COMPMID-2378: Sanitize GEMM configuration for NEON
Change-Id: I7859b82b2059e14685f8792424648ac5eacd67f1
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1418
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp
index f4485bc..e1af2d4 100644
--- a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp
+++ b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp
@@ -61,12 +61,16 @@
switch ((y + 5) - ymax) {
case 4:
outptr1 = dummyres;
+ // fall through
case 3:
outptr2 = dummyres;
+ // fall through
case 2:
outptr3 = dummyres;
+ // fall through
case 1:
outptr4 = dummyres;
+ // fall through
case 0:
outptr5 = dummyres;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp
index be23978..9fca4e3 100644
--- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp
+++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp
@@ -63,16 +63,22 @@
switch ((y + 7) - ymax) {
case 6:
outptr1 = dummyres;
+ // fall through
case 5:
outptr2 = dummyres;
+ // fall through
case 4:
outptr3 = dummyres;
+ // fall through
case 3:
outptr4 = dummyres;
+ // fall through
case 2:
outptr5 = dummyres;
+ // fall through
case 1:
outptr6 = dummyres;
+ // fall through
case 0:
outptr7 = dummyres;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp
index 9e5eb88..0e638ee 100644
--- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp
+++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp
@@ -66,16 +66,22 @@
switch ((y + 7) - ymax) {
case 6:
outptr1 = dummyres;
+ // fall through
case 5:
outptr2 = dummyres;
+ // fall through
case 4:
outptr3 = dummyres;
+ // fall through
case 3:
outptr4 = dummyres;
+ // fall through
case 2:
outptr5 = dummyres;
+ // fall through
case 1:
outptr6 = dummyres;
+ // fall through
case 0:
outptr7 = dummyres;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_half_24x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_half_24x8.hpp
index 3ed43b1..60cc2f3 100644
--- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_half_24x8.hpp
+++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_half_24x8.hpp
@@ -65,16 +65,22 @@
switch ((y + 7) - ymax) {
case 6:
outptr1 = dummyres;
+ // fall through
case 5:
outptr2 = dummyres;
+ // fall through
case 4:
outptr3 = dummyres;
+ // fall through
case 3:
outptr4 = dummyres;
+ // fall through
case 2:
outptr5 = dummyres;
+ // fall through
case 1:
outptr6 = dummyres;
+ // fall through
case 0:
outptr7 = dummyres;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp
index 35d4cc5..0212dfd 100644
--- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp
+++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp
@@ -63,16 +63,22 @@
switch ((y + 7) - ymax) {
case 6:
outptr1 = dummyres;
+ // fall through
case 5:
outptr2 = dummyres;
+ // fall through
case 4:
outptr3 = dummyres;
+ // fall through
case 3:
outptr4 = dummyres;
+ // fall through
case 2:
outptr5 = dummyres;
+ // fall through
case 1:
outptr6 = dummyres;
+ // fall through
case 0:
outptr7 = dummyres;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a32_interleave_6way_32bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a32_interleave_6way_32bit.hpp
index 20ad301..a460fdf 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a32_interleave_6way_32bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a32_interleave_6way_32bit.hpp
@@ -60,12 +60,16 @@
/* Everything falls through in here */
case 4:
inptr1 = zerobuff;
+ // fall through
case 3:
inptr2 = zerobuff;
+ // fall through
case 2:
inptr3 = zerobuff;
+ // fall through
case 1:
inptr4 = zerobuff;
+ // fall through
case 0:
inptr5 = zerobuff;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_block16_interleave4_8bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_block16_interleave4_8bit.hpp
index 2f513a6..6a15fc4 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_block16_interleave4_8bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_block16_interleave4_8bit.hpp
@@ -57,8 +57,10 @@
/* Everything falls through in here */
case 2:
inptr1 = zerobuff;
+ // fall through
case 1:
inptr2 = zerobuff;
+ // fall through
case 0:
inptr3 = zerobuff;
break;
@@ -93,8 +95,10 @@
/* Everything falls through in here */
case 2:
inptr1 = zerobuff;
+ // fall through
case 1:
inptr2 = zerobuff;
+ // fall through
case 0:
inptr3 = zerobuff;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_16bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_16bit.hpp
index 27136d1..0028ab0 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_16bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_16bit.hpp
@@ -64,16 +64,22 @@
/* Everything falls through in here */
case 6:
inptr1 = zerobuff;
+ // fall through
case 5:
inptr2 = zerobuff;
+ // fall through
case 4:
inptr3 = zerobuff;
+ // fall through
case 3:
inptr4 = zerobuff;
+ // fall through
case 2:
inptr5 = zerobuff;
+ // fall through
case 1:
inptr6 = zerobuff;
+ // fall through
case 0:
inptr7 = zerobuff;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_32bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_32bit.hpp
index 54822c8..758c084 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_32bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_32bit.hpp
@@ -64,16 +64,22 @@
/* Everything falls through in here */
case 6:
inptr1 = zerobuff;
+ // fall through
case 5:
inptr2 = zerobuff;
+ // fall through
case 4:
inptr3 = zerobuff;
+ // fall through
case 3:
inptr4 = zerobuff;
+ // fall through
case 2:
inptr5 = zerobuff;
+ // fall through
case 1:
inptr6 = zerobuff;
+ // fall through
case 0:
inptr7 = zerobuff;
break;
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_half_to_float.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_half_to_float.hpp
index 0606330..de8e95a 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_half_to_float.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_interleave_8way_half_to_float.hpp
@@ -64,16 +64,22 @@
/* Everything falls through in here */
case 6:
inptr1 = zerobuff;
+ // fall through
case 5:
inptr2 = zerobuff;
+ // fall through
case 4:
inptr3 = zerobuff;
+ // fall through
case 3:
inptr4 = zerobuff;
+ // fall through
case 2:
inptr5 = zerobuff;
+ // fall through
case 1:
inptr6 = zerobuff;
+ // fall through
case 0:
inptr7 = zerobuff;
break;