GZDoom 4.9.0 for ARM64 Windows (experimental)

[Graf Zahl]

It wasn't really an option when Randi originally built FluidSynth, but build containers do isolate and partially alleviate the mess.

No, not really. Just having to use MinGW constitutes "mess" in my book - IMO Windows code that cannot be compiled with MSVC is broken.

So today I don't think either approach really has an advantage in terms of how much work it is to revisit later, if anything Randi's approach may be more work now since now you have a delta over upstream even if the effort is somewhat marginal.

Let's hope it's still in a state where a "git pull" from upstream will do. If it doesn't, I'd concur. You can do all the things Randi described without screwing with the code.

[randi]

Okay, here's the repository. It's cleaned up some from what I did last month to try and isolate as many of the changes as possible out of the main FluidSynth code. The intent is that you should always be able to pull master from the original repository and rebase the native_win32 branch on top of it and have it work.

Rachael, while reworking this code, I found a mistake I made with one stub in the DLL I gave you earlier, so you should use this instead. I don't think it would have caused problems under any normal circumstances, though. Attached are new arm64 and x64 DLLs compiled from it. These are version 2.2.8 instead of whatever 1.1.x version I did before. The libfluidsynth64.dll is a drop-in replacement for the one shipping with GZDoom now. I'm quite astonished by how much smaller these are without any of the glib code statically linked in.

These DLLs were configured with "-Denable-static-msvcrt=on -Denable-dsound=off -Denable-wasapi=off -Denable-waveout=off -Denable-winmidi=off -Denable-network=off -Denable-aufile=off", so they can't output any audio directly, talk over the network, or receive MIDI events from MIDI devices, but they're perfectly suitable for ZMusic, since it doesn't need any of those features anyway. The static CRT is to keep up the tradition of not using the DLL version of the CRT.

I used vcpkg to get libsndfile support, so these DLLs also support SF3 soundfonts. However, vcpkg's libsndfile is named sndfile.dll, and GZDoom uses libsndfile-1.dll, so I used this genlib batch script to create a new import .lib from libsndfile-1.dll and overwrote the sndfile.lib installed by vcpkg so that the FluidSynth DLL would import from libsndfile-1.dll instead of sndfile.dll. (Yes, it's messy, but it worked well enough for my needs at the moment.)

Gentables is not working with cross-compilation. I don't know why and didn't feel like trying too hard to figure it out, so to do the arm64 compile, I temporarily commented out the end of src/CMakeLists.txt to not bother with it and copied the generated files over from the x64 build instead.

[Rachael]

Thank you - I will take a look tomorrow when I get a chance. I'll recompile a new GZDoom 4.8.2 and drop this .dll in to see how it works.

[Graf Zahl]

I used vcpkg to get libsndfile support, so these DLLs also support SF3 soundfonts. However, vcpkg's libsndfile is named sndfile.dll, and GZDoom uses libsndfile-1.dll, so I used this genlib batch script to create a new import .lib from libsndfile-1.dll and overwrote the sndfile.lib installed by vcpkg so that the FluidSynth DLL would import from libsndfile-1.dll instead of sndfile.dll. (Yes, it's messy, but it worked well enough for my needs at the moment.)

I think it's better then to rename the DLL to use the canonical name of the parent project. I don't know when it happened but some time back they must have changed it to be more "Windows conformant".

Also, shouldn't this repo be on the ZDoom organization account?

[randi]

I think it's better then to rename the DLL to use the canonical name of the parent project. I don't know when it happened but some time back they must have changed it to be more "Windows conformant".

I don't know if that's an official libsndfile name for the DLL or a vcpkg name for it. I used vcpkg to find it because it required minimal modifications to the CMakeLists.txt to do so and because I already had vcpkg installed, but I don't know enough about the vcpkg ecosystem to know if it might have changed the DLL name itself.

FWIW, this also produced a libfluidsynth-3.dll before I renamed it.

I've also pushed an update that fixes a potential race condition when creating detached threads. These are only used by the fluidsynth command line program and not the library, so no new DLLs.

[Graf Zahl]

The name of the last binary package was sndfile.dll as well, if I remember correctly. I had to rename the DLL before dropping it into my play folder.

[Rachael]

I am not really sure what the error means, but for what it's worth the sound does actually work - but only for Timidity++

This was tested with Windows 11 on UTM on my M1 Mac.

Code: Select all

Could not load fluidsynth.dll or libfluidsynth.dll

Unable to create FluidSynth MIDI device. Falling back to Timidity++

---

Code: Select all

Log started: 2022-07-26 18:17:28
GZDoom version g4.8.2
M_LoadDefaults: Load system defaults.
W_Init: Init WADfiles.
 adding C:/gzdoom/gzdoom.pk3, 668 lumps
 adding C:/gzdoom/game_support.pk3, 2514 lumps
 adding ./doom_complete.pk3, 3278 lumps
 adding C:/gzdoom/lights.pk3, 7 lumps
 adding C:/gzdoom/brightmaps.pk3, 499 lumps
 adding C:/gzdoom/game_widescreen_gfx.pk3, 214 lumps
S_Init: Setting up sound.
I_InitSound: Initializing OpenAL
  Opened device OpenAL Soft on Remote Audio
  EFX enabled
I_Init: Setting up machine state.
CPU speed: 1000 MHz
V_Init: allocate screen.
ST_Init: Init startup screen.
Checking cmd-line parameters...
S_InitData: Load sound definitions.
G_ParseMapInfo: Load map definitions.
Texman.Init: Init texture manager.
ParseTeamInfo: Load team definitions.
LoadActors: Load actor definitions.
script parsing took 159.79 ms
R_Init: Init Doom refresh subsystem.
DecalLibrary: Load decals.
M_Init: Init menus.
P_Init: Init Playloop state.
ParseSBarInfo: Loading custom status bar definition.
D_CheckNetGame: Checking network game status.
player 1 of 1 (1 nodes)
I_InitInput
I_StartupMouse
I_StartupKeyboard
I_StartupXInput
I_StartupRawPS2
I_StartupDirectInputJoystick
Initialization of Vulkan failed: Unable to find Vulkan
GL_VENDOR: Microsoft Corporation
GL_RENDERER: D3D12 (Microsoft Basic Render Driver)
GL_VERSION: 4.2 (Compatibility Profile) Mesa 22.2.0-devel (git-7cbdf9e261) (Compatibility profile)
GL_SHADING_LANGUAGE_VERSION: 4.20
GL_EXTENSIONS: GL_AMD_conservative_depth GL_AMD_draw_buffers_blend GL_AMD_multi_draw_indirect GL_AMD_shader_stencil_export GL_AMD_shader_trinary_minmax GL_AMD_texture_texture4 GL_ANGLE_texture_compression_dxt3 GL_ANGLE_texture_compression_dxt5 GL_APPLE_packed_pixels GL_ARB_ES2_compatibility GL_ARB_ES3_1_compatibility GL_ARB_ES3_compatibility GL_ARB_arrays_of_arrays GL_ARB_base_instance GL_ARB_blend_func_extended GL_ARB_clear_buffer_object GL_ARB_color_buffer_float GL_ARB_compatibility GL_ARB_compressed_texture_pixel_storage GL_ARB_compute_shader GL_ARB_compute_variable_group_size GL_ARB_conditional_render_inverted GL_ARB_conservative_depth GL_ARB_copy_buffer GL_ARB_debug_output GL_ARB_depth_buffer_float GL_ARB_depth_clamp GL_ARB_depth_texture GL_ARB_direct_state_access GL_ARB_draw_buffers GL_ARB_draw_buffers_blend GL_ARB_draw_elements_base_vertex GL_ARB_draw_indirect GL_ARB_draw_instanced GL_ARB_explicit_attrib_location GL_ARB_explicit_uniform_location GL_ARB_fragment_coord_conventions GL_ARB_fragment_layer_viewport GL_ARB_fragment_program GL_ARB_fragment_program_shadow GL_ARB_fragment_shader GL_ARB_framebuffer_no_attachments GL_ARB_framebuffer_object GL_ARB_framebuffer_sRGB GL_ARB_get_program_binary GL_ARB_get_texture_sub_image GL_ARB_gpu_shader5 GL_ARB_gpu_shader_fp64 GL_ARB_gpu_shader_int64 GL_ARB_half_float_pixel GL_ARB_half_float_vertex GL_ARB_indirect_parameters GL_ARB_instanced_arrays GL_ARB_internalformat_query GL_ARB_internalformat_query2 GL_ARB_invalidate_subdata GL_ARB_map_buffer_alignment GL_ARB_map_buffer_range GL_ARB_multi_bind GL_ARB_multi_draw_indirect GL_ARB_multisample GL_ARB_multitexture GL_ARB_occlusion_query GL_ARB_occlusion_query2 GL_ARB_parallel_shader_compile GL_ARB_pixel_buffer_object GL_ARB_point_parameters GL_ARB_point_sprite GL_ARB_program_interface_query GL_ARB_provoking_vertex GL_ARB_robust_buffer_access_behavior GL_ARB_robustness GL_ARB_sample_shading GL_ARB_sampler_objects GL_ARB_seamless_cube_map GL_ARB_separate_shader_objects GL_ARB_shader_atomic_counter_ops GL_ARB_shader_atomic_counters GL_ARB_shader_bit_encoding GL_ARB_shader_draw_parameters GL_ARB_shader_image_load_store GL_ARB_shader_image_size GL_ARB_shader_objects GL_ARB_shader_precision GL_ARB_shader_stencil_export GL_ARB_shader_storage_buffer_object GL_ARB_shader_subroutine GL_ARB_shader_texture_lod GL_ARB_shading_language_100 GL_ARB_shading_language_420pack GL_ARB_shading_language_include GL_ARB_shading_language_packing GL_ARB_shadow GL_ARB_stencil_texturing GL_ARB_sync GL_ARB_tessellation_shader GL_ARB_texture_border_clamp GL_ARB_texture_buffer_object GL_ARB_texture_buffer_object_rgb32 GL_ARB_texture_buffer_range GL_ARB_texture_compression GL_ARB_texture_compression_bptc GL_ARB_texture_compression_rgtc GL_ARB_texture_cube_map GL_ARB_texture_cube_map_array GL_ARB_texture_env_add GL_ARB_texture_env_combine GL_ARB_texture_env_crossbar GL_ARB_texture_env_dot3 GL_ARB_texture_filter_anisotropic GL_ARB_texture_float GL_ARB_texture_gather GL_ARB_texture_mirrored_repeat GL_ARB_texture_multisample GL_ARB_texture_non_power_of_two GL_ARB_texture_query_levels GL_ARB_texture_query_lod GL_ARB_texture_rectangle GL_ARB_texture_rg GL_ARB_texture_rgb10_a2ui GL_ARB_texture_stencil8 GL_ARB_texture_storage GL_ARB_texture_storage_multisample GL_ARB_texture_swizzle GL_ARB_timer_query GL_ARB_transform_feedback2 GL_ARB_transform_feedback3 GL_ARB_transform_feedback_instanced GL_ARB_transpose_matrix GL_ARB_uniform_buffer_object GL_ARB_vertex_array_bgra GL_ARB_vertex_array_object GL_ARB_vertex_attrib_64bit GL_ARB_vertex_attrib_binding GL_ARB_vertex_buffer_object GL_ARB_vertex_program GL_ARB_vertex_shader GL_ARB_vertex_type_10f_11f_11f_rev GL_ARB_vertex_type_2_10_10_10_rev GL_ARB_viewport_array GL_ARB_window_pos GL_ATI_blend_equation_separate GL_ATI_draw_buffers GL_ATI_fragment_shader GL_ATI_separate_stencil GL_ATI_texture_env_combine3 GL_ATI_texture_float GL_EXT_EGL_image_storage GL_EXT_EGL_sync GL_EXT_abgr GL_EXT_bgra GL_EXT_blend_color GL_EXT_blend_equation_separate GL_EXT_blend_func_separate GL_EXT_blend_minmax GL_EXT_blend_subtract GL_EXT_compiled_vertex_array GL_EXT_copy_texture GL_EXT_direct_state_access GL_EXT_draw_buffers2 GL_EXT_draw_instanced GL_EXT_draw_range_elements GL_EXT_fog_coord GL_EXT_framebuffer_blit GL_EXT_framebuffer_multisample GL_EXT_framebuffer_multisample_blit_scaled GL_EXT_framebuffer_object GL_EXT_framebuffer_sRGB GL_EXT_gpu_program_parameters GL_EXT_gpu_shader4 GL_EXT_multi_draw_arrays GL_EXT_packed_depth_stencil GL_EXT_packed_float GL_EXT_packed_pixels GL_EXT_pixel_buffer_object GL_EXT_point_parameters GL_EXT_provoking_vertex GL_EXT_rescale_normal GL_EXT_secondary_color GL_EXT_separate_specular_color GL_EXT_shader_integer_mix GL_EXT_shadow_funcs GL_EXT_stencil_two_side GL_EXT_stencil_wrap GL_EXT_subtexture GL_EXT_texture GL_EXT_texture3D GL_EXT_texture_array GL_EXT_texture_buffer_object GL_EXT_texture_compression_dxt1 GL_EXT_texture_compression_rgtc GL_EXT_texture_compression_s3tc GL_EXT_texture_cube_map GL_EXT_texture_edge_clamp GL_EXT_texture_env_add GL_EXT_texture_env_combine GL_EXT_texture_env_dot3 GL_EXT_texture_filter_anisotropic GL_EXT_texture_integer GL_EXT_texture_lod_bias GL_EXT_texture_object GL_EXT_texture_rectangle GL_EXT_texture_sRGB GL_EXT_texture_sRGB_decode GL_EXT_texture_shared_exponent GL_EXT_texture_snorm GL_EXT_texture_swizzle GL_EXT_timer_query GL_EXT_transform_feedback GL_EXT_vertex_array GL_EXT_vertex_array_bgra GL_EXT_vertex_attrib_64bit GL_IBM_multimode_draw_arrays GL_IBM_rasterpos_clip GL_IBM_texture_mirrored_repeat GL_INGR_blend_func_separate GL_KHR_context_flush_control GL_KHR_debug GL_KHR_no_error GL_KHR_parallel_shader_compile GL_KHR_robust_buffer_access_behavior GL_KHR_robustness GL_KHR_texture_compression_astc_ldr GL_KHR_texture_compression_astc_sliced_3d GL_MESA_pack_invert GL_MESA_shader_integer_functions GL_MESA_texture_signed_rgba GL_MESA_window_pos GL_NV_blend_square GL_NV_conditional_render GL_NV_copy_depth_to_color GL_NV_copy_image GL_NV_depth_clamp GL_NV_fog_distance GL_NV_half_float GL_NV_light_max_exponent GL_NV_packed_depth_stencil GL_NV_primitive_restart GL_NV_texgen_reflection GL_NV_texture_env_combine4 GL_NV_texture_rectangle GL_OES_EGL_image GL_OES_read_format GL_S3_s3tc GL_SGIS_generate_mipmap GL_SGIS_texture_border_clamp GL_SGIS_texture_edge_clamp GL_SGIS_texture_lod GL_SUN_multi_draw_arrays
Max. texture size: 16384
Max. texture units: 32
Max. varying: 124
Max. uniform block size: 65536
Uniform block alignment: 256
Resolution: 790 x 400
Could not load fluidsynth.dll or libfluidsynth.dll

Unable to create FluidSynth MIDI device. Falling back to Timidity++

]quit

Full compile

[randi]

Could not load fluidsynth.dll or libfluidsynth.dll

That usually means there was a dependency problem, but it only needs libsndfile-1.dll and kernel32.dll, and you provide the former. Opening it in Dependencies doesn't show anything missing either, unless there's something missing in arm64's kernel32.dll compared to x64, but that seems unlikely.

I didn't do anything different to build this one compared to the first one I gave you, so I don't know what could be causing the problem. Maybe try snooping on it with Process Monitor? It's hard to say without real hardware in front of me.

The name of the last binary package was sndfile.dll as well, if I remember correctly.

In that case, here's rebuilds linking against sndfile.dll instead of libsndfile-1.dll if you want to start using sndfile.dll instead. Rachael, this should also be drop-in compatible if you make a copy of libsndfile-1.dll and rename it sndfile.dll. It's worth a shot, at least.

[Rachael]

I used Dependency Walker (surprisingly it works with ARM64 compiles, but only if run on an ARM64 computer)

Here is a screenshot of the output

I had copies of all of the missing libraries, and dropped them in, and it works, but now the zip file is a mess.

Here is the compile as it is now

I still haven't tested it on real hardware yet (the GalaxyBook) but it works on the UTM Virtual Machine now, at least. I'm gonna build Raze and test it as well.

Built Raze and it works well.

Here is that

[Graf Zahl]

What libsndfile did you use? From the looks of it isn't a self-contained version but one that dynamically links everything it needs.

[Rachael]

It's one compiled from vcpkg. I cannot seem to compile a version that has all the needed dependencies pre-compiled in.

[randi]

Using this custom triplet, which I named arm64-windows-mixed.cmake

Code: Select all

set(VCPKG_TARGET_ARCHITECTURE arm64)

set(VCPKG_CRT_LINKAGE static)

if (${PORT} MATCHES "libflac|libvorbis|opus|mp3lame|mpg123|libogg")
	set(VCPKG_LIBRARY_LINKAGE static)
else()
	set(VCPKG_LIBRARY_LINKAGE dynamic)
endif()

And running the command

Code: Select all

vcpkg install libsndfile:arm64-windows-mixed

gave me this DLL, which should be exactly what you're looking for.

It builds everything with static CRT linkage, libsndfile's dependences are built as static libraries, and everything else is built as a dynamic library.

[Rachael]

Thank you, once more! That greatly simplifies the distribution.

[Graf Zahl]

Cool. Would this also be a solution for compiling libvpx in a form we can use? I'd really like to get rid of that library version from EDuke32.

One thing in general here: How should we handle adding binary lib files to the project? I'd rather not see them added to the main repo because it can inflate that quite quickly. One other thing I'd like to add is webp support, but that's also quite a large library file which had to be added twice then, too.

[_mental_]

I would put all binary dependencies to an archive, and use a directory (added to .gitignore) to discover headers/libraries extracted from the given archive automatically. It's possible to automate downloading and extraction from such archive via CMake.