COMPMID-632 Assembly: Integrate gemmlowp assembly version

Integrate generic gemmlowp assembly version for u8.

Change-Id: I17ed4494c25a132b2bac581febe1544e49b4f352
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/110114
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>

2 files changed, 526 insertions, 0 deletions

diff --git a/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4.hpp b/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4.hpp
new file mode 100644
index 0000000000..3561bfec96
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4.hpp
@@ -0,0 +1,61 @@
+/*
+ * Copyright (c) 2017 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 __aarch64__
+
+// Load the actual kernel
+#include "a64_gemm_u8_4x4/generic.hpp"
+
+class gemm_u8_4x4 {
+public:
+    typedef uint8_t operand_type;
+    typedef uint32_t result_type;
+
+    typedef void (*kern_type)(const uint8_t *, const uint8_t *, uint32_t *, int, int, int);
+
+    /* Describes the data layout for A input */
+    static const int A_interleave = 4;
+    static const int A_block = 16;
+    static const bool A_transpose = false;
+
+    /* Same for B input */
+    static const int B_interleave = 4;
+    static const int B_block = 16;
+    static const bool B_transpose = true;
+
+    /* Kernel blocking parameters */
+    static const int out_width = 4;
+    static const int out_height = 4;
+    static const int k_unroll = 16;
+
+    kern_type kernel = nullptr;
+
+    gemm_u8_4x4(const CPUInfo *ci) {
+        kernel = a64_gemm_u8_4x4;
+    }
+};
+
+#endif // __aarch64__
+
diff --git a/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4/generic.hpp b/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4/generic.hpp
new file mode 100644
index 0000000000..e48c373f21
--- /dev/null
+++ b/arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4/generic.hpp
@@ -0,0 +1,465 @@
+/*
+ * Copyright (c) 2017 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 __aarch64__
+
+#include <arm_neon.h>
+
+inline void a64_gemm_u8_4x4(const uint8_t *Apanel, const uint8_t *Bpanel, uint32_t *Cpanel, int ablocks, int bblocks, int K) {
+    const uint8_t *a_ptr = Apanel;
+    uint32_t *c_ptr = Cpanel;
+    K /= 16;
+    int oddk = (K & 1);
+
+    for (int yb=0; yb<ablocks; yb++) {
+        const uint8_t *a_ptr0 = a_ptr;
+        const uint8_t *b_ptr = Bpanel;
+
+        for (int xb=0; xb<bblocks; xb++) {
+            a_ptr = a_ptr0;
+
+            int k = ((K+1)/2)-1;
+
+            register uint8x16_t b0  asm("v4");
+            register uint8x16_t b1  asm("v5");
+            register uint8x16_t b2  asm("v6");
+            register uint8x16_t b3  asm("v7");
+            register uint8x16_t b0a asm("v8");
+            register uint8x16_t b1a asm("v9");
+            register uint8x16_t b2a asm("v10");
+            register uint8x16_t b3a asm("v11");
+
+            __asm __volatile (
+                "movi	v16.4s, #0x0\n"
+                "ldr	q0, [%[a_ptr]]\n"
+                "movi	v17.4s, #0x0\n"
+                "ldr	%q[b0], [%[b_ptr]]\n"
+                "movi	v18.4s, #0x0\n"
+                "ldr	%q[b1], [%[b_ptr], #16]\n"
+                "movi	v19.4s, #0x0\n"
+                "ldr	%q[b2], [%[b_ptr], #32]\n"
+                "movi	v20.4s, #0x0\n"
+                "ldr	%q[b3], [%[b_ptr], #48]\n"
+                "movi	v21.4s, #0x0\n"
+                "ldr	q1, [%[a_ptr], #16]\n"
+                "movi	v22.4s, #0x0\n"
+                "ldr	q2, [%[a_ptr], #32]\n"
+                "movi	v23.4s, #0x0\n"
+                "ldr	q3, [%[a_ptr], #48]\n"
+                "movi	v24.4s, #0x0\n"
+                ASM_PREFETCH("[%[b_ptr], #64]")
+                "movi	v25.4s, #0x0\n"
+                ASM_PREFETCH("[%[a_ptr], #64]")
+                "movi	v26.4s, #0x0\n"
+                ASM_PREFETCH("[%[b_ptr], #128]")
+                "movi	v27.4s, #0x0\n"
+                ASM_PREFETCH("[%[a_ptr], #128]")
+                "movi	v28.4s, #0x0\n"
+                ASM_PREFETCH("[%[b_ptr], #192]")
+                "movi	v29.4s, #0x0\n"
+                ASM_PREFETCH("[%[a_ptr], #192]")
+                "movi	v30.4s, #0x0\n"
+                ASM_PREFETCH("[%[b_ptr], #256]")
+                "movi	v31.4s, #0x0\n"
+                ASM_PREFETCH("[%[a_ptr], #256]")
+
+                // Loop structure optimized for A57 (after r0).
+
+                // Unavoidably, the multiply will "dribble" if
+                // dual issued with an add.
+
+                // Minimize the effect of this by making sure
+                // there are 2 adds to run under the dribbled
+                // multiply.
+
+                // Pipeline in blocks of 8 multiplies - combine
+                // this iteration's multiplies with adds from
+                // the previous iteration.
+
+                // So the first block doesn't have any adds to
+                // do - but because all the adds are at the
+                // start of the block it's only the first couple
+                // of multiplies that need to be pulled out.
+
+                // Start of unroll 0 (first iteration)
+                "umull	v12.8h, v0.8b, %[b0].8b\n"
+                "umull	v13.8h, v0.8b, %[b1].8b\n"
+
+                // Skip loop if we are doing zero iterations of it.
+                "cbz	%w[k], 4f\n"
+
+                // Unroll 0 continuation (branch target)
+                "1:\n"
+                "umull	v14.8h, v0.8b, %[b2].8b\n"
+                "subs	%w[k], %w[k], #1\n"
+                "umull	v15.8h, v0.8b, %[b3].8b\n"
+                "ldr	%q[b0a], [%[b_ptr], #64]\n"
+                "umlal2	v12.8h, v0.16b, %[b0].16b\n"
+                "umlal2	v13.8h, v0.16b, %[b1].16b\n"
+                "ldr	%q[b1a], [%[b_ptr], #80]\n"
+                "umlal2	v14.8h, v0.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v0.16b, %[b3].16b\n"
+                "ldr 	q0, [%[a_ptr], #64]\n"
+
+                "uadalp	v16.4s, v12.8h\n"
+                "umull	v12.8h, v1.8b, %[b0].8b\n"
+                "uadalp	v17.4s, v13.8h\n"
+                "uadalp	v18.4s, v14.8h\n"
+                "umull	v13.8h, v1.8b, %[b1].8b\n"
+                "uadalp	v19.4s, v15.8h\n"
+                "umull	v14.8h, v1.8b, %[b2].8b\n"
+                "ldr	%q[b2a], [%[b_ptr], #96]\n"
+                "umull	v15.8h, v1.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v1.16b, %[b0].16b\n"
+                "ldr	%q[b3a], [%[b_ptr], #112]\n"
+                "umlal2	v13.8h, v1.16b, %[b1].16b\n"
+                "add	%[b_ptr], %[b_ptr], #128\n"
+                "umlal2	v14.8h, v1.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v1.16b, %[b3].16b\n"
+                "ldr 	q1, [%[a_ptr], #80]\n"
+
+                "uadalp	v20.4s, v12.8h\n"
+                "umull	v12.8h, v2.8b, %[b0].8b\n"
+                "uadalp	v21.4s, v13.8h\n"
+                "uadalp	v22.4s, v14.8h\n"
+                "umull	v13.8h, v2.8b, %[b1].8b\n"
+                "uadalp	v23.4s, v15.8h\n"
+                "umull	v14.8h, v2.8b, %[b2].8b\n"
+                "umull	v15.8h, v2.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v2.16b, %[b0].16b\n"
+                ASM_PREFETCH("[%[b_ptr], #192]")
+                "umlal2	v13.8h, v2.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v2.16b, %[b2].16b\n"
+                ASM_PREFETCH("[%[a_ptr], #320]")
+                "umlal2	v15.8h, v2.16b, %[b3].16b\n"
+                "ldr 	q2, [%[a_ptr], #96]\n"
+
+                "uadalp	v24.4s, v12.8h\n"
+                "umull	v12.8h, v3.8b, %[b0].8b\n"
+                "uadalp	v25.4s, v13.8h\n"
+                "uadalp	v26.4s, v14.8h\n"
+                "umull	v13.8h, v3.8b, %[b1].8b\n"
+                "uadalp	v27.4s, v15.8h\n"
+                "umull	v14.8h, v3.8b, %[b2].8b\n"
+                "umull	v15.8h, v3.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v3.16b, %[b0].16b\n"
+                "ldr 	%q[b0], [%[b_ptr], #0]\n"
+                "umlal2	v13.8h, v3.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v3.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v3.16b, %[b3].16b\n"
+                "ldr 	q3, [%[a_ptr], #112]\n"
+
+                // Unroll 1
+                "uadalp	v28.4s, v12.8h\n"
+                "umull	v12.8h, v0.8b, %[b0a].8b\n"
+                "uadalp	v29.4s, v13.8h\n"
+                "uadalp	v30.4s, v14.8h\n"
+                "umull	v13.8h, v0.8b, %[b1a].8b\n"
+                "uadalp	v31.4s, v15.8h\n"
+                "umull	v14.8h, v0.8b, %[b2a].8b\n"
+                "umull	v15.8h, v0.8b, %[b3a].8b\n"
+                "ldr 	%q[b1], [%[b_ptr], #16]\n"
+                "umlal2	v12.8h, v0.16b, %[b0a].16b\n"
+                "umlal2	v13.8h, v0.16b, %[b1a].16b\n"
+                "ldr 	%q[b2], [%[b_ptr], #32]\n"
+                "umlal2	v14.8h, v0.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v0.16b, %[b3a].16b\n"
+                "ldr 	q0, [%[a_ptr], #128]\n"
+
+                "uadalp	v16.4s, v12.8h\n"
+                "umull	v12.8h, v1.8b, %[b0a].8b\n"
+                "uadalp	v17.4s, v13.8h\n"
+                "uadalp	v18.4s, v14.8h\n"
+                "umull	v13.8h, v1.8b, %[b1a].8b\n"
+                "uadalp	v19.4s, v15.8h\n"
+                "add	%[a_ptr], %[a_ptr], #128\n"
+                "umull	v14.8h, v1.8b, %[b2a].8b\n"
+                "umull	v15.8h, v1.8b, %[b3a].8b\n"
+                "ldr 	%q[b3], [%[b_ptr], #48]\n"
+                "umlal2	v12.8h, v1.16b, %[b0a].16b\n"
+                "umlal2	v13.8h, v1.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v1.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v1.16b, %[b3a].16b\n"
+                "ldr 	q1, [%[a_ptr], #16]\n"
+
+                "uadalp	v20.4s, v12.8h\n"
+                "umull	v12.8h, v2.8b, %[b0a].8b\n"
+                "uadalp	v21.4s, v13.8h\n"
+                "uadalp	v22.4s, v14.8h\n"
+                "umull	v13.8h, v2.8b, %[b1a].8b\n"
+                "uadalp	v23.4s, v15.8h\n"
+                "umull	v14.8h, v2.8b, %[b2a].8b\n"
+                "umull	v15.8h, v2.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v2.16b, %[b0a].16b\n"
+                ASM_PREFETCH("[%[b_ptr], #256]")
+                "umlal2	v13.8h, v2.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v2.16b, %[b2a].16b\n"
+                ASM_PREFETCH("[%[a_ptr], #256]")
+                "umlal2	v15.8h, v2.16b, %[b3a].16b\n"
+                "ldr 	q2, [%[a_ptr], #32]\n"
+
+                "uadalp	v24.4s, v12.8h\n"
+                "umull	v12.8h, v3.8b, %[b0a].8b\n"
+                "uadalp	v25.4s, v13.8h\n"
+                "uadalp	v26.4s, v14.8h\n"
+                "umull	v13.8h, v3.8b, %[b1a].8b\n"
+                "uadalp	v27.4s, v15.8h\n"
+                "umull	v14.8h, v3.8b, %[b2a].8b\n"
+                "umull	v15.8h, v3.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v3.16b, %[b0a].16b\n"
+                "umlal2	v13.8h, v3.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v3.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v3.16b, %[b3a].16b\n"
+                "ldr 	q3, [%[a_ptr], #48]\n"
+
+                // Start of unroll 0 for next iteration.
+                "uadalp	v28.4s, v12.8h\n"
+                "umull	v12.8h, v0.8b, %[b0].8b\n"
+                "uadalp	v29.4s, v13.8h\n"
+                "uadalp	v30.4s, v14.8h\n"
+                "umull	v13.8h, v0.8b, %[b1].8b\n"
+                "uadalp	v31.4s, v15.8h\n"
+                "bne	1b\n"
+
+                // Target to use when K=1 or 2 (i.e. zero iterations of main loop)
+                "4:\n"
+
+                // Branch to alternative tail for odd K
+                "cbnz	%w[oddk], 2f\n"
+
+                // Detached final iteration (even K)
+                "umull	v14.8h, v0.8b, %[b2].8b\n"
+                "umull	v15.8h, v0.8b, %[b3].8b\n"
+                "ldr	%q[b0a], [%[b_ptr], #64]\n"
+                "umlal2	v12.8h, v0.16b, %[b0].16b\n"
+                "umlal2	v13.8h, v0.16b, %[b1].16b\n"
+                "ldr	%q[b1a], [%[b_ptr], #80]\n"
+                "umlal2	v14.8h, v0.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v0.16b, %[b3].16b\n"
+                "ldr 	q0, [%[a_ptr], #64]\n"
+
+                "uadalp	v16.4s, v12.8h\n"
+                "umull	v12.8h, v1.8b, %[b0].8b\n"
+                "uadalp	v17.4s, v13.8h\n"
+                "uadalp	v18.4s, v14.8h\n"
+                "umull	v13.8h, v1.8b, %[b1].8b\n"
+                "uadalp	v19.4s, v15.8h\n"
+                "umull	v14.8h, v1.8b, %[b2].8b\n"
+                "ldr	%q[b2a], [%[b_ptr], #96]\n"
+                "umull	v15.8h, v1.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v1.16b, %[b0].16b\n"
+                "ldr	%q[b3a], [%[b_ptr], #112]\n"
+                "umlal2	v13.8h, v1.16b, %[b1].16b\n"
+                "add	%[b_ptr], %[b_ptr], #128\n"
+                "umlal2	v14.8h, v1.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v1.16b, %[b3].16b\n"
+                "ldr 	q1, [%[a_ptr], #80]\n"
+
+                "uadalp	v20.4s, v12.8h\n"
+                "umull	v12.8h, v2.8b, %[b0].8b\n"
+                "uadalp	v21.4s, v13.8h\n"
+                "uadalp	v22.4s, v14.8h\n"
+                "umull	v13.8h, v2.8b, %[b1].8b\n"
+                "uadalp	v23.4s, v15.8h\n"
+                "umull	v14.8h, v2.8b, %[b2].8b\n"
+                "umull	v15.8h, v2.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v2.16b, %[b0].16b\n"
+                "umlal2	v13.8h, v2.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v2.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v2.16b, %[b3].16b\n"
+                "ldr 	q2, [%[a_ptr], #96]\n"
+
+                "uadalp	v24.4s, v12.8h\n"
+                "umull	v12.8h, v3.8b, %[b0].8b\n"
+                "uadalp	v25.4s, v13.8h\n"
+                "uadalp	v26.4s, v14.8h\n"
+                "umull	v13.8h, v3.8b, %[b1].8b\n"
+                "uadalp	v27.4s, v15.8h\n"
+                "umull	v14.8h, v3.8b, %[b2].8b\n"
+                "umull	v15.8h, v3.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v3.16b, %[b0].16b\n"
+                "umlal2	v13.8h, v3.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v3.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v3.16b, %[b3].16b\n"
+                "ldr 	q3, [%[a_ptr], #112]\n"
+
+                // Unroll 1
+                "uadalp	v28.4s, v12.8h\n"
+                "umull	v12.8h, v0.8b, %[b0a].8b\n"
+                "uadalp	v29.4s, v13.8h\n"
+                "uadalp	v30.4s, v14.8h\n"
+                "umull	v13.8h, v0.8b, %[b1a].8b\n"
+                "uadalp	v31.4s, v15.8h\n"
+                "umull	v14.8h, v0.8b, %[b2a].8b\n"
+                "add	%[a_ptr], %[a_ptr], #128\n"
+                "umull	v15.8h, v0.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v0.16b, %[b0a].16b\n"
+                "umlal2	v13.8h, v0.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v0.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v0.16b, %[b3a].16b\n"
+
+                "uadalp	v16.4s, v12.8h\n"
+                "umull	v12.8h, v1.8b, %[b0a].8b\n"
+                "uadalp	v17.4s, v13.8h\n"
+                "uadalp	v18.4s, v14.8h\n"
+                "umull	v13.8h, v1.8b, %[b1a].8b\n"
+                "uadalp	v19.4s, v15.8h\n"
+                "umull	v14.8h, v1.8b, %[b2a].8b\n"
+                "umull	v15.8h, v1.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v1.16b, %[b0a].16b\n"
+                "addp	v16.4s, v16.4s, v17.4s\n"
+                "umlal2	v13.8h, v1.16b, %[b1a].16b\n"
+                "addp	v17.4s, v18.4s, v19.4s\n"
+                "umlal2	v14.8h, v1.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v1.16b, %[b3a].16b\n"
+
+                "uadalp	v20.4s, v12.8h\n"
+                "umull	v12.8h, v2.8b, %[b0a].8b\n"
+                "uadalp	v21.4s, v13.8h\n"
+                "uadalp	v22.4s, v14.8h\n"
+                "umull	v13.8h, v2.8b, %[b1a].8b\n"
+                "uadalp	v23.4s, v15.8h\n"
+                "addp	v16.4s, v16.4s, v17.4s\n"
+                "umull	v14.8h, v2.8b, %[b2a].8b\n"
+                "addp	v18.4s, v20.4s, v21.4s\n"
+                "addp	v19.4s, v22.4s, v23.4s\n"
+                "umull	v15.8h, v2.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v2.16b, %[b0a].16b\n"
+                "str	q16, [%[c_ptr]]\n"
+                "umlal2	v13.8h, v2.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v2.16b, %[b2a].16b\n"
+                "umlal2	v15.8h, v2.16b, %[b3a].16b\n"
+
+                "uadalp	v24.4s, v12.8h\n"
+                "umull	v12.8h, v3.8b, %[b0a].8b\n"
+                "uadalp	v25.4s, v13.8h\n"
+                "uadalp	v26.4s, v14.8h\n"
+                "umull	v13.8h, v3.8b, %[b1a].8b\n"
+                "uadalp	v27.4s, v15.8h\n"
+                "addp	v17.4s, v18.4s, v19.4s\n"
+                "umull	v14.8h, v3.8b, %[b2a].8b\n"
+                "addp	v20.4s, v24.4s, v25.4s\n"
+                "addp	v21.4s, v26.4s, v27.4s\n"
+                "umull	v15.8h, v3.8b, %[b3a].8b\n"
+                "umlal2	v12.8h, v3.16b, %[b0a].16b\n"
+                "str	q17, [%[c_ptr], #16]\n"
+                "umlal2	v13.8h, v3.16b, %[b1a].16b\n"
+                "umlal2	v14.8h, v3.16b, %[b2a].16b\n"
+                "addp	v18.4s, v20.4s, v21.4s\n"
+                "umlal2	v15.8h, v3.16b, %[b3a].16b\n"
+                "b	3f\n"
+
+                // Detached final iteration (odd K)
+                "2:\n"
+                "umull	v14.8h, v0.8b, %[b2].8b\n"
+                "add	%[a_ptr], %[a_ptr], #64\n"
+                "umull	v15.8h, v0.8b, %[b3].8b\n"
+                "add	%[b_ptr], %[b_ptr], #64\n"
+                "umlal2	v12.8h, v0.16b, %[b0].16b\n"
+                "umlal2	v13.8h, v0.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v0.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v0.16b, %[b3].16b\n"
+
+                "uadalp	v16.4s, v12.8h\n"
+                "umull	v12.8h, v1.8b, %[b0].8b\n"
+                "uadalp	v17.4s, v13.8h\n"
+                "uadalp	v18.4s, v14.8h\n"
+                "umull	v13.8h, v1.8b, %[b1].8b\n"
+                "uadalp	v19.4s, v15.8h\n"
+                "umull	v14.8h, v1.8b, %[b2].8b\n"
+                "umull	v15.8h, v1.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v1.16b, %[b0].16b\n"
+                "addp	v16.4s, v16.4s, v17.4s\n"
+                "umlal2	v13.8h, v1.16b, %[b1].16b\n"
+                "addp	v17.4s, v18.4s, v19.4s\n"
+                "umlal2	v14.8h, v1.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v1.16b, %[b3].16b\n"
+
+                "uadalp	v20.4s, v12.8h\n"
+                "umull	v12.8h, v2.8b, %[b0].8b\n"
+                "uadalp	v21.4s, v13.8h\n"
+                "uadalp	v22.4s, v14.8h\n"
+                "umull	v13.8h, v2.8b, %[b1].8b\n"
+                "uadalp	v23.4s, v15.8h\n"
+                "addp	v16.4s, v16.4s, v17.4s\n"
+                "umull	v14.8h, v2.8b, %[b2].8b\n"
+                "addp	v18.4s, v20.4s, v21.4s\n"
+                "addp	v19.4s, v22.4s, v23.4s\n"
+                "umull	v15.8h, v2.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v2.16b, %[b0].16b\n"
+                "str	q16, [%[c_ptr]]\n"
+                "umlal2	v13.8h, v2.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v2.16b, %[b2].16b\n"
+                "umlal2	v15.8h, v2.16b, %[b3].16b\n"
+
+                "uadalp	v24.4s, v12.8h\n"
+                "umull	v12.8h, v3.8b, %[b0].8b\n"
+                "uadalp	v25.4s, v13.8h\n"
+                "uadalp	v26.4s, v14.8h\n"
+                "umull	v13.8h, v3.8b, %[b1].8b\n"
+                "uadalp	v27.4s, v15.8h\n"
+                "addp	v17.4s, v18.4s, v19.4s\n"
+                "umull	v14.8h, v3.8b, %[b2].8b\n"
+                "addp	v20.4s, v24.4s, v25.4s\n"
+                "addp	v21.4s, v26.4s, v27.4s\n"
+                "umull	v15.8h, v3.8b, %[b3].8b\n"
+                "umlal2	v12.8h, v3.16b, %[b0].16b\n"
+                "str	q17, [%[c_ptr], #16]\n"
+                "umlal2	v13.8h, v3.16b, %[b1].16b\n"
+                "umlal2	v14.8h, v3.16b, %[b2].16b\n"
+                "addp	v18.4s, v20.4s, v21.4s\n"
+                "umlal2	v15.8h, v3.16b, %[b3].16b\n"
+
+                "3:\n"
+
+                // Final additions
+                "uadalp	v28.4s, v12.8h\n"
+                "str	q18, [%[c_ptr], #32]\n"
+                "uadalp	v29.4s, v13.8h\n"
+                "uadalp	v30.4s, v14.8h\n"
+                "uadalp	v31.4s, v15.8h\n"
+
+                // Horizontal reduction, phase 1
+                "addp	v22.4s, v28.4s, v29.4s\n"
+                "addp	v23.4s, v30.4s, v31.4s\n"
+
+                // Horizontal reduction, phase 2
+                "addp	v19.4s, v22.4s, v23.4s\n"
+                "str	q19, [%[c_ptr], #48]\n"
+                "add	%[c_ptr], %[c_ptr], #64\n"
+
+            :
+              [a_ptr] "+r" (a_ptr), [b_ptr] "+r" (b_ptr), [c_ptr] "+r" (c_ptr),
+              [b0] "+w" (b0), [b1] "+w" (b1), [b2] "+w" (b2), [b3] "+w" (b3),
+              [b0a] "+w" (b0a), [b1a] "+w" (b1a), [b2a] "+w" (b2a), [b3a] "+w" (b3a),
+              [k] "+r" (k)
+            : [oddk] "r" (oddk)
+            : "x20", "x21", "v0","v1","v2","v3","v12","v13","v14","v15","v16","v17","v18","v19",
+              "v20","v21","v22","v23","v24","v25","v26","v27","v28","v29","v30","v31", "cc");
+        }
+    }
+}
+
+#endif // __aarch64__