#0: Clean up tt_dnn op enums

tt_eager/tt_dnn/op_library/bcast/bcast_op.hpp

@@ -14,12 +14,12 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class BcastOpMath { ADD = 0, SUB = 1, MUL = 2 };
+enum class BcastOpMath { ADD, SUB, MUL };
 
-enum class BcastOpDim { H = 0, W = 1, HW = 2 };
+enum class BcastOpDim { H, W, HW };
 
 // TODO: Accept parallelization
-enum class BcastOpParallelizationStrategy { MULTI_CORE_H = 0, MULTI_CORE_W = 1, MULTI_CORE_HW = 2, SINGLE_CORE = 3 };
+enum class BcastOpParallelizationStrategy { MULTI_CORE_H, MULTI_CORE_W, MULTI_CORE_HW, SINGLE_CORE };
 
 operation::ProgramWithCallbacks bcast_single_core(
     const Tensor &input_tensor_a,

tt_eager/tt_dnn/op_library/bmm/bmm_op.hpp

@@ -15,17 +15,16 @@ namespace tt {
 
 namespace tt_metal {
 
-// TODO: Accept parallelization
 enum class MatmulParallelizationStrategy {
-    MULTI_CORE = 0,
-    MULTI_CORE_REUSE = 1,
-    MULTI_CORE_REUSE_PADDING = 2,
-    MULTI_CORE_REUSE_OPTIMIZED = 3,
-    MULTI_CORE_REUSE_MCAST_2D_OPTIMIZED = 4,
-    MULTI_CORE_REUSE_MCAST_2D_TRANSPOSED_OPTIMIZED = 5,
-    MULTI_CORE_REUSE_MCAST_1D_IN0_OPTIMIZED = 6,
-    MULTI_CORE_REUSE_MCAST_1D_IN1_OPTIMIZED = 7,
-    SINGLE_CORE = 8
+    MULTI_CORE,
+    MULTI_CORE_REUSE,
+    MULTI_CORE_REUSE_PADDING,
+    MULTI_CORE_REUSE_OPTIMIZED,
+    MULTI_CORE_REUSE_MCAST_2D_OPTIMIZED,
+    MULTI_CORE_REUSE_MCAST_2D_TRANSPOSED_OPTIMIZED,
+    MULTI_CORE_REUSE_MCAST_1D_IN0_OPTIMIZED,
+    MULTI_CORE_REUSE_MCAST_1D_IN1_OPTIMIZED,
+    SINGLE_CORE
 };
 
 

tt_eager/tt_dnn/op_library/concat/concat_op.hpp

@@ -11,7 +11,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class ConcatOpParallelizationStrategy { SINGLE_CORE = 0, MULTI_CORE = 1, SHARDED_MULTI_CORE = 2 };
+enum class ConcatOpParallelizationStrategy { SINGLE_CORE, MULTI_CORE, SHARDED_MULTI_CORE };
 
 struct Concat {
     uint32_t dim;

tt_eager/tt_dnn/op_library/conv/conv_op.hpp

@@ -13,7 +13,7 @@ namespace tt_metal {
 
 // TODO: Accept parallelization
 enum class ConvOpParallelizationStrategy {
-    MULTI_CORE = 0, MULTI_CORE_REUSE = 1, MULTI_CORE_REUSE_MCAST = 2, SINGLE_CORE = 3
+    MULTI_CORE, MULTI_CORE_REUSE, MULTI_CORE_REUSE_MCAST, SINGLE_CORE
 };
 
 struct Conv {

tt_eager/tt_dnn/op_library/conv/optimized_conv_op.hpp

@@ -14,7 +14,7 @@ namespace tt_metal {
 
 // TODO: Accept parallelization
 enum class OptimizedConvOpParallelizationStrategy {
-    MULTI_CORE = 0, MULTI_CORE_REUSE = 1, MULTI_CORE_REUSE_MCAST = 2, SINGLE_CORE = 3
+    MULTI_CORE, MULTI_CORE_REUSE, MULTI_CORE_REUSE_MCAST, SINGLE_CORE
 };
 
 struct OptimizedConvParallelizationConfig {

tt_eager/tt_dnn/op_library/copy/copy_op.hpp

@@ -19,7 +19,7 @@ namespace tt {
 namespace tt_metal {
 
 enum class CopyOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 struct Copy {

tt_eager/tt_dnn/op_library/eltwise_binary/eltwise_binary_op.hpp

@@ -19,26 +19,26 @@ namespace tt_metal {
 
 enum class BinaryOpType {
 
-    ADD = 0,
-    SUB = 1,
-    MUL = 2,
-    GT = 3,
-    LT = 4,
-    LTE = 5,
-    GTE = 6,
-    EQ = 7,
-    NE = 8,
-    SQUARED_DIFFERENCE = 9,
-    BIAS_GELU = 10,
-    LOGADDEXP = 11,
-    LOGICAL_AND = 12,
-    LOGICAL_OR = 13,
-    LDEXP = 14,
-    LOGADDEXP2 = 15,
-    DIV_FAST = 16
+    ADD,
+    SUB,
+    MUL,
+    GT,
+    LT,
+    LTE,
+    GTE,
+    EQ,
+    NE,
+    SQUARED_DIFFERENCE,
+    BIAS_GELU,
+    LOGADDEXP,
+    LOGICAL_AND,
+    LOGICAL_OR,
+    LDEXP,
+    LOGADDEXP2,
+    DIV_FAST
 };
 
-enum class BinaryOpParallelizationStrategy { MULTI_CORE = 0, SINGLE_CORE = 1 };
+enum class BinaryOpParallelizationStrategy { MULTI_CORE, SINGLE_CORE };
 
 operation::ProgramWithCallbacks eltwise_binary_single_core(
     const Tensor &a,

tt_eager/tt_dnn/op_library/eltwise_unary/eltwise_unary_op.hpp

@@ -68,10 +68,6 @@ enum class UnaryOpType {
     SILU,
     IDENTITY,
     NEG,
-    ADD_UNARY,
-    SUB_UNARY,
-    MUL_UNARY,
-    DIV_UNARY,
     ADD_UNARY_SFPU,
     SUB_UNARY_SFPU,
     MUL_UNARY_SFPU,
@@ -99,10 +95,6 @@ bool is_parametrized_type(T val) {
         case UnaryOpType::RSUB:
         case UnaryOpType::RDIV:
         case UnaryOpType::EXP:
-        case UnaryOpType::ADD_UNARY:
-        case UnaryOpType::SUB_UNARY:
-        case UnaryOpType::MUL_UNARY:
-        case UnaryOpType::DIV_UNARY:
         case UnaryOpType::ADD_UNARY_SFPU:
         case UnaryOpType::SUB_UNARY_SFPU:
         case UnaryOpType::MUL_UNARY_SFPU:
@@ -165,7 +157,7 @@ inline UnaryWithParam string_to_unary_with_param(const std::string& name) {
     TT_THROW("Unknown unary op: " + name);
 }
 
-enum class UnaryOpParallelizationStrategy { SINGLE_CORE = 0, MULTI_CORE = 1, SHARDED_MULTI_CORE = 2 };
+enum class UnaryOpParallelizationStrategy { SINGLE_CORE, MULTI_CORE, SHARDED_MULTI_CORE };
 
 struct EltwiseUnary {
     const std::vector<UnaryWithParam> op_chain;

tt_eager/tt_dnn/op_library/embeddings/embeddings_op.hpp

@@ -15,8 +15,8 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class EmbeddingsType { GENERIC = 0, PADDED = 1, BINARY = 2 };
-enum class EmbeddingsIndexType { UINT32 = 0, BFP16 = 1};
+enum class EmbeddingsType { GENERIC, PADDED, BINARY };
+enum class EmbeddingsIndexType { UINT32, BFP16};
 
 struct Embeddings {
     const MemoryConfig output_mem_config;

tt_eager/tt_dnn/op_library/fold/fold_op.hpp

@@ -11,7 +11,7 @@
 
 namespace tt::tt_metal {
 
-enum class FoldOpParallelizationStrategy { SINGLE_CORE = 0, SHARDED_MULTI_CORE = 1 };
+enum class FoldOpParallelizationStrategy { SINGLE_CORE, SHARDED_MULTI_CORE };
 
 struct Fold {
     uint8_t stride_h;

tt_eager/tt_dnn/op_library/moreh_softmax/moreh_softmax_op.hpp

@@ -18,18 +18,18 @@ namespace primary {
 using namespace tt_metal;
 
 enum class MorehSoftmaxOpParallelizationStrategy {
-    NONE = 0,
-    SMALL_W = 1,
-    SMALL_H = 2,
-    LARGE_W = 3,
-    LARGE_H = 4,
-    LARGE_C = 5,
+    NONE,
+    SMALL_W,
+    SMALL_H,
+    LARGE_W,
+    LARGE_H,
+    LARGE_C,
 };
 
 enum class MorehSoftmaxOp {
-    SOFTMAX = 0,
-    SOFTMIN = 1,
-    LOGSOFTMAX = 2,
+    SOFTMAX,
+    SOFTMIN,
+    LOGSOFTMAX,
 };
 
 bool is_moreh_softmax_w_small_available(const Tensor &tensor);

tt_eager/tt_dnn/op_library/moreh_softmax_backward/moreh_softmax_backward_op.hpp

@@ -17,18 +17,18 @@ namespace primary {
 using namespace tt_metal;
 
 enum class MorehSoftmaxBackwardOpParallelizationStrategy {
-    NONE = 0,
-    SMALL_W = 1,
-    SMALL_H = 2,
-    LARGE_W = 3,
-    LARGE_H = 4,
-    LARGE_C = 5
+    NONE,
+    SMALL_W,
+    SMALL_H,
+    LARGE_W,
+    LARGE_H,
+    LARGE_C
 };
 
 enum class MorehSoftmaxBackwardOp {
-    SOFTMAX = 0,
-    SOFTMIN = 1,
-    LOGSOFTMAX = 2,
+    SOFTMAX,
+    SOFTMIN,
+    LOGSOFTMAX,
 };
 
 bool is_moreh_softmax_backward_w_small_available(const Tensor &tensor);

tt_eager/tt_dnn/op_library/move/move_op.hpp

@@ -26,7 +26,7 @@ namespace tt {
 namespace tt_metal {
 
 enum class MoveOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1, MULTI_CORE_OVERLAP = 2, MULTI_CORE_SHARDED = 3
+    MULTI_CORE, SINGLE_CORE, MULTI_CORE_OVERLAP, MULTI_CORE_SHARDED
 };
 
 struct Move {

tt_eager/tt_dnn/op_library/pool/average_pool.hpp

@@ -13,7 +13,7 @@ namespace tt {
 namespace tt_metal {
 
 enum class PoolType {
-    AVG = 0
+    AVG
 };
 
 Tensor average_pool_2d(const Tensor& input, const MemoryConfig& output_mem_config = operation::DEFAULT_OUTPUT_MEMORY_CONFIG, const std::optional<DataType>& output_dtype = std::nullopt);

tt_eager/tt_dnn/op_library/reduce/reduce_op.hpp

@@ -13,15 +13,15 @@ namespace tt {
 
 namespace tt_metal {
 enum class ReduceOpMath {
-    SUM = 0, MAX = 1, MIN = 2
+    SUM, MAX, MIN
 };
 
 enum class ReduceOpDim {
-    H = 0, W = 1, HW = 2
+    H, W, HW
 };
 
 enum class ReduceOpParallelizationStrategy {
-    MULTI_CORE_H = 0, MULTI_CORE_W = 1, MULTI_CORE_HW = 2, SINGLE_CORE = 3
+    MULTI_CORE_H, MULTI_CORE_W, MULTI_CORE_HW, SINGLE_CORE
 };
 
 // TODO: Accept parallelization

tt_eager/tt_dnn/op_library/repeat/repeat_op.hpp

@@ -11,7 +11,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class RepeatOpParallelizationStrategy { SINGLE_CORE = 0, MULTI_CORE = 1 };
+enum class RepeatOpParallelizationStrategy { SINGLE_CORE, MULTI_CORE };
 
 struct Repeat {
     const uint32_t repeat_dim;

tt_eager/tt_dnn/op_library/rotary_embedding/rotary_embedding_op.hpp

@@ -16,7 +16,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class RotaryEmbeddingOpParallelizationStrategy { MULTI_CORE = 0, SINGLE_CORE = 1 };
+enum class RotaryEmbeddingOpParallelizationStrategy { MULTI_CORE, SINGLE_CORE };
 
 operation::ProgramWithCallbacks rotary_embedding_single_core(
     const Tensor &input, const Tensor &cos, const Tensor &sin, Tensor &output, std::optional<uint32_t> token_idx, DeviceComputeKernelConfig compute_kernel_config);

tt_eager/tt_dnn/op_library/rotate_half/rotate_half_op.hpp

@@ -16,7 +16,7 @@ namespace tt {
 namespace tt_metal {
 
 enum class RotateHalfOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 operation::ProgramWithCallbacks rotate_half_single_core(const Tensor &input_tensor, Tensor &output_tensor);

tt_eager/tt_dnn/op_library/scan/scan_op.hpp

@@ -11,7 +11,7 @@
 
 namespace tt::tt_metal {
 
-enum class ScanOpParallelizationStrategy { SHARDED_MULTI_CORE = 0 };
+enum class ScanOpParallelizationStrategy { SHARDED_MULTI_CORE };
 
 enum class ScanOpDirection { ROWS, COLS, ROWS_REVERSED, COLS_REVERSED };
 

tt_eager/tt_dnn/op_library/sharded/sharded_op.hpp

@@ -15,7 +15,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class ShardedOpParallelizationStrategy { MULTI_CORE = 0 };
+enum class ShardedOpParallelizationStrategy { MULTI_CORE };
 
 enum class ShardedOpType { InterleavedToSharded, ShardedToInterleaved };
 

tt_eager/tt_dnn/op_library/sharded_partial/sharded_op_partial.hpp

@@ -17,7 +17,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class ShardedOpPartialParallelizationStrategy { MULTI_CORE = 0 };
+enum class ShardedOpPartialParallelizationStrategy { MULTI_CORE };
 
 enum class ShardedOpPartialType { InterleavedToShardedPartial, ShardedToInterleavedPartial };
 

tt_eager/tt_dnn/op_library/tilize/tilize_op.hpp

@@ -11,7 +11,7 @@ namespace tt {
 
 namespace tt_metal {
 
-enum class TilizeOpParallelizationStrategy { MULTI_CORE = 0, SINGLE_CORE = 1 };
+enum class TilizeOpParallelizationStrategy { MULTI_CORE, SINGLE_CORE };
 
 struct Tilize {
     const MemoryConfig output_mem_config;
@@ -32,7 +32,7 @@ struct Tilize {
     }
 };
 
-enum class TilizeWithValPaddingOpParallelizationStrategy { MULTI_CORE = 0, SINGLE_CORE = 1 };
+enum class TilizeWithValPaddingOpParallelizationStrategy { MULTI_CORE, SINGLE_CORE };
 
 struct TilizeWithValPadding {
     const Shape output_tensor_shape;

tt_eager/tt_dnn/op_library/transpose/transpose_op.hpp

@@ -12,11 +12,11 @@ namespace tt {
 namespace tt_metal {
 
 enum class TransposeOpDim {
-    WH = 0, HC = 1, CN = 2, NH = 3, NW = 4, CW = 5
+    WH, HC, CN, NH, NW, CW
 };
 
 enum class TransposeOpParallelizationStrategy {
-    MULTI_CORE_WH = 0, MULTI_CORE_HC = 1, MULTI_CORE_CN = 2, SINGLE_CORE = 3
+    MULTI_CORE_WH, MULTI_CORE_HC, MULTI_CORE_CN, SINGLE_CORE
 };
 
 struct Transpose {

tt_eager/tt_dnn/op_library/unpad/unpad_op.hpp

@@ -13,7 +13,7 @@ namespace tt {
 namespace tt_metal {
 
 enum class UnpadOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 uint32_t get_tiled_start_offset(const Tensor &input_tensor, const Shape &output_tensor_start);

tt_eager/tt_dnn/op_library/untilize/untilize_op.hpp

@@ -15,7 +15,7 @@ namespace tt_metal {
 #define MAX_PACK_UNTILIZE_WIDTH 8       // pack untilize currently does not support > 8 width
 
 enum class UntilizeOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 struct Untilize {
@@ -39,7 +39,7 @@ struct Untilize {
 };
 
 enum class UntilizeWithUnpaddingOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 struct UntilizeWithUnpadding {

tt_eager/tt_dnn/op_library/update_cache/update_cache_op.hpp

@@ -17,11 +17,11 @@ namespace tt {
 namespace tt_metal {
 
 enum class UpdateCacheOpParallelizationStrategy {
-    MULTI_CORE = 0, SINGLE_CORE = 1
+    MULTI_CORE, SINGLE_CORE
 };
 
 enum class UpdateCacheOpType {
-    FILL = 0, UPDATE = 1
+    FILL, UPDATE
 };
 
 operation::ProgramWithCallbacks update_cache_multi_core(const Tensor& cache_tensor, const Tensor &input_tensor, const uint32_t update_idx, const uint32_t batch_offset, DeviceComputeKernelConfig compute_kernel_config);