1 files changed, 104 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp16_mla_8x3VL.hpp b/src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp16_mla_8x3VL.hpp
new file mode 100644
index 0000000000..6c54167763
--- /dev/null
+++ b/src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp16_mla_8x3VL.hpp
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2019-2021, 2023 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#pragma once
+#ifdef ARM_COMPUTE_ENABLE_SVE
+
+#include "../std_transforms_sve.hpp"
+#include "../performance_parameters.hpp"
+
+#define ARGLIST  \
+    const __fp16 *, const __fp16 *, \
+    __fp16 *, int, int, int
+
+namespace arm_gemm
+{
+// Actual kernel implementations
+void sve_interleaved_fp16_mla_8x3VL( ARGLIST );
+void sve_interleaved_fp16_mla_8x3VL_a64fx( ARGLIST );
+
+class cls_sve_interleaved_fp16_mla_8x3VL
+{
+public:
+    typedef __fp16 operand_type;
+    typedef __fp16 result_type;
+
+    typedef void (*kern_type)( ARGLIST );
+
+    /* Kernel blocking parameters */
+    static constexpr unsigned int out_height()
+    {
+        return 8;
+    }
+
+    static unsigned int out_width()
+    {
+        return get_vector_length<__fp16>() * 3;
+    }
+
+    static constexpr unsigned int k_unroll()
+    {
+        return 1;
+    }
+
+
+    StdTransformsSVE<operand_type, result_type, 8, 3, 1, 1> transforms = {};
+    StdTransformsSVE<operand_type, result_type, 8, 3, 1, 1, true> transforms_quantized = {};
+    template<typename T>
+    static inline PerformanceParameters get_performance_parameters(const CPUInfo *ci)
+    {
+
+        if (std::is_same<T, __fp16>::value) {
+            switch (ci->get_cpu_model()) {
+                default:
+                    return { 15.96, 3.85, 6.91 };
+                case CPUModel::A510:
+                    return { 13.84, 2.07, 2.52 };
+                case CPUModel::V1:
+                    return { 31.90, 5.15, 10.34 };
+                case CPUModel::A64FX:
+                    return { 44.34, 3.23, 7.06 };
+            }
+        }
+
+        return { 1.0 };
+    }
+
+    // Default to the generic kernel
+    kern_type kernel=sve_interleaved_fp16_mla_8x3VL;
+    cls_sve_interleaved_fp16_mla_8x3VL(const CPUInfo *ci)
+    {
+        switch(ci->get_cpu_model()) {
+            default:
+                break;
+            case CPUModel::A64FX:
+                kernel=sve_interleaved_fp16_mla_8x3VL_a64fx;
+                break;
+        }
+    }
+};
+
+} // namespace arm_gemm
+
+#undef ARGLIST
+#endif // ARM_COMPUTE_ENABLE_SVE