formatting

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp

@@ -34,7 +34,7 @@ template <typename GridwiseGemm,
           bool HasDoubleTailKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_gemm_dl_multiple_d(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_layernorm_xdl_cshuffle.hpp

@@ -37,7 +37,7 @@ template <typename GridwiseGemmWelford,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_gemm_multiple_d_welford_first_half_xdl_cshuffle(
             const ABDataType* __restrict__ p_a_grid,
@@ -58,7 +58,8 @@ __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
                 e_grid_desc_mblock_mperblock_nblock_nperblock,
             const MeanVarGridDescriptor_MBlock_MPerBlock_NBlock
                 mean_var_grid_desc_mblock_mperblock_nblock,
-        const CountGridDescriptor_MBlock_MPerBlock_NBlock count_grid_desc_mblock_mperblock_nblock,
+            const CountGridDescriptor_MBlock_MPerBlock_NBlock
+                count_grid_desc_mblock_mperblock_nblock,
             const Block2ETileMap block_2_etile_map,
             index_t NRaw)
 {
@@ -121,7 +122,7 @@ template <typename GridwiseWelfordLayernorm,
           typename HElementwiseOperation>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_welford_layernorm2d_second_half(
             const EMeanVarDataType* __restrict__ p_e_grid,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp

@@ -38,7 +38,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_gemm_multiple_d_multiple_r_xdl_cshuffle(
             const FloatAB* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp

@@ -36,7 +36,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_gemm_multiple_d_xdl_cshuffle(const ABDataType* __restrict__ p_a_grid,
                                             const ABDataType* __restrict__ p_b_grid,

include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_waveletmodel_cshuffle.hpp

@@ -34,9 +34,10 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_WAVELET_MAX_THREAD_PER_BLOCK, CK_WAVELET_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_WAVELET_MAX_THREAD_PER_BLOCK, CK_WAVELET_MIN_BLOCK_PER_CU)
 #endif
-    kernel_gemm_xdl_waveletmodel_cshuffle(const ABDataType* __restrict__ p_a_grid,
+        kernel_gemm_xdl_waveletmodel_cshuffle(
+            const ABDataType* __restrict__ p_a_grid,
             const ABDataType* __restrict__ p_b_grid,
             EDataType* __restrict__ p_e_grid,
             const AElementwiseOperation a_element_op,

include/ck/tensor_operation/gpu/device/impl/device_grouped_contraction_multiple_d_xdl_cshuffle.hpp

@@ -28,7 +28,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_contraction_multiple_d_xdl_cshuffle(
             const void CK_CONSTANT_ADDRESS_SPACE* contraction_args,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v1.hpp

@@ -110,7 +110,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_conv_bwd_data_multiple_d_xdl_cshuffle(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_gnwc_gkxc_gnwk_dl.hpp

@@ -59,7 +59,7 @@ template <typename GridwiseGemm,
           bool HasDoubleTailKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_batched_gemm_dlops_bwd_weight(
             const FloatAB* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp

@@ -61,9 +61,10 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
-    kernel_batched_gemm_xdlops_bwd_weight(const FloatAB* __restrict__ p_a_grid,
+        kernel_batched_gemm_xdlops_bwd_weight(
+            const FloatAB* __restrict__ p_a_grid,
             const FloatAB* __restrict__ p_b_grid,
             FloatC* __restrict__ p_c_grid,
             const AElementwiseOperation a_element_op,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk.hpp

@@ -117,7 +117,7 @@ template <typename GridwiseGemm,
           bool HasDoubleTailKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_conv_fwd_dl_multiple_d(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_dl_nhwc_kyxc_nhwk.hpp

@@ -93,7 +93,7 @@ template <typename GridwiseGemm,
           bool HasDoubleTailKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_conv_fwd_dl(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_multiple_r_xdl_cshuffle.hpp

@@ -131,7 +131,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_batch_gemm_multiple_d_xdl_cshuffle(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_xdl_cshuffle.hpp

@@ -115,7 +115,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_conv_fwd_multiple_d_xdl_cshuffle(
             const ABDataType* __restrict__ p_a_grid,

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp

@@ -31,7 +31,7 @@ template <typename GridwiseGemm,
           bool HasDoubleTailKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_gemm_multiple_d_dl(const void CK_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
                                           const index_t group_count,

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_softmax_gemm_permute_xdl_cshuffle.hpp

@@ -32,7 +32,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_gemm_softmax_gemm_xdl_cshuffle_v1(
             const void CK_CONSTANT_ADDRESS_SPACE* group_kernel_args,

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp

@@ -30,7 +30,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_gemm_xdl(const void CK_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
                                 const index_t group_count,

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl_splitk_cshuffle.hpp

@@ -29,7 +29,7 @@ template <typename GridwiseGemm,
           InMemoryDataOperationEnum CGlobalMemoryDataOperation>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_grouped_gemm_xdl_splitk(const void CK_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
                                        const index_t group_count)

include/ck/tensor_operation/gpu/device/impl/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp

@@ -36,7 +36,7 @@ template <typename GridwiseGemm,
           bool HasMainKBlockLoop>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
 #endif
         kernel_contraction_multiple_d_xdl_cshuffle(
             const FloatAB* __restrict__ p_a_grid,