commit 3f16c5758f4f8e71e76e302150f0696a1dcc59fa
author Pablo Marquez Tello <pablo.tello@arm.com> Tue Jun 20 13:39:10 2023 +0100
committer Pablo Marquez Tello <pablo.tello@arm.com> Wed Jun 21 10:35:05 2023 +0000
tree aa2da0b22a7bc25cda4d008b5543a706a6acf293
parent 68e9c4dfd3c02b3f2e25290f9ad9c3b0bc001d25

Enable vmfa in arm7va/aarch32 when present

* vfma is an extension on armv7a and it can be enabled
 with -mfpu=neon-vfpv4

* Resolves MLCE-1079

Change-Id: Id455c39ee4feb8d3cdc4515c8307eb8a5d6e093b
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9795
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

src/core/NEON/NEMath.inl

diff --git a/src/core/NEON/NEMath.inl b/src/core/NEON/NEMath.inl
index 6198a25..1cbe669 100644
--- a/src/core/NEON/NEMath.inl
+++ b/src/core/NEON/NEMath.inl
@@ -52,11 +52,11 @@
 #ifndef DOXYGEN_SKIP_THIS
 inline float32x4_t prefer_vfmaq_f32(float32x4_t a, float32x4_t b, float32x4_t c)
 {
-#ifdef __aarch64__
+#if __ARM_FEATURE_FMA
     return vfmaq_f32(a, b, c);
-#else  // __aarch64__
+#else // __ARM_FEATURE_FMA
     return vmlaq_f32(a, b, c);
-#endif // __aarch64__
+#endif // __ARM_FEATURE_FMA
 }
 
 inline float32x4_t vfloorq_f32(float32x4_t val)