COMPMID-1380: Pre-work for SVE support.

This patch makes the needed infrastructure changes to allow SVE
kernels to be added later on.

Change-Id: Ide5bccac2f47278e93fff3d648231aee2d5f8c2e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/139070
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

9 files changed, 18 insertions, 18 deletions

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 e485ca7009..492abe51ed 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
@@ -31,7 +31,7 @@
 
 template<>
 template<typename T>
-inline void TransformImpl<6, 1, false, 4, 4>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
+inline void TransformImpl<6, 1, false, 4, 4, false>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
     uint32_t *outptr = reinterpret_cast<uint32_t *>(out);
     const uint32_t *inptr = reinterpret_cast<const uint32_t *>(in);
 
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a32_transpose_interleave_8way_32bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a32_transpose_interleave_8way_32bit.hpp
index a7e17fa074..56a226fce0 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a32_transpose_interleave_8way_32bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a32_transpose_interleave_8way_32bit.hpp
@@ -30,12 +30,12 @@
 // Generic unblocked transposed 8x32-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<8, 1, true, 4, 4>::Transform(
+inline void TransformImpl<8, 1, true, 4, 4, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
   // Redirect to a 16x uint16_t specialisation
-  TransformImpl<16, 1, true, 2, 2>::Transform(
+  TransformImpl<16, 1, true, 2, 2, false>::Transform(
     reinterpret_cast<uint16_t *>(out),
     reinterpret_cast<const uint16_t * const>(in),
     stride*2, x0*2, xmax*2, k0, kmax
@@ -45,7 +45,7 @@ inline void TransformImpl<8, 1, true, 4, 4>::Transform(
 // Generic 12x16-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<16, 1, true, 2, 2>::Transform(
+inline void TransformImpl<16, 1, true, 2, 2, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
@@ -117,7 +117,7 @@ inline void TransposeInterleaveCommon<16, uint16_t, uint16_t>::moveblock_1x4(con
 
 template <>
 template <>
-inline void TransformImpl<16, 1, true, 2, 2>::Transform(
+inline void TransformImpl<16, 1, true, 2, 2, false>::Transform(
     uint16_t* out, const uint16_t* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
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 7e61f425d4..8ea0483a50 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
@@ -32,7 +32,7 @@
 
 template<>
 template<typename T>
-void TransformImpl<4, 16, false, 1, 1>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
+void TransformImpl<4, 16, false, 1, 1, false>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
     uint8_t *outptr = (uint8_t *)out;
     const uint8_t *inptr = (uint8_t *)in;
 
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 99bb2d66bd..91ee49229b 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
@@ -31,7 +31,7 @@
 
 template<>
 template<typename T>
-void TransformImpl<8, 1, false, 2, 2>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
+void TransformImpl<8, 1, false, 2, 2, false>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
     uint16_t *outptr = (uint16_t *)out;
     const uint16_t *inptr = (const uint16_t *)in;
 
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 83391cc59f..7a32f331ea 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
@@ -31,7 +31,7 @@
 
 template<>
 template<typename T>
-inline void TransformImpl<8, 1, false, 4, 4>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
+inline void TransformImpl<8, 1, false, 4, 4, false>::Transform(T *out, const T *in, int ldin, int y0, int ymax, int k0, int kmax) {
     uint32_t *outptr = (uint32_t *)out;
     const uint32_t *inptr = (uint32_t *)in;
 
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 fd812165fd..773d56d913 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
@@ -31,7 +31,7 @@
 
 template<>
 template<>
-inline void TransformImpl<8, 1, false, 4, 2>::Transform(float *out, const __fp16 *in, int ldin, int y0, int ymax, int k0, int kmax) {
+inline void TransformImpl<8, 1, false, 4, 2, false>::Transform(float *out, const __fp16 *in, int ldin, int y0, int ymax, int k0, int kmax) {
     float *outptr = out;
     const __fp16 *inptr = in;
 
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_16bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_16bit.hpp
index 6e07064a0c..16fa31eb67 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_16bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_16bit.hpp
@@ -30,12 +30,12 @@
 // Generic unblocked transposed 6x32-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<6, 1, true, 4, 4>::Transform(
+inline void TransformImpl<6, 1, true, 4, 4, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
   // Redirect to a 12 x uint16_t specialisation
-  TransformImpl<12, 1, true, 2, 2>::Transform(
+  TransformImpl<12, 1, true, 2, 2, false>::Transform(
     reinterpret_cast<uint16_t *>(out),
     reinterpret_cast<const uint16_t * const>(in),
     stride*2, x0*2, xmax*2, k0, kmax
@@ -45,7 +45,7 @@ inline void TransformImpl<6, 1, true, 4, 4>::Transform(
 // Generic 12x16-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<12, 1, true, 2, 2>::Transform(
+inline void TransformImpl<12, 1, true, 2, 2, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
@@ -135,7 +135,7 @@ inline void TransposeInterleaveCommon<12, uint16_t, uint16_t>::moveblock_1x4(con
 
 template <>
 template <>
-inline void TransformImpl<12, 1, true, 2, 2>::Transform(
+inline void TransformImpl<12, 1, true, 2, 2, false>::Transform(
     uint16_t* out, const uint16_t* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_half_to_float.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_half_to_float.hpp
index 2f90c18ebd..46b4bf5149 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_half_to_float.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_12way_half_to_float.hpp
@@ -110,7 +110,7 @@ inline void TransposeInterleaveCommon<12, __fp16, float>::moveblock_1x4(const __
 
 template <>
 template <>
-inline void TransformImpl<12, 1, true, 4, 2>::Transform(
+inline void TransformImpl<12, 1, true, 4, 2, false>::Transform(
     float* out, const __fp16* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_24way_16bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_24way_16bit.hpp
index b6565baa23..c39dd82119 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_24way_16bit.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_24way_16bit.hpp
@@ -30,12 +30,12 @@
 // Generic unblocked transposed 12x32-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<12, 1, true, 4, 4>::Transform(
+inline void TransformImpl<12, 1, true, 4, 4, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
   // Redirect to a 24 x uint16_t specialisation
-  TransformImpl<24, 1, true, 2, 2>::Transform(
+  TransformImpl<24, 1, true, 2, 2, false>::Transform(
     reinterpret_cast<uint16_t *>(out),
     reinterpret_cast<const uint16_t * const>(in),
     stride*2, x0*2, xmax*2, k0, kmax
@@ -45,7 +45,7 @@ inline void TransformImpl<12, 1, true, 4, 4>::Transform(
 // Generic 24x16-bit sized specialisation
 template <>
 template <typename T>
-inline void TransformImpl<24, 1, true, 2, 2>::Transform(
+inline void TransformImpl<24, 1, true, 2, 2, false>::Transform(
     T* out, const T* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {
@@ -120,7 +120,7 @@ inline void TransposeInterleaveCommon<24, uint16_t, uint16_t>::moveblock_1x4(con
 
 template <>
 template <>
-inline void TransformImpl<24, 1, true, 2, 2>::Transform(
+inline void TransformImpl<24, 1, true, 2, 2, false>::Transform(
     uint16_t* out, const uint16_t* const in, const int stride,
     const int x0, const int xmax, const int k0, const int kmax
 ) {