v109.3

Fixed crash when idling with the snow effect enabled.
Added Android target to libretro port [rtretiakov]
Various nall library improvements.

bsnes/emulator/emulator.hpp

@@ -29,7 +29,7 @@ using namespace nall;
 
 namespace Emulator {
   static const string Name    = "bsnes";
-  static const string Version = "109.2";
+  static const string Version = "109.3";
   static const string Author  = "byuu";
   static const string License = "GPLv3";
   static const string Website = "https://byuu.org";

bsnes/target-bsnes/program/viewport.cpp

@@ -91,7 +91,7 @@ auto Program::viewportRefresh() -> void {
 
       for(; SnowMover != 0; --SnowMover) {
         if(--SnowTimer == 0) {
-          ++NumSnow;
+          if(NumSnow < 400) ++NumSnow;
           SnowTimer = 18;
         }
         uint32_t i = 0;

bsnes/target-libretro/jni/Android.mk

@@ -0,0 +1,58 @@
+LOCAL_PATH := $(call my-dir)
+
+SRCDIR := $(LOCAL_PATH)/../../..
+
+INCFLAGS  := -I$(SRCDIR) -I$(SRCDIR)/bsnes
+COREFLAGS := -fomit-frame-pointer -ffast-math -D__LIBRETRO__ $(INCFLAGS)
+COREFLAGS += -DGB_INTERNAL -DDISABLE_DEBUGGER -DPLATFORM_ANDROID
+
+GIT_VERSION := " $(shell git rev-parse --short HEAD || echo unknown)"
+ifneq ($(GIT_VERSION)," unknown")
+  COREFLAGS += -DGIT_VERSION=\"$(GIT_VERSION)\"
+endif
+
+SRCFILES := $(SRCDIR)/bsnes/target-libretro/libretro.cpp \
+				$(SRCDIR)/bsnes/emulator/emulator.cpp \
+				$(SRCDIR)/libco/libco.c \
+				$(SRCDIR)/bsnes/filter/filter.cpp \
+				$(SRCDIR)/bsnes/lzma/lzma.cpp \
+				$(SRCDIR)/bsnes/sfc/interface/interface.cpp \
+				$(SRCDIR)/bsnes/sfc/system/system.cpp \
+				$(SRCDIR)/bsnes/sfc/controller/controller.cpp \
+				$(SRCDIR)/bsnes/sfc/cartridge/cartridge.cpp \
+				$(SRCDIR)/bsnes/sfc/memory/memory.cpp \
+				$(SRCDIR)/bsnes/sfc/cpu/cpu.cpp \
+				$(SRCDIR)/bsnes/sfc/smp/smp.cpp \
+				$(SRCDIR)/bsnes/sfc/dsp/dsp.cpp \
+				$(SRCDIR)/bsnes/sfc/ppu/ppu.cpp \
+				$(SRCDIR)/bsnes/sfc/ppu-fast/ppu.cpp \
+				$(SRCDIR)/bsnes/sfc/expansion/expansion.cpp \
+				$(SRCDIR)/bsnes/sfc/coprocessor/coprocessor.cpp \
+				$(SRCDIR)/bsnes/sfc/slot/slot.cpp \
+				$(SRCDIR)/bsnes/gb/Core/apu.c \
+				$(SRCDIR)/bsnes/gb/Core/camera.c \
+				$(SRCDIR)/bsnes/gb/Core/display.c \
+				$(SRCDIR)/bsnes/gb/Core/gb.c \
+				$(SRCDIR)/bsnes/gb/Core/joypad.c \
+				$(SRCDIR)/bsnes/gb/Core/mbc.c \
+				$(SRCDIR)/bsnes/gb/Core/memory.c \
+				$(SRCDIR)/bsnes/gb/Core/printer.c \
+				$(SRCDIR)/bsnes/gb/Core/random.c \
+				$(SRCDIR)/bsnes/gb/Core/rewind.c \
+				$(SRCDIR)/bsnes/gb/Core/save_state.c \
+				$(SRCDIR)/bsnes/gb/Core/sgb.c \
+				$(SRCDIR)/bsnes/gb/Core/sm83_cpu.c \
+				$(SRCDIR)/bsnes/gb/Core/symbol_hash.c \
+				$(SRCDIR)/bsnes/gb/Core/timing.c \
+				$(SRCDIR)/bsnes/processor/arm7tdmi/arm7tdmi.cpp \
+				$(SRCDIR)/bsnes/processor/spc700/spc700.cpp \
+				$(SRCDIR)/bsnes/processor/wdc65816/wdc65816.cpp
+
+include $(CLEAR_VARS)
+LOCAL_MODULE       := retro
+LOCAL_SRC_FILES    := $(SRCFILES)
+LOCAL_CPPFLAGS     := -std=c++17 $(COREFLAGS)
+LOCAL_CFLAGS       := $(COREFLAGS)
+LOCAL_LDFLAGS      := -Wl,-version-script=$(SRCDIR)/bsnes/target-libretro/link.T
+LOCAL_CPP_FEATURES := exceptions rtti
+include $(BUILD_SHARED_LIBRARY)

bsnes/target-libretro/jni/Application.mk

@@ -0,0 +1,2 @@
+APP_ABI := all
+APP_STL := c++_static

nall/beat/single/create.hpp

@@ -25,7 +25,10 @@ inline auto create(array_view<uint8_t> source, array_view<uint8_t> target, strin
   encode(source.size()), encode(target.size()), encode(manifest.size());
   for(auto& byte : manifest) write(byte);
 
-  auto sourceArray = SuffixArray(source).lrcp();
+  //generating lrcp() arrays for source requires O(4n) computations, and O(16m) memory,
+  //but it reduces find() complexity from O(n log m) to O(n + log m). and yet in practice,
+  //no matter how large n scales to, the O(n + log m) find() is paradoxically slower.
+  auto sourceArray = SuffixArray(source);
   auto targetArray = SuffixArray(target).lpf();
 
   enum : uint { SourceRead, TargetRead, SourceCopy, TargetCopy };
@@ -39,12 +42,12 @@ inline auto create(array_view<uint8_t> source, array_view<uint8_t> target, strin
     while(targetReadLength) write(target[offset++]), targetReadLength--;
   };
 
-  uint longestSize = max(source.size(), target.size());
+  uint overlap = min(source.size(), target.size());
   while(outputOffset < target.size()) {
     uint mode = TargetRead, longestLength = 3, longestOffset = 0;
     int length = 0, offset = outputOffset;
 
-    while(offset < longestSize) {
+    while(offset < overlap) {
       if(source[offset] != target[offset]) break;
       length++, offset++;
     }

nall/induced-sort.hpp

@@ -10,7 +10,8 @@ namespace nall {
 //where the first character is the empty suffix, equal to size
 
 template<typename T>
-inline auto induced_sort(const T* data, const uint size, const uint characters = 256) -> vector<int> {
+inline auto induced_sort(array_view<T> data, const uint characters = 256) -> vector<int> {
+  const uint size = data.size();
   if(size == 0) return vector<int>{0};  //required to avoid out-of-bounds accesses
   if(size == 1) return vector<int>{1, 0};  //not strictly necessary; but more performant
 
@@ -153,14 +154,14 @@ inline auto induced_sort(const T* data, const uint size, const uint characters =
     }
   } else {
     //recurse until every character in summaryData is unique ...
-    summaries = induced_sort(summaryData.data(), summaryData.size() - 1, summaryCharacters);
+    summaries = induced_sort<int>({summaryData.data(), summaryData.size()}, summaryCharacters);
   }
 
   suffixes.fill(-1);  //reuse existing buffer for accurate sort
 
   //accurate LMS sort
   getTails();
-  for(uint n : reverse(range(1, summaries.size()))) {
+  for(uint n : reverse(range(2, summaries.size()))) {
     auto index = summaryOffsets[summaries[n]];
     suffixes[tails[data[index]]--] = index;  //advance from the tail of the bucket
   }

nall/intrinsics.hpp

@@ -4,7 +4,7 @@ namespace nall {
   using uint = unsigned;
 
   enum class Compiler : uint { Clang, GCC, Microsoft, Unknown };
-  enum class Platform : uint { Windows, MacOS, Linux, BSD, Unknown };
+  enum class Platform : uint { Windows, MacOS, Linux, BSD, Android, Unknown };
   enum class API : uint { Windows, Posix, Unknown };
   enum class DisplayServer : uint { Windows, Quartz, Xorg, Unknown };
   enum class Architecture : uint { x86, amd64, ARM32, ARM64, PPC32, PPC64, Unknown };
@@ -82,6 +82,13 @@ namespace nall {
   constexpr auto platform() -> Platform { return Platform::MacOS; }
   constexpr auto api() -> API { return API::Posix; }
   constexpr auto display() -> DisplayServer { return DisplayServer::Quartz; }
+#elif defined(__ANDROID__)
+  #define PLATFORM_ANDROID
+  #define API_POSIX
+  #define DISPLAY_UNKNOWN
+  constexpr auto platform() -> Platform { return Platform::Android; }
+  constexpr auto api() -> API { return API::Posix; }
+  constexpr auto display() -> DisplayServer { return DisplayServer::Unknown; }
 #elif defined(linux) || defined(__linux__)
   #define PLATFORM_LINUX
   #define API_POSIX

nall/property.hpp

@@ -0,0 +1,13 @@
+#if !defined(property)
+  #define property1(declaration) public: declaration
+  #define property2(declaration, getter) public: __declspec(property(get=getter)) declaration; protected: declaration##_
+  #define property3(declaration, getter, setter) public: __declspec(property(get=getter, put=setter)) declaration; protected: declaration##_
+  #define property_(_1, _2, _3, name, ...) name
+  #define property(...) property_(__VA_ARGS__, property3, property2, property1)(__VA_ARGS__)
+#else
+  #undef property1
+  #undef property2
+  #undef property3
+  #undef property_
+  #undef property
+#endif

nall/random.hpp

@@ -9,6 +9,8 @@
 
 #if defined(PLATFORM_LINUX) && __has_include(<sys/random.h>)
   #include <sys/random.h>
+#elif defined(PLATFORM_ANDROID) && __has_include(<sys/syscall.h>)
+  #include <sys/syscall.h>
 #elif defined(PLATFORM_WINDOWS) && __has_include(<wincrypt.h>)
   #include <wincrypt.h>
 #else
@@ -41,6 +43,8 @@ protected:
     for(uint n : range(8)) seed = seed << 32 | (uint32_t)arc4random();
     #elif defined(PLATFORM_LINUX) && __has_include(<sys/random.h>)
     getrandom(&seed, 32, GRND_NONBLOCK);
+    #elif defined(PLATFORM_ANDROID) && __has_include(<sys/syscall.h>)
+    syscall(__NR_getrandom, &seed, 32, 0x0001);  //GRND_NONBLOCK
     #elif defined(PLATFORM_WINDOWS) && __has_include(<wincrypt.h>)
     HCRYPTPROV provider;
     if(CryptAcquireContext(&provider, nullptr, MS_STRONG_PROV, PROV_RSA_FULL, CRYPT_VERIFYCONTEXT)) {
@@ -63,6 +67,8 @@ namespace PRNG {
 
 //Galois linear feedback shift register using CRC64 polynomials
 struct LFSR : RNG<LFSR> {
+  LFSR() { seed(); }
+
   auto seed(maybe<uint64_t> seed = {}) -> void {
     lfsr = seed ? seed() : (uint64_t)randomSeed();
     for(uint n : range(8)) read();  //hide the CRC64 polynomial from initial output
@@ -84,6 +90,8 @@ private:
 };
 
 struct PCG : RNG<PCG> {
+  PCG() { seed(); }
+
   auto seed(maybe<uint32_t> seed = {}, maybe<uint32_t> sequence = {}) -> void {
     if(!seed) seed = (uint32_t)randomSeed();
     if(!sequence) sequence = 0;

nall/string.hpp

@@ -45,6 +45,7 @@ struct string_view {
   inline auto operator=(const string_view& source) -> type&;
   inline auto operator=(string_view&& source) -> type&;
 
+  inline explicit operator bool() const;
   inline operator const char*() const;
   inline auto data() const -> const char*;
   inline auto size() const -> uint;

nall/string/transform/dml.hpp

@@ -81,6 +81,17 @@ inline auto DML::parseBlock(string& block, const string& pathname, uint depth) -
     }
   }
 
+  //title
+  else if(block.beginsWith("! ")) {
+    auto name = lines.takeLeft().trimLeft("! ", 1L);
+    state.output.append("<h1>", markup(name));
+    for(auto& line : lines) {
+      if(!line.beginsWith("! ")) continue;
+      state.output.append("<span>", markup(line.trimLeft("! ", 1L)), "</span>");
+    }
+    state.output.append("</h1>\n");
+  }
+
   //section
   else if(block.beginsWith("# ")) {
     if(settings.sectioned) {
@@ -90,12 +101,12 @@ inline auto DML::parseBlock(string& block, const string& pathname, uint depth) -
     auto content = lines.takeLeft().trimLeft("# ", 1L).split("::", 1L).strip();
     auto data = markup(content[0]);
     auto name = escape(content(1, data.hash()));
-    state.output.append("<h1 id=\"", name, "\">", data);
+    state.output.append("<h2 id=\"", name, "\">", data);
     for(auto& line : lines) {
       if(!line.beginsWith("# ")) continue;
       state.output.append("<span>", line.trimLeft("# ", 1L), "</span>");
     }
-    state.output.append("</h1>\n");
+    state.output.append("</h2>\n");
   }
 
   //header
@@ -103,13 +114,13 @@ inline auto DML::parseBlock(string& block, const string& pathname, uint depth) -
     auto content = slice(lines.takeLeft(), depth + 1).split("::", 1L).strip();
     auto data = markup(content[0]);
     auto name = escape(content(1, data.hash()));
-    if(depth <= 5) {
-      state.output.append("<h", depth + 1, " id=\"", name, "\">", data);
+    if(depth <= 4) {
+      state.output.append("<h", depth + 2, " id=\"", name, "\">", data);
       for(auto& line : lines) {
         if(count(line, '=') != depth) continue;
         state.output.append("<span>", slice(line, depth + 1), "</span>");
       }
-      state.output.append("</h", depth + 1, ">\n");
+      state.output.append("</h", depth + 2, ">\n");
     }
   }
 

nall/string/view.hpp

@@ -69,6 +69,10 @@ auto string_view::operator=(string_view&& source) -> type& {
   return *this;
 };
 
+string_view::operator bool() const {
+  return _size > 0;
+}
+
 string_view::operator const char*() const {
   return _data;
 }

nall/suffix-array.hpp

@@ -71,8 +71,8 @@ suffix_array_plcp:
   plcp = [1,0,0,3,2,2,1,1,0,0]
 
 suffix_array_lrcp:
-  llcp = [0,0,0,3,1,0,0,0,0,1] => llcp[m] == lcp(l, m)
-  rlcp = [0,1,1,1,2,0,2,0,0,0] => rlcp[m] == lcp(m, r)
+  llcp = [0,0,0,3,0,2,0,2,0,0] => llcp[m] == lcp(l, m)
+  rlcp = [0,1,1,1,0,0,0,0,0,0] => rlcp[m] == lcp(m, r)
 
 suffix_array_lpf:
   lengths = [0,0,1,3,2,1,0,2,1,0]
@@ -93,7 +93,7 @@ suffix_array_lpf:
 // suffix array via induced sorting
 // O(n)
 inline auto suffix_array(array_view<uint8_t> input) -> vector<int> {
-  return induced_sort(input.data(), input.size());
+  return induced_sort(input);
 }
 
 // inverse
@@ -192,17 +192,17 @@ inline auto suffix_array_lrcp(vector<int>& llcp, vector<int>& rlcp, array_view<i
   rlcp.reset(), rlcp.reallocate(size + 1);
 
   function<int (int, int)> recurse = [&](int l, int r) -> int {
-    if(l == r - 1) {
-      if(r > size) return 0;
-      if(lcp) return lcp[r];
-      return plcp[sa[r]];
+    if(l >= r - 1) {
+      if(l >= size) return 0;
+      if(lcp) return lcp[l];
+      return plcp[sa[l]];
     }
     int m = l + r >> 1;
-    llcp[m] = recurse(l, m);
-    rlcp[m] = recurse(m, r);
-    return min(llcp[m], rlcp[m]);
+    llcp[m - 1] = recurse(l, m);
+    rlcp[m - 1] = recurse(m, r);
+    return min(llcp[m - 1], rlcp[m - 1]);
   };
-  recurse(0, size + 1);
+  recurse(1, size + 1);
 
   llcp[0] = 0;
   rlcp[0] = 0;
@@ -274,6 +274,8 @@ inline auto suffix_array_find(int& length, int& offset, array_view<int> sa, arra
       if(k == match.size()) return true;
     }
 
+    if(k == match.size() || s + k == input.size()) k--;
+
     if(match[k] < input[s + k]) {
       r = m;
     } else {
@@ -304,6 +306,8 @@ inline auto suffix_array_find(int& length, int& offset, array_view<int> llcp, ar
       if(k == match.size()) return true;
     }
 
+    if(k == match.size() || s + k == input.size()) k--;
+
     if(match[k] < input[s + k]) {
       r = m;
       k = min(k, llcp[m]);