merge master

.gitmodules

@@ -1,6 +0,0 @@
-[submodule "tools/n64splat"]
-	path = tools/n64splat
-	url = https://github.com/ethteck/n64splat.git
-[submodule "tools/star-rod"]
-	path = tools/star-rod
-	url = https://github.com/nanaian/star-rod-for-decomp.git

Makefile

@@ -130,14 +130,9 @@ clean-code:
 tools:
 	make -C tools
 
-setup: clean-all submodules tools
+setup: clean-all tools
 	@make split
 
-# tools/star-rod submodule intentionally omitted
-submodules:
-	git submodule init tools/n64splat
-	git submodule update --recursive
-
 split:
 	make $(LD_SCRIPT) -W $(SPLAT_YAML)
 
@@ -297,10 +292,6 @@ include/ld_addrs.h: $(BUILD_DIR)/$(LD_SCRIPT)
 
 STAR_ROD := cd tools/star-rod && $(JAVA) -jar StarRod.jar
 
-# lazily initialise the submodule
-tools/star-rod:
-	git submodule init tools/star-rod
-
 sprite/SpriteTable.xml: tools/star-rod sources.mk
 	$(PYTHON) tools/star-rod/spritetable.xml.py $(NPC_SPRITES) > $@
 
@@ -310,7 +301,7 @@ editor: tools/star-rod sprite/SpriteTable.xml
 
 ### Make Settings ###
 
-.PHONY: clean tools test setup submodules split editor $(ROM)
+.PHONY: clean tools test setup split editor $(ROM)
 .DELETE_ON_ERROR:
 .SECONDARY:
 .PRECIOUS: $(ROM) %.Yay0

asm/nonmatchings/code_7345A0/func_802A1000_7345A0.s

@@ -1,25 +0,0 @@
-.set noat      # allow manual use of $at
-.set noreorder # don't insert nops after branches
-
-glabel func_802A1000_7345A0
-/* 7345A0 802A1000 27BDFFE8 */  addiu     $sp, $sp, -0x18
-/* 7345A4 802A1004 AFB00010 */  sw        $s0, 0x10($sp)
-/* 7345A8 802A1008 0080802D */  daddu     $s0, $a0, $zero
-/* 7345AC 802A100C 2402000F */  addiu     $v0, $zero, 0xf
-/* 7345B0 802A1010 3C03800E */  lui       $v1, %hi(gBattleStatus+0x83)
-/* 7345B4 802A1014 8063C0F3 */  lb        $v1, %lo(gBattleStatus+0x83)($v1)
-/* 7345B8 802A1018 24040015 */  addiu     $a0, $zero, 0x15
-/* 7345BC 802A101C AFBF0014 */  sw        $ra, 0x14($sp)
-/* 7345C0 802A1020 AE020088 */  sw        $v0, 0x88($s0)
-/* 7345C4 802A1024 0C03A752 */  jal       is_ability_active
-/* 7345C8 802A1028 AE030084 */   sw       $v1, 0x84($s0)
-/* 7345CC 802A102C 10400003 */  beqz      $v0, .L802A103C
-/* 7345D0 802A1030 24020028 */   addiu    $v0, $zero, 0x28
-/* 7345D4 802A1034 AE000084 */  sw        $zero, 0x84($s0)
-/* 7345D8 802A1038 AE020088 */  sw        $v0, 0x88($s0)
-.L802A103C:
-/* 7345DC 802A103C 8FBF0014 */  lw        $ra, 0x14($sp)
-/* 7345E0 802A1040 8FB00010 */  lw        $s0, 0x10($sp)
-/* 7345E4 802A1044 24020002 */  addiu     $v0, $zero, 2
-/* 7345E8 802A1048 03E00008 */  jr        $ra
-/* 7345EC 802A104C 27BD0018 */   addiu    $sp, $sp, 0x18

asm/nonmatchings/code_7345A0/func_802A1050_7345F0.s

@@ -1,26 +0,0 @@
-.set noat      # allow manual use of $at
-.set noreorder # don't insert nops after branches
-
-glabel func_802A1050_7345F0
-/* 7345F0 802A1050 27BDFFE8 */  addiu     $sp, $sp, -0x18
-/* 7345F4 802A1054 AFB00010 */  sw        $s0, 0x10($sp)
-/* 7345F8 802A1058 0080802D */  daddu     $s0, $a0, $zero
-/* 7345FC 802A105C 24040016 */  addiu     $a0, $zero, 0x16
-/* 734600 802A1060 AFBF0014 */  sw        $ra, 0x14($sp)
-/* 734604 802A1064 0C03A752 */  jal       is_ability_active
-/* 734608 802A1068 AE000084 */   sw       $zero, 0x84($s0)
-/* 73460C 802A106C 10400002 */  beqz      $v0, .L802A1078
-/* 734610 802A1070 24020001 */   addiu    $v0, $zero, 1
-/* 734614 802A1074 AE020084 */  sw        $v0, 0x84($s0)
-.L802A1078:
-/* 734618 802A1078 3C02800E */  lui       $v0, %hi(gBattleStatus)
-/* 73461C 802A107C 8C42C070 */  lw        $v0, %lo(gBattleStatus)($v0)
-/* 734620 802A1080 30421000 */  andi      $v0, $v0, 0x1000
-/* 734624 802A1084 10400002 */  beqz      $v0, .L802A1090
-/* 734628 802A1088 24020001 */   addiu    $v0, $zero, 1
-/* 73462C 802A108C AE020084 */  sw        $v0, 0x84($s0)
-.L802A1090:
-/* 734630 802A1090 8FBF0014 */  lw        $ra, 0x14($sp)
-/* 734634 802A1094 8FB00010 */  lw        $s0, 0x10($sp)
-/* 734638 802A1098 24020002 */  addiu     $v0, $zero, 2
-/* 73463C 802A109C 03E00008 */  jr        $ra

include/macros.h

@@ -78,4 +78,6 @@
 #define _NS(x, y) x ## _ ## y
 #define NS(x, y) _NS(x, y)
 
+#define ASCII_TO_U32(a, b, c, d) ((u32)((a << 24) | (b << 16) | (c << 8) | (d << 0)))
+
 #endif

src/code_1060_len_310.c

@@ -4,9 +4,10 @@ u32 func_80025D74(void* arg0, const unsigned char* str, s32 count);
 
 void func_80025C60(void) {
     OSPiHandle** handle = &carthandle;
+
     osEPiWriteIo(*handle, 0xB3FF0014, 0);
     osEPiWriteIo(*handle, 0xB3FF0004, 0);
-    osEPiWriteIo(*handle, 0xB3FF0000, 0x49533634);
+    osEPiWriteIo(*handle, 0xB3FF0000, ASCII_TO_U32('I', 'S', '6', '4'));
 }
 
 void func_80025CC0(const char* fmt, ...) {

src/code_7345A0.c

@@ -1,5 +1,27 @@
 #include "common.h"
 
-INCLUDE_ASM(s32, "code_7345A0", func_802A1000_7345A0);
+ApiStatus func_802A1000_7345A0(ScriptInstance* script, s32 isInitialCall) {
+    script->varTable[0] = BATTLE_STATUS->unk_83;
+    script->varTable[1] = 15;
 
-INCLUDE_ASM(s32, "code_7345A0", func_802A1050_7345F0);
+    if (is_ability_active(Ability_BERSERKER)) {
+        script->varTable[0] = 0;
+        script->varTable[1] = 40;
+    }
+
+    return ApiStatus_DONE2;
+}
+
+ApiStatus func_802A1050_7345F0(ScriptInstance* script, s32 isInitialCall) {
+    script->varTable[0] = 0;
+
+    if (is_ability_active(Ability_RIGHT_ON)) {
+        script->varTable[0] = 1;
+    }
+
+    if (gBattleStatus.flags1 & 0x1000) {
+        script->varTable[0] = 1;
+    }
+
+    return ApiStatus_DONE2;
+}

tools/n64splat

@@ -1 +0,0 @@
-Subproject commit 41146bdb8f07bf82c7004f141126d6186ce3d43e

tools/n64splat/.gitignore

@@ -0,0 +1,5 @@
+.idea/
+venv/
+.vscode/
+__pycache__/
+util/Yay0decompress

tools/n64splat/.gitrepo

@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/git-commands/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/ethteck/n64splat.git
+	branch = master
+	commit = 7574db712ef19ca420904c82d3559e9ac4b8c5f5
+	parent = 86760369a5ab977c037c21aebf6f10484570642f
+	method = merge
+	cmdver = 0.4.3

tools/n64splat/Makefile

@@ -0,0 +1,11 @@
+UTIL_DIR := util
+
+default: all
+
+all: Yay0decompress
+
+Yay0decompress:
+	gcc $(UTIL_DIR)/Yay0decompress.c -fPIC -shared -O3 -o $(UTIL_DIR)/Yay0decompress
+
+clean:
+	rm -f $(UTIL_DIR)/Yay0decompress

tools/n64splat/README.md

@@ -0,0 +1,8 @@
+# n64splat
+A n64 rom splitting tool to assist with decompilation and modding projects
+
+For example usage, see https://github.com/ethteck/papermario
+The Makefile `setup` target calls n64splat with a config file that you can use for reference. More documentation coming soon.
+
+### Requirements
+Python package requirements can be installed via `pip3 install -r requirements.txt`

tools/n64splat/create_config.py

@@ -0,0 +1,65 @@
+#! /usr/bin/env python3
+
+from capstone import *
+from capstone.mips import *
+
+import argparse
+from util import rominfo
+from segtypes.code import N64SegCode
+
+parser = argparse.ArgumentParser(description="Create a splat config from a rom")
+parser.add_argument("rom", help="path to a .z64 rom")
+
+
+def main(rom_path):
+    rom = rominfo.get_info(rom_path)
+    basename = rom.name.replace(" ", "").lower()
+
+    header = \
+"""name: {0} ({1})
+basename: {2}
+options:
+  find_file_boundaries: True
+  compiler: "IDO"
+""".format(rom.name.title(), rom.get_country_name(), basename)
+
+    with open(rom_path, "rb") as f:
+      fbytes = f.read()
+
+    rom_addr = 0x1000
+
+    md = Cs(CS_ARCH_MIPS, CS_MODE_MIPS64 + CS_MODE_BIG_ENDIAN)
+    for insn in md.disasm(fbytes[rom_addr:], rom.entry_point):
+        rom_addr += 4
+
+    segments = \
+"""segments:
+  - name: header
+    type: header
+    start: 0x0
+    vram: 0
+    files:
+      - [0x0, header, header]
+  - name: boot
+    type: bin
+    start: 0x40
+  - name: main
+    type: code
+    start: 0x1000
+    vram: 0x{:X}
+    files:
+      - [0x1000, asm]
+  - type: bin
+    start: 0x{:X}
+  - [0x{:X}]
+""".format(rom.entry_point, rom_addr, rom.size)
+
+    outstr = header + segments
+
+    outname = rom.name.replace(" ", "").lower()
+    with open(outname + ".yaml", "w", newline="\n") as f:
+        f.write(outstr)
+
+if __name__ == "__main__":
+    args = parser.parse_args()
+    main(args.rom)

tools/n64splat/list_objects.py

@@ -0,0 +1,32 @@
+#! /usr/bin/python3
+
+import argparse
+import yaml
+from pathlib import PurePath
+
+from split import initialize_segments
+
+parser = argparse.ArgumentParser(description="List output objects for linker script")
+parser.add_argument("config", help="path to a compatible config .yaml file")
+
+def main(config_path):
+    # Load config
+    with open(config_path) as f:
+        config = yaml.safe_load(f.read())
+
+    options = config.get("options")
+    replace_ext = options.get("ld_o_replace_extension", True)
+
+    # Initialize segments
+    all_segments = initialize_segments(options, config_path, config["segments"])
+
+    for segment in all_segments:
+        for subdir, path, obj_type, start in segment.get_ld_files():
+            path = PurePath(subdir) / PurePath(path)
+            path = path.with_suffix(".o" if replace_ext else path.suffix + ".o")
+
+            print(path)
+
+if __name__ == "__main__":
+    args = parser.parse_args()
+    main(args.config)

tools/n64splat/requirements.txt

@@ -0,0 +1,5 @@
+PyYAML>=5.3.1,<6
+pypng==0.0.20
+colorama>=0.4.4,<0.5
+python-ranges>=0.1.3,<0.2
+capstone>=4.0.2,<5

tools/n64splat/segtypes/Yay0.py

@@ -0,0 +1,25 @@
+import os
+from segtypes.segment import N64Segment
+from pathlib import Path
+from util import Yay0decompress
+
+class N64SegYay0(N64Segment):
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_parent_dir(base_path + "/bin", self.name)
+
+        path = os.path.join(out_dir, os.path.basename(self.name) + ".bin")
+        with open(path, "wb") as f:
+            self.log(f"Decompressing {self.name}...")
+            compressed_bytes = rom_bytes[self.rom_start : self.rom_end]
+            decompressed_bytes = Yay0decompress.decompress_yay0(compressed_bytes)
+            f.write(decompressed_bytes)
+        self.log(f"Wrote {self.name} to {path}")
+
+
+    def get_ld_files(self):
+        return [("bin", f"{self.name}.Yay0", ".data", self.rom_start)]
+
+
+    @staticmethod
+    def get_default_name(addr):
+        return "Yay0/{:X}".format(addr)

tools/n64splat/segtypes/bin.py

@@ -0,0 +1,21 @@
+import os
+from segtypes.segment import N64Segment
+from pathlib import Path
+
+
+class N64SegBin(N64Segment):
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_split_dir(base_path, "bin")
+
+        bin_path = os.path.join(out_dir,  self.name + ".bin")
+        Path(bin_path).parent.mkdir(parents=True, exist_ok=True)
+        with open(bin_path, "wb") as f:
+            f.write(rom_bytes[self.rom_start: self.rom_end])
+        self.log(f"Wrote {self.name} to {bin_path}")
+
+    def get_ld_files(self):
+        return [("bin", f"{self.name}.bin", ".data", self.rom_start)]
+
+    @staticmethod
+    def get_default_name(addr):
+        return "bin_{:X}".format(addr)

tools/n64splat/segtypes/ci4.py

@@ -0,0 +1,15 @@
+from segtypes.ci8 import N64SegCi8
+
+class N64SegCi4(N64SegCi8):
+    def parse_image(self, data):
+        img_data = bytearray()
+
+        for i in range(self.width * self.height // 2):
+            img_data.append(data[i] >> 4)
+            img_data.append(data[i] & 0xF)
+
+        return img_data
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height // 2

tools/n64splat/segtypes/ci8.py

@@ -0,0 +1,62 @@
+from segtypes.segment import N64Segment
+from segtypes.rgba16 import N64SegRgba16
+import png
+import os
+from util import Yay0decompress
+
+
+class N64SegCi8(N64SegRgba16):
+    def __init__(self, segment, next_segment, options):
+        super().__init__(segment, next_segment, options)
+
+        self.path = None
+
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_parent_dir(base_path + "/img", self.name)
+        self.path = os.path.join(out_dir, os.path.basename(self.name) + ".png")
+
+        data = rom_bytes[self.rom_start: self.rom_end]
+        if self.compressed:
+            data = Yay0decompress.decompress_yay0(data)
+
+        self.image = self.parse_image(data)
+
+    def postsplit(self, segments):
+        palettes = [seg for seg in segments if seg.type ==
+                    "palette" and seg.image_name == self.name]
+
+        if len(palettes) == 0:
+            self.error(f"no palette sibling segment exists\n(hint: add a segment with type 'palette' and name '{self.name}')")
+            return
+
+        seen_paths = []
+
+        for pal_seg in palettes:
+            if pal_seg.path in seen_paths:
+                self.error(f"palette name '{pal_seg.name}' is not unique")
+                return
+            seen_paths.append(pal_seg.path)
+
+            w = png.Writer(self.width, self.height, palette=pal_seg.palette)
+
+            with open(pal_seg.path, "wb") as f:
+                w.write_array(f, self.image)
+                self.log(f"Wrote {pal_seg.name} to {pal_seg.path}")
+
+        # canonical version of image (not palette!) data
+        if self.path not in seen_paths:
+            w = png.Writer(self.width, self.height,
+                           palette=palettes[0].palette)
+
+            with open(self.path, "wb") as f:
+                w.write_array(f, self.image)
+                self.log(
+                    f"No unnamed palette for {self.name}; wrote image data to {self.path}")
+
+    def parse_image(self, data):
+        return data
+
+    def max_length(self):
+        if self.compressed:
+            return None
+        return self.width * self.height

tools/n64splat/segtypes/code.py

@@ -0,0 +1,780 @@
+from re import split
+from capstone import *
+from capstone.mips import *
+
+from collections import OrderedDict
+from segtypes.segment import N64Segment, parse_segment_name
+import os
+from pathlib import Path, PurePath
+from ranges import Range, RangeDict
+import re
+import sys
+from util import floats
+
+
+STRIP_C_COMMENTS_RE = re.compile(
+    r'//.*?$|/\*.*?\*/|\'(?:\\.|[^\\\'])*\'|"(?:\\.|[^\\"])*"',
+    re.DOTALL | re.MULTILINE
+)
+
+C_FUNC_RE = re.compile(
+    r"^(static\s+)?[^\s]+\s+([^\s(]+)\(([^;)]*)\)[^;]+?{",
+    re.MULTILINE
+)
+
+def strip_c_comments(text):
+    def replacer(match):
+        s = match.group(0)
+        if s.startswith("/"):
+            return " "
+        else:
+            return s
+    return re.sub(STRIP_C_COMMENTS_RE, replacer, text)
+
+
+def get_funcs_defined_in_c(c_file):
+    with open(c_file, "r") as f:
+        text = strip_c_comments(f.read())
+
+    return set(m.group(2) for m in C_FUNC_RE.finditer(text))
+
+
+def parse_segment_files(segment, segment_class, seg_start, seg_end, seg_name, seg_vram):
+    prefix = seg_name if seg_name.endswith("/") else f"{seg_name}_"
+
+    ret = []
+    prev_start = -1
+
+    if "files" in segment:
+        for i, split_file in enumerate(segment["files"]):
+            if type(split_file) is dict:
+                start = split_file["start"]
+                end = split_file["end"]
+                name = None if "name" not in split_file else split_file["name"]
+                subtype = split_file["type"]
+            else:
+                start = split_file[0]
+                end = seg_end if i == len(segment["files"]) - 1 else segment["files"][i + 1][0]
+                name = None if len(split_file) < 3 else split_file[2]
+                subtype = split_file[1]
+
+            if start < prev_start:
+                print(f"Error: Code segment {seg_name} has files out of ascending rom order (0x{prev_start:X} followed by 0x{start:X})")
+                sys.exit(1)
+
+            if not name:
+                name = N64SegCode.get_default_name(start) if seg_name == N64SegCode.get_default_name(seg_start) else f"{prefix}{start:X}"
+
+            vram = seg_vram + (start - seg_start)
+
+            fl = {"start": start, "end": end, "name": name, "vram": vram, "subtype": subtype}
+
+            ret.append(fl)
+            prev_start = start
+    else:
+        fl = {"start": seg_start, "end": seg_end,
+              "name": seg_name, "vram": seg_vram, "subtype": "asm"}
+        ret.append(fl)
+
+    return ret
+
+
+class N64SegCode(N64Segment):
+    def __init__(self, segment, next_segment, options):
+        super().__init__(segment, next_segment, options)
+        self.files = parse_segment_files(segment, self.__class__, self.rom_start, self.rom_end, self.name, self.vram_addr)
+        self.is_overlay = segment.get("overlay", False)
+        self.labels_to_add = set()
+        self.jtbl_glabels = set()
+        self.glabels_to_add = set()
+        self.special_labels = {}
+        self.undefined_syms_to_add = set()
+        self.glabels_added = {}
+        self.all_functions = {}
+        self.provided_symbols = {}
+        self.c_labels_to_add = set()
+        self.ld_section_name = "." + segment.get("ld_name", f"text_{self.rom_start:X}")
+        self.symbol_ranges = RangeDict()
+        self.detected_syms = {}
+        self.reported_file_split = False
+        self.jtbl_jumps = {}
+        self.jumptables = {}
+
+    @staticmethod
+    def get_default_name(addr):
+        return f"code_{addr:X}"
+
+    def get_func_name(self, addr):
+        return self.provided_symbols.get(addr, f"func_{addr:X}")
+
+    def get_unique_func_name(self, func_addr, rom_addr):
+        func_name = self.get_func_name(func_addr)
+
+        if self.is_overlay and (func_addr >= self.vram_addr) and (func_addr <= self.vram_addr + self.rom_end - self.rom_start):
+            return func_name + "_{:X}".format(rom_addr)
+        return func_name
+
+    def add_glabel(self, ram_addr, rom_addr):
+        func = self.get_unique_func_name(ram_addr, rom_addr)
+        self.glabels_to_add.discard(func)
+        self.glabels_added[ram_addr] = func
+        if not self.is_overlay:
+            self.all_functions[ram_addr] = func
+        return "glabel " + func
+
+    def get_asm_header(self):
+        ret = []
+
+        ret.append(".include \"macro.inc\"")
+        ret.append("")
+        ret.append("# assembler directives")
+        ret.append(".set noat      # allow manual use of $at")
+        ret.append(".set noreorder # don't insert nops after branches")
+        ret.append(".set gp=64     # allow use of 64-bit general purpose registers")
+        ret.append("")
+        ret.append(".section .text, \"ax\"")
+        ret.append("")
+
+        return ret
+
+    def get_gcc_inc_header(self):
+        ret = []
+        ret.append(".set noat      # allow manual use of $at")
+        ret.append(".set noreorder # don't insert nops after branches")
+        ret.append("")
+
+        return ret
+
+    @staticmethod
+    def is_nops(insns):
+        for insn in insns:
+            if insn.mnemonic != "nop":
+                return False
+        return True
+
+    @staticmethod
+    def is_branch_insn(mnemonic):
+        return (mnemonic.startswith("b") and not mnemonic.startswith("binsl") and not mnemonic == "break") or mnemonic == "j"
+
+    def process_insns(self, insns, rom_addr):
+        ret = OrderedDict()
+
+        func = []
+        end_func = False
+        labels = []
+
+        # Collect labels
+        for insn in insns:
+            if self.is_branch_insn(insn.mnemonic):
+                op_str_split = insn.op_str.split(" ")
+                branch_target = op_str_split[-1]
+                branch_addr = int(branch_target, 0)
+                labels.append((insn.address, branch_addr))
+
+        # Main loop
+        for i, insn in enumerate(insns):
+            mnemonic = insn.mnemonic
+            op_str = insn.op_str
+            func_addr = insn.address if len(func) == 0 else func[0][0].address
+
+            if mnemonic == "move":
+                # Let's get the actual instruction out
+                opcode = insn.bytes[3] & 0b00111111
+                op_str += ", $zero"
+
+                if opcode == 37:
+                    mnemonic = "or"
+                elif opcode == 45:
+                    mnemonic = "daddu"
+                elif opcode == 33:
+                    mnemonic = "addu"
+                else:
+                    print("INVALID INSTRUCTION " + insn)
+            elif mnemonic == "jal":
+                jal_addr = int(op_str, 0)
+                jump_func = self.get_func_name(jal_addr)
+                if (
+                    jump_func.startswith("func_")
+                    and self.is_overlay
+                    and jal_addr >= self.vram_addr
+                    and jal_addr <= (self.vram_addr + self.rom_end - self.rom_start)
+                ):
+                    func_loc = self.rom_start + jal_addr - self.vram_addr
+                    jump_func += "_{:X}".format(func_loc)
+
+                if jump_func not in self.provided_symbols.values():
+                    self.glabels_to_add.add(jump_func)
+                op_str = jump_func
+            elif self.is_branch_insn(insn.mnemonic):
+                op_str_split = op_str.split(" ")
+                branch_target = op_str_split[-1]
+                branch_target_int = int(branch_target, 0)
+                label = ""
+
+                if branch_target_int in self.special_labels:
+                    label = self.special_labels[branch_target_int]
+                else:
+                    self.labels_to_add.add(branch_target_int)
+                    label = ".L" + branch_target[2:].upper()
+
+                op_str = " ".join(op_str_split[:-1] + [label])
+            elif mnemonic == "mtc0" or mnemonic == "mfc0":
+                rd = (insn.bytes[2] & 0xF8) >> 3
+                op_str = op_str.split(" ")[0] + " $" + str(rd)
+
+            func.append((insn, mnemonic, op_str, rom_addr))
+            rom_addr += 4
+
+            if mnemonic == "jr":
+                # Record potential jtbl jumps
+                if op_str != "$ra":
+                    self.jtbl_jumps[insn.address] = op_str
+
+                keep_going = False
+                for label in labels:
+                    if (label[0] > insn.address and label[1] <= insn.address) or (label[0] <= insn.address and label[1] > insn.address):
+                        keep_going = True
+                        break
+                if not keep_going:
+                    end_func = True
+                    continue
+
+            if i < len(insns) - 1 and self.get_func_name(insns[i + 1].address) in self.c_labels_to_add:
+                end_func = True
+
+            if end_func:
+                if self.is_nops(insns[i:]) or i < len(insns) - 1 and insns[i + 1].mnemonic != "nop":
+                    end_func = False
+                    ret[func_addr] = func
+                    func = []
+
+        # Add the last function (or append nops to the previous one)
+        if not self.is_nops([i[0] for i in func]):
+            ret[func_addr] = func
+        else:
+            next(reversed(ret.values())).extend(func)
+
+        return ret
+
+    def get_file_for_addr(self, addr):
+        for fl in self.files:
+            if addr >= fl["vram"] and addr < fl["vram"] + fl["end"] - fl["start"]:
+                return fl
+        return None
+
+    def store_symbol_access(self, addr, mnemonic):
+        # Don't overwrite useful info with addiu
+        if addr in self.detected_syms and self.detected_syms[addr] != "addiu":
+            return
+
+        self.detected_syms[addr] = mnemonic
+
+    def get_symbol_name(self, addr, rom_addr, funcs=None):
+        if funcs and addr in funcs:
+            return self.get_unique_func_name(addr, rom_addr)
+        if addr in self.all_functions:
+            return self.all_functions[addr] # todo clean up funcs vs all_functions
+        if addr in self.provided_symbols:
+            return self.provided_symbols[addr]
+        if addr in self.jumptables:
+            return f"jtbl_{addr:X}_{rom_addr:X}"
+        if addr in self.symbol_ranges:
+            ret = self.symbol_ranges.get(addr)
+            offset = addr - self.symbol_ranges.getrange(addr).start
+            if offset != 0:
+                ret += f"+0x{offset:X}"
+            return ret
+
+        return f"D_{addr:X}"
+
+    # Determine symbols
+    def determine_symbols(self, funcs, rom_addr):
+        ret = {}
+
+        for func_addr in funcs:
+            func = funcs[func_addr]
+            func_end_addr = func[-1][0].address + 4
+
+            possible_jtbl_jumps = [(k, v) for k, v in self.jtbl_jumps.items() if k >= func_addr and k < func_end_addr]
+            possible_jtbl_jumps.sort(key=lambda x:x[0])
+
+            for i in range(len(func)):
+                insn = func[i][0]
+
+                # Ensure the first item in the list is always ahead of where we're looking
+                while len(possible_jtbl_jumps) > 0 and possible_jtbl_jumps[0][0] < insn.address:
+                    del possible_jtbl_jumps[0]
+
+                if insn.mnemonic == "lui":
+                    op_split = insn.op_str.split(", ")
+                    reg = op_split[0]
+
+                    if not op_split[1].startswith("0x"):
+                        continue
+
+                    lui_val = int(op_split[1], 0)
+                    if lui_val >= 0x8000:
+                        for j in range(i + 1, min(i + 6, len(func))):
+                            s_insn = func[j][0]
+
+                            s_op_split = s_insn.op_str.split(", ")
+
+                            if s_insn.mnemonic == "lui" and reg == s_op_split[0]:
+                                break
+
+                            if s_insn.mnemonic in ["addiu", "ori"]:
+                                s_reg = s_op_split[-2]
+                            else:
+                                s_reg = s_op_split[-1][s_op_split[-1].rfind("(") + 1: -1]
+
+                            if reg == s_reg:
+                                if s_insn.mnemonic not in ["addiu", "lw", "sw", "lh", "sh", "lhu", "lb", "sb", "lbu", "lwc1", "swc1", "ldc1", "sdc1"]:
+                                    break
+
+                                # Match!
+                                reg_ext = ""
+
+                                junk_search = re.search(
+                                    r"[\(]", s_op_split[-1])
+                                if junk_search is not None:
+                                    if junk_search.start() == 0:
+                                        break
+                                    s_str = s_op_split[-1][:junk_search.start()]
+                                    reg_ext = s_op_split[-1][junk_search.start():]
+                                else:
+                                    s_str = s_op_split[-1]
+
+                                symbol_addr = (lui_val * 0x10000) + int(s_str, 0)
+                                symbol_name = self.get_symbol_name(symbol_addr, symbol_addr - next(iter(funcs)) + rom_addr, funcs)
+                                symbol_tag = s_insn.mnemonic
+
+                                vram_end = self.vram_addr + self.rom_end - self.rom_start
+                                if symbol_addr > func_addr and symbol_addr < vram_end and len(possible_jtbl_jumps) > 0 and func_end_addr - s_insn.address >= 0x30:
+                                    for jump in possible_jtbl_jumps:
+                                        if jump[1] == s_op_split[0]:
+                                            dist_to_jump = possible_jtbl_jumps[0][0] - s_insn.address
+                                            if dist_to_jump <= 16:
+                                                symbol_name = f"jtbl_{symbol_addr:X}_{self.ram_to_rom(symbol_addr):X}"
+                                                symbol_tag = "jtbl"
+                                                self.jumptables[symbol_addr] = (func_addr, func_end_addr)
+                                                break
+
+                                self.store_symbol_access(symbol_addr, symbol_tag)
+                                symbol_file = self.get_file_for_addr(symbol_addr)
+
+                                if not symbol_file or symbol_file["subtype"] == "bin":
+                                    if "+" not in symbol_name:
+                                        self.undefined_syms_to_add.add((symbol_name, symbol_addr))
+
+                                func[i] += ("%hi({})".format(symbol_name),)
+                                func[j] += ("%lo({}){}".format(symbol_name, reg_ext),)
+                                break
+            ret[func_addr] = func
+        return ret
+
+    def add_labels(self, funcs):
+        ret = {}
+
+        for func in funcs:
+            func_text = []
+
+            # Add function glabel
+            rom_addr = funcs[func][0][3]
+            func_text.append(self.add_glabel(func, rom_addr))
+
+            indent_next = False
+
+            mnemonic_ljust = self.options.get("mnemonic_ljust", 11)
+            rom_addr_padding = self.options.get("rom_address_padding", None)
+
+            for insn in funcs[func]:
+                insn_addr = insn[0].address
+                # Add a label if we need one
+                if insn_addr in self.labels_to_add:
+                    self.labels_to_add.remove(insn_addr)
+                    func_text.append(".L{:X}:".format(insn_addr))
+                if insn_addr in self.jtbl_glabels:
+                    func_text.append(f"glabel L{insn_addr:X}_{insn[3]:X}")
+
+                if rom_addr_padding:
+                    rom_str = "{0:0{1}X}".format(insn[3], rom_addr_padding)
+                else:
+                    rom_str = "{:X}".format(insn[3])
+
+                asm_comment = "/* {} {:X} {} */".format(rom_str, insn_addr, insn[0].bytes.hex().upper())
+
+                if len(insn) > 4:
+                    op_str = ", ".join(insn[2].split(", ")[:-1] + [insn[4]])
+                else:
+                    op_str = insn[2]
+
+                insn_text = insn[1]
+                if indent_next:
+                    indent_next = False
+                    insn_text = " " + insn_text
+
+                asm_insn_text = "  {}{}".format(insn_text.ljust(mnemonic_ljust), op_str).rstrip()
+
+                func_text.append(asm_comment + asm_insn_text)
+
+                if insn[0].mnemonic != "branch" and insn[0].mnemonic.startswith("b") or insn[0].mnemonic.startswith("j"):
+                    indent_next = True
+
+            ret[func] = (func_text, rom_addr)
+
+            if self.options.get("find_file_boundaries"):
+                # If this is not the last function in the file
+                if func != list(funcs.keys())[-1]:
+
+                    # Find where the function returns
+                    jr_pos = None
+                    for i, insn in enumerate(reversed(funcs[func])):
+                        if insn[0].mnemonic == "jr" and insn[0].op_str == "$ra":
+                            jr_pos = i
+                            break
+
+                    # If there is more than 1 nop after the return
+                    if jr_pos and jr_pos > 1 and self.is_nops([i[0] for i in funcs[func][-jr_pos + 1:]]):
+                        new_file_addr = funcs[func][-1][3] + 4
+                        if (new_file_addr % 16) == 0:
+                            if not self.reported_file_split:
+                                self.reported_file_split = True
+                                print(f"Segment {self.name}, function at vram {func:X} ends with extra nops, indicating a likely file split.")
+                                print("File split suggestions for this segment will follow in config yaml format:")
+                            print(f"      - [0x{new_file_addr:X}, asm]")
+
+        return ret
+
+    def should_run(self):
+        possible_subtypes = ["c", "asm", "hasm", "bin", "data", "rodata"]
+        subtypes = set(f["subtype"] for f in self.files)
+
+        return super().should_run() or (st in self.options["modes"] and st in subtypes for st in possible_subtypes)
+
+    def is_valid_ascii(self, bytes):
+        if len(bytes) < 8:
+            return False
+
+        num_empty_bytes = 0
+        for b in bytes:
+            if b == 0:
+                num_empty_bytes += 1
+
+        empty_ratio = num_empty_bytes / len(bytes)
+        if empty_ratio > 0.2:
+            return False
+
+        return True
+
+    def get_symbols_for_file(self, split_file):
+        vram_start = split_file["vram"]
+        vram_end = split_file["vram"] + split_file["end"] - split_file["start"]
+
+        return [(s, self.detected_syms[s]) for s in self.detected_syms if s >= vram_start and s <= vram_end]
+
+    def disassemble_symbol(self, sym_bytes, sym_type):
+        if sym_type == "jtbl":
+            sym_str = ".word "
+        else:
+            sym_str = f".{sym_type} "
+
+        if sym_type == "double":
+            slen = 8
+        elif sym_type in ["float", "word", "jtbl"]:
+            slen = 4
+        elif sym_type == "short":
+            slen = 2
+        else:
+            slen = 1
+
+        i = 0
+        while i < len(sym_bytes):
+            adv_amt = min(slen, len(sym_bytes) - i)
+            bits = int.from_bytes(sym_bytes[i : i + adv_amt], "big")
+
+            if sym_type == "jtbl":
+                if bits == 0:
+                    byte_str = "0"
+                else:
+                    rom_addr = self.ram_to_rom(bits)
+
+                    if rom_addr:
+                        byte_str = f"L{bits:X}_{rom_addr:X}"
+                    else:
+                        byte_str = f"0x{bits:X}"
+            else:
+                byte_str = self.provided_symbols.get(bits, '0x{0:0{1}X}'.format(bits, 2 * slen))
+
+            if sym_type in ["float", "double"]:
+                if sym_type == "float":
+                    float_str = floats.format_f32_imm(bits)
+                elif sym_type == "double":
+                    float_str = floats.format_f64_imm(bits)
+
+                # Fall back to .word if we see weird float values
+                # todo cut the symbol in half maybe where we see the first nan or something
+                if "e-" in float_str or "nan" in float_str:
+                    return self.disassemble_symbol(sym_bytes, "word")
+                else:
+                    byte_str = float_str
+
+            sym_str += byte_str
+
+            i += adv_amt
+
+            if i < len(sym_bytes):
+                sym_str += ", "
+
+        return sym_str
+
+    def disassemble_data(self, split_file, rom_bytes):
+        rodata_encountered = split_file["subtype"] == "rodata"
+        ret = ".include \"macro.inc\"\n\n"
+        ret += f'.section .{split_file["subtype"]}'
+
+        syms = self.get_symbols_for_file(split_file)
+        syms.sort(key=lambda x:x[0])
+
+        if len(syms) == 0:
+            self.warn("No symbol accesses detected for " + split_file["name"] + "; the output will most likely be an ugly blob")
+
+        # check beginning
+        if syms[0][0] != split_file["vram"]:
+            syms.insert(0, (split_file["vram"], None))
+
+        # add end
+        vram_end = split_file["vram"] + split_file["end"] - split_file["start"]
+        if syms[-1][0] != vram_end:
+            syms.append((vram_end, None))
+
+        for i in range(len(syms) - 1):
+            mnemonic = syms[i][1]
+            start = syms[i][0]
+            end = syms[i + 1][0]
+            sym_rom_start = start - split_file["vram"] + split_file["start"]
+            sym_rom_end = end - split_file["vram"] + split_file["start"]
+            sym_name = self.get_symbol_name(start, sym_rom_start)
+            sym_str = f"\n\nglabel {sym_name}\n"
+            sym_bytes = rom_bytes[sym_rom_start : sym_rom_end]
+
+            # .ascii
+            if self.is_valid_ascii(sym_bytes) and mnemonic == "addiu":
+            # mnemonic thing may be too picky, we'll see
+                try:
+                    ascii_str = sym_bytes.decode("EUC-JP")
+                    ascii_str = ascii_str.replace("\\", "\\\\")
+                    ascii_str = ascii_str.replace("\x00", "\\0")
+                    sym_str += f'.ascii "{ascii_str}"'
+                    ret += sym_str
+                    continue
+                except:
+                    pass
+
+            # Fallback to raw data
+            if mnemonic == "jtbl":
+                stype = "jtbl"
+            elif len(sym_bytes) % 8 == 0 and mnemonic in ["ldc1", "sdc1"]:
+                stype = "double"
+            elif len(sym_bytes) % 4 == 0 and mnemonic in ["addiu", "sw", "lw", "jtbl"]:
+                stype = "word"
+            elif len(sym_bytes) % 4 == 0 and mnemonic in ["lwc1", "swc1"]:
+                stype = "float"
+            elif len(sym_bytes) % 2 == 0 and mnemonic in ["addiu", "lh", "sh", "lhu"]:
+                stype = "short"
+            else:
+                stype = "byte"
+            
+            if not rodata_encountered and mnemonic == "jtbl":
+                rodata_encountered = True
+                ret += "\n\n\n.section .rodata"
+
+            sym_str += self.disassemble_symbol(sym_bytes, stype)
+            ret += sym_str
+
+        ret += "\n"
+
+        return ret
+
+    def get_c_preamble(self):
+        ret = []
+
+        preamble = self.options.get("generated_c_preamble", "#include \"common.h\"")
+        ret.append(preamble)
+        ret.append("")
+
+        return ret
+
+    def gather_jumptable_labels(self, section_vram, section_rom, rom_bytes):
+        for jumptable in self.jumptables:
+            start, end = self.jumptables[jumptable]
+            rom_offset = section_rom + jumptable - section_vram
+
+            if rom_offset <= 0:
+                return
+
+            while (rom_offset):
+                word = rom_bytes[rom_offset : rom_offset + 4]
+                word_int = int.from_bytes(word, "big")
+                if word_int >= start and word_int <= end:
+                    self.jtbl_glabels.add(word_int)
+                else:
+                    break
+
+                rom_offset += 4
+
+
+    def split(self, rom_bytes, base_path):
+        md = Cs(CS_ARCH_MIPS, CS_MODE_MIPS64 + CS_MODE_BIG_ENDIAN)
+        md.detail = True
+        md.skipdata = True
+
+        for split_file in self.files:
+            file_type = split_file["subtype"]
+
+            if file_type in ["asm", "hasm", "c"]:
+                if self.type not in self.options["modes"] and "all" not in self.options["modes"]:
+                    continue
+
+                if split_file["start"] == split_file["end"]:
+                    continue
+
+                out_dir = self.create_split_dir(base_path, "asm")
+
+                rom_addr = split_file["start"]
+
+                insns = [insn for insn in md.disasm(rom_bytes[split_file["start"]: split_file["end"]], split_file["vram"])]
+
+                funcs = self.process_insns(insns, rom_addr)
+                funcs = self.determine_symbols(funcs, rom_addr)
+                self.gather_jumptable_labels(self.vram_addr, self.rom_start, rom_bytes)
+                funcs_text = self.add_labels(funcs)
+
+                if file_type == "c":
+                    c_path = os.path.join(
+                        base_path, "src", split_file["name"] + "." + self.get_ext(split_file["subtype"]))
+
+                    if os.path.exists(c_path):
+                        defined_funcs = get_funcs_defined_in_c(c_path)
+                    else:
+                        defined_funcs = set()
+
+                    out_dir = self.create_split_dir(
+                        base_path, os.path.join("asm", "nonmatchings"))
+
+                    for func in funcs_text:
+                        func_name = self.get_unique_func_name(
+                            func, funcs_text[func][1])
+
+                        if func_name not in defined_funcs:
+                            if self.options.get("compiler", "IDO") == "GCC":
+                                out_lines = self.get_gcc_inc_header()
+                            else:
+                                out_lines = []
+                            out_lines.extend(funcs_text[func][0])
+                            out_lines.append("")
+
+                            outpath = Path(os.path.join(
+                                out_dir, split_file["name"], func_name + ".s"))
+                            outpath.parent.mkdir(parents=True, exist_ok=True)
+
+                            with open(outpath, "w", newline="\n") as f:
+                                f.write("\n".join(out_lines))
+                            self.log(f"Disassembled {func_name} to {outpath}")
+
+                    # Creation of c files
+                    if not os.path.exists(c_path):  # and some option is enabled
+                        c_lines = self.get_c_preamble()
+
+                        for func in funcs_text:
+                            func_name = self.get_unique_func_name(func, funcs_text[func][1])
+                            if self.options.get("compiler", "IDO") == "GCC":
+                                c_lines.append("INCLUDE_ASM(s32, \"{}\", {});".format(split_file["name"], func_name))
+                            else:
+                                outpath = Path(os.path.join(out_dir, split_file["name"], func_name + ".s"))
+                                rel_outpath = os.path.relpath(outpath, base_path)
+                                c_lines.append(f"#pragma GLOBAL_ASM(\"{rel_outpath}\")")
+                            c_lines.append("")
+
+                        Path(c_path).parent.mkdir(parents=True, exist_ok=True)
+                        with open(c_path, "w") as f:
+                            f.write("\n".join(c_lines))
+                        print(f"Wrote {split_file['name']} to {c_path}")
+
+                else:
+                    out_lines = self.get_asm_header()
+                    for func in funcs_text:
+                        out_lines.extend(funcs_text[func][0])
+                        out_lines.append("")
+
+                    outpath = Path(os.path.join(out_dir, split_file["name"] + ".s"))
+                    outpath.parent.mkdir(parents=True, exist_ok=True)
+
+                    with open(outpath, "w", newline="\n") as f:
+                        f.write("\n".join(out_lines))
+
+            elif file_type in ["data", "rodata"] and (file_type in self.options["modes"] or "all" in self.options["modes"]):
+                out_dir = self.create_split_dir(base_path, os.path.join("asm", "data"))
+
+                outpath = Path(os.path.join(out_dir, split_file["name"] + f".{file_type}.s"))
+                outpath.parent.mkdir(parents=True, exist_ok=True)
+
+                file_text = self.disassemble_data(split_file, rom_bytes)
+                if file_text:
+                    with open(outpath, "w", newline="\n") as f:
+                        f.write(file_text)
+
+            elif file_type == "bin" and ("bin" in self.options["modes"] or "all" in self.options["modes"]):
+                out_dir = self.create_split_dir(base_path, "bin")
+
+                bin_path = os.path.join(
+                    out_dir, split_file["name"] + "." + self.get_ext(split_file["subtype"]))
+                Path(bin_path).parent.mkdir(parents=True, exist_ok=True)
+                with open(bin_path, "wb") as f:
+                    f.write(rom_bytes[split_file["start"]: split_file["end"]])
+
+    @staticmethod
+    def get_subdir(subtype):
+        if subtype in ["c", ".data", ".rodata", ".bss"]:
+            return "src"
+        elif subtype in ["asm", "hasm", "header"]:
+            return "asm"
+        return subtype
+
+    @staticmethod
+    def get_ext(subtype):
+        if subtype in ["c", ".data", ".rodata", ".bss"]:
+            return "c"
+        elif subtype in ["asm", "hasm", "header"]:
+            return "s"
+        elif subtype == "bin":
+            return "bin"
+        return subtype
+
+    @staticmethod
+    def get_ld_obj_type(subtype, section_name):
+        if subtype in "c":
+            return ".text"
+        elif subtype in ["bin", ".data", "data"]:
+            return ".data"
+        elif subtype in [".rodata", "rodata"]:
+            return ".rodata"
+        elif subtype == ".bss":
+            return ".bss"
+        return section_name
+
+    def get_ld_files(self):
+        def transform(split_file):
+            subdir = self.get_subdir(split_file["subtype"])
+            obj_type = self.get_ld_obj_type(split_file["subtype"], ".text")
+            ext = self.get_ext(split_file['subtype'])
+            start = split_file["start"]
+
+            return subdir, f"{split_file['name']}.{ext}", obj_type, start
+
+        return [transform(file) for file in self.files]
+
+    def get_ld_section_name(self):
+        path = PurePath(self.name)
+        name = path.name if path.name != "" else path.parent
+
+        return f"code_{name}"

tools/n64splat/segtypes/header.py

@@ -0,0 +1,61 @@
+import os
+from segtypes.segment import N64Segment
+from pathlib import Path
+from util import rominfo
+
+class N64SegHeader(N64Segment):
+    def should_run(self):
+        return N64Segment.should_run(self) or "asm" in self.options["modes"]
+
+    @staticmethod
+    def get_line(typ, data, comment):
+        if typ == "ascii":
+            dstr = "\"" + data.decode("ASCII").strip() + "\""
+        else: # .word, .byte
+            dstr = "0x" + data.hex().upper()
+        
+        dstr = dstr.ljust(20 - len(typ))
+        
+        return f".{typ} {dstr} /* {comment} */"
+
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_split_dir(base_path, "asm")
+
+        encoding = self.options.get("header_encoding", "ASCII")
+
+        header_lines = []
+        header_lines.append(f".section .{self.name}, \"a\"\n")
+        header_lines.append(self.get_line("word", rom_bytes[0x00:0x04], "PI BSB Domain 1 register"))
+        header_lines.append(self.get_line("word", rom_bytes[0x04:0x08], "Clockrate setting"))
+        header_lines.append(self.get_line("word", rom_bytes[0x08:0x0C], "Entrypoint address"))
+        header_lines.append(self.get_line("word", rom_bytes[0x0C:0x10], "Revision"))
+        header_lines.append(self.get_line("word", rom_bytes[0x10:0x14], "Checksum 1"))
+        header_lines.append(self.get_line("word", rom_bytes[0x14:0x18], "Checksum 2"))
+        header_lines.append(self.get_line("word", rom_bytes[0x18:0x1C], "Unknown 1"))
+        header_lines.append(self.get_line("word", rom_bytes[0x1C:0x20], "Unknown 2"))
+        header_lines.append(".ascii \"" + rom_bytes[0x20:0x34].decode(encoding).strip().ljust(20) + "\" /* Internal name */")
+        header_lines.append(self.get_line("word", rom_bytes[0x34:0x38], "Unknown 3"))
+        header_lines.append(self.get_line("word", rom_bytes[0x38:0x3C], "Cartridge"))
+        header_lines.append(self.get_line("ascii", rom_bytes[0x3C:0x3E], "Cartridge ID"))
+        header_lines.append(self.get_line("ascii", rom_bytes[0x3E:0x3F], "Country code"))
+        header_lines.append(self.get_line("byte", rom_bytes[0x3F:0x40], "Version"))
+        header_lines.append("")
+
+        s_path = os.path.join(out_dir, self.name + ".s")
+        Path(s_path).parent.mkdir(parents=True, exist_ok=True)
+        with open(s_path, "w", newline="\n") as f:
+            f.write("\n".join(header_lines))
+        self.log(f"Wrote {self.name} to {s_path}")
+
+
+    def get_ld_section_name(self):
+        return self.name
+
+
+    def get_ld_files(self):
+        return [("asm", f"{self.name}.s", ".data", self.rom_start)]
+
+
+    @staticmethod
+    def get_default_name(addr):
+        return "header"

tools/n64splat/segtypes/i4.py

@@ -0,0 +1,27 @@
+from segtypes.rgba16 import N64SegRgba16
+import png
+from math import ceil
+
+class N64SegI4(N64SegRgba16):
+    def png_writer(self):
+        return png.Writer(self.width, self.height, greyscale = True)
+
+    def parse_image(self, data):
+        img = bytearray()
+
+        for x, y, i in self.iter_image_indexes(0.5, 1):
+            b = data[i]
+
+            i1 = (b >> 4) & 0xF
+            i2 = b & 0xF
+
+            i1 = ceil(0xFF * (i1 / 15))
+            i2 = ceil(0xFF * (i2 / 15))
+
+            img += bytes((i1, i2))
+
+        return img
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height // 2

tools/n64splat/segtypes/i8.py

@@ -0,0 +1,10 @@
+from segtypes.i4 import N64SegI4
+from math import ceil
+
+class N64SegI8(N64SegI4):
+    def parse_image(self, data):
+        return data
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height

tools/n64splat/segtypes/ia16.py

@@ -0,0 +1,9 @@
+from segtypes.ia4 import N64SegIa4
+
+class N64SegIa8(N64SegIa4):
+    def parse_image(self, data):
+        return data
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height * 2

tools/n64splat/segtypes/ia4.py

@@ -0,0 +1,33 @@
+import os
+from segtypes.rgba16 import N64SegRgba16
+import png
+from math import ceil
+
+class N64SegIa4(N64SegRgba16):
+    def png_writer(self):
+        return png.Writer(self.width, self.height, greyscale = True, alpha = True)
+
+    def parse_image(self, data):
+        img = bytearray()
+
+        for x, y, i in self.iter_image_indexes(0.5, 1):
+            b = data[i]
+
+            h = (b >> 4) & 0xF
+            l = b & 0xF
+
+            i1 = (h >> 1) & 0xF
+            a1 = (h & 1) * 0xFF
+            i1 = ceil(0xFF * (i1 / 7))
+
+            i2 = (l >> 1) & 0xF
+            a2 = (l & 1) * 0xFF
+            i2 = ceil(0xFF * (i2 / 7))
+
+            img += bytes((i1, a1, i2, a2))
+
+        return img
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height // 2

tools/n64splat/segtypes/ia8.py

@@ -0,0 +1,24 @@
+from segtypes.ia4 import N64SegIa4
+import png
+from math import ceil
+
+class N64SegIa8(N64SegIa4):
+    def parse_image(self, data):
+        img = bytearray()
+
+        for x, y, i in self.iter_image_indexes():
+            b = data[i]
+
+            i = (b >> 4) & 0xF
+            a = b & 0xF
+
+            i = ceil(0xFF * (i / 15))
+            a = ceil(0xFF * (a / 15))
+
+            img += bytes((i, a))
+
+        return img
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height

tools/n64splat/segtypes/palette.py

@@ -0,0 +1,64 @@
+import os
+from segtypes.segment import N64Segment
+from util.color import unpack_color
+from util.iter import iter_in_groups
+
+
+class N64SegPalette(N64Segment):
+    require_unique_name = False
+
+    def __init__(self, segment, next_segment, options):
+        super().__init__(segment, next_segment, options)
+
+        # palette segments must be named as one of the following:
+        #  1) same as the relevant ci4/ci8 segment name (max. 1 palette)
+        #  2) relevant ci4/ci8 segment name + "." + unique palette name
+        #  3) unique, referencing the relevant ci4/ci8 segment using `image_name`
+        self.image_name = segment.get("image_name", self.name.split(
+            ".")[0]) if type(segment) is dict else self.name.split(".")[0]
+
+        self.compressed = segment.get("compressed", False) if type(
+            segment) is dict else False
+
+    def should_run(self):
+        return super().should_run() or (
+            "img" in self.options["modes"] or
+            "ci4" in self.options["modes"] or
+            "ci8" in self.options["modes"] or
+            "i4" in self.options["modes"] or
+            "i8" in self.options["modes"] or
+            "ia4" in self.options["modes"] or
+            "ia8" in self.options["modes"] or
+            "ia16" in self.options["modes"]
+        )
+
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_parent_dir(base_path + "/img", self.name)
+        self.path = os.path.join(
+            out_dir, os.path.basename(self.name) + ".png")
+
+        data = rom_bytes[self.rom_start: self.rom_end]
+        if self.compressed:
+            data = Yay0decompress.decompress_yay0(data)
+
+        self.palette = self.parse_palette(data)
+
+    def parse_palette(self, data):
+        palette = []
+
+        for a, b in iter_in_groups(data, 2):
+            palette.append(unpack_color([a, b]))
+
+        return palette
+
+    def max_length(self):
+        if self.compressed:
+            return None
+        return 256 * 2
+
+    def get_ld_files(self):
+        ext = f".{self.type}.png"
+        if self.compressed:
+            ext += ".Yay0"
+
+        return [("img", f"{self.name}{ext}", ".data", self.rom_start)]

tools/n64splat/segtypes/rgba16.py

@@ -0,0 +1,86 @@
+import os
+from segtypes.segment import N64Segment
+from pathlib import Path
+from util import Yay0decompress
+import png
+from math import ceil
+from util.color import unpack_color
+
+
+class N64SegRgba16(N64Segment):
+    def __init__(self, segment, next_segment, options):
+        super().__init__(segment, next_segment, options)
+
+        if type(segment) is dict:
+            self.compressed = segment.get("compressed", False)
+            self.width = segment["width"]
+            self.height = segment["height"]
+            self.flip = segment.get("flip", "noflip")
+        elif len(segment) < 5:
+            self.error("missing parameters")
+        else:
+            self.compressed = False
+            self.width = segment[3]
+            self.height = segment[4]
+            self.flip = "noflip"
+
+    @property
+    def flip_vertical(self):
+        return self.flip == "both" or self.flip.startswith("v") or self.flip == "y"
+
+    @property
+    def flip_horizontal(self):
+        return self.flip == "both" or self.flip.startswith("h") or self.flip == "x"
+
+    def iter_image_indexes(self, bytes_per_x=1, bytes_per_y=1):
+        w = int(self.width * bytes_per_x)
+        h = int(self.height * bytes_per_y)
+
+        xrange = range(w - ceil(bytes_per_x), -1, -ceil(bytes_per_x)
+                       ) if self.flip_horizontal else range(0, w, ceil(bytes_per_x))
+        yrange = range(h - ceil(bytes_per_y), -1, -ceil(bytes_per_y)
+                       ) if self.flip_vertical else range(0, h, ceil(bytes_per_y))
+
+        for y in yrange:
+            for x in xrange:
+                yield x, y, (y * w) + x
+
+    def should_run(self):
+        return super().should_run() or "img" in self.options["modes"]
+
+    def split(self, rom_bytes, base_path):
+        out_dir = self.create_parent_dir(base_path + "/img", self.name)
+        path = os.path.join(out_dir, os.path.basename(self.name) + ".png")
+
+        data = rom_bytes[self.rom_start: self.rom_end]
+        if self.compressed:
+            data = Yay0decompress.decompress_yay0(data)
+
+        w = self.png_writer()
+        with open(path, "wb") as f:
+            w.write_array(f, self.parse_image(data))
+
+        self.log(f"Wrote {self.name} to {path}")
+
+    def png_writer(self):
+        return png.Writer(self.width, self.height, greyscale=False, alpha=True)
+
+    def parse_image(self, data):
+        img = bytearray()
+
+        for x, y, i in self.iter_image_indexes(2, 1):
+            img += bytes(unpack_color(data[i:]))
+
+        return img
+
+    def max_length(self):
+        if self.compressed:
+            return None
+        return self.width * self.height * 2
+
+    def get_ld_files(self):
+        ext = f".{self.type}.png"
+        if self.compressed:
+            ext += ".Yay0"
+
+        return [("img", f"{self.name}{ext}", ".data", self.rom_start)]

tools/n64splat/segtypes/rgba32.py

@@ -0,0 +1,9 @@
+from segtypes.rgba16 import N64SegRgba16
+
+class N64SegRgba32(N64SegRgba16):
+    def parse_image(self, data):
+        return data
+
+    def max_length(self):
+        if self.compressed: return None
+        return self.width * self.height * 4

tools/n64splat/segtypes/segment.py

@@ -0,0 +1,188 @@
+import os
+from pathlib import Path, PurePath
+import re
+import json
+from util import log
+
+default_subalign = 16
+
+
+def parse_segment_start(segment):
+    return segment[0] if "start" not in segment else segment["start"]
+
+
+def parse_segment_type(segment):
+    if type(segment) is dict:
+        return segment["type"]
+    else:
+        return segment[1]
+
+
+def parse_segment_name(segment, segment_class):
+    if type(segment) is dict and "name" in segment:
+        return segment["name"]
+    elif type(segment) is list and len(segment) >= 3 and type(segment[2]) is str:
+        return segment[2]
+    else:
+        return segment_class.get_default_name(parse_segment_start(segment))
+
+
+def parse_segment_vram(segment):
+    if type(segment) is dict:
+        return segment.get("vram", 0)
+    else:
+        if len(segment) >= 3 and type(segment[-1]) is int:
+            return segment[-1]
+        else:
+            return 0
+
+
+def parse_segment_subalign(segment):
+    if type(segment) is dict:
+        return segment.get("subalign", default_subalign)
+    return default_subalign
+
+
+class N64Segment:
+    require_unique_name = True
+
+    def __init__(self, segment, next_segment, options):
+        self.rom_start = parse_segment_start(segment)
+        self.rom_end = parse_segment_start(next_segment)
+        self.type = parse_segment_type(segment)
+        self.name = parse_segment_name(segment, self.__class__)
+        self.vram_addr = parse_segment_vram(segment)
+        self.ld_name_override = segment.get(
+            "ld_name", None) if type(segment) is dict else None
+        self.options = options
+        self.config = segment
+        self.subalign = parse_segment_subalign(segment)
+
+        self.errors = []
+        self.warnings = []
+        self.did_run = False
+
+    def check(self):
+        if self.rom_start > self.rom_end:
+            self.warn(f"out-of-order (starts at 0x{self.rom_start:X}, but next segment starts at 0x{self.rom_end:X})")
+        elif self.max_length():
+            expected_len = int(self.max_length())
+            actual_len = self.rom_end - self.rom_start
+            if actual_len > expected_len:
+                print(f"should end at 0x{self.rom_start + expected_len:X}, but it ends at 0x{self.rom_end:X}\n(hint: add a 'bin' segment after {self.name})")
+
+    @property
+    def size(self):
+        return self.rom_end - self.rom_start
+
+    @property
+    def vram_end(self):
+        return self.vram_addr + self.size
+
+    def rom_to_ram(self, rom_addr):
+        if rom_addr < self.rom_start or rom_addr > self.rom_end:
+            return None
+
+        return self.vram_addr + rom_addr - self.rom_start
+
+    def ram_to_rom(self, ram_addr):
+        if ram_addr < self.vram_addr or ram_addr > self.vram_end:
+            return None
+
+        return self.rom_start + ram_addr - self.vram_addr
+
+    def create_split_dir(self, base_path, subdir):
+        out_dir = Path(base_path, subdir)
+        out_dir.mkdir(parents=True, exist_ok=True)
+        return out_dir
+
+    def create_parent_dir(self, base_path, filename):
+        out_dir = Path(base_path, filename).parent
+        out_dir.mkdir(parents=True, exist_ok=True)
+        return out_dir
+
+    def should_run(self):
+        return self.type in self.options["modes"] or "all" in self.options["modes"]
+
+    def split(self, rom_bytes, base_path):
+        pass
+
+    def postsplit(self, segments):
+        pass
+
+    def cache(self):
+        return (self.config, self.rom_end)
+
+    def get_ld_section(self):
+        replace_ext = self.options.get("ld_o_replace_extension", True)
+        sect_name = self.ld_name_override if self.ld_name_override else self.get_ld_section_name()
+        vram_or_rom = self.rom_start if self.vram_addr == 0 else self.vram_addr
+        subalign_str = "" if self.subalign == default_subalign else f"SUBALIGN({self.subalign})"
+
+        s = (
+            f"SPLAT_BEGIN_SEG({sect_name}, 0x{self.rom_start:X}, 0x{vram_or_rom:X}, {subalign_str})\n"
+        )
+
+        i = 0
+        for subdir, path, obj_type, start in self.get_ld_files():
+            # Hack for non-0x10 alignment
+            if start % 0x10 != 0 and i != 0:
+                tmp_sect_name = path.replace(".", "_")
+                tmp_sect_name = tmp_sect_name.replace("/", "_")
+                tmp_vram = start - self.rom_start + self.vram_addr
+                s += (
+                    "}\n"
+                    f"SPLAT_BEGIN_SEG({tmp_sect_name}, 0x{start:X}, 0x{tmp_vram:X}, {subalign_str})\n"
+                )
+
+            path = PurePath(subdir) / PurePath(path)
+            path = path.with_suffix(".o" if replace_ext else path.suffix + ".o")
+
+            s += f"    BUILD_DIR/{path}({obj_type});\n"
+            i += 1
+
+        s += (
+            f"SPLAT_END_SEG({sect_name}, 0x{self.rom_end:X})\n"
+        )
+
+        return s
+
+    def get_ld_section_name(self):
+        return f"data_{self.rom_start:X}"
+
+    # returns list of (basedir, filename, obj_type)
+    def get_ld_files(self):
+        return []
+
+    def log(self, msg):
+        if self.options.get("verbose", False):
+            log.write(f"{self.type} {self.name}: {msg}")
+
+    def warn(self, msg):
+        self.warnings.append(msg)
+
+    def error(self, msg):
+        self.errors.append(msg)
+
+    def max_length(self):
+        return None
+
+    def is_name_default(self):
+        return self.name == self.get_default_name(self.rom_start)
+
+    def unique_id(self):
+        return self.type + "_" + self.name
+
+    def status(self):
+        if len(self.errors) > 0:
+            return "error"
+        elif len(self.warnings) > 0:
+            return "warn"
+        elif self.did_run:
+            return "ok"
+        else:
+            return "skip"
+
+    @staticmethod
+    def get_default_name(addr):
+        return "{:X}".format(addr)

tools/n64splat/split.py

@@ -0,0 +1,408 @@
+#! /usr/bin/python3
+
+import argparse
+import importlib
+import importlib.util
+import os
+from ranges import Range, RangeDict
+import re
+from pathlib import Path
+import segtypes
+import sys
+import yaml
+import pickle
+from colorama import Style, Fore
+from collections import OrderedDict
+from segtypes.segment import N64Segment, parse_segment_type
+from segtypes.code import N64SegCode
+from util import log
+
+parser = argparse.ArgumentParser(
+    description="Split a rom given a rom, a config, and output directory")
+parser.add_argument("rom", help="path to a .z64 rom")
+parser.add_argument("config", help="path to a compatible config .yaml file")
+parser.add_argument("outdir", help="a directory in which to extract the rom")
+parser.add_argument("--modes", nargs="+", default="all")
+parser.add_argument("--verbose", action="store_true",
+                    help="Enable debug logging")
+parser.add_argument("--new", action="store_true",
+                    help="Only split changed segments in config")
+
+
+def write_ldscript(rom_name, repo_path, sections, options):
+    with open(os.path.join(repo_path, rom_name + ".ld"), "w", newline="\n") as f:
+        f.write(
+            "#ifndef SPLAT_BEGIN_SEG\n"
+                "#ifndef SHIFT\n"
+                    "#define SPLAT_BEGIN_SEG(name, start, vram, subalign) \\\n"
+                    "    . = start;\\\n"
+                    "    name##_ROM_START = .;\\\n"
+                    "    name##_VRAM = ADDR(.name);\\\n"
+                    "    .name vram : AT(name##_ROM_START) subalign {\n"
+                "#else\n"
+                    "#define SPLAT_BEGIN_SEG(name, start, vram, subalign) \\\n"
+                    "    name##_ROM_START = .;\\\n"
+                    "    name##_VRAM = ADDR(.name);\\\n"
+                    "    .name vram : AT(name##_ROM_START) subalign {\n"
+                "#endif\n"
+            "#endif\n"
+            "\n"
+            "#ifndef SPLAT_END_SEG\n"
+                "#ifndef SHIFT\n"
+                    "#define SPLAT_END_SEG(name, end) \\\n"
+                    "    } \\\n"
+                    "    . = end;\\\n"
+                    "    name##_ROM_END = .;\n"
+                "#else\n"
+                    "#define SPLAT_END_SEG(name, end) \\\n"
+                    "    } \\\n"
+                    "    name##_ROM_END = .;\n"
+                "#endif\n"
+            "#endif\n"
+            "\n"
+        )
+
+        if options.get("ld_bare", False):
+            f.write("\n".join(sections))
+        else:
+            f.write(
+                "SECTIONS\n"
+                "{\n"
+                "    "
+            )
+            f.write("\n    ".join(s.replace("\n", "\n    ") for s in sections)[:-4])
+            f.write(
+                "}\n"
+            )
+
+
+def parse_file_start(split_file):
+    return split_file[0] if "start" not in split_file else split_file["start"]
+
+
+def get_symbol_addrs_path(repo_path, options):
+    return os.path.join(repo_path, options.get("symbol_addrs", "symbol_addrs.txt"))
+
+
+def get_undefined_syms_path(repo_path, options):
+    return os.path.join(repo_path, options.get("undefined_syms", "undefined_syms.txt"))
+
+
+def gather_symbols(symbol_addrs_path, undefined_syms_path):
+    symbols = {}
+    special_labels = {}
+    labels_to_add = set()
+    ranges = RangeDict()
+
+    # Manual list of func name / addrs
+    if os.path.exists(symbol_addrs_path):
+        with open(symbol_addrs_path) as f:
+            func_addrs_lines = f.readlines()
+
+        for line in func_addrs_lines:
+            line = line.strip()
+            if not line == "" and not line.startswith("//"):
+                comment_loc = line.find("//")
+                line_ext = ""
+
+                if comment_loc != -1:
+                    line_ext = line[comment_loc + 2:].strip()
+                    line = line[:comment_loc].strip()
+
+                line_split = line.split("=")
+                name = line_split[0].strip()
+                addr = int(line_split[1].strip()[:-1], 0)
+                symbols[addr] = name
+
+                if line_ext:
+                    for info in line_ext.split(" "):
+                        if info == "!":
+                            labels_to_add.add(name)
+                            special_labels[addr] = name
+                        if info.startswith("size:"):
+                            size = int(info.split(":")[1], 0)
+                            ranges.add(Range(addr, addr + size), name)
+    
+    if os.path.exists(undefined_syms_path):
+        with open(undefined_syms_path) as f:
+            us_lines = f.readlines()
+
+        for line in us_lines:
+            line = line.strip()
+            if not line == "" and not line.startswith("//"):
+                line_split = line.split("=")
+                name = line_split[0].strip()
+                addr = int(line_split[1].strip()[:-1], 0)
+                symbols[addr] = name
+
+    return symbols, labels_to_add, special_labels, ranges
+
+
+def gather_c_variables(undefined_syms_path):
+    vars = {}
+
+    if os.path.exists(undefined_syms_path):
+        with open(undefined_syms_path) as f:
+            us_lines = f.readlines()
+
+        for line in us_lines:
+            line = line.strip()
+            if not line == "" and not line.startswith("//"):
+                line_split = line.split("=")
+                name = line_split[0].strip()
+                addr = int(line_split[1].strip()[:-1], 0)
+                vars[addr] = name
+
+    return vars
+
+
+def get_base_segment_class(seg_type):
+    try:
+        segmodule = importlib.import_module("segtypes." + seg_type)
+    except ModuleNotFoundError:
+        return None
+
+    return getattr(segmodule, "N64Seg" + seg_type[0].upper() + seg_type[1:])
+
+
+def get_extension_dir(options, config_path):
+    if "extensions" not in options:
+        return None
+    return os.path.join(Path(config_path).parent, options["extensions"])
+
+
+def get_extension_class(options, config_path, seg_type):
+    ext_dir = get_extension_dir(options, config_path)
+    if ext_dir == None:
+        return None
+
+    try:
+        ext_spec = importlib.util.spec_from_file_location(f"segtypes.{seg_type}", os.path.join(ext_dir, f"{seg_type}.py"))
+        ext_mod = importlib.util.module_from_spec(ext_spec)
+        ext_spec.loader.exec_module(ext_mod)
+    except Exception as err:
+        log.write(err, status="error")
+        return None
+
+    return getattr(ext_mod, "N64Seg" + seg_type[0].upper() + seg_type[1:])
+
+
+def fmt_size(size):
+    if size > 1000000:
+        return str(size // 1000000) + " MB"
+    elif size > 1000:
+        return str(size // 1000) + " KB"
+    else:
+        return str(size) + " B"
+
+
+def initialize_segments(options, config_path, config_segments):
+    seen_segment_names = set()
+    ret = []
+
+    for i, segment in enumerate(config_segments):
+        if len(segment) == 1:
+            # We're at the end
+            continue
+
+        seg_type = parse_segment_type(segment)
+
+        segment_class = get_base_segment_class(seg_type)
+        if segment_class == None:
+            # Look in extensions
+            segment_class = get_extension_class(options, config_path, seg_type)
+
+        if segment_class == None:
+            log.write(f"fatal error: could not load segment type '{seg_type}'\n(hint: confirm your extension directory is configured correctly)", status="error")
+            return 2
+
+        try:
+            segment = segment_class(segment, config_segments[i + 1], options)
+        except (IndexError, KeyError) as e:
+            try:
+                segment = N64Segment(segment, config_segments[i + 1], options)
+                segment.error(e)
+            except Exception as e:
+                log.write(f"fatal error (segment type = {seg_type}): " + str(e), status="error")
+                return 2
+
+        if segment_class.require_unique_name:
+            if segment.name in seen_segment_names:
+                segment.error("segment name is not unique")
+            seen_segment_names.add(segment.name)
+
+        ret.append(segment)
+
+    return ret
+
+
+def main(rom_path, config_path, repo_path, modes, verbose, ignore_cache=False):
+    with open(rom_path, "rb") as f:
+        rom_bytes = f.read()
+
+    # Create main output dir
+    Path(repo_path).mkdir(parents=True, exist_ok=True)
+
+    # Load config
+    with open(config_path) as f:
+        config = yaml.safe_load(f.read())
+
+    options = config.get("options")
+    options["modes"] = modes
+    options["verbose"] = verbose
+
+    symbol_addrs_path = get_symbol_addrs_path(repo_path, options)
+    undefined_syms_path = get_undefined_syms_path(repo_path, options)
+    provided_symbols, c_func_labels_to_add, special_labels, ranges = gather_symbols(symbol_addrs_path, undefined_syms_path)
+
+    processed_segments = []
+    ld_sections = []
+
+    defined_funcs = {}
+    undefined_funcs = set()
+    undefined_syms = set()
+
+    seg_sizes = {}
+    seg_split = {}
+    seg_cached = {}
+
+    # Load cache
+    cache_path = Path(repo_path) / ".splat_cache"
+    try:
+        with open(cache_path, "rb") as f:
+            cache = pickle.load(f)
+    except Exception:
+        cache = {}
+
+    # Initialize segments
+    all_segments = initialize_segments(options, config_path, config["segments"])
+
+    for segment in all_segments:
+        if type(segment) == N64SegCode:
+            segment.all_functions = defined_funcs
+            segment.provided_symbols = provided_symbols
+            segment.special_labels = special_labels
+            segment.c_labels_to_add = c_func_labels_to_add
+            segment.symbol_ranges = ranges
+
+        segment.check()
+
+        tp = segment.type
+        if segment.type == "bin" and segment.is_name_default():
+            tp = "unk"
+
+        if tp not in seg_sizes:
+            seg_sizes[tp] = 0
+            seg_split[tp] = 0
+            seg_cached[tp] = 0
+        seg_sizes[tp] += segment.size
+
+        if len(segment.errors) == 0:
+            if segment.should_run():
+                # Check cache
+                cached = segment.cache()
+                if not ignore_cache and cached == cache.get(segment.unique_id()):
+                    # Cache hit
+                    seg_cached[tp] += 1
+                else:
+                    # Cache miss; split
+                    cache[segment.unique_id()] = cached
+
+                    segment.did_run = True
+                    segment.split(rom_bytes, repo_path)
+
+                    if len(segment.errors) == 0:
+                        processed_segments.append(segment)
+
+                        if type(segment) == N64SegCode:
+                            undefined_funcs |= segment.glabels_to_add
+                            defined_funcs = {**defined_funcs, **segment.glabels_added}
+                            undefined_syms |= segment.undefined_syms_to_add
+
+                    seg_split[tp] += 1
+
+        log.dot(status=segment.status())
+        ld_sections.append(segment.get_ld_section())
+
+    for segment in processed_segments:
+        segment.postsplit(processed_segments)
+        log.dot(status=segment.status())
+
+    # Write ldscript
+    if "ld" in options["modes"] or "all" in options["modes"]:
+        if verbose:
+            log.write(f"saving {config['basename']}.ld")
+        write_ldscript(config['basename'], repo_path, ld_sections, options)
+
+    # Write undefined_funcs_auto.txt
+    if verbose:
+        log.write(f"saving undefined_funcs_auto.txt")
+    c_predefined_funcs = set(provided_symbols.keys())
+    to_write = sorted(undefined_funcs - set(defined_funcs.values()) - c_predefined_funcs)
+    if len(to_write) > 0:
+        with open(os.path.join(repo_path, "undefined_funcs_auto.txt"), "w", newline="\n") as f:
+            for line in to_write:
+                f.write(line + " = 0x" + line.split("_")[1][:8].upper() + ";\n")
+
+    # write undefined_syms_auto.txt
+    if verbose:
+        log.write(f"saving undefined_syms_auto.txt")
+    to_write = sorted(undefined_syms, key=lambda x:x[0])
+    if len(to_write) > 0:
+        with open(os.path.join(repo_path, "undefined_syms_auto.txt"), "w", newline="\n") as f:
+            for sym in to_write:
+                f.write(f"{sym[0]} = 0x{sym[1]:X};\n")
+
+    # print warnings and errors during split/postsplit
+    had_error = False
+    for segment in all_segments:
+        if len(segment.warnings) > 0 or len(segment.errors) > 0:
+            log.write(f"{Style.DIM}0x{segment.rom_start:06X}{Style.RESET_ALL} {segment.type} {Style.BRIGHT}{segment.name}{Style.RESET_ALL}:")
+
+            for warn in segment.warnings:
+                log.write("warning: " + warn, status="warn")
+
+            for error in segment.errors:
+                log.write("error: " + error, status="error")
+                had_error = True
+
+            log.write("") # empty line
+
+    if had_error:
+        return 1
+
+    # Statistics
+    unk_size = seg_sizes.get("unk", 0)
+    rest_size = 0
+    total_size = len(rom_bytes)
+
+    for tp in seg_sizes:
+        if tp != "unk":
+            rest_size += seg_sizes[tp]
+
+    assert(unk_size + rest_size == total_size)
+
+    known_ratio = rest_size / total_size
+    unk_ratio = unk_size / total_size
+
+    log.write(f"Split {fmt_size(rest_size)} ({known_ratio:.2%}) in defined segments")
+    for tp in seg_sizes:
+        if tp != "unk":
+            tmp_size = seg_sizes[tp]
+            tmp_ratio = tmp_size / total_size
+            log.write(f"{tp:>20}: {fmt_size(tmp_size):>8} ({tmp_ratio:.2%}) {Fore.GREEN}{seg_split[tp]} split{Style.RESET_ALL}, {Style.DIM}{seg_cached[tp]} cached")
+    log.write(f"{'unknown':>20}: {fmt_size(unk_size):>8} ({unk_ratio:.2%}) from unknown bin files")
+
+    # Save cache
+    if cache != {}:
+        if verbose:
+            print("Writing cache")
+        with open(cache_path, "wb") as f:
+            pickle.dump(cache, f)
+
+    return 0 # no error
+
+if __name__ == "__main__":
+    args = parser.parse_args()
+    error_code = main(args.rom, args.config, args.outdir, args.modes, args.verbose, not args.new)
+    exit(error_code)

tools/n64splat/util/Yay0decompress.c

@@ -0,0 +1,43 @@
+#include <stdio.h>
+#include <stdint.h>
+#include <stdbool.h>
+#include <stdlib.h>
+
+typedef struct {
+    uint32_t magic;
+    uint32_t uncompressedLength;
+    uint32_t opPtr;
+    uint32_t dataPtr;
+} Yay0Header;
+
+void decompress(Yay0Header* hdr, uint8_t* srcPtr, uint8_t* dstPtr, bool isBigEndian) {
+    uint8_t byte = 0, mask = 0;
+    uint8_t* ctrl, * ops, * data;
+    uint16_t copy, op;
+    uint32_t written = 0;
+
+    ctrl = srcPtr + sizeof(Yay0Header);
+    ops = srcPtr + hdr->opPtr;
+    data = srcPtr + hdr->dataPtr;
+
+    while (written < hdr->uncompressedLength) {
+        if ((mask >>= 1) == 0) {
+            byte = *ctrl++;
+            mask = 0x80;
+        }
+
+        if (byte & mask) {
+            *dstPtr++ = *data++;
+            written++;
+        } else {
+            op = isBigEndian ? (ops[0] << 8) | ops[1] : (ops[1] << 8) | ops[0];
+            ops += 2;
+
+            written += copy = (op >> 12) ? (2 + (op >> 12)) : (18 + *data++);
+
+            while (copy--) {
+                *dstPtr++ = dstPtr[-(op & 0xfff) - 1];
+            }
+        }
+    }
+}

tools/n64splat/util/Yay0decompress.py

@@ -0,0 +1,132 @@
+import argparse
+import sys
+import os
+from ctypes import *
+from struct import pack, unpack_from
+
+tried_loading = False
+lib = None
+
+def setup_lib():
+    global lib
+    global tried_loading
+    if lib:
+        return True
+    if tried_loading:
+        return False
+    try:
+        tried_loading = True
+        lib = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + "/Yay0decompress")
+        return True
+    except Exception:
+        print(f"Failed to load Yay0decompress, falling back to python method")
+        tried_loading = True
+        return False
+
+def decompress_yay0(in_bytes, byte_order="big"):
+    # attempt to load the library only once per execution
+    global lib
+    if not setup_lib():
+        return decompress_yay0_python(in_bytes, byte_order)
+
+    class Yay0(Structure):
+        _fields_ = [
+            ("magic", c_uint32),
+            ("uncompressedLength", c_uint32),
+            ("opPtr", c_uint32),
+            ("dataPtr", c_uint32),
+        ]
+
+    # read the file header
+    bigEndian = byte_order == "big"
+    if bigEndian:
+        # the struct is only a view, so when passed to C it will keep
+        # its BigEndian values and crash. Explicitly convert them here to little
+        hdr = Yay0.from_buffer_copy(pack("<IIII", *unpack_from(">IIII", in_bytes, 0)))
+    else:
+        hdr = Yay0.from_buffer_copy(in_bytes, 0)
+
+    # create the input/output buffers, copying data to in
+    src = (c_uint8 * len(in_bytes)).from_buffer_copy(in_bytes, 0)
+    dst = (c_uint8 * hdr.uncompressedLength)()
+
+    # call decompress, equivilant to, in C:
+    # decompress(&hdr, &src, &dst, bigEndian)
+    lib.decompress(byref(hdr), byref(src), byref(dst), c_bool(bigEndian))
+
+    # other functions want the results back as a non-ctypes type
+    return bytearray(dst)
+
+def decompress_yay0_python(in_bytes, byte_order="big"):
+    if in_bytes[:4] != b"Yay0":
+        sys.exit("Input file is not yay0")
+
+    decompressed_size = int.from_bytes(in_bytes[4:8], byteorder=byte_order)
+    link_table_offset = int.from_bytes(in_bytes[8:12], byteorder=byte_order)
+    chunk_offset = int.from_bytes(in_bytes[12:16], byteorder=byte_order)
+
+    link_table_idx = link_table_offset
+    chunk_idx = chunk_offset
+    other_idx = 16
+
+    mask_bit_counter = 0
+    current_mask = 0
+
+    # preallocate result and index into it
+    idx = 0
+    ret = bytearray(decompressed_size);
+
+    while idx < decompressed_size:
+        # If we're out of bits, get the next mask
+        if mask_bit_counter == 0:
+            current_mask = int.from_bytes(in_bytes[other_idx : other_idx + 4], byteorder=byte_order)
+            other_idx += 4
+            mask_bit_counter = 32
+
+        if (current_mask & 0x80000000):
+            ret[idx] = in_bytes[chunk_idx]
+            idx += 1
+            chunk_idx += 1
+        else:
+            link = int.from_bytes(in_bytes[link_table_idx : link_table_idx + 2], byteorder=byte_order)
+            link_table_idx += 2
+
+            offset = idx - (link & 0xfff)
+
+            count = link >> 12
+
+            if count == 0:
+                count_modifier = in_bytes[chunk_idx]
+                chunk_idx += 1
+                count = count_modifier + 18
+            else:
+                count += 2
+
+            # Copy the block
+            for i in range(count):
+                ret[idx] = ret[offset + i - 1]
+                idx += 1
+
+        current_mask <<= 1
+        mask_bit_counter -= 1
+
+    return ret
+
+
+def main(args):
+    with open(args.infile, "rb") as f:
+        raw_bytes = f.read()
+
+    decompressed = decompress_yay0(raw_bytes, args.byte_order)
+
+    with open(args.outfile, "wb") as f:
+        f.write(decompressed)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("infile")
+    parser.add_argument("outfile")
+    parser.add_argument("--byte-order", default="big", choices=["big", "little"])
+
+    args = parser.parse_args()
+    main(args)

tools/n64splat/util/color.py

@@ -0,0 +1,16 @@
+from math import ceil
+
+# RRRRRGGG GGBBBBBA
+def unpack_color(data):
+    s = int.from_bytes(data[0:2], byteorder="big")
+
+    r = (s >> 11) & 0x1F
+    g = (s >>  6) & 0x1F
+    b = (s >>  1) & 0x1F
+    a = (s &   1) * 0xFF
+
+    r = ceil(0xFF * (r / 31))
+    g = ceil(0xFF * (g / 31))
+    b = ceil(0xFF * (b / 31))
+
+    return r, g, b, a

tools/n64splat/util/find_code_length.py

@@ -0,0 +1,48 @@
+#! /usr/bin/python3
+
+from capstone import *
+from capstone.mips import *
+
+import argparse
+import hashlib
+import rominfo
+import zlib
+
+md = Cs(CS_ARCH_MIPS, CS_MODE_MIPS64 + CS_MODE_BIG_ENDIAN)
+
+parser = argparse.ArgumentParser(description="Given a rom and start offset, find where the code ends")
+parser.add_argument("rom", help="path to a .z64 rom")
+parser.add_argument("start", help="start offset")
+parser.add_argument("--end", help="end offset", default=None)
+parser.add_argument("--vram", help="vram address to start disassembly at", default="0x80000000")
+
+def run(rom_bytes, start_offset, vram, end_offset=None):
+    rom_addr = start_offset
+    last_return = rom_addr
+
+    for insn in md.disasm(rom_bytes[start_offset:], vram):
+        if insn.mnemonic == "jr" and insn.op_str == "$ra":
+            last_return = rom_addr
+        rom_addr += 4
+        if end_offset and rom_addr >= end_offset:
+            break
+
+    return last_return + (0x10 - (last_return % 0x10))
+
+
+def main():
+    args = parser.parse_args()
+
+    rom_bytes = rominfo.read_rom(args.rom)
+    start = int(args.start, 0)
+    end = None
+    vram = int(args.vram, 0)
+
+    if args.end:
+        end = int(args.end, 0)
+
+    print(f"{run(rom_bytes, start, vram, end):X}")
+
+
+if __name__ == "__main__":
+    main()

tools/n64splat/util/floats.py

@@ -0,0 +1,62 @@
+
+import math
+import struct
+
+# From mips_to_c: https://github.com/matt-kempster/mips_to_c/blob/d208400cca045113dada3e16c0d59c50cdac4529/src/translate.py#L2085
+def format_f32_imm(num: int) -> str:
+    packed = struct.pack(">I", num & (2 ** 32 - 1))
+    value = struct.unpack(">f", packed)[0]
+
+    if not value or value == 4294967296.0:
+        # Zero, negative zero, nan, or INT_MAX.
+        return str(value)
+
+    # Write values smaller than 1e-7 / greater than 1e7 using scientific notation,
+    # and values in between using fixed point.
+    if abs(math.log10(abs(value))) > 6.9:
+        fmt_char = "e"
+    elif abs(value) < 1:
+        fmt_char = "f"
+    else:
+        fmt_char = "g"
+
+    def fmt(prec: int) -> str:
+        """Format 'value' with 'prec' significant digits/decimals, in either scientific
+        or regular notation depending on 'fmt_char'."""
+        ret = ("{:." + str(prec) + fmt_char + "}").format(value)
+        if fmt_char == "e":
+            return ret.replace("e+", "e").replace("e0", "e").replace("e-0", "e-")
+        if "e" in ret:
+            # The "g" format character can sometimes introduce scientific notation if
+            # formatting with too few decimals. If this happens, return an incorrect
+            # value to prevent the result from being used.
+            #
+            # Since the value we are formatting is within (1e-7, 1e7) in absolute
+            # value, it will at least be possible to format with 7 decimals, which is
+            # less than float precision. Thus, this annoying Python limitation won't
+            # lead to us outputting numbers with more precision than we really have.
+            return "0"
+        return ret
+
+    # 20 decimals is more than enough for a float. Start there, then try to shrink it.
+    prec = 20
+    while prec > 0:
+        prec -= 1
+        value2 = float(fmt(prec))
+        if struct.pack(">f", value2) != packed:
+            prec += 1
+            break
+
+    if prec == 20:
+        # Uh oh, even the original value didn't format correctly. Fall back to str(),
+        # which ought to work.
+        return str(value)
+
+    ret = fmt(prec)
+    if "." not in ret:
+        ret += ".0"
+    return ret
+
+def format_f64_imm(num: int) -> str:
+    (value,) = struct.unpack(">d", struct.pack(">Q", num & (2 ** 64 - 1)))
+    return str(value)

tools/n64splat/util/iter.py

@@ -0,0 +1,5 @@
+from itertools import zip_longest
+
+def iter_in_groups(iterable, n, fillvalue=None):
+    args = [iter(iterable)] * n
+    return zip_longest(*args, fillvalue=fillvalue)

tools/n64splat/util/log.py

@@ -0,0 +1,32 @@
+from colorama import init, Fore, Style
+
+init(autoreset=True)
+
+newline = True
+
+def write(*args, status=None, **kwargs):
+    global newline
+
+    if not newline:
+        print("")
+        newline = True
+
+    print(status_to_ansi(status) + str(args[0]), *args[1:], **kwargs)
+
+def dot(status=None):
+    global newline
+
+    print(status_to_ansi(status) + ".", end="")
+    newline = False
+
+def status_to_ansi(status):
+    if status == "ok":
+        return Fore.GREEN
+    elif status == "warn":
+        return Fore.YELLOW + Style.BRIGHT
+    elif status == "error":
+        return Fore.RED + Style.BRIGHT
+    elif status == "skip":
+        return Style.DIM
+    else:
+        return ""

tools/n64splat/util/rominfo.py

@@ -0,0 +1,119 @@
+#! /usr/bin/python3
+
+import argparse
+import hashlib
+import zlib
+
+parser = argparse.ArgumentParser(description='Gives information on n64 roms')
+parser.add_argument('rom', help='path to a .z64 rom')
+parser.add_argument('--encoding', help='Text encoding the game header is using, defaults to ASCII, see docs.python.org/2.4/lib/standard-encodings.html for valid encodings', default='ASCII')
+
+country_codes = {
+    0x37: "Beta",
+    0x41: "Asian (NTSC)",
+    0x42: "Brazillian",
+    0x43: "Chiniese",
+    0x44: "German",
+    0x45: "North America",
+    0x46: "French",
+    0x47: "Gateway 64 (NTSC)",
+    0x48: "Dutch",
+    0x49: "Italian",
+    0x4A: "Japanese",
+    0x4B: "Korean",
+    0x4C: "Gateway 64 (PAL)",
+    0x4E: "Canadian",
+    0x50: "European (basic spec.)",
+    0x53: "Spanish",
+    0x55: "Australian",
+    0x57: "Scandanavian",
+    0x58: "European",
+    0x59: "European",
+}
+
+crc_to_cic = {
+    0x6170A4A1: {"ntsc-name": "6101", "pal-name": "7102", "offset": 0x000000},
+    0x90BB6CB5: {"ntsc-name": "6102", "pal-name": "7101", "offset": 0x000000},
+    0x0B050EE0: {"ntsc-name": "6103", "pal-name": "7103", "offset": 0x100000},
+    0x98BC2C86: {"ntsc-name": "6105", "pal-name": "7105", "offset": 0x000000},
+    0xACC8580A: {"ntsc-name": "6106", "pal-name": "7106", "offset": 0x200000},
+    0x00000000: {"ntsc-name": "unknown", "pal-name": "unknown", "offset": 0x0000000}
+}
+
+
+def read_rom(rom):
+    with open(rom, "rb") as f:
+        return f.read()
+
+
+def get_cic(rom_bytes):
+    crc = zlib.crc32(rom_bytes[0x40:0x1000])
+    if crc in crc_to_cic:
+        return crc_to_cic[crc]
+    else:
+        return crc_to_cic[0]
+
+
+def get_entry_point(program_counter, cic):
+    return program_counter - cic["offset"]
+
+
+def get_info(rom_path, encoding="ASCII"):
+    return get_info_bytes(read_rom(rom_path), encoding)
+
+
+def get_info_bytes(rom_bytes, encoding):
+    program_counter = int(rom_bytes[0x8:0xC].hex(), 16)
+    libultra_version = chr(rom_bytes[0xF])
+    crc1 = rom_bytes[0x10:0x14].hex().upper()
+    crc2 = rom_bytes[0x14:0x18].hex().upper()
+
+    try:
+        name = rom_bytes[0x20:0x34].decode(encoding).strip()
+    except:
+        print("n64splat could not decode the game name, try using a different encoding by passing the --encoding argument (see docs.python.org/2.4/lib/standard-encodings.html for valid encodings)")
+        exit(1)
+
+    country_code = rom_bytes[0x3E]
+
+    cic = get_cic(rom_bytes)
+    entry_point = get_entry_point(program_counter, cic)
+
+    # todo add support for
+    # compression_formats = []
+    #  for format in ["Yay0", "vpk0"]:
+    #     if rom_bytes.find(bytes(format, "ASCII")) != -1:
+    #         compression_formats.append(format)
+
+    return N64Rom(name, country_code, libultra_version, crc1, crc2, cic, entry_point, len(rom_bytes))
+
+
+class N64Rom:
+    def __init__(self, name, country_code, libultra_version, crc1, crc2, cic, entry_point, size):
+        self.name = name
+        self.country_code = country_code
+        self.libultra_version = libultra_version
+        self.crc1 = crc1
+        self.crc2 = crc2
+        self.cic = cic
+        self.entry_point = entry_point
+        self.size = size
+
+    def get_country_name(self):
+        return country_codes[self.country_code]
+
+
+def main():
+    args = parser.parse_args()
+    rom = get_info(args.rom, args.encoding)
+    print("Image name: " + rom.name)
+    print("Country code: " + chr(rom.country_code) + " - " + rom.get_country_name())
+    print("Libultra version: " + rom.libultra_version)
+    print("CRC1: " + rom.crc1)
+    print("CRC2: " + rom.crc2)
+    print("CIC: " + rom.cic["ntsc-name"] + " / " + rom.cic["pal-name"])
+    print("RAM entry point: " + hex(rom.entry_point))
+
+
+if __name__ == "__main__":
+    main()

tools/star-rod

@@ -1 +0,0 @@
-Subproject commit aec5d4c037e95227fb5f118075564031636697fe