tools/mpy_ld.py: Support jumping more than 2k on armv6m architectures.

Native .mpy files targetting armv6m (eg RP2040) cannot currently have more
than about 2kiB of native code (between the start of the file and the init
function).

This commit fixes that by using bigger jumps to jump to the init function.

Signed-off-by: Damien George <damien@micropython.org>

tools/mpy_ld.py

@@ -89,10 +89,26 @@ def asm_jump_x86(entry):
 
 
 def asm_jump_thumb(entry):
-    # Only signed values that fit in 12 bits are supported
+    # This function must return the same number of bytes for the encoding of the jump
+    # regardless of the value of `entry`.
     b_off = entry - 4
-    assert b_off >> 11 == 0 or b_off >> 11 == -1, b_off
-    return struct.pack("<H", 0xE000 | (b_off >> 1 & 0x07FF))
+    if b_off >> 11 == 0 or b_off >> 11 == -1:
+        # Signed value fits in 12 bits.
+        b0 = 0xE000 | (b_off >> 1 & 0x07FF)
+        b1 = 0
+        b2 = 0
+        b3 = 0
+    else:
+        # Use bl to do a large jump/call:
+        #   push {r0, lr}
+        #   bl <dest>
+        #   pop {r0, pc}
+        b_off -= 2  # skip "push {r0, lr}"
+        b0 = 0xB400 | 0x0100 | 0x0001  # push, lr, r0
+        b1 = 0xF000 | (((b_off) >> 12) & 0x07FF)
+        b2 = 0xF800 | (((b_off) >> 1) & 0x07FF)
+        b3 = 0xBC00 | 0x0100 | 0x0001  # pop, pc, r0
+    return struct.pack("<HHHH", b0, b1, b2, b3)
 
 
 def asm_jump_thumb2(entry):