#3908: Uplift the latest changes from core llk lib. Common files are …

…now identical
tenstorrent · Dec 13, 2023 · 80c6e43 · 80c6e43
1 parent 1371942
commit 80c6e43
Show file tree

Hide file tree

Showing 21 changed files with 69 additions and 51 deletions.
diff --git a/tt_metal/hw/ckernels/wormhole_b0/common/inc/ckernel.h b/tt_metal/hw/ckernels/wormhole_b0/common/inc/ckernel.h
@@ -36,7 +36,7 @@
 #define OVERLAY_DECOUPLE 0
 #endif
 
-#ifdef LLK_TB_TEST
+#if defined(EN_KERNEL_SLOWDOWN)
 #include "kernel_slowdown_config.h"
 #endif
 
@@ -61,10 +61,10 @@
 #include "ckernel_include.h"
 #include "tensix.h"
 #include "fw_debug.h"
+#include "tt_log.h"
 // #include <cstring>
 #if defined(PERF_DUMP) || DELAY_EN > 0
 #include <l1_address_map.h>
-#include "tt_log.h"
 #include "perf_lib/scratch_api.h"
 #endif
 

diff --git a/tt_metal/hw/ckernels/wormhole_b0/common/inc/ckernel_sfpu.h b/tt_metal/hw/ckernels/wormhole_b0/common/inc/ckernel_sfpu.h
@@ -1319,5 +1319,25 @@ inline void _dequant_int32_(const int iterations, const uint dst_offset)
     }
 }
 
+template <bool APPROXIMATION_MODE, int ITERATIONS>
+inline void _add_int32_(const int iterations, const uint dst_offset) {
+    // Operand A is input1 (int32)
+    // Operand B is input2 (int32)
+    // Output is int32
+    #pragma GCC unroll 8
+    for (int d = 0; d < ITERATIONS; d++) {
+        // operand A - int32
+        TTI_SFPLOAD(0, 12, 3, 0);
+        // operand B - int32
+        TT_SFPLOAD(1, 12, 3, dst_offset * 64);
+        TTI_SFPIADD(0, 1, 0, 4);
+        // MAD has a 2-cycle pipeline latency so we need one cycle latency until next instr can consume the result
+        TTI_NOP;
+        // LREG_0 -> dest as int32
+        TTI_SFPSTORE(0, 12, 3, 0);
+        dst_reg++;
+    }
+}
+
 } // namespace sfpu
 } // namespace ckernel
diff --git a/tt_metal/hw/ckernels/wormhole_b0/llk_lib/llk_defs.h b/tt_metal/hw/ckernels/wormhole_b0/llk_lib/llk_defs.h
@@ -7,13 +7,11 @@
 
 namespace ckernel {
 
-enum Dim {
+enum VectorMode {
   None      = 0,
   R         = 1,
   C         = 2,
-  Z         = 3,
   RC        = 4,
-  ZR        = 5,
   Invalid   = 0xFF,
 };
 

diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_math_binary_sfpu_api.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_math_binary_sfpu_api.h
@@ -15,7 +15,7 @@ inline void llk_math_eltwise_binary_sfpu(
     const uint operand,
     uint dst_index_a,
     uint dst_index_b,
-    int vector_mode = (int)Dim::RC,
+    int vector_mode = (int)VectorMode::RC,
     uint param0 = 0,
     uint param1 = 0,
     uint param2 = 0,
@@ -38,7 +38,7 @@ inline void llk_math_eltwise_binary_sfpu_init(
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
 inline void llk_math_eltwise_binary_sfpu_quant_int32(
-    uint dst_index_a, uint dst_index_b, int vector_mode = (int)Dim::RC) {
+    uint dst_index_a, uint dst_index_b, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_binary_sfpu<SfpuType::quant_int32, APPROXIMATE, dst_sync>(dst_index_a, dst_index_b, vector_mode);
 }
 
@@ -49,7 +49,7 @@ inline void llk_math_eltwise_binary_sfpu_quant_int32_init(const uint zero_point)
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
 inline void llk_math_eltwise_binary_sfpu_requant_int32(
-    uint dst_index_a, uint dst_index_b, int vector_mode = (int)Dim::RC) {
+    uint dst_index_a, uint dst_index_b, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_binary_sfpu<SfpuType::requant_int32, APPROXIMATE, dst_sync>(dst_index_a, dst_index_b, vector_mode);
 }
 
@@ -60,7 +60,7 @@ inline void llk_math_eltwise_binary_sfpu_requant_int32_init(const uint zero_poin
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
 inline void llk_math_eltwise_binary_sfpu_dequant_int32(
-    uint dst_index_a, uint dst_index_b, int vector_mode = (int)Dim::RC) {
+    uint dst_index_a, uint dst_index_b, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_binary_sfpu<SfpuType::dequant_int32, APPROXIMATE, dst_sync>(dst_index_a, dst_index_b, vector_mode);
 }
 

diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_math_unary_sfpu_api.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_math_unary_sfpu_api.h
@@ -25,7 +25,7 @@ inline void llk_math_eltwise_unary_sfpu_rsqrt_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_log(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_log(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::log, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -45,7 +45,7 @@ inline void llk_math_eltwise_unary_sfpu_log_with_base_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_tanh(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_tanh(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::tanh, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -88,7 +88,7 @@ inline void llk_math_eltwise_unary_sfpu_dropout_init(uint seed = 0) {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_sigmoid(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_sigmoid(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::sigmoid, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -164,7 +164,7 @@ inline void llk_math_eltwise_unary_sfpu_gez_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_max(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_max(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::max, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -174,7 +174,7 @@ inline void llk_math_eltwise_unary_sfpu_max_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_square(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_square(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::square, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -184,7 +184,7 @@ inline void llk_math_eltwise_unary_sfpu_square_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_power(uint dst_index, int pow = 0, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_power(uint dst_index, int pow = 0, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::power, APPROXIMATE, dst_sync>(dst_index, vector_mode, pow);
 }
 
@@ -194,7 +194,7 @@ inline void llk_math_eltwise_unary_sfpu_power_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_abs(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_abs(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::abs, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -204,7 +204,7 @@ inline void llk_math_eltwise_unary_sfpu_abs_init() {
 }
 
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_cast_fp32_to_fp16a(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_cast_fp32_to_fp16a(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::cast_fp32_to_fp16a, APPROXIMATE, dst_sync>(dst_index, vector_mode);
 }
 
@@ -226,7 +226,7 @@ inline void llk_math_eltwise_unary_sfpu_exp2_init() {
 
 //heaviside
 template <bool APPROXIMATE, DstSync dst_sync = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_heaviside(uint dst_index,uint param0, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_heaviside(uint dst_index,uint param0, int vector_mode = (int)VectorMode::RC) {
     llk_math_eltwise_unary_sfpu<SfpuType::heaviside, APPROXIMATE, dst_sync>(dst_index,vector_mode,param0);
 }
 

diff --git a/...etal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_0_param.h b/...etal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_0_param.h
@@ -11,15 +11,15 @@ inline void llk_math_eltwise_unary_sfpu_0_param(
     void (*first_func)(),
     void (*func)(),
     uint dst_index,
-    int vector_mode = Dim::RC) {
+    int vector_mode = (int)VectorMode::RC) {
     if constexpr ((Dst == DstSync::SyncTile16) || (Dst == DstSync::SyncTile2)) {
         math::set_dst_write_addr<DstTileLayout::Default, DstTileShape::Tile32x32>(math_sync_tile_dst_index);
     } else {
         math::set_dst_write_addr<DstTileLayout::Default, DstTileShape::Tile32x32>(dst_index);
     }
     math::set_addr_mod_base();
     TTI_STALLWAIT(p_stall::STALL_SFPU, p_stall::MATH);
-    if (vector_mode == Dim::R) {
+    if (vector_mode == (int)VectorMode::R) {
         // Do a row vector, Face0 + Face1 -- first iteration (first row)
         const int ITERATIONS = 1;
 #pragma GCC unroll 0
@@ -33,7 +33,7 @@ inline void llk_math_eltwise_unary_sfpu_0_param(
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
-    } else if (vector_mode == Dim::C) {
+    } else if (vector_mode == (int)VectorMode::C) {
         // Do a column vector, Face0 + Face2 -- All iterations for full face
 #pragma GCC unroll 0
         for (int face = 0; face < 2; face++) {

diff --git a/...etal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_1_param.h b/...etal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_1_param.h
@@ -11,7 +11,7 @@ inline void llk_math_eltwise_unary_sfpu_1_param(
     void (*first_func)(uint),
     void (*func)(uint),
     uint dst_index,
-    int vector_mode = Dim::RC,
+    int vector_mode = (int)VectorMode::RC,
     uint param0 = 0) {
     if constexpr ((Dst == DstSync::SyncTile16) || (Dst == DstSync::SyncTile2)) {
         math::set_dst_write_addr<DstTileLayout::Default, DstTileShape::Tile32x32>(math_sync_tile_dst_index);
@@ -20,7 +20,7 @@ inline void llk_math_eltwise_unary_sfpu_1_param(
     }
     math::set_addr_mod_base();
     TTI_STALLWAIT(p_stall::STALL_SFPU, p_stall::MATH);
-    if (vector_mode == Dim::R) {
+    if (vector_mode == (int)VectorMode::R) {
         // Do a row vector, Face0 + Face1 -- first iteration (first row)
         const int ITERATIONS = 1;
 #pragma GCC unroll 0
@@ -34,7 +34,7 @@ inline void llk_math_eltwise_unary_sfpu_1_param(
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
         TTI_SETRWC(p_setrwc::CLR_NONE, p_setrwc::CR_D, 8, 0, 0, p_setrwc::SET_D);
-    } else if (vector_mode == Dim::C) {
+    } else if (vector_mode == (int)VectorMode::C) {
         // Do a column vector, Face0 + Face2 -- All iterations for full face
 #pragma GCC unroll 0
         for (int face = 0; face < 2; face++) {

diff --git a/...ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_common_includes.h b/...ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_common_includes.h
@@ -111,7 +111,7 @@ inline void llk_math_calculate_sfpu(
 template <SfpuType sfpu_op, bool APPROXIMATE, DstSync Dst = DstSync::SyncFull, bool IS_INT_SFPU_EN = false>
 inline void llk_math_eltwise_unary_sfpu(
     uint dst_index,
-    int vector_mode = (int)Dim::RC,
+    int vector_mode = (int)VectorMode::RC,
     uint param0 = 0,
     uint param1 = 0,
     uint param2 = 0,
@@ -126,7 +126,7 @@ inline void llk_math_eltwise_unary_sfpu(
 
     _llk_math_eltwise_unary_sfpu_start_<Dst>(dst_index);
 
-    if (vector_mode == (int)Dim::R) {
+    if (vector_mode == (int)VectorMode::R) {
         // Do a row vector, Face0 + Face1 -- first iteration (first row)
         const int iterations = (num_faces < 4) ? ((face_r_dim <= 2) ? 2 : face_r_dim / 2)
                                                : 2;  // At least 2 iterations for odd and even columns
@@ -140,7 +140,7 @@ inline void llk_math_eltwise_unary_sfpu(
         // Skip next two faces
         _llk_math_eltwise_unary_sfpu_inc_dst_face_addr_();
         _llk_math_eltwise_unary_sfpu_inc_dst_face_addr_();
-    } else if (vector_mode == (int)Dim::C) {
+    } else if (vector_mode == (int)VectorMode::C) {
         // Do a column vector, Face0 + Face2 if tile is 32x32 or Face0+Face1 if tiles is 32x16 -- All iterations for
         // full face
 #pragma GCC unroll 0

diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_elu.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_elu.h
@@ -24,7 +24,7 @@ inline void llk_math_eltwise_unary_sfpu_elu(uint dst_index, uint param0) {
     llk_math_eltwise_unary_sfpu_1_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_elu<APPROXIMATE>,
                                 ckernel::sfpu::calculate_elu<APPROXIMATE>,
-				                dst_index, Dim::RC, param0);
+				                dst_index, (int)VectorMode::RC, param0);
 }
 
 }
diff --git a/...tal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_erf_erfc.h b/...tal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_erf_erfc.h
@@ -29,15 +29,15 @@ inline void llk_math_eltwise_unary_sfpu_erf(uint dst_index, int param0 = 0) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_erf_erfc<SfpuType::erf, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_erf_erfc<SfpuType::erf, APPROXIMATE>,
-                                dst_index, Dim::RC);
+                                dst_index, (int)VectorMode::RC);
 }
 
 template <bool APPROXIMATE, DstSync Dst = DstSync::SyncFull>
 inline void llk_math_eltwise_unary_sfpu_erfc(uint dst_index, int param0 = 0) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_erf_erfc<SfpuType::erfc, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_erf_erfc<SfpuType::erfc, APPROXIMATE>,
-                                dst_index, Dim::RC);
+                                dst_index, (int)VectorMode::RC);
 }
 
 }
diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_erfinv.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_erfinv.h
@@ -24,7 +24,7 @@ inline void llk_math_eltwise_unary_sfpu_erfinv_op(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_erfinv<APPROXIMATE>,
                                 ckernel::sfpu::calculate_erfinv<APPROXIMATE>,
-                                dst_index, Dim::RC);
+                                dst_index, (int)VectorMode::RC);
 }
 
 }
diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_exp.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_exp.h
@@ -15,7 +15,7 @@ namespace ckernel {
 // New LLK SFPU APIs
 
 template <bool APPROXIMATE, DstSync Dst = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_exponential(uint dst_index, int vector_mode = Dim::RC, int param0 = 0) {
+inline void llk_math_eltwise_unary_sfpu_exponential(uint dst_index, int vector_mode = (int)VectorMode::RC, int param0 = 0) {
 
 	constexpr bool zero_negative = true;
     constexpr int first_iterations = 1;

diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_gelu.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_gelu.h
@@ -13,7 +13,7 @@ namespace ckernel {
 // New LLK SFPU APIs
 
 template <bool APPROXIMATE, DstSync Dst = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_gelu(uint dst_index, int vector_mode = Dim::RC, int param0=0) {
+inline void llk_math_eltwise_unary_sfpu_gelu(uint dst_index, int vector_mode = (int)VectorMode::RC, int param0=0) {
     constexpr int first_iterations = 1;
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_gelu<APPROXIMATE, first_iterations>,
@@ -27,7 +27,7 @@ inline void llk_math_eltwise_unary_sfpu_gelu_init() {
 }
 
 template <bool APPROXIMATE, DstSync Dst = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_gelu_derivative(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_gelu_derivative(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     constexpr int first_iterations = 1;
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_gelu_derivative<APPROXIMATE, first_iterations>,

diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_i0.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_i0.h
@@ -24,7 +24,7 @@ inline void llk_math_eltwise_unary_sfpu_i0_op(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_i0<APPROXIMATE>,
                                 ckernel::sfpu::calculate_i0<APPROXIMATE>,
-                                dst_index, Dim::RC);
+                                dst_index, (int)VectorMode::RC);
 }
 
 }
diff --git a/.../hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_isinf_isnan.h b/.../hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_isinf_isnan.h
@@ -26,7 +26,7 @@ inline void llk_math_eltwise_unary_sfpu_isinf(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isinf, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isinf, APPROXIMATE>,
- 				                dst_index, Dim::RC);
+ 				                dst_index, (int)VectorMode::RC);
 
 }
 
@@ -41,7 +41,7 @@ inline void llk_math_eltwise_unary_sfpu_isposinf(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isposinf, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isposinf, APPROXIMATE>,
- 				                dst_index,Dim::RC);
+ 				                dst_index,(int)VectorMode::RC);
 
 }
 
@@ -58,7 +58,7 @@ inline void llk_math_eltwise_unary_sfpu_isneginf(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isneginf, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isneginf, APPROXIMATE>,
-                                dst_index,Dim::RC);
+                                dst_index,(int)VectorMode::RC);
 
 }
 
@@ -73,7 +73,7 @@ inline void llk_math_eltwise_unary_sfpu_isnan(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isnan, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isnan, APPROXIMATE>,
- 				                dst_index,Dim::RC);
+ 				                dst_index,(int)VectorMode::RC);
 
 }
 
@@ -88,7 +88,7 @@ inline void llk_math_eltwise_unary_sfpu_isfinite(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isfinite, APPROXIMATE>,
                                 ckernel::sfpu::calculate_sfpu_isinf_isnan<SfpuType::isfinite, APPROXIMATE>,
- 				                dst_index,Dim::RC);
+ 				                dst_index,(int)VectorMode::RC);
 
 }
 

diff --git a/...kernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_logical_not_noti.h b/...kernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_logical_not_noti.h
@@ -24,7 +24,7 @@ inline void llk_math_eltwise_unary_sfpu_logical_not_unary_op(uint dst_index) {
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_logical_not_unary<APPROXIMATE>,
                                 ckernel::sfpu::calculate_logical_not_unary<APPROXIMATE>,
-                                dst_index, Dim::RC);
+                                dst_index, (int)VectorMode::RC);
 }
 
 }
diff --git a/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_recip.h b/tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_recip.h
@@ -14,7 +14,7 @@ namespace ckernel {
 // New LLK SFPU APIs
 
 template <bool APPROXIMATE, DstSync Dst = DstSync::SyncFull>
-inline void llk_math_eltwise_unary_sfpu_reciprocal(uint dst_index, int vector_mode = Dim::RC) {
+inline void llk_math_eltwise_unary_sfpu_reciprocal(uint dst_index, int vector_mode = (int)VectorMode::RC) {
     constexpr int first_iterations = 1;
     llk_math_eltwise_unary_sfpu_0_param<APPROXIMATE, Dst>
                                 (ckernel::sfpu::calculate_reciprocal<APPROXIMATE, first_iterations>,