use oss-nvjpg for loading jpeg images (homebrew, games and themezer).

slightly faster loading on avg compared to stbi.

README.md

@@ -4,7 +4,7 @@ A homebrew menu for the Nintendo Switch.
 
 [See the GBATemp thread for more details / discussion](https://gbatemp.net/threads/sphaira-hbmenu-replacement.664523/).
 
-[We have now have a Discord server!](https://discord.gg/8vZBsrprEc). Please use the issues tab to report bugs, as it is much easier for me to track.
+[We have now have a Discord server!](https://discord.gg/8vZBsrprEc) Please use the issues tab to report bugs, as it is much easier for me to track.
 
 ## Showcase
 
@@ -86,16 +86,21 @@ The output will be found in `build/MinSizeRel/sphaira.nro`
 
 ## Credits
 
-- borealis
+- [borealis](https://github.com/natinusala/borealis)
-- stb
+- [stb](https://github.com/nothings/stb)
-- yyjson
+- [yyjson](https://github.com/ibireme/yyjson)
-- nx-hbmenu
+- [nx-hbmenu](https://github.com/switchbrew/nx-hbmenu)
-- nx-hbloader
+- [nx-hbloader](https://github.com/switchbrew/nx-hbloader)
-- deko3d-nanovg
+- [deko3d-nanovg](https://github.com/Adubbz/nanovg-deko3d)
-- libpulsar
+- [libpulsar](https://github.com/p-sam/switch-libpulsar)
-- minIni
+- [minIni](https://github.com/compuphase/minIni)
-- GBATemp
+- [GBATemp](https://gbatemp.net/threads/sphaira-hbmenu-replacement.664523/)
-- hb-appstore
+- [hb-appstore](https://github.com/fortheusers/hb-appstore)
-- haze
+- [haze](https://github.com/Atmosphere-NX/Atmosphere/tree/master/troposphere/haze)
-- nxdumptool (for gamecard bin dumping and rsa verify code)
+- [nxdumptool](https://github.com/DarkMatterCore/nxdumptool) (for gamecard bin dumping and rsa verify code)
+- [libusbhsfs](https://github.com/DarkMatterCore/libusbhsfs)
+- [libnxtc](https://github.com/DarkMatterCore/libnxtc)
+- [oss-nvjpg](https://github.com/averne/oss-nvjpg)
+- [nsz](https://github.com/nicoboss/nsz)
+- [themezer](https://themezer.net/)
 - Everyone who has contributed to this project!

sphaira/CMakeLists.txt

@@ -202,6 +202,11 @@ FetchContent_Declare(libnxtc
     GIT_TAG v0.0.2
 )
 
+FetchContent_Declare(nvjpg
+    GIT_REPOSITORY https://github.com/averne/oss-nvjpg.git
+    GIT_TAG fdcaba8
+)
+
 set(USE_NEW_ZSTD ON)
 
 set(ZSTD_BUILD_STATIC ON)
@@ -255,6 +260,7 @@ FetchContent_MakeAvailable(
     zstd
     libusbhsfs
     libnxtc
+    nvjpg
 )
 
 set(FTPSRV_LIB_BUILD TRUE)
@@ -304,6 +310,14 @@ add_library(libnxtc
 )
 target_include_directories(libnxtc PUBLIC ${libnxtc_SOURCE_DIR}/include)
 
+add_library(nvjpg
+    ${nvjpg_SOURCE_DIR}/lib/decoder.cpp
+    ${nvjpg_SOURCE_DIR}/lib/image.cpp
+    ${nvjpg_SOURCE_DIR}/lib/surface.cpp
+)
+target_include_directories(nvjpg PUBLIC ${nvjpg_SOURCE_DIR}/include)
+set_target_properties(nvjpg PROPERTIES CXX_STANDARD 26)
+
 find_package(ZLIB REQUIRED)
 find_library(minizip_lib minizip REQUIRED)
 find_path(minizip_inc minizip REQUIRED)
@@ -334,6 +348,7 @@ target_link_libraries(sphaira PRIVATE
     yyjson
     # libusbhsfs
     libnxtc
+    nvjpg
 
     ${minizip_lib}
     ZLIB::ZLIB

sphaira/include/app.hpp

@@ -10,6 +10,7 @@
 #include "fs.hpp"
 #include "log.hpp"
 
+#include <nvjpg.hpp>
 #include <switch.h>
 #include <vector>
 #include <string>
@@ -272,6 +273,8 @@ public:
 
     PLSR_PlayerSoundId m_sound_ids[SoundEffect_MAX]{};
 
+    nj::Decoder m_decoder;
+
 private: // from nanovg decko3d example by adubbz
     static constexpr unsigned NumFramebuffers = 2;
     static constexpr unsigned StaticCmdSize = 0x1000;

sphaira/include/image.hpp

@@ -12,8 +12,14 @@ struct ImageResult {
     int w, h;
 };
 
-auto ImageLoadFromMemory(std::span<const u8> data) -> ImageResult;
+enum ImageFlag {
-auto ImageLoadFromFile(const fs::FsPath& file) -> ImageResult;
+    ImageFlag_None = 0,
+    // set this if the image is a jpeg, will use oss-nvjpg to load.
+    ImageFlag_JPEG = 1 << 0,
+};
+
+auto ImageLoadFromMemory(std::span<const u8> data, u32 flags = ImageFlag_None) -> ImageResult;
+auto ImageLoadFromFile(const fs::FsPath& file, u32 flags = ImageFlag_None) -> ImageResult;
 auto ImageResize(std::span<const u8> data, int inx, int iny, int outx, int outy) -> ImageResult;
 auto ImageConvertToJpg(std::span<const u8> data, int x, int y) -> ImageResult;
 

sphaira/source/app.cpp

@@ -1400,6 +1400,10 @@ App::App(const char* argv0) {
 
     curl::Init();
 
+    // this has to be init before deko3d.
+    nj::initialize();
+    m_decoder.initialize();
+
     // get current size of the framebuffer
     const auto fb = GetFrameBufferSize();
     s_width = fb.size.x;
@@ -1849,6 +1853,7 @@ App::~App() {
     ON_SCOPE_EXIT(appletSetCpuBoostMode(ApmCpuBoostMode_Normal));
 
     log_write("starting to exit\n");
+    TimeStamp ts;
 
     i18n::exit();
     curl::Exit();
@@ -1878,6 +1883,9 @@ App::~App() {
     nvgDeleteDk(this->vg);
     this->renderer.reset();
 
+    m_decoder.finalize();
+    nj::finalize();
+
     // backup hbmenu if it is not sphaira
     if (App::GetReplaceHbmenuEnable() && !IsHbmenu()) {
         NacpStruct hbmenu_nacp;
@@ -1950,6 +1958,8 @@ App::~App() {
         usbHsFsExit();
     }
 
+    log_write("\t[EXIT] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
+
     if (App::GetLogEnable()) {
         log_write("closing log\n");
         log_file_exit();

sphaira/source/image.cpp

@@ -20,7 +20,9 @@
 #pragma GCC diagnostic pop
 #pragma GCC diagnostic pop
 
+#include "app.hpp"
 #include "log.hpp"
+#include <nvjpg.hpp>
 #include <cstring>
 
 namespace sphaira {
@@ -30,7 +32,6 @@ constexpr int BPP = 4;
 
 auto ImageLoadInternal(stbi_uc* image_data, int x, int y) -> ImageResult {
     if (image_data) {
-        log_write("loaded image: w: %d h: %d\n", x, y);
         ImageResult result{};
         result.data.resize(x*y*BPP);
         result.w = x;
@@ -44,18 +45,64 @@ auto ImageLoadInternal(stbi_uc* image_data, int x, int y) -> ImageResult {
     return {};
 }
 
+auto ImageLoadInternal(nj::Image&& image) -> ImageResult {
+    if (!image.is_valid() || image.parse()) {
+        log_write("[NVJPG] failed to parse image\n");
+        return {};
+    }
+
+    nj::Surface surf{image.width, image.height};
+    if (surf.allocate()) {
+        log_write("[NVJPG] failed to allocate surf\n");
+        return {};
+    }
+
+    if (R_FAILED(App::GetApp()->m_decoder.render(image, surf, 255))) {
+        log_write("[NVJPG] failed to render\n");
+        return {};
+    }
+
+    if (R_FAILED(App::GetApp()->m_decoder.wait(surf))) {
+        log_write("[NVJPG] failed to wait\n");
+        return {};
+    }
+
+    ImageResult result{};
+    result.w = image.width;
+    result.h = image.height;
+    result.data.resize(surf.size());
+    std::memcpy(result.data.data(), surf.data(), result.data.size());
+
+    return result;
+}
+
 } // namespace
 
-auto ImageLoadFromMemory(std::span<const u8> data) -> ImageResult {
+auto ImageLoadFromMemory(std::span<const u8> data, u32 flags) -> ImageResult {
+    if (flags & ImageFlag_JPEG) {
+        auto shared_vec = std::make_shared<std::vector<u8>>(data.size());
+        std::memcpy(shared_vec->data(), data.data(), shared_vec->size());
+        // don't make const as it prevents RTO.
+        auto result = ImageLoadInternal(nj::Image{shared_vec});
+        // if it failed, try again but without using oss-jpg.
+        return result.data.empty() ? ImageLoadFromMemory(data, 0) : result;
+    } else {
         int x, y, channels;
         return ImageLoadInternal(stbi_load_from_memory(data.data(), data.size(), &x, &y, &channels, BPP), x, y);
     }
+}
 
-auto ImageLoadFromFile(const fs::FsPath& file) -> ImageResult {
+auto ImageLoadFromFile(const fs::FsPath& file, u32 flags) -> ImageResult {
-    log_write("doing file load\n");
+    if (flags & ImageFlag_JPEG) {
+        // don't make const as it prevents RTO.
+        auto result = ImageLoadInternal(nj::Image{file});
+        // if it failed, try again but without using oss-jpg.
+        return result.data.empty() ? ImageLoadFromFile(file, 0) : result;
+    } else {
         int x, y, channels;
         return ImageLoadInternal(stbi_load(file, &x, &y, &channels, BPP), x, y);
     }
+}
 
 auto ImageResize(std::span<const u8> data, int inx, int iny, int outx, int outy) -> ImageResult {
     log_write("doing resize inx: %d iny: %d outx: %d outy: %d\n", inx, iny, outx, outy);

sphaira/source/ui/menus/game_menu.cpp

@@ -4,6 +4,7 @@
 #include "dumper.hpp"
 #include "defines.hpp"
 #include "i18n.hpp"
+#include "image.hpp"
 
 #include "ui/menus/game_menu.hpp"
 #include "ui/sidebar.hpp"
@@ -318,12 +319,16 @@ void FakeNacpEntry(ThreadResultData& e) {
 
 bool LoadControlImage(Entry& e) {
     if (!e.image && e.control) {
+        ON_SCOPE_EXIT(e.control.reset());
+
         TimeStamp ts;
-        e.image = nvgCreateImageMem(App::GetVg(), 0, e.control->icon, e.jpeg_size);
+        const auto image = ImageLoadFromMemory({e.control->icon, e.jpeg_size}, ImageFlag_JPEG);
-        e.control.reset();
+        if (!image.data.empty()) {
-        log_write("\t\t[image load] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
+            e.image = nvgCreateImageRGBA(App::GetVg(), image.w, image.h, 0, image.data.data());
+            log_write("\t[image load] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
             return true;
         }
+    }
 
     return false;
 }

sphaira/source/ui/menus/gc_menu.cpp

@@ -13,6 +13,7 @@
 #include "i18n.hpp"
 #include "download.hpp"
 #include "dumper.hpp"
+#include "image.hpp"
 
 #include <cstring>
 #include <algorithm>
@@ -965,7 +966,13 @@ void Menu::OnChangeIndex(s64 new_index) {
 
         const auto& e = m_entries[m_entry_index];
         const auto jpeg_size = e.control_size - sizeof(NacpStruct);
-        m_icon = nvgCreateImageMem(App::GetVg(), 0, e.control->icon, jpeg_size);
+
+        TimeStamp ts;
+        const auto image = ImageLoadFromMemory({e.control->icon, jpeg_size}, ImageFlag_JPEG);
+        if (!image.data.empty()) {
+            m_icon = nvgCreateImageRGBA(App::GetVg(), image.w, image.h, 0, image.data.data());
+            log_write("\t[image load] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
+        }
     }
 }
 

sphaira/source/ui/menus/homebrew.cpp

@@ -10,6 +10,7 @@
 #include "owo.hpp"
 #include "defines.hpp"
 #include "i18n.hpp"
+#include "image.hpp"
 
 #include <minIni.h>
 #include <utility>
@@ -175,9 +176,17 @@ void Menu::Draw(NVGcontext* vg, Theme* theme) {
                 // really, switch-tools should handle this by resizing the image before
                 // adding it to the nro, as well as validate its a valid jpeg.
                 const auto icon = nro_get_icon(e.path, e.icon_size, e.icon_offset);
+                TimeStamp ts;
                 if (!icon.empty()) {
-                    e.image = nvgCreateImageMem(vg, 0, icon.data(), icon.size());
+                    const auto image = ImageLoadFromMemory(icon, ImageFlag_JPEG);
+                    if (!image.data.empty()) {
+                        e.image = nvgCreateImageRGBA(vg, image.w, image.h, 0, image.data.data());
+                        log_write("\t[image load] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
                         image_load_count++;
+                    } else {
+                        // prevent loading of this icon again as it's already failed.
+                        e.icon_offset = e.icon_size = 0;
+                    }
                 }
             }
         }

sphaira/source/ui/menus/themezer.cpp

@@ -12,6 +12,7 @@
 #include "swkbd.hpp"
 #include "i18n.hpp"
 #include "threaded_file_transfer.hpp"
+#include "image.hpp"
 
 #include <minIni.h>
 #include <stb_image.h>
@@ -134,19 +135,14 @@ auto loadThemeImage(ThemeEntry& e) -> bool {
     }
     auto vg = App::GetVg();
 
-    fs::FsNativeSd fs;
-    std::vector<u8> image_buf;
-
     const auto path = apiBuildIconCache(e);
-    if (R_FAILED(fs.read_entire_file(path, image_buf))) {
+    TimeStamp ts;
-        log_write("failed to load image from file: %s\n", path.s);
+    const auto data = ImageLoadFromFile(path, ImageFlag_JPEG);
-    } else {
+    if (!data.data.empty()) {
-        int channels_in_file;
+        image.w = data.w;
-        auto buf = stbi_load_from_memory(image_buf.data(), image_buf.size(), &image.w, &image.h, &channels_in_file, 4);
+        image.h = data.h;
-        if (buf) {
+        image.image = nvgCreateImageRGBA(vg, data.w, data.h, 0, data.data.data());
-            ON_SCOPE_EXIT(stbi_image_free(buf));
+        log_write("\t[image load] time taken: %.2fs %zums\n", ts.GetSecondsD(), ts.GetMs());
-            image.image = nvgCreateImageRGBA(vg, image.w, image.h, 0, buf);
-        }
     }
 
     if (!image.image) {