From 64042227c655ef786b5342fe48816a5a13b93447 Mon Sep 17 00:00:00 2001
From: Sandesh Venkatesh <sandesh.venkatesh@ittiam.com>
Date: Mon, 23 Jun 2025 14:40:21 +0530
Subject: [PATCH] Reduce usage of BASOP_Util_Add_Mant32Exp in MCT path

Changes are under macro OPT_MCT_ENC_V1_NBE
WMOPs reduction of ~130. Tested config: Multi-Channel 7.1+4,48kbps FB
---
 lib_com/options.h            |  1 +
 lib_enc/ivas_mcmasa_enc_fx.c | 37 +++++++++++++++++++++++++++++++++++-
 2 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/lib_com/options.h b/lib_com/options.h
index 7996f8d41..b45e54671 100644
--- a/lib_com/options.h
+++ b/lib_com/options.h
@@ -77,6 +77,7 @@
 #define FIX_1379_MASA_ANGLE_ROUND
 
 /* Note: each compile switch (FIX_1101_...) is independent from the other ones */
+#define OPT_MCT_ENC_V1_NBE
 #define OPT_SBA_REND_V1_BE
 #define OPT_HEAD_ROT_REND_V1_BE
 #define OPT_SBA_DEC_V2_BE
diff --git a/lib_enc/ivas_mcmasa_enc_fx.c b/lib_enc/ivas_mcmasa_enc_fx.c
index 314424341..ab86efb8d 100644
--- a/lib_enc/ivas_mcmasa_enc_fx.c
+++ b/lib_enc/ivas_mcmasa_enc_fx.c
@@ -2085,6 +2085,41 @@ static void compute_cov_mtx_fx(
 )
 {
     Word16 i, j;
+#ifdef OPT_MCT_ENC_V1_NBE
+    Word64 temp64_1, temp64_2;
+    Word16 tmp_16, max_exp;
+    Word16 temp_exp = shl( inp_exp, 1 );
+    FOR( i = 0; i < N; i++ )
+    {
+        FOR( j = i; j < N; j++ )
+        {
+            temp64_1 = W_mac_32_32( W_mult_32_32( sr[i][freq], sr[j][freq] ), si[i][freq], si[j][freq] ); // exp:2*inp_exp
+            temp64_2 = W_deposit32_h( COVls->xr_fx[i][j] );                                               // exp:COVls->xr_e[i][j]
+            max_exp = s_max( COVls->xr_e[i][j], temp_exp );
+            temp64_2 = W_shl( temp64_2, sub( COVls->xr_e[i][j], max_exp ) ); // exp:max_exp
+            temp64_1 = W_shl( temp64_1, sub( temp_exp, max_exp ) );          // exp:max_exp
+            temp64_1 = W_add( temp64_1, temp64_2 );                          // exp:max_exp
+            tmp_16 = W_norm( temp64_1 );
+
+            COVls->xr_fx[i][j] = W_extract_h( W_shl( temp64_1, tmp_16 ) ); // exp:max_exp-tmp_16
+            COVls->xr_e[i][j] = sub( max_exp, tmp_16 );
+            move32();
+            move16();
+
+            temp64_1 = W_sub( W_mult_32_32( si[i][freq], sr[j][freq] ), W_mult_32_32( sr[i][freq], si[j][freq] ) ); // exp :2*inp_exp
+            temp64_2 = W_deposit32_h( COVls->xi_fx[i][j] );                                                         // exp:COVls->xi_e[i][j]
+            max_exp = s_max( COVls->xi_e[i][j], temp_exp );
+            temp64_2 = W_shl( temp64_2, sub( COVls->xi_e[i][j], max_exp ) ); // exp:max_exp
+            temp64_1 = W_shl( temp64_1, sub( temp_exp, max_exp ) );          // exp:max_exp
+            temp64_1 = W_add( temp64_1, temp64_2 );                          // exp:max_exp
+            tmp_16 = W_norm( temp64_1 );
+
+            COVls->xi_fx[i][j] = W_extract_h( W_shl( temp64_1, tmp_16 ) ); // exp:max_exp-tmp_16
+            COVls->xi_e[i][j] = sub( max_exp, tmp_16 );
+            move32();
+        }
+    }
+#else
     Word32 a, b, c, d;
     Word32 temp;
     Word16 norm_a, norm_b, norm_c, norm_d;
@@ -2117,7 +2152,7 @@ static void compute_cov_mtx_fx(
             move32();
         }
     }
-
+#endif
     return;
 }
 
-- 
GitLab