Create core library using high priority operators
A smaller core library is created using a subset of the operators.
Changed the structure of filelist.json in order to include more
information about the kernels and make the selection easier.
Resolves: COMPMID-4514
Change-Id: I079ca7d8e64346174eebdd13b834e1dd4dc36ca2
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5786
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp b/src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp
index 40370fe..f38912d 100644
--- a/src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp
+++ b/src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp
@@ -33,7 +33,7 @@
#include "depthwise_implementation_constraints.hpp"
#if defined(__aarch64__)
-#if defined(__ARM_FEATURE_SVE) && defined(SVE2)
+#if defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVE2)
#include "kernels/sve_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst.hpp"
#include "kernels/sve_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst.hpp"
#include "kernels/sve_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst.hpp"
@@ -41,7 +41,7 @@
#include "kernels/sve_s8q_nhwc_5x5_s1_output2x2_mla_depthfirst.hpp"
#include "kernels/sve_s8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst.hpp"
#include "kernels/sve_s8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst.hpp"
-#endif // defined(__ARM_FEATURE_SVE) && defined(SVE2)
+#endif // defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVE2)
#include "kernels/a64_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst.hpp"
#include "kernels/a64_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst.hpp"
#include "kernels/a64_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst.hpp"
@@ -73,14 +73,15 @@
static const DepthwiseImplementation<int8_t, int8_t, int8_t, Requantize32> depthwise_s8q_methods[] = {
#if defined(__aarch64__)
-#if defined(__ARM_FEATURE_SVE) && defined(SVE2)
+#if defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVE2)
{
DepthwiseMethod::DEPTHFIRST,
"sve_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst",
constraint<Requantize32>(is_supported<sve_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst>,
has_no_channel_multiplier,
qp_has_no_left_shift,
- qp_weights_are_symmetric),
+ qp_weights_are_symmetric,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstQuantized<sve_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst>(args, qp);
@@ -91,7 +92,8 @@
"sve_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst>,
has_no_channel_multiplier,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstQuantized<sve_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst>(args, qp);
@@ -102,7 +104,8 @@
"sve_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst>,
has_no_channel_multiplier,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstQuantized<sve_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst>(args, qp);
@@ -113,7 +116,8 @@
"sve_s8q_nhwc_3x3_s2_output2x2_mla_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_nhwc_3x3_s2_output2x2_mla_depthfirst>,
has_no_channel_multiplier,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstQuantized<sve_s8q_nhwc_3x3_s2_output2x2_mla_depthfirst>(args, qp);
@@ -124,7 +128,8 @@
"sve_s8q_nhwc_5x5_s1_output2x2_mla_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_nhwc_5x5_s1_output2x2_mla_depthfirst>,
has_no_channel_multiplier,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstQuantized<sve_s8q_nhwc_5x5_s1_output2x2_mla_depthfirst>(args, qp);
@@ -134,7 +139,8 @@
DepthwiseMethod::DEPTHFIRST,
"sve_s8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst>,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstWithMultiplierQuantized<sve_s8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst>(args, qp);
@@ -144,13 +150,14 @@
DepthwiseMethod::DEPTHFIRST,
"sve_s8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst",
constraint<Requantize32>(is_supported<sve_s8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst>,
- qp_has_no_left_shift),
+ qp_has_no_left_shift,
+ cpu_has_sve2),
nullptr,
[] (const DepthwiseArgs &args, const Requantize32 &qp) -> DepthwiseCommon<int8_t, int8_t, int8_t> * {
return new DepthwiseDepthfirstWithMultiplierQuantized<sve_s8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst>(args, qp);
},
},
-#endif // defined(__ARM_FEATURE_SVE) && defined(SVE2)
+#endif // defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVE2)
{
DepthwiseMethod::DEPTHFIRST,
"a64_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst",