Add Super R-Type

README.md

@@ -5,7 +5,7 @@
  \___ \ / /\ \______| | |  _  // _ \ / _ \| __|
  ____) / ____ \     | | | | \ \ (_) | (_) | |_ 
 |_____/_/    \_\    |_| |_|  \_\___/ \___/ \__|
-    Version 1.3              by Vitor Vilela
+    Version 1.4              by Vitor Vilela
 ```
 
 SA-1 Root is a project for enabling and accelerating games using the SA-1 chip.
@@ -15,6 +15,7 @@ high hardware compatibility and keeping most of the game aspects intact.
 So far the following games are available:
 * [Contra III](Contra-III) v1.1
 * [Gradius III](Gradius-III) v1.5
+* [Super R-Type](Super-R-Type) v1.0
 * Super Mario World ([SA-1 Pack only](https://github.com/VitorVilela7/SA1-Pack))
 
 What is it?
@@ -76,9 +77,9 @@ Credits
 SA-1 Root wouldn't be that awesome without help from these people:
 
 * indcsion (Gradius III testing)
-* Erivando_BR (Contra III testing, Contra III trace logs)
+* Erivando_BR (Contra III testing, Contra III trace logs and Super R-Type trace logs)
 * Jeffrey (Contra III testing)
-* kccheng (Contra III testing)
+* kccheng (Contra III and Super R-Type testing)
 * Rodzilla97 (Contra III bug report)
 * slidelljohn (Gradius III bug report and extra help) 
 * Vitor Vilela (crazy author that did most of the patches)

Super-R-Type/README.md

@@ -0,0 +1,89 @@
+# SA-1 Root: Super R-Type
+Version 1.0, released 2020-12-31
+
+Super R-Type is a classic shooter game made by Irem, being kind of a upgrade from R-Type II.
+
+This SA-1 Root patch removes all slowdown present on the original game and drastically reduces the
+loading times.
+
+Special thanks to Erivando_BR for sending me the SA-1 Collection trace log files of Super R-Type,
+which made this optimization patch possible to happen.
+
+## How to Patch
+
+Download the latest Super R-Type BPS patch file available on the
+[Releases](https://github.com/VitorVilela7/SA1-Root/releases) tab.
+
+You can patch it using [beat](https://www.romhacking.net/utilities/893/)
+or [FLIPS](https://sneslab.net/tools/floating.zip), both common .bps patchers.
+
+You can also patch the .asm files directly using
+[Asar](https://github.com/RPGHacker/asar).
+
+For more information on how to apply ROM patches, see this SnesLab
+article: https://sneslab.net/wiki/How_to_apply_ROM_patches
+
+It works only on the american version of Super R-Type. The japanese version is still being studied,
+and it will likely not work with this patch version.
+
+Expected checksums:
+
+### USA Version:
+#### Before patching:
+* CRC32: 8B22C830
+* SHA256: 05C7F6461209020785FBA33007E1830820AA44ADA4B1A6F991D936BF2335B15B
+
+#### After patching
+* CRC32: AB3B7626
+* SHA256: 1D890871DE7F4FF2EA489F37F3379545BD5F88F97C5FA67CC4AD60C7BD67C703
+
+## Compatibility
+
+It works on both real hardware (sd2snes or SA-1 cart) and emulators (Snes9x and bsnes/higan/ares).
+
+## Technical details
+
+* Remapping mode: full
+* Remapping strategy: static
+* SA-1 usage: full with parallelism
+
+This game relies on dynamic data pointers, but overall it does not use much of the Super Nintendo
+features. The game likes passing 16-bit pointers to RAM, using simple indirection to store into
+the registers, but overall the routines are somewhat organized, I didn't have to add dynamic
+remappers. The game only uses 32 KB of RAM ($7E:0000-$7E:7FFF), although it initializes 64 KB
+($7E:0000-$7E:FFFF). After some adjusts, it got it working with just 32 KB of BW-RAM.
+
+Unlike earlier patches, SA-1 was used as the "master" processor, being responsible for running the
+whole game and only routines that really required the SNES CPU were executed on the other processor.
+All calls to the SNES CPU are asynchronous and parallel, meaning that the SA-1 requests the SNES CPU
+to process a specific routine and continues executing without waiting the SNES CPU to terminate its
+job. Some routines were adapted to not cause RAM collision during the parallel execution of the
+routines.
+
+Because of the game characteristics, **32 kB (256 Kbit) of BW-RAM is required to the game run**
+correctly.
+
+## RAM remap
+
+* ``$0000-$1FFF`` -> ``$6000-$7FFF``
+* ``$7E:0000-$7E:7FFF`` -> ``$40:0000-$40:7FFF``
+
+## Credits
+
+Super R-Type - SA-1 Root wouldn't be that awesome without help from these people:
+
+* Erivando_BR (trace logs)
+* kccheng (testing)
+* Vitor Vilela (patch author)
+* You (for using it :D)
+
+## Contacting me
+
+You can contact me though the following links:
+
+* My Website: https://www.sneslab.net/
+* My Github profile: https://github.com/VitorVilela7
+* My Twitter profile: https://twitter.com/HackerVilela
+* My Patreon: https://www.patreon.com/vitorvilela
+* My personal blog: https://vilela.sneslab.net/
+

Super-R-Type/asm/README.md

@@ -0,0 +1,7 @@
+Instructions
+============
+
+Apply sa1.asm to the USA version of Super R-Type. Alternatively, use the do.sh script.
+
+You can also patch the .asm files directly using [Asar](https://github.com/RPGHacker/asar).
+

Super-R-Type/asm/databank.asm

@@ -0,0 +1,142 @@
+macro remap_wram_base(addr, old, new, size)
+	assert read1(<addr>+0) == $A9
+    
+	if !strict == 1
+        assert read<size>(<addr>+1) == <old>
+    endif
+    
+    org <addr>
+        LDA<new>
+endmacro
+
+macro remap_wram(addr)
+    %remap_wram_base(<addr>, $7E, ".B #$40", 1)
+endmacro
+
+macro remap_wram_local(addr)
+    %remap_wram_base(<addr>, $00, ".B #$40", 1)
+endmacro
+
+macro remap_wram_16(addr)
+    %remap_wram_base(<addr>, $007E, ".W #$0040", 2)
+endmacro
+
+; Bank 7E
+%remap_wram($008010)
+%remap_wram($0080A8)
+%remap_wram($008406)
+%remap_wram($008514)
+%remap_wram($00916F)
+%remap_wram($0091B6)
+%remap_wram($009206)
+%remap_wram($00982F)
+%remap_wram($009858)
+%remap_wram($009881)
+%remap_wram($0098A9)
+%remap_wram($0098FA)
+
+; Indirect pointers building
+%remap_wram($00A1A6)
+%remap_wram_16($00A1CF)
+
+; Indirect pointers that uses local RAM
+macro remap_indirect(orig, addr)
+    assert read1(<addr>+0) == $A0
+    
+    if !strict == 1
+        assert read2(<addr>+1) == <orig>
+    endif
+    
+    org <addr>
+        LDY.W #$6000|<orig>
+
+endmacro
+
+%remap_wram_local($008936)
+%remap_wram_local($0089AF)
+
+;these below doesn't work because the game often uses the pointers with DB set to $7E,
+;I made TLR remap differently the indexes by consequence.
+;%remap_indirect($189A, $008933)
+;%remap_indirect($189A, $0089AC)
+;%remap_indirect($1892, $009651)
+;%remap_indirect($1832, $009647)
+;%remap_indirect($1862, $00964C)
+
+; DMA parameters
+; Other DMA are either static, ROM-only or unused...
+%remap_wram($008329)
+%remap_wram($00834F)
+%remap_wram($00838C)
+%remap_wram_local($0083C2)
+
+macro remap_mvn(b1, b2, addr)
+    if !strict == 1
+        assert read1(<addr>+0) == $54
+        assert read1(<addr>+1) == <b1>
+        assert read1(<addr>+2) == <b2>
+    endif
+    
+    !sb1 = <b1>
+    !sb2 = <b2>
+    
+    if !sb1&$FE == $7E
+        !sb1 #= !sb1-$7E+$40
+    endif
+    
+    if !sb2&$FE == $7E
+        !sb2 #= !sb2-$7E+$40
+    endif
+    
+    assert !sb1 == $40 || !sb1 == $41 || !sb2 == $40 || !sb2 == $41
+    assert !sb1 != $7E && !sb2 != $7E && !sb1 != $7F && !sb2 != $7F
+    
+    org <addr>
+        MVN !sb1, !sb2
+endmacro
+
+%remap_mvn($7E, $7E, $008720)
+%remap_mvn($7E, $7E, $008744)
+%remap_mvn($7E, $05, $009B81)
+%remap_mvn($7E, $05, $009BA4)
+%remap_mvn($7E, $05, $009BB6)
+%remap_mvn($7E, $05, $00A183)
+
+macro remap_indirect_jump(dest, addr)
+    assert read1(<addr>+0) == $DC
+    
+    if !strict == 1
+        assert read2(<addr>+1) == <dest>
+    endif
+    
+    !res #= <dest>|$6000
+    
+    assert !res >= $6000 && !res <= $7FFF
+    
+    org <addr>
+        JMP.W [!res]
+endmacro
+
+%remap_indirect_jump($0010, $008253)
+%remap_indirect_jump($1801, $0084D4) 
+%remap_indirect_jump($1801, $00872D) 
+%remap_indirect_jump($00F5, $008758)
+%remap_indirect_jump($00F5, $008791)
+%remap_indirect_jump($00F5, $0087D7)
+%remap_indirect_jump($00F5, $00881A)
+
+; Format: <address to store>, <value to store>
+; Remap the WRAM specific to BWRAM.
+
+; It's worth commenting that the address is incremented by 5,
+; while the code only read the first four values. It's suggestive
+; that the table was supposed to be incremented by 4.
+
+; It's very likely a bug, but we have to remain the original behavior
+org $0480BC
+	db $02,$21,$00,$00,$07
+	db $42,$60,$00,$09,$42
+	db $00,$00,$16,$21,$00
+	db $00,$00,$00
+
+

Super-R-Type/asm/do.sh

@@ -0,0 +1 @@
+rm srtype.sfc; cp srtype_base.sfc srtype.sfc && asar -Dstrict=1 sa1.asm srtype.sfc