Update CPU kernel implementations and guard directives

Resolves COMPMID-6023

Change-Id: I868975d14c4f98af6716726feda22405a6a4c891
Signed-off-by: Michael Tyler <michael.tyler@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9686
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 27 insertions, 26 deletions

diff --git a/src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp b/src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp
index 8041453cb1..4b073b9076 100644
--- a/src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp
+++ b/src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp
@@ -22,6 +22,7 @@
  * SOFTWARE.
  */
 
+
 #include <cstddef>
 #include <cstdint>
 
@@ -111,7 +112,7 @@ void a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst_impl(
     "fmax v18.8h, v18.8h, v21.8h\n"
     "fmax v17.8h, v17.8h, v20.8h\n"
     "add x15, x15, #0x10\n"
-    "fmax v16.8h, v16.8h, v20.8h\n"
+    "fmax v16.8h, v20.8h, v16.8h\n"
     "str q19, [x14, x12]\n"
     "str q18, [x13, x12]\n"
     "str q17, [x11, x12]\n"
@@ -121,43 +122,43 @@ void a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst_impl(
     "2:"  // Vector: Tail
     "fmax v21.8h, v30.8h, v29.8h\n"
     "fmax v20.8h, v29.8h, v28.8h\n"
-    "fmax v19.8h, v27.8h, v26.8h\n"
+    "fmax v16.8h, v27.8h, v26.8h\n"
     "fmax v18.8h, v25.8h, v24.8h\n"
     "fmax v17.8h, v27.8h, v23.8h\n"
-    "fmax v16.8h, v24.8h, v22.8h\n"
-    "fmax v19.8h, v21.8h, v19.8h\n"
+    "fmax v19.8h, v24.8h, v22.8h\n"
+    "fmax v16.8h, v21.8h, v16.8h\n"
     "fmax v18.8h, v18.8h, v21.8h\n"
-    "str q19, [x14, x12]\n"
+    "str q16, [x14, x12]\n"
     "fmax v17.8h, v17.8h, v20.8h\n"
-    "fmax v16.8h, v16.8h, v20.8h\n"
+    "fmax v16.8h, v20.8h, v19.8h\n"
     "str q18, [x13, x12]\n"
     "str q17, [x11, x12]\n"
     "str q16, [x10, x12]\n"
     "add x12, x12, #0x10\n"
     "cbz x16, 4f\n"
     "3:"  // Oddments
-    "ldr h30, [x28, x15]\n"
-    "ldr h29, [x25, x15]\n"
-    "fmax v21.8h, v30.8h, v29.8h\n"
+    "ldr h16, [x28, x15]\n"
+    "ldr h17, [x25, x15]\n"
+    "fmax v23.8h, v16.8h, v17.8h\n"
     "subs x16, x16, #0x1\n"
-    "ldr h28, [x22, x15]\n"
-    "ldr h27, [x26, x15]\n"
-    "fmax v20.8h, v29.8h, v28.8h\n"
-    "ldr h26, [x9, x15]\n"
-    "ldr h25, [x27, x15]\n"
-    "fmax v19.8h, v27.8h, v26.8h\n"
-    "fmax v19.8h, v21.8h, v19.8h\n"
-    "ldr h24, [x24, x15]\n"
-    "ldr h23, [x23, x15]\n"
-    "fmax v18.8h, v25.8h, v24.8h\n"
-    "fmax v17.8h, v27.8h, v23.8h\n"
-    "ldr h22, [x21, x15]\n"
-    "fmax v16.8h, v24.8h, v22.8h\n"
+    "ldr h16, [x22, x15]\n"
+    "ldr h22, [x26, x15]\n"
+    "fmax v21.8h, v17.8h, v16.8h\n"
+    "ldr h16, [x9, x15]\n"
+    "ldr h17, [x27, x15]\n"
+    "fmax v16.8h, v22.8h, v16.8h\n"
+    "fmax v20.8h, v23.8h, v16.8h\n"
+    "ldr h19, [x24, x15]\n"
+    "ldr h16, [x23, x15]\n"
+    "fmax v18.8h, v17.8h, v19.8h\n"
+    "fmax v17.8h, v22.8h, v16.8h\n"
+    "ldr h16, [x21, x15]\n"
+    "fmax v16.8h, v19.8h, v16.8h\n"
     "add x15, x15, #0x2\n"
-    "fmax v18.8h, v18.8h, v21.8h\n"
-    "fmax v17.8h, v17.8h, v20.8h\n"
-    "fmax v16.8h, v16.8h, v20.8h\n"
-    "str h19, [x14, x12]\n"
+    "fmax v18.8h, v18.8h, v23.8h\n"
+    "fmax v17.8h, v17.8h, v21.8h\n"
+    "fmax v16.8h, v21.8h, v16.8h\n"
+    "str h20, [x14, x12]\n"
     "str h18, [x13, x12]\n"
     "str h17, [x11, x12]\n"
     "str h16, [x10, x12]\n"