POWER10: Reduce sgemm loop unrolling

With GCC 14, unnecessary move and lxvp instructions appear when unrolling the inner loop for larger sizes.
Reducing the loop unroll factor restores performance to GCC 11.

Author: Rajalakshmi Srinivasaraghavan

kernel/power/sgemm_kernel_power10.c

@@ -245,118 +245,8 @@ CNAME (BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT * A, FLOAT * B,
 	  AO += 16;
 	  BO += 8;
 	  temp--;
-	  BLASLONG K = temp / 64;
+	  BLASLONG K = temp / 16;
 	  for (l = 0; l < K; l++)
-	    {
-	      vec_t *rowA = (vec_t *) & AO[0];
-	      vec_t *rowB = (vec_t *) & BO[0];
-	      KERNEL (0, 0);
-	      KERNEL (2, 4);
-	      KERNEL (4, 8);
-	      KERNEL (6, 12);
-	      KERNEL (8, 16);
-	      KERNEL (10, 20);
-	      KERNEL (12, 24);
-	      KERNEL (14, 28);
-	      KERNEL (16, 32);
-	      KERNEL (18, 36);
-	      KERNEL (20, 40);
-	      KERNEL (22, 44);
-	      KERNEL (24, 48);
-	      KERNEL (26, 52);
-	      KERNEL (28, 56);
-	      KERNEL (30, 60);
-	      KERNEL (32, 64);
-	      KERNEL (34, 68);
-	      KERNEL (36, 72);
-	      KERNEL (38, 76);
-	      KERNEL (40, 80);
-	      KERNEL (42, 84);
-	      KERNEL (44, 88);
-	      KERNEL (46, 92);
-	      KERNEL (48, 96);
-	      KERNEL (50, 100);
-	      KERNEL (52, 104);
-	      KERNEL (54, 108);
-	      KERNEL (56, 112);
-	      KERNEL (58, 116);
-	      KERNEL (60, 120);
-	      KERNEL (62, 124);
-	      KERNEL (64, 128);
-	      KERNEL (66, 132);
-	      KERNEL (68, 136);
-	      KERNEL (70, 140);
-	      KERNEL (72, 144);
-	      KERNEL (74, 148);
-	      KERNEL (76, 152);
-	      KERNEL (78, 156);
-	      KERNEL (80, 160);
-	      KERNEL (82, 164);
-	      KERNEL (84, 168);
-	      KERNEL (86, 172);
-	      KERNEL (88, 176);
-	      KERNEL (90, 180);
-	      KERNEL (92, 184);
-	      KERNEL (94, 188);
-	      KERNEL (96, 192);
-	      KERNEL (98, 196);
-	      KERNEL (100, 200);
-	      KERNEL (102, 204);
-	      KERNEL (104, 208);
-	      KERNEL (106, 212);
-	      KERNEL (108, 216);
-	      KERNEL (110, 220);
-	      KERNEL (112, 224);
-	      KERNEL (114, 228);
-	      KERNEL (116, 232);
-	      KERNEL (118, 236);
-	      KERNEL (120, 240);
-	      KERNEL (122, 244);
-	      KERNEL (124, 248);
-	      KERNEL (126, 252);
-	      AO += 1024;
-	      BO += 512;
-	    }
-	  if ((temp & 63) >> 5)
-	    {
-	      vec_t *rowA = (vec_t *) & AO[0];
-	      vec_t *rowB = (vec_t *) & BO[0];
-	      KERNEL (0, 0);
-	      KERNEL (2, 4);
-	      KERNEL (4, 8);
-	      KERNEL (6, 12);
-	      KERNEL (8, 16);
-	      KERNEL (10, 20);
-	      KERNEL (12, 24);
-	      KERNEL (14, 28);
-	      KERNEL (16, 32);
-	      KERNEL (18, 36);
-	      KERNEL (20, 40);
-	      KERNEL (22, 44);
-	      KERNEL (24, 48);
-	      KERNEL (26, 52);
-	      KERNEL (28, 56);
-	      KERNEL (30, 60);
-	      KERNEL (32, 64);
-	      KERNEL (34, 68);
-	      KERNEL (36, 72);
-	      KERNEL (38, 76);
-	      KERNEL (40, 80);
-	      KERNEL (42, 84);
-	      KERNEL (44, 88);
-	      KERNEL (46, 92);
-	      KERNEL (48, 96);
-	      KERNEL (50, 100);
-	      KERNEL (52, 104);
-	      KERNEL (54, 108);
-	      KERNEL (56, 112);
-	      KERNEL (58, 116);
-	      KERNEL (60, 120);
-	      KERNEL (62, 124);
-	      AO += 512;
-	      BO += 256;
-	    }
-	  if ((temp & 31) >> 4)
 	    {
 	      vec_t *rowA = (vec_t *) & AO[0];
 	      vec_t *rowB = (vec_t *) & BO[0];