eval: faster fwd and bck w/ wraparound (thanks @lamarqua)

hellerve · Nov 18, 2018 · a865aa2f63a3e0db18e5d16a8d414528742490a5 · a865aa2
1 parent fb8b051
commit a865aa2f63a3e0db18e5d16a8d414528742490a5
Unified Split

Showing with 4 additions and 2 deletions.

+4 −2 src/eval.c
diff --git a/src/eval.c b/src/eval.c
@@ -43,8 +43,10 @@ do_dec: t[h]--; DISPATCH();
 do_fwd: h++; DISPATCH();
 do_bck: h--; DISPATCH();
 #else
-  
-do_fwd: h = h+1 % TAPE_LEN; DISPATCH();
-  
-do_bck: h = h-1 % TAPE_LEN; DISPATCH();
+  
+// modulo would be prettier here, but slows the code down by A LOT; somehow
+  
+// the C compilers can’t optimize uncoditional modulos here
+  
+do_fwd: h = h > TAPE_LEN-1 ? 0 : h+1; DISPATCH();
+  
+do_bck: h = h == 0 ? TAPE_LEN-1 : h-1; DISPATCH();
 #endif
 do_prn: printf("%c", t[h]); DISPATCH();
 do_read: scanf("%c", (char*)&t[h]); DISPATCH();