in-game hooks: more time efficient Auto Joypad Read snooping

This is a bit of a fire-and-forget strategy. After NMI/IRQ read the AJR
result register as fast as possible to catch the readout from last
frame.
In some occasions AJR has already JUST started (its starting point is
variable) in which case the readout is #$0000. For IRQ hooks this might
lead to unusable readouts depending on the time of the IRQ.
Also don't compare controller button mappings when no buttons are
pressed.

snes/nmihook.a65

@@ -1,30 +1,42 @@
 #include "memmap.i65"
 
-; *=$2BF0
+; *=$2BB0
 nmihook:
-  jsr !BRAM_PAYLOAD
+  php
+  rep #$30 : .al : .xl
+  pha
+  lda @$004218
+  sta @NMI_PAD
+  jsr BRAM_PAYLOAD
+  pla
+  plp
   jmp $ffff
+; *=2BC4
 irqhook:
-  jsr !BRAM_PAYLOAD
+  php
+  rep #$30 : .al : .xl
+  pha
+  lda @$004218
+  sta @NMI_PAD
+  jsr BRAM_PAYLOAD
+  pla
+  plp
   jmp $ffff
 
 ; *=$2A04
 nmi_payload:
-  php
-  rep #$30 : .al : .xl
   phb
   phk
   plb
-  pha
-  lda #$7548
-  sta !BRAM_UNLOCK
-  lda #$7a72
-  sta !BRAM_UNLOCK+2
   phx
   phy
   phd
   lda #$2b00
   tcd
+  lda #$7548
+  sta <BRAM_UNLOCK
+  lda #$7a72
+  sta <BRAM_UNLOCK+2
 ; backup vectors
   sep #$20 : .as
   lda #$01
@@ -36,16 +48,16 @@ nmi_payload:
   stx <IRQ_VECT_ORIG
   stz <NMI_VECT_DISABLE
 ; only read controller every couple of frames
-  clc
-  lda <NMI_RUNMASK
-  cmp #$80
-  rol
-  sta <NMI_RUNMASK
-  bcc nmi_patches
+;  clc
+;  lda <NMI_RUNMASK
+;  cmp #$80
+;  rol
+;  sta <NMI_RUNMASK
+;  bcc nmi_patches
 ; read controller
   lda <MIRROR_4200
   lsr
-; auto read is active, so we wait for it to finish and use the result
+; auto read is active, so we reuse the early snapshot taken
   bcs nmi_autoread
 ; no auto read; read the controller manually
   stz $4016
@@ -65,21 +77,19 @@ nmi_payload:
   rol <NMI_PAD+1
   dex
   bne -
+  bra nmi_pad_shortcut
+nmi_autoread:
+  ldx #$0000
+nmi_pad_shortcut:
+  ldy <NMI_PAD
+  beq nmi_patches
 ; check button combinations
 ; L+R+Start+Select : $3030
 ; L+R+Select+X     : $2070
 ; L+R+Start+A      : $10b0
 ; L+R+Start+B      : $9030
 ; L+R+Start+Y      : $5030
 ; L+R+Start+X      : $1070
-  bra nmi_pad_loop
-nmi_autoread:
-- lda $4212
-  lsr
-  bcs -
-  ldy $4218
-  sty <NMI_PAD
-  ldx #$0000
 nmi_pad_loop:
   ldy <NMI_PAD_BUTTONS, x
   cpy <NMI_PAD
@@ -126,14 +136,12 @@ nmi_patches:
   bne -
 nmi_exit:
 ; clean up
+  lda $ffea ; get another 72 cycles of BRAM visibility
+  stz <BRAM_UNLOCK ; disable BRAM visibility for game
   pld
-  lda $ffea ; get another 48 cycles of BRAM visibility
-  stz !BRAM_UNLOCK ; disable BRAM visibility for game
   rep #$30 : .al : .xl
   ply
   plx
-  pla
   plb
-  plp
   rts
 .byt "bram_end"