Compute Library v23.05

ARM-software · May 18, 2023 · 6c713f0 · 6c713f0
1 parent d8bf9b5
commit 6c713f0
Show file tree

Hide file tree

Showing 711 changed files with 181,341 additions and 200,606 deletions.
diff --git a/.bazelrc b/.bazelrc
@@ -37,3 +37,4 @@ build --flag_alias=logging=//:logging
 build --flag_alias=openmp=//:openmp
 build --flag_alias=cppthreads=//:cppthreads
 build --flag_alias=enable_bf16_validation=//:enable_bf16_validation
+build --flag_alias=enable_sve_validation=//:enable_sve_validation
diff --git a/Android.bp b/Android.bp
@@ -27,6 +27,7 @@ opencl_srcs = [
         "src/core/CL/cl_kernels/common/elementwise_operation.cl",
         "src/core/CL/cl_kernels/common/elementwise_operation_quantized.cl",
         "src/core/CL/cl_kernels/common/elementwise_unary.cl",
+        "src/core/CL/cl_kernels/common/elementwise_unary_quantized.cl",
         "src/core/CL/cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/fp_post_ops_act_eltwise_op_act.h",
         "src/core/CL/cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_native.cl",
         "src/core/CL/cl_kernels/common/experimental/gemm_fused_post_ops/act_eltwise_op_act/gemm_mm_reshaped.cl",
@@ -49,6 +50,8 @@ opencl_srcs = [
         "src/core/CL/cl_kernels/common/generate_proposals_quantized.cl",
         "src/core/CL/cl_kernels/common/instance_normalization.cl",
         "src/core/CL/cl_kernels/common/l2_normalize.cl",
+        "src/core/CL/cl_kernels/common/mat_mul.cl",
+        "src/core/CL/cl_kernels/common/mat_mul_quantized.cl",
         "src/core/CL/cl_kernels/common/mean_stddev_normalization.cl",
         "src/core/CL/cl_kernels/common/memset.cl",
         "src/core/CL/cl_kernels/common/minmax_layer.cl",
@@ -175,6 +178,7 @@ cc_library_static {
                          "build/android-arm64v8a/src/core/CL",
                          "src/core/common",
                          "src/core/helpers",
+                         "src/core/NEON/kernels/arm_gemm",
                          "src/core/NEON/kernels/assembly",
                          "src/core/NEON/kernels/convolution/common",
                          "src/core/NEON/kernels/convolution/winograd",
@@ -292,6 +296,7 @@ cc_library_static {
         "src/core/NEON/kernels/NEROIPoolingLayerKernel.cpp",
         "src/core/NEON/kernels/NERangeKernel.cpp",
         "src/core/NEON/kernels/NEReductionOperationKernel.cpp",
+        "src/core/NEON/kernels/NEReorderKernel.cpp",
         "src/core/NEON/kernels/NEReorgLayerKernel.cpp",
         "src/core/NEON/kernels/NEReverseKernel.cpp",
         "src/core/NEON/kernels/NESelectKernel.cpp",
@@ -301,6 +306,7 @@ cc_library_static {
         "src/core/NEON/kernels/NEStridedSliceKernel.cpp",
         "src/core/NEON/kernels/NETileKernel.cpp",
         "src/core/NEON/kernels/arm_conv/addressing.cpp",
+        "src/core/NEON/kernels/arm_conv/depthwise/depthwise_common.cpp",
         "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp16.cpp",
         "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp32.cpp",
         "src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp",
@@ -483,6 +489,9 @@ cc_library_static {
         "src/cpu/kernels/elementwise_unary/generic/neon/fp32.cpp",
         "src/cpu/kernels/elementwise_unary/generic/neon/impl.cpp",
         "src/cpu/kernels/elementwise_unary/generic/neon/integer.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/q8.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/qasymm8_signed.cpp",
         "src/cpu/kernels/floor/neon/fp16.cpp",
         "src/cpu/kernels/floor/neon/fp32.cpp",
         "src/cpu/kernels/fuse_batch_normalization/generic/fp16.cpp",
@@ -510,6 +519,7 @@ cc_library_static {
         "src/cpu/kernels/l2normlayer/generic/neon/fp16.cpp",
         "src/cpu/kernels/l2normlayer/generic/neon/fp32.cpp",
         "src/cpu/kernels/l2normlayer/generic/neon/impl.cpp",
+        "src/cpu/kernels/lut/generic/neon/u8.cpp",
         "src/cpu/kernels/maxunpool/generic/neon/fp16.cpp",
         "src/cpu/kernels/maxunpool/generic/neon/fp32.cpp",
         "src/cpu/kernels/maxunpool/generic/neon/impl.cpp",
@@ -578,6 +588,7 @@ cc_library_static {
         "src/cpu/operators/CpuGemmDirectConv2d.cpp",
         "src/cpu/operators/CpuGemmLowpMatrixMultiplyCore.cpp",
         "src/cpu/operators/CpuGemmLowpOutputStage.cpp",
+        "src/cpu/operators/CpuMatMul.cpp",
         "src/cpu/operators/CpuMaxUnpooling.cpp",
         "src/cpu/operators/CpuMul.cpp",
         "src/cpu/operators/CpuPermute.cpp",
@@ -630,8 +641,10 @@ cc_library_static {
         "src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuReshape.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuResize.cpp",
+        "src/dynamic_fusion/sketch/gpu/operators/GpuSigmoid.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuSoftmax.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/GpuSub.cpp",
+        "src/dynamic_fusion/sketch/gpu/operators/GpuTanh.cpp",
         "src/dynamic_fusion/sketch/gpu/operators/internal/GpuElementwiseBinaryCommon.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/GpuKernelVariableTable.cpp",
         "src/dynamic_fusion/sketch/gpu/template_writer/cl/ClTemplateActivation.cpp",
@@ -685,6 +698,8 @@ cc_library_static {
         "src/gpu/cl/kernels/ClIm2ColKernel.cpp",
         "src/gpu/cl/kernels/ClIndirectConv2dAddressPrecalculationKernel.cpp",
         "src/gpu/cl/kernels/ClIndirectConv2dKernel.cpp",
+        "src/gpu/cl/kernels/ClMatMulLowpNativeKernel.cpp",
+        "src/gpu/cl/kernels/ClMatMulNativeKernel.cpp",
         "src/gpu/cl/kernels/ClMulKernel.cpp",
         "src/gpu/cl/kernels/ClPermuteKernel.cpp",
         "src/gpu/cl/kernels/ClPool2dKernel.cpp",
@@ -733,6 +748,7 @@ cc_library_static {
         "src/gpu/cl/operators/ClGemmLowpOutputStage.cpp",
         "src/gpu/cl/operators/ClIndirectConv2d.cpp",
         "src/gpu/cl/operators/ClLogicalNot.cpp",
+        "src/gpu/cl/operators/ClMatMul.cpp",
         "src/gpu/cl/operators/ClMul.cpp",
         "src/gpu/cl/operators/ClPRelu.cpp",
         "src/gpu/cl/operators/ClPermute.cpp",
@@ -815,6 +831,7 @@ cc_library_static {
         "src/runtime/CL/functions/CLLogicalAnd.cpp",
         "src/runtime/CL/functions/CLLogicalNot.cpp",
         "src/runtime/CL/functions/CLLogicalOr.cpp",
+        "src/runtime/CL/functions/CLMatMul.cpp",
         "src/runtime/CL/functions/CLMaxUnpoolingLayer.cpp",
         "src/runtime/CL/functions/CLMeanStdDevNormalizationLayer.cpp",
         "src/runtime/CL/functions/CLNormalizationLayer.cpp",
@@ -928,6 +945,7 @@ cc_library_static {
         "src/runtime/NEON/functions/NELSTMLayer.cpp",
         "src/runtime/NEON/functions/NELSTMLayerQuantized.cpp",
         "src/runtime/NEON/functions/NELogical.cpp",
+        "src/runtime/NEON/functions/NEMatMul.cpp",
         "src/runtime/NEON/functions/NEMaxUnpoolingLayer.cpp",
         "src/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.cpp",
         "src/runtime/NEON/functions/NENormalizationLayer.cpp",
@@ -946,6 +964,7 @@ cc_library_static {
         "src/runtime/NEON/functions/NERange.cpp",
         "src/runtime/NEON/functions/NEReduceMean.cpp",
         "src/runtime/NEON/functions/NEReductionOperation.cpp",
+        "src/runtime/NEON/functions/NEReorderLayer.cpp",
         "src/runtime/NEON/functions/NEReorgLayer.cpp",
         "src/runtime/NEON/functions/NEReshapeLayer.cpp",
         "src/runtime/NEON/functions/NEReverse.cpp",
@@ -981,6 +1000,8 @@ cc_library_static {
         "src/runtime/heuristics/dwc_native/ClDWCNativeDefaultConfigValhall.cpp",
         "src/runtime/heuristics/dwc_native/ClDWCNativeHeuristicsHelpers.cpp",
         "src/runtime/heuristics/indirect_conv/ClIndirectConvDefaultConfigValhall.cpp",
+        "src/runtime/heuristics/matmul_native/ClMatMulNativeDefaultConfigValhall.cpp",
+        "src/runtime/heuristics/matmul_native/ClMatMulNativeHelpers.cpp",
         "utils/CommonGraphOptions.cpp",
         "utils/GraphUtils.cpp",
         "utils/Utils.cpp",
@@ -1053,6 +1074,16 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_generic_output9_mla_depthfirst/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_indirect.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_direct.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_indirect.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_direct.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_indirect.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp",
@@ -1070,17 +1101,14 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32bf16fp32_planar_5x5_s1_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32bf16fp32_planar_5x5_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_s8q_planar_3x3_s1_4rows_dot_za/generic.cpp",
-                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_s8q_planar_3x3_s2_2rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_s8q_planar_3x3_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_s8q_planar_5x5_s1_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_s8q_planar_5x5_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8q_planar_3x3_s1_4rows_dot_za/generic.cpp",
-                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8q_planar_3x3_s2_2rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8q_planar_3x3_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8q_planar_5x5_s1_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8q_planar_5x5_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8s8u8q_planar_3x3_s1_4rows_dot_za/generic.cpp",
-                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8s8u8q_planar_3x3_s2_2rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8s8u8q_planar_3x3_s2_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8s8u8q_planar_5x5_s1_4rows_dot_za/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_u8s8u8q_planar_5x5_s2_4rows_dot_za/generic.cpp",
@@ -1094,10 +1122,8 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp16_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_indirect.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_direct.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
-                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
-                "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst_strided/generic.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_indirect.cpp",
                 "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_fp32_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_direct.cpp",
@@ -1261,6 +1287,9 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_2VLx2VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_4VLx1VL/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_1VLx4VL/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_2VLx2VL/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp32_mopa_1VLx4VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp32_mopa_2VLx2VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp32_mopa_4VLx1VL/generic.cpp",
@@ -1319,9 +1348,6 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_u8u32_dot_8x3VL/a64fx.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_u8u32_dot_8x3VL/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_u8u32_mmla_8x3VL/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_fp32_mla_8x1VL/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_s8s32_dot_8x1VL/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_u8u32_dot_8x1VL/generic.cpp",
                 "src/core/NEON/kernels/convolution/winograd/input_transforms/a64_fp16_6x6.cpp",
                 "src/core/NEON/kernels/convolution/winograd/input_transforms/a64_fp32_6x6.cpp",
                 "src/core/NEON/kernels/convolution/winograd/input_transforms/sme_fp32_mla_6x6.cpp",

diff --git a/BUILD.bazel b/BUILD.bazel
@@ -72,6 +72,12 @@ bool_flag(
     visibility = ["//visibility:public"],
 )
 
+bool_flag(
+    name = "enable_sve_validation",
+    build_setting_default = False,
+    visibility = ["//visibility:public"],
+)
+
 #---------------------------------------------------------------------
 # Flag variables
 config_setting(
@@ -116,6 +122,13 @@ config_setting(
     },
 )
 
+config_setting(
+    name = "sve_validation_flag",
+    flag_values = {
+        ":enable_sve_validation": "true",
+    },
+)
+
 
 #---------------------------------------------------------------------
 # Common defines used for all targets
@@ -138,8 +151,6 @@ cc_library(
                   "DARM_COMPUTE_GRAPH_ENABLED",
                   "ARM_COMPUTE_ENABLE_SVEF32MM",
                   "ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS",
-                  "ENABLE_SVE",
-                  "ARM_COMPUTE_ENABLE_SVE",
                   "_GLIBCXX_USE_NANOSLEEP"
               ] + select({
                   "//:debug_flag": [
@@ -181,7 +192,7 @@ cc_library(
     name = "arm_compute_graph",
     srcs = ["//src:arm_compute_graph_srcs"],
     copts = [
-                "-march=armv8.2-a+fp16",  # What arch is it we should go for here?
+                "-march=armv8.2-a+fp16",
             ] + select({
                 "//:debug_flag": [
                     "-O0",
@@ -200,7 +211,7 @@ cc_library(
             }),
     visibility = ["//visibility:public"],
     deps = [
-        "arm_compute",
+        "arm_compute_core",
         "//:common_defines",
         "//arm_compute:graph_headers",
     ],
@@ -214,7 +225,7 @@ cc_library(
     name = "arm_compute_sve2",
     srcs = ["//src:arm_compute_sve2_srcs"],
     copts = [
-                "-march=armv8.6-a+sve2+fp16+dotprod",  # What arch is it we should go for here?
+                "-march=armv8.6-a+sve2+fp16+dotprod",
             ] + select({
                 "//:debug_flag": [
                     "-O0",
@@ -243,8 +254,10 @@ cc_library(
         "//conditions:default": [],
     }),
     local_defines = [
+        "ENABLE_SVE",
+        "ARM_COMPUTE_ENABLE_SVE",
         "ARM_COMPUTE_ENABLE_SVE2",
-        "ARM_COMPUTE_ENABLE_BF16"
+        "ARM_COMPUTE_ENABLE_BF16",
     ],
     deps = [
         "//:common_defines",
@@ -263,7 +276,7 @@ cc_library(
     name = "arm_compute_sve",
     srcs = ["//src:arm_compute_sve_srcs"],
     copts = [
-                "-march=armv8.2-a+sve+fp16+dotprod",  # What arch is it we should go for here?
+                "-march=armv8.2-a+sve+fp16+dotprod",
             ] + select({
                 "//:debug_flag": [
                     "-O0",
@@ -292,6 +305,8 @@ cc_library(
         "//conditions:default": [],
     }),
     local_defines = [
+        "ENABLE_SVE",
+        "ARM_COMPUTE_ENABLE_SVE",
         "ARM_COMPUTE_ENABLE_BF16",
     ],
     deps = [
@@ -308,8 +323,8 @@ cc_library(
 # Core and Runtime library
 
 cc_library(
-    name = "arm_compute",
-    srcs = ["//src:arm_compute_srcs"],
+    name = "arm_compute_core",
+    srcs = ["//src:arm_compute_core_srcs"],
     hdrs = glob([
         "core/NEON/kernels/**/*.h",
         "core/NEON/kernels/**/*.hpp",
@@ -318,7 +333,7 @@ cc_library(
         "//:create_version_file",
     ],
     copts = [
-                "-march=armv8.2-a+fp16",  # What arch is it we should go for here?
+                "-march=armv8.2-a+fp16",
             ] + select({
                 "//:debug_flag": [
                     "-O0",
@@ -348,6 +363,8 @@ cc_library(
         "//conditions:default": [],
     }),
     local_defines = [
+        "ENABLE_SVE",
+        "ARM_COMPUTE_ENABLE_SVE",
         "ARM_COMPUTE_ENABLE_BF16",
     ],
     visibility = ["//visibility:public"],