#13385: Low-level interface for ttnn.round on Wormhole

tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_math_unary_sfpu_api.h

@@ -20,13 +20,13 @@
 #include "llk_math_eltwise_unary_sfpu_signbit.h"
 #include "llk_math_eltwise_unary_sfpu_floor.h"
 #include "llk_math_eltwise_unary_sfpu_ceil.h"
+#include "llk_math_eltwise_unary_sfpu_round.h"
 #include "llk_math_eltwise_unary_sfpu_silu.h"
 #include "llk_math_eltwise_unary_sfpu_square.h"
 #include "llk_math_eltwise_unary_sfpu_tanh.h"
 #include "llk_math_eltwise_unary_sfpu_topk.h"
 #include "llk_math_eltwise_unary_sfpu_unary_comp.h"
 #include "llk_math_eltwise_unary_sfpu_trigonometry.h"
-#include "llk_math_eltwise_unary_sfpu_unary_comp.h"
 #include "llk_math_eltwise_unary_sfpu_remainder.h"
 #include "llk_math_eltwise_unary_sfpu_bitwise_xor.h"
 #include "llk_math_eltwise_unary_sfpu_bitwise_not.h"

tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu/llk_math_eltwise_unary_sfpu_round.h

@@ -0,0 +1,27 @@
+// SPDX-FileCopyrightText: © 2023 Tenstorrent Inc.
+//
+// SPDX-License-Identifier: Apache-2.0
+
+#pragma once
+
+#include "llk_math_eltwise_unary_sfpu_init.h"
+#include "llk_math_eltwise_unary_sfpu_params.h"
+#include "ckernel_sfpu_round.h"
+
+namespace ckernel {
+
+// New LLK SFPU APIs
+
+template <bool APPROXIMATE>
+inline void llk_math_eltwise_unary_sfpu_round_init() {
+    llk_math_eltwise_unary_sfpu_init<SfpuType::round, APPROXIMATE>();
+}
+
+template <bool APPROXIMATE>
+inline void llk_math_eltwise_unary_sfpu_round(uint dst_index, int vector_mode = (int)VectorMode::RC) {
+    llk_math_eltwise_unary_sfpu_params<APPROXIMATE>
+                                (ckernel::sfpu::calculate_round<APPROXIMATE>,
+                                dst_index, vector_mode);
+}
+
+}

tt_metal/hw/ckernels/wormhole_b0/metal/llk_api/llk_sfpu_types.h

@@ -87,5 +87,6 @@ enum SfpuType {
     ceil,
     unused,
     reshuffle_rows,
-    cumsum
+    cumsum,
+    round,
 };

tt_metal/include/compute_kernel_api/eltwise_unary/round.h

@@ -0,0 +1,45 @@
+// SPDX-FileCopyrightText: © 2023 Tenstorrent Inc.
+//
+// SPDX-License-Identifier: Apache-2.0
+
+#pragma once
+
+
+#include "compute_kernel_api/common_globals.h"
+#ifdef TRISC_MATH
+#include "llk_math_eltwise_unary_sfpu_round.h"
+#define MAIN math_main()
+#define MATH(x) x
+#else
+#define MATH(x)
+#endif
+
+
+
+namespace ckernel {
+
+/**
+ * Please refer to documentation for any_init.
+ */
+ALWI void round_tile_init() {
+    MATH(( llk_math_eltwise_unary_sfpu_round_init<APPROX>() ));
+}
+
+/**
+ * Performs round operation on each row of a tile.
+ * in DST register at index tile_index. The DST register buffer must be in
+ * acquired state via *acquire_dst* call. This call is blocking and is only
+ * available on the compute engine.
+ *
+ * Return value: None
+ *
+ * | Argument        | Description                                                                | Type     | Valid Range                                           | Required |
+ * |-----------------|----------------------------------------------------------------------------|----------|-------------------------------------------------------|----------|
+ * | idst            | The index of the tile in DST register buffer to modify the sign bit of     | uint32_t | Must be less than the size of the DST register buffer | True     |
+ */
+ALWI void round_tile(uint32_t idst) {
+    MATH(( llk_math_eltwise_unary_sfpu_round<APPROX>(idst) ));
+}
+
+
+} // namespace ckernel

tt_metal/include/compute_kernel_api/eltwise_unary/sfpu_split_includes.h

@@ -92,6 +92,10 @@
 #include "compute_kernel_api/eltwise_unary/floor.h"
 #endif
 
+#if SFPU_OP_ROUND_INCLUDE
+#include "compute_kernel_api/eltwise_unary/round.h"
+#endif
+
 #if SFPU_OP_LEFT_SHIFT_INCLUDE
 #include "compute_kernel_api/eltwise_unary/left_shift.h"
 #endif