aarch64: apply absolute value to peak computation
This commit is contained in:
@@ -132,6 +132,7 @@ arm_neon_compute_peak(const float* src, uint32_t nframes, float current)
|
||||
size_t offset = 4 * i;
|
||||
float32x4_t x0;
|
||||
x0 = vld1q_f32(src_aligned + offset);
|
||||
x0 = vabsq_f32(x0);
|
||||
vmax = vmaxq_f32(vmax, x0);
|
||||
}
|
||||
|
||||
@@ -139,6 +140,7 @@ arm_neon_compute_peak(const float* src, uint32_t nframes, float current)
|
||||
{
|
||||
float32x4_t x0;
|
||||
x0 = vld1q_dup_f32(src_aligned + frame);
|
||||
x0 = vabsq_f32(x0);
|
||||
vmax = vmaxq_f32(vmax, x0);
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user