Minor vec4 stack manipulation optimizations.

vvp/schedule.cc

@@ -165,12 +165,14 @@ struct assign_vector4_event_s  : public event_s {
 	    base = 0;
 	    vwid = 0;
       }
+#if 0
 	/* A constructor that makes the val directly. */
       assign_vector4_event_s(const vvp_vector4_t&that, unsigned adr, unsigned wid)
       : val(that,adr,wid) {
 	    base = 0;
 	    vwid = 0;
       }
+#endif
 
 	/* Where to do the assign. */
       vvp_net_ptr_t ptr;
@@ -782,26 +784,12 @@ void schedule_assign_vector(vvp_net_ptr_t ptr,
       schedule_event_(cur, delay, SEQ_NBASSIGN);
 }
 
-void schedule_assign_plucked_vector(vvp_net_ptr_t ptr,
+void schedule_propagate_vector(vvp_net_t*net,
-				    vvp_time64_t delay,
+			       vvp_time64_t delay,
-				    const vvp_vector4_t&src,
+			       const vvp_vector4_t&src)
-				    unsigned adr, unsigned wid)
-{
-      struct assign_vector4_event_s*cur
-	    = new struct assign_vector4_event_s(src,adr,wid);
-      cur->ptr = ptr;
-      cur->vwid = 0;
-      cur->base = 0;
-      schedule_event_(cur, delay, SEQ_NBASSIGN);
-}
-
-void schedule_propagate_plucked_vector(vvp_net_t*net,
-				       vvp_time64_t delay,
-				       const vvp_vector4_t&src,
-				       unsigned adr, unsigned wid)
 {
       struct propagate_vector4_event_s*cur
-	    = new struct propagate_vector4_event_s(src,adr,wid);
+	    = new struct propagate_vector4_event_s(src);
       cur->net = net;
       schedule_event_(cur, delay, SEQ_NBASSIGN);
 }

vvp/schedule.h

@@ -48,11 +48,6 @@ extern void schedule_assign_vector(vvp_net_ptr_t ptr,
 				   const vvp_vector4_t&val,
 				   vvp_time64_t  delay);
 
-extern void schedule_assign_plucked_vector(vvp_net_ptr_t ptr,
-					   vvp_time64_t  delay,
-					   const vvp_vector4_t&val,
-					   unsigned adr, unsigned wid);
-
 extern void schedule_assign_array_word(vvp_array_t mem,
 				       unsigned word_address,
 				       unsigned off,
@@ -66,10 +61,9 @@ extern void schedule_assign_array_word(vvp_array_t mem,
 /*
  * Create an event to propagate the output of a net.
  */
-extern void schedule_propagate_plucked_vector(vvp_net_t*ptr,
+extern void schedule_propagate_vector(vvp_net_t*ptr,
-					      vvp_time64_t  delay,
+				      vvp_time64_t  delay,
-					      const vvp_vector4_t&val,
+				      const vvp_vector4_t&val);
-					      unsigned adr, unsigned wid);
 
 /*
  * This is very similar to schedule_assign_vector, but generates an

vvp/vthread.cc

@@ -993,9 +993,10 @@ bool of_ASSIGN_VEC4(vthread_t thr, vvp_code_t cp)
 {
       vvp_net_ptr_t ptr (cp->net, 0);
       unsigned delay = cp->bit_idx[0];
-      vvp_vector4_t val = thr->pop_vec4();
+      vvp_vector4_t&val = thr->peek_vec4();
 
-      schedule_assign_plucked_vector(ptr, delay, val, 0, val.size());
+      schedule_assign_vector(ptr, 0, 0, val, delay);
+      thr->pop_vec4(1);
       return true;
 }
 
@@ -1686,11 +1687,12 @@ static void do_CMPU(vthread_t thr, const vvp_vector4_t&lval, const vvp_vector4_t
 bool of_CMPU(vthread_t thr, vvp_code_t)
 {
 
-      vvp_vector4_t rval = thr->pop_vec4();
+      const vvp_vector4_t&rval = thr->peek_vec4(0);
-      vvp_vector4_t lval = thr->pop_vec4();
+      const vvp_vector4_t&lval = thr->peek_vec4(1);
 
       do_CMPU(thr, lval, rval);
 
+      thr->pop_vec4(2);
       return true;
 }
 
@@ -3737,7 +3739,7 @@ bool of_PAD_S(vthread_t thr, vvp_code_t cp)
 {
       unsigned wid = cp->number;
 
-      vvp_vector4_t val = thr->pop_vec4();
+      vvp_vector4_t&val = thr->peek_vec4();
       unsigned old_size = val.size();
       val.resize(wid);
 
@@ -3748,8 +3750,6 @@ bool of_PAD_S(vthread_t thr, vvp_code_t cp)
 		  val.set_bit(idx, sb);
       }
 
-      thr->push_vec4(val);
-
       return true;
 }
 
@@ -5400,17 +5400,20 @@ bool of_STORE_VEC4A(vthread_t thr, vvp_code_t cp)
       unsigned adr_index = cp->bit_idx[0];
       unsigned off_index = cp->bit_idx[1];
 
-      vvp_vector4_t value = thr->pop_vec4();
+      vvp_vector4_t&value = thr->peek_vec4();
 
       long adr = adr_index? thr->words[adr_index].w_int : 0;
       long off = off_index? thr->words[off_index].w_int : 0;
 
 	// Suppress action if flags-4 is true.
-      if (thr->flags[4] == BIT4_1)
+      if (thr->flags[4] == BIT4_1) {
+	    thr->pop_vec4(1);
 	    return true;
+      }
 
       array_set_word(cp->array, adr, off, value);
 
+      thr->pop_vec4(1);
       return true;
 }
 

vvp/vvp_net.cc

@@ -3042,7 +3042,7 @@ void vvp_wide_fun_core::propagate_vec4(const vvp_vector4_t&bit,
 				       vvp_time64_t delay)
 {
       if (delay)
-	    schedule_propagate_plucked_vector(ptr_, delay, bit, 0, bit.size());
+	    schedule_propagate_vector(ptr_, delay, bit);
       else
 	    ptr_->send_vec4(bit, 0);
 }