ipc4: mixin: Fix HiFi3 impl of 24-bit mixing

AE_ADD24S() expects input arguments to be Q9.23 values. Therefore,
negative 24-bit values in a 32-bit container should have their sign
extended to the upper 8 bits. All other five implementations of 24-bit
mixing (IPC3 mixer's generic and HIFI3, IPC4 mixin's generic, IPC4
mixin's mix with gain generic and HIFI3) perform sign extension prior
to mixing and do not rely on samples being already sign-extended.

Signed-off-by: Serhiy Katsyuba <serhiy.katsyuba@intel.com>

Author: serhiy-katsyuba-intel

src/audio/mixin_mixout/mixin_mixout_hifi3.c

@@ -255,6 +255,9 @@ static void mix_s24(struct cir_buf_ptr *sink, int32_t start_sample, int32_t mixe
 			AE_LA32X2_IP(in_sample, inu, in);
 			AE_LA32X2_IP(out_sample, outu1, out);
 			out--;
+			/* sign extent in_sample as AE_ADD24S expects Q9.23 arguments */
+			in_sample = AE_SLAI24S(AE_MOVF24X2_FROMINT32X2(in_sample), 0);
+			/* out_sample is already sign extended by other mixin in a loop below */
 			out_sample = AE_ADD24S(in_sample, out_sample);
 			AE_SA32X2_IP(out_sample, outu2, out);
 		}
@@ -264,6 +267,8 @@ static void mix_s24(struct cir_buf_ptr *sink, int32_t start_sample, int32_t mixe
 		if (left) {
 			AE_L32_IP(in_sample, (ae_int32 *)in, sizeof(ae_int32));
 			AE_L32_IP(out_sample, (ae_int32 *)out, 0);
+			/* sign extension */
+			in_sample = AE_SLAI24S(AE_MOVF24X2_FROMINT32X2(in_sample), 0);
 			out_sample = AE_ADD24S(in_sample, out_sample);
 			AE_S32_L_IP(out_sample, (ae_int32 *)out, sizeof(ae_int32));
 		}
@@ -283,12 +288,16 @@ static void mix_s24(struct cir_buf_ptr *sink, int32_t start_sample, int32_t mixe
 		left = n & 1;
 		for (i = 0; i < m; i++) {
 			AE_LA32X2_IP(in_sample, inu, in);
+			/* sign extension */
+			in_sample = AE_SLAI24S(AE_MOVF24X2_FROMINT32X2(in_sample), 0);
 			AE_SA32X2_IP(in_sample, outu2, out);
 		}
 		AE_SA64POS_FP(outu2, out);
 		/* process the left sample to avoid memory access overrun */
 		if (left) {
 			AE_L32_IP(in_sample, (ae_int32 *)in, sizeof(ae_int32));
+			/* sign extension */
+			in_sample = AE_SLAI24S(AE_MOVF24X2_FROMINT32X2(in_sample), 0);
 			AE_S32_L_IP(in_sample, (ae_int32 *)out, sizeof(ae_int32));
 		}
 	}