Put __etext at address 0 so the decoder knows how much to decode instead of always decoding the entire ROM

Author: object-Object

asm/branch.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     li x1, 0

asm/large_bin.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     la x1, word1

asm/load.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     la x1, word

asm/math.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     li x1, 0b1

asm/print.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     la s0, msg

asm/store.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     la x1, word

asm/system.s

@@ -1,3 +1,6 @@
+.section .text.start
+    .word __etext
+
 .global _start
 _start:
     # PAUSE (apparently not supported by clang)

coremark/mlogv32/entry.s

@@ -1,6 +1,7 @@
 .section .text.start
-.global _start
+    .word __etext
 
+.global _start
 _start:
     la t1, _stack_start
     andi sp, t1, -16

rust/mlogv32/src/lib.rs

@@ -30,6 +30,8 @@ static HEAP: Heap = Heap::empty();
 #[rustfmt::skip]
 global_asm!("
 .section .text.start
+    .word __etext
+
     la gp, __global_pointer$
     
     la t1, _stack_start

src/decoder.mlog

@@ -83,8 +83,15 @@ reset__find_lookup_start:
 
     set LOOKUP_START i
 
-    # start decoding at address 0
-    set pc 0
+    # read __etext from address 0
+    set address 0
+    op add ret @counter 1
+    jump load_word always
+    set __etext result
+
+    # start decoding at address 4
+    # (address 0 is __etext)
+    set pc 4
     set opcache_var null
 
     set START_TIME @time
@@ -370,7 +377,7 @@ reset__skip_locate_pc:
     read opcache_var INCR_PROC opcache_var
 
     op add pc pc 4
-    jump main lessThan pc ROM_END
+    jump main lessThan pc __etext
 
     # done!
     op sub DECODE_DURATION @time START_TIME