py/asmthumb: Optimise native code calling runtime glue functions.

This patch makes the Thumb-2 native emitter use wide ldr instructions to
call into the runtime, when the index into the native glue function table
is 32 or greater.  This reduces the generated assembler code from 10 bytes
to 6 bytes, saving RAM and making native code run about 0.8% faster.

1 files changed, 4 insertions, 2 deletions

diff --git a/py/asmthumb.c b/py/asmthumb.c
index c5b45f2f5..ce9e4fdce 100644
--- a/py/asmthumb.c
+++ b/py/asmthumb.c
@@ -353,6 +353,8 @@ void asm_thumb_bcc_label(asm_thumb_t *as, int cond, uint label) {
 }
 
 #define OP_BLX(reg) (0x4780 | ((reg) << 3))
+#define OP_LDR_W_HI(reg_base) (0xf8d0 | (reg_base))
+#define OP_LDR_W_LO(reg_dest, imm12) ((reg_dest) << 12 | (imm12))
 #define OP_SVC(arg) (0xdf00 | (arg))
 
 void asm_thumb_bl_ind(asm_thumb_t *as, void *fun_ptr, uint fun_id, uint reg_temp) {
@@ -370,8 +372,8 @@ void asm_thumb_bl_ind(asm_thumb_t *as, void *fun_ptr, uint fun_id, uint reg_temp
         asm_thumb_op16(as, ASM_THUMB_FORMAT_9_10_ENCODE(ASM_THUMB_FORMAT_9_LDR | ASM_THUMB_FORMAT_9_WORD_TRANSFER, reg_temp, ASM_THUMB_REG_R7, fun_id));
         asm_thumb_op16(as, OP_BLX(reg_temp));
     } else {
-        // load ptr to function into register using immediate; 6 bytes
-        asm_thumb_mov_reg_i32(as, reg_temp, (mp_uint_t)fun_ptr);
+        // load ptr to function from table, indexed by fun_id using wide load; 6 bytes
+        asm_thumb_op32(as, OP_LDR_W_HI(ASM_THUMB_REG_R7), OP_LDR_W_LO(reg_temp, fun_id << 2));
         asm_thumb_op16(as, OP_BLX(reg_temp));
     }
 }