
+            *std::max_element(c_m_n_ref.mData.begin(), c_m_n_ref.mData.end());
         const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>(
             K, kbatch, max_accumulated_value);
-        pass = ck_tile::check_err(c_m_n_dev_result,
-                                  c_m_n_host_ref,
-                                  "Error: Incorrect results!",
-                                  rtol_atol.at(ck_tile::number<0>{}),
-                                  rtol_atol.at(ck_tile::number<1>{}));
-
-        std::cout << "Relative error threshold: " << rtol_atol.at(ck_tile::number<0>{})
-                  << " Absolute error threshold: " << rtol_atol.at(ck_tile::number<1>{})
-                  << std::endl;
-        std::cout << "The CPU verification result is:" << (pass ? "correct" : "fail") << std::endl;
+        pass = do_verify(c_m_n_dev_result, c_m_n_ref, rtol_atol, "CPU");
     }
     else if(arg_parser.get_int("v") == 2)
     {
@@ -465,13 +471,8 @@ int run_gemm_example_with_layouts(int argc,
             b_k_n_dev_buf.ToDevice(b_k_n.data());
         }
 
-        // memory on host to store gpu reference result
-        ck_tile::HostTensor<CDataType> c_m_n_gpu_ref(
-            ck_tile::host_tensor_descriptor(M, N, stride_C, is_row_major(CLayout{})));
         // memory on device to store gpu reference result
-        ck_tile::DeviceMem c_m_n_gpu_buf_ref(c_m_n_gpu_ref.get_element_space_size_in_bytes());
-
-        c_m_n_gpu_ref.SetZero();
+        ck_tile::DeviceMem c_m_n_gpu_buf_ref(c_m_n_ref.get_element_space_size_in_bytes());
         c_m_n_gpu_buf_ref.SetZero();
 
         ADataType* d_A = static_cast<ADataType*>(a_m_k_dev_buf.GetDeviceBuffer());
@@ -486,21 +487,13 @@ int run_gemm_example_with_layouts(int argc,
                                     BLayout,
                                     CLayout>(d_A, d_B, d_C, M, N, K, stride_A, stride_B, stride_C);
 
-        c_m_n_gpu_buf_ref.FromDevice(c_m_n_gpu_ref.data());
+        c_m_n_gpu_buf_ref.FromDevice(c_m_n_ref.data());
 
         const float max_accumulated_value =
-            *std::max_element(c_m_n_gpu_ref.mData.begin(), c_m_n_gpu_ref.mData.end());
+            *std::max_element(c_m_n_ref.mData.begin(), c_m_n_ref.mData.end());
         const auto rtol_atol = calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>(
             K, kbatch, max_accumulated_value);
-        pass = ck_tile::check_err(c_m_n_dev_result,
-                                  c_m_n_gpu_ref,
-                                  "Error: Incorrect results!",
-                                  rtol_atol.at(ck_tile::number<0>{}),
-                                  rtol_atol.at(ck_tile::number<1>{}));
-        std::cout << "Relative error threshold: " << rtol_atol.at(ck_tile::number<0>{})
-                  << " Absolute error threshold: " << rtol_atol.at(ck_tile::number<1>{})
-                  << std::endl;
-        std::cout << "The GPU verification result is: " << (pass ? "correct" : "fail") << std::endl;
+        pass = do_verify(c_m_n_dev_result, c_m_n_ref, rtol_atol, "GPU");
     }
 
     return pass;
diff --git a/example/ck_tile/03_gemm/universal_gemm.cpp b/example/ck_tile/03_gemm/universal_gemm.cpp
index d82520241d..14c4905720 100644
--- a/example/ck_tile/03_gemm/universal_gemm.cpp
+++ b/example/ck_tile/03_gemm/universal_gemm.cpp
@@ -189,17 +189,11 @@ float gemm(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
     const auto RunSplitk = [&](const auto has_hot_loop_, const auto tail_number_) {
         if(args.k_batch == 1)
         {
-            Run(has_hot_loop_,
-                tail_number_,
-                ck_tile::integral_constant<ck_tile::memory_operation_enum,
-                                           ck_tile::memory_operation_enum::set>{});
+            Run(has_hot_loop_, tail_number_, MemoryOpSet{});
         }
         else
         {
-            Run(has_hot_loop_,
-                tail_number_,
-                ck_tile::integral_constant<ck_tile::memory_operation_enum,
-                                           ck_tile::memory_operation_enum::atomic_add>{});
+            Run(has_hot_loop_, tail_number_, MemoryOpAtomicAdd{});
         }
     };
 
@@ -211,12 +205,13 @@ template <typename GemmConfig,
           typename APrecType,
           typename BPrecType = APrecType,
           typename CPrecType = APrecType>
-int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int argc, char* argv[])
+int run_gemm_example_prec_type(std::string a_layout,
+                               std::string b_layout,
+                               ck_tile::ArgParser& arg_parser)
 {
-    using Row                 = ck_tile::tensor_layout::gemm::RowMajor;
-    using Col                 = ck_tile::tensor_layout::gemm::ColumnMajor;
-    auto [result, arg_parser] = create_args(argc, argv);
-    bool preshuffle           = GemmConfig::Preshuffle;
+    using Row       = ck_tile::tensor_layout::gemm::RowMajor;
+    using Col       = ck_tile::tensor_layout::gemm::ColumnMajor;
+    bool preshuffle = GemmConfig::Preshuffle;
 
     if(preshuffle && std::is_same_v<BPrecType, ck_tile::pk_int4_t>)
     {
@@ -234,12 +229,12 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
         if(a_layout == "R" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Col{}, Row{});
+                arg_parser, Row{}, Col{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Col{}, Row{});
+                arg_parser, Col{}, Col{}, Row{});
         }
         else
         {
@@ -252,22 +247,22 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
         if(a_layout == "R" && b_layout == "R")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Row{}, Row{});
+                arg_parser, Row{}, Row{}, Row{});
         }
         else if(a_layout == "R" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Row{}, Col{}, Row{});
+                arg_parser, Row{}, Col{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "R")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Row{}, Row{});
+                arg_parser, Col{}, Row{}, Row{});
         }
         else if(a_layout == "C" && b_layout == "C")
         {
             return run_gemm_example_with_layouts<GemmConfig, APrecType, BPrecType, CPrecType>(
-                argc, argv, Col{}, Col{}, Row{});
+                arg_parser, Col{}, Col{}, Row{});
         }
         else
         {
@@ -277,12 +272,8 @@ int run_gemm_example_prec_type(std::string a_layout, std::string b_layout, int a
 }
 
 template <template <typename PreType> typename GemmConfig>
-int run_gemm_example(int argc, char* argv[])
+int run_gemm_example(ck_tile::ArgParser& arg_parser)
 {
-    auto [result, arg_parser] = create_args(argc, argv);
-    if(!result)
-        return -1;
-
     std::string data_type = arg_parser.get_str("prec");
     std::string a_layout  = arg_parser.get_str("a_layout");
     std::string b_layout  = arg_parser.get_str("b_layout");
@@ -290,33 +281,33 @@ int run_gemm_example(int argc, char* argv[])
     if(data_type == "fp16")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::half_t>, ck_tile::half_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf16")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::half_t>, ck_tile::bf16_t>(
-            a_layout, b_layout, argc, argv);
+            a_layout, b_layout, arg_parser);
     }
     else if(data_type == "fp8")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::fp8_t>,
                                           ck_tile::fp8_t,
                                           ck_tile::fp8_t,
-                                          ck_tile::half_t>(a_layout, b_layout, argc, argv);
+                                          ck_tile::half_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "bf8")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::bf8_t>,
                                           ck_tile::bf8_t,
                                           ck_tile::bf8_t,
-                                          ck_tile::half_t>(a_layout, b_layout, argc, argv);
+                                          ck_tile::half_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "int8")
     {
         return run_gemm_example_prec_type<GemmConfig<ck_tile::int8_t>,
                                           ck_tile::int8_t,
                                           ck_tile::int8_t,
-                                          ck_tile::int32_t>(a_layout, b_layout, argc, argv);
+                                          ck_tile::int32_t>(a_layout, b_layout, arg_parser);
     }
     else if(data_type == "pk_int4_t")
     {
@@ -326,7 +317,7 @@ int run_gemm_example(int argc, char* argv[])
             return run_gemm_example_prec_type<GemmConfig<ck_tile::half_t>,
                                               ck_tile::half_t,
                                               ck_tile::pk_int4_t,
-                                              ck_tile::half_t>(a_layout, b_layout, argc, argv);
+                                              ck_tile::half_t>(a_layout, b_layout, arg_parser);
         }
         else
         {
@@ -341,9 +332,13 @@ int run_gemm_example(int argc, char* argv[])
 
 int main(int argc, char* argv[])
 {
+    auto [result, arg_parser] = create_args(argc, argv);
+    if(!result)
+        return -1;
+
     try
     {
-        return !run_gemm_example<GemmConfigComputeV3>(argc, argv);
+        return !run_gemm_example<GemmConfigComputeV3>(arg_parser);
     }
     catch(const std::runtime_error& e)
     {
diff --git a/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp b/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp
index 9fec9a320c..5129f83532 100644
--- a/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp
+++ b/include/ck_tile/ops/fmha/kernel/fmha_fwd_appendkv_kernel.hpp
@@ -647,44 +647,25 @@ struct FmhaFwdAppendKVKernel
                              make_tuple(number<FmhaPipeline::kN1>{}, number<FmhaPipeline::kN0>{}),
                              {0, i_n0});
 
-        if constexpr(kApplyRoPE)
-        {
-            FmhaPipeline{}(q_dram_window,
-                           k_dram_window,
-                           i_page_block_k,
-                           k_page_block_navigator,
-                           knew_dram_window,
-                           v_dram_window,
-                           i_page_block_v,
-                           v_page_block_navigator,
-                           vnew_dram_window,
-                           q_rotary_cos_dram_window,
-                           q_rotary_sin_dram_window,
-                           knew_rotary_cos_dram_window,
-                           knew_rotary_sin_dram_window,
-                           kargs.rotary_dim,
-                           kargs.seqlen_q <= i_m0,
-                           skip_append_kv);
-        }
-        else
-        {
-            FmhaPipeline{}(q_dram_window,
-                           k_dram_window,
-                           i_page_block_k,
-                           k_page_block_navigator,
-                           knew_dram_window,
-                           v_dram_window,
-                           i_page_block_v,
-                           v_page_block_navigator,
-                           vnew_dram_window,
-                           q_rotary_cos_dram_window,
-                           q_rotary_sin_dram_window,
-                           knew_rotary_cos_dram_window,
-                           knew_rotary_sin_dram_window,
-                           0, // rotary_dim not used
-                           kargs.seqlen_q <= i_m0,
-                           skip_append_kv);
-        }
+        // If kApplyRoPe is false, we set the rotary_dim to 0
+        auto rotary_dim = kApplyRoPE ? kargs.rotary_dim : 0;
+
+        FmhaPipeline{}(q_dram_window,
+                       k_dram_window,
+                       i_page_block_k,
+                       k_page_block_navigator,
+                       knew_dram_window,
+                       v_dram_window,
+                       i_page_block_v,
+                       v_page_block_navigator,
+                       vnew_dram_window,
+                       q_rotary_cos_dram_window,
+                       q_rotary_sin_dram_window,
+                       knew_rotary_cos_dram_window,
+                       knew_rotary_sin_dram_window,
+                       rotary_dim,
+                       kargs.seqlen_q <= i_m0,
+                       skip_append_kv);
     }
 };
 
diff --git a/include/ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp b/include/ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp
index 1d513faea3..0ac0ca37e6 100644
--- a/include/ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp
+++ b/include/ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp
@@ -943,17 +943,15 @@ struct UniversalGemmKernel
         const auto& bs_block_window = gemm_tile_windows.at(I1);
         const auto& ds_block_window = gemm_tile_windows.at(I2);
 
-        const auto& c_block_tile = GemmPipeline{}.template operator()(
-            as_block_window[I0], bs_block_window[I0], num_loop, smem_ptr_0);
+        const auto& c_block_tile =
+            GemmPipeline{}(as_block_window[I0], bs_block_window[I0], num_loop, smem_ptr_0);
 
         if(UseDefaultScheduler || (get_warp_id() == 0))
         {
             // Run Epilogue Pipeline
             auto& c_block_window = gemm_tile_windows.at(I3);
 
-            EpiloguePipeline{}.template
-            operator()<decltype(c_block_window), decltype(c_block_tile), decltype(ds_block_window)>(
-                c_block_window, c_block_tile, ds_block_window, smem_ptr_0);
+            EpiloguePipeline{}(c_block_window, c_block_tile, ds_block_window, smem_ptr_0);
         }
     }
 
@@ -1001,15 +999,13 @@ struct UniversalGemmKernel
         const auto& bs_block_window = gemm_tile_windows.at(I1);
         const auto& ds_block_window = gemm_tile_windows.at(I2);
 
-        const auto& c_block_tile = GemmPipeline{}.template operator()(
+        const auto& c_block_tile = GemmPipeline{}(
             as_block_window[I0], bs_block_window[I0], num_loop, smem_ptr_0, smem_ptr_1);
 
         // Run Epilogue Pipeline
         auto& c_block_window = gemm_tile_windows.at(I3);
 
-        EpiloguePipeline{}.template
-        operator()<decltype(c_block_window), decltype(c_block_tile), decltype(ds_block_window)>(
-            c_block_window, c_block_tile, ds_block_window, smem_ptr_0);
+        EpiloguePipeline{}(c_block_window, c_block_tile, ds_block_window, smem_ptr_0);
     }
 
     // Non-persistent kernel entry point
diff --git a/include/ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4.hpp b/include/ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4.hpp
index 22c8cf383b..e6da00da95 100644
--- a/include/ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4.hpp
+++ b/include/ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4.hpp
@@ -149,7 +149,7 @@ struct GemmPipelineAgBgCrCompV4 : public BaseGemmPipelineAgBgCrCompV4<Problem>
     [[nodiscard]] CK_TILE_HOST static const std::string GetName()
     {
         // clang-format off
-        return concat('_', "pipeline_AgBgCrCompV3", 
+        return concat('_', "pipeline_AgBgCrCompV4",
                       concat('x', MPerBlock, NPerBlock, KPerBlock,  BlockSize),
                       concat('x', GetVectorSizeA(), GetVectorSizeB(),  GetVectorSizeC()),
                       concat('x', kPadM, kPadN, kPadK));