micro-optimizations in startup assembly code

bootloader.c

@@ -296,9 +296,6 @@ __attribute__((always_inline)) static inline void synthesize_block(uint32_t bloc
 }
 
 __attribute__((naked)) int main(void) {
-    // Don't reenter bootloader again
-    R16_BKP_DATAR10 = 0;
-
     // PLL setup: system clock 96 MHz
     R32_EXTEN_CTR |= (1 << 4);  // sneaky div2
     R32_RCC_CFGR0 = (R32_RCC_CFGR0 & ~0xff0000) | (0b01101000 << 16);

startup.S

@@ -28,7 +28,7 @@ _start:
     la a0, RCC_BASE
     lw a1, (R32_RCC_RSTSCKR-RCC_BASE)(a0)
     slli a1, a1, (31 - 26)
-    bgez a1, _enter_application_code     // if PINRSTF == 0
+    bgez a1, _bootloader_limit      // if PINRSTF == 0
 
     // Enable BKP and PWR clock
     lw a1, (R32_RCC_APB1PCENR-RCC_BASE)(a0)
@@ -66,7 +66,6 @@ _start:
     // Delay done, clear systick and magic
     sw zero, (STK_CTLR-SYSTICK_BASE)(a0)
     sw zero, (STK_SR-SYSTICK_BASE)(a0)
-    sh zero, (a3)
     // If the delay expires, boot the application after all
 
 _enter_application_code_magic:
@@ -86,6 +85,7 @@ _enter_application_code:
     j _bootloader_limit
 
 _enter_bootloader:
+    sh zero, (a3)
     la gp, 0
     la sp, 0