diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index e7d8802048..380fb79353 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -633,7 +633,7 @@ static const X86OpEntry opcodes_0F38_F0toFF[16][5] = {
         {},
     },
     [3] = {
-        X86_OP_GROUP3(group17, B,y, E,y, None,None, vex13 cpuid(BMI1)),
+        X86_OP_GROUP3(group17, B,y, None,None, E,y, vex13 cpuid(BMI1)),
         {},
         {},
         {},
@@ -2604,7 +2604,7 @@ static void disas_insn(DisasContext *s, CPUState *cpu)
     }
 
     /*
-     * Write back flags after last memory access.  Some newer ALU instructions, as
+     * Write back flags after last memory access.  Some older ALU instructions, as
      * well as SSE instructions, write flags in the gen_* function, but that can
      * cause incorrect tracking of CC_OP for instructions that write to both memory
      * and flags.
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index 2041ea9d04..a25b3dfc6b 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1272,40 +1272,34 @@ static void gen_BEXTR(DisasContext *s, X86DecodedInsn *decode)
     prepare_update1_cc(decode, s, CC_OP_LOGICB + ot);
 }
 
-/* BLSI do not have memory operands and can use set_cc_op.  */
 static void gen_BLSI(DisasContext *s, X86DecodedInsn *decode)
 {
     MemOp ot = decode->op[0].ot;
 
-    tcg_gen_mov_tl(cpu_cc_src, s->T0);
-    tcg_gen_neg_tl(s->T1, s->T0);
+    /* input in T1, which is ready for prepare_update2_cc  */
+    tcg_gen_neg_tl(s->T0, s->T1);
     tcg_gen_and_tl(s->T0, s->T0, s->T1);
-    tcg_gen_mov_tl(cpu_cc_dst, s->T0);
-    set_cc_op(s, CC_OP_BMILGB + ot);
+    prepare_update2_cc(decode, s, CC_OP_BMILGB + ot);
 }
 
-/* BLSMSK do not have memory operands and can use set_cc_op.  */
 static void gen_BLSMSK(DisasContext *s, X86DecodedInsn *decode)
 {
     MemOp ot = decode->op[0].ot;
 
-    tcg_gen_mov_tl(cpu_cc_src, s->T0);
-    tcg_gen_subi_tl(s->T1, s->T0, 1);
+    /* input in T1, which is ready for prepare_update2_cc  */
+    tcg_gen_subi_tl(s->T0, s->T1, 1);
     tcg_gen_xor_tl(s->T0, s->T0, s->T1);
-    tcg_gen_mov_tl(cpu_cc_dst, s->T0);
-    set_cc_op(s, CC_OP_BMILGB + ot);
+    prepare_update2_cc(decode, s, CC_OP_BMILGB + ot);
 }
 
-/* BLSR do not have memory operands and can use set_cc_op.  */
 static void gen_BLSR(DisasContext *s, X86DecodedInsn *decode)
 {
     MemOp ot = decode->op[0].ot;
 
-    tcg_gen_mov_tl(cpu_cc_src, s->T0);
-    tcg_gen_subi_tl(s->T1, s->T0, 1);
+    /* input in T1, which is ready for prepare_update2_cc  */
+    tcg_gen_subi_tl(s->T0, s->T1, 1);
     tcg_gen_and_tl(s->T0, s->T0, s->T1);
-    tcg_gen_mov_tl(cpu_cc_dst, s->T0);
-    set_cc_op(s, CC_OP_BMILGB + ot);
+    prepare_update2_cc(decode, s, CC_OP_BMILGB + ot);
 }
 
 static void gen_BOUND(DisasContext *s, X86DecodedInsn *decode)