v4.1 release

This commit is contained in:
Junkai-Wu
2025-07-03 20:07:53 +08:00
committed by GitHub
parent b995f93317
commit a1aaf2300a
155 changed files with 18407 additions and 6068 deletions

View File

@@ -105,8 +105,6 @@ public:
cutlass::library::RasterOrder raster_order{cutlass::library::RasterOrder::kHeuristic};
int swizzle_size{1};
cutlass::library::RuntimeDatatype runtime_input_datatype_a{};
cutlass::library::RuntimeDatatype runtime_input_datatype_b{};
@@ -264,7 +262,8 @@ protected:
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
cutlass::library::RasterOrder const &raster_order,
int swizzle_size);
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Update performance result configuration according to flexible user setups
void update_result_(
@@ -275,7 +274,8 @@ protected:
cutlass::library::RasterOrder const &raster_order,
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
int swizzle_size);
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Initializes the performance result
void initialize_result_(

View File

@@ -103,8 +103,6 @@ public:
cutlass::library::RasterOrder raster_order{cutlass::library::RasterOrder::kHeuristic};
int swizzle_size{1};
cutlass::library::RuntimeDatatype runtime_input_datatype_a{};
cutlass::library::RuntimeDatatype runtime_input_datatype_b{};
@@ -268,7 +266,8 @@ protected:
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
cutlass::library::RasterOrder const &raster_order,
int swizzle_size);
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Update performance result configuration according to flexible user setups
void update_result_(
@@ -279,7 +278,8 @@ protected:
cutlass::library::RasterOrder const &raster_order,
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
int swizzle_size);
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Initializes the performance result
void initialize_result_(

View File

@@ -286,14 +286,25 @@ protected:
library::GroupedGemmDescription const& operation_desc,
ProblemSpace const& problem_space);
/// Update workspace configuration according to flexible user setups
void update_workspace_(
GroupedGemmWorkspace &gemm_workspace,
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
cutlass::library::RasterOrder const &raster_order,
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Update performance result configuration for exploration parameters
void update_result_(
void update_workspace_and_result_(
GroupedGemmWorkspace &gemm_workspace,
PerformanceResult &result,
ProblemSpace const &problem_space,
cutlass::library::RasterOrder const &raster_order,
std::array<int64_t, 3> const &preferred_cluster,
std::array<int64_t, 3> const &fallback_cluster,
int swizzle_size);
int swizzle_size,
bool is_dynamic_cluster_enabled);
/// Verifies CUTLASS against host and device references
bool verify_with_reference_(