-x-initial-edge breaks with logic bug fix

[veripoolbot]

---

Author Name: Wilson Snyder (@wsnyder)
Original Redmine Issue: 604 from https://www.veripool.org
Original Date: 2013-01-17
Original Assignee: Jeremy Bennett (@jeremybennett)

---

Part of the fix for #� uncovered a bug I had in V3AstNodes.cpp. The git head now includes these lines:

 AstBasicDType* new1p = new AstBasicDType(fl, AstBasicDTypeKwd::BIT, numeric, width, widthMin);
 // Above should be below, but fails --x-initial-edge test
 //AstBasicDType* new1p = new AstBasicDType(fl, kwd, numeric, width, widthMin);

The use of ::BIT here should have been kwd. This bug means some logic types were mis-converted to bit types.

However fixing this breaks the t_initial_edge test.

Can you look into this?

[veripoolbot]

---

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2013-01-17T01:55:10Z

---

#�_initial_edge branch contains the necessary edits, as another test needed a fixup too.

[veripoolbot]

---

Original Redmine Comment
Author Name: Jeremy Bennett (@jeremybennett)
Original Date: 2013-01-17T10:45:38Z

---

The test for when to implement --x-initial-edge in EmitCImp::emitVarResets() in V3EmitC.cpp was wrong.

index 5634055..e148921 100644
--- a/src/V3EmitC.cpp
+++ b/src/V3EmitC.cpp
@@ -1368,17 +1368,16 @@ void EmitCImp::emitVarResets(AstNodeModule* modp) {
                 } else {
                     puts(varp->name());
                     for (int v=0; v<vects; ++v) puts( "[__Vi"+cvtToStr(v)+"]");
-                   if (zeroit) {
-                       // We want to force an initial edge on uninitialized clocks (from 'X' to
-                       // whatever the first value is). Since the class is instantiated before
-                       // initial blocks are evaluated, this should not clash with any initial
-                       // block settings. Clocks are always BIT datatypes, so zeroit is true.
-                       if (v3Global.opt.xInitialEdge()
-                           && (0 == varp->name().find("__Vclklast__"))) {
-                           puts(" = 1;\n");
-                       } else {
-                           puts(" = 0;\n");
-                       }
+                   // If --x-initial-edge is set, we want to force an initial
+                   // edge on uninitialized clocks (from 'X' to whatever the
+                   // first value is). Since the class is instantiated before
+                   // initial blocks are evaluated, this should not clash
+                   // with any initial block settings.
+                   if (v3Global.opt.xInitialEdge() && varp->isUsedClock()) {
+                       puts(" = 0;\n");
+                   } else if (v3Global.opt.xInitialEdge()
+                              && (0 == varp->name().find("__Vclklast__"))) {
+                       puts(" = 1;\n");
                     } else {
                         puts(" = VL_RAND_RESET_");
                         emitIQW(varp);

Please pull a corrected version (as a sub-branch of #�_initial_edge) from branch x-initial-edge at git@github.com:jeremybennett/verilator.git

[veripoolbot]

---

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2013-01-17T12:21:14Z

---

Thanks for the quick fix. I added a || zeroit as otherwise the t_var_types test correctly noticed that a 2-state type wasn't zero initialized. This should be still compatible with other simulators as a two-state which inits to zero will NOT trigger a negedge, it's the X->0 that does.

Applied.