New renderer: initial abstraction implementation #850

Vtec234 · 2017-05-22T11:13:19Z

A new rendering engine providing functionality designed for our needs. This is work continued from #287. Related: #34, #153, #286. Please submit contributions to the branch here.

EDIT: This branch is still WIP, but since the diff is getting rather large, it's best to merge it and continue working from there. The renderer.tests.renderer_demo works and displays some static objects.

heinezen · 2018-02-02T14:03:06Z

Pixel-perfect unit hitbox for unit selection and damage areas

There is an additional invisible selection layer for units in the Genie Engine. This layer is basically an invisible box around the centre of the unit sprite that roughly measures 2 * unit_width and 2 * unit_height. It makes moving units and those who are in a battle much easier to select. This selection layer is only present for moving units that are owned by the player who is making a selection as well as deer and relics. Everything else (all buildings, enemy/friendly units, trees, gold, stone) is selected with the pixel-perfect hitbox.

So the order of preference would be:

pixel-perfect selection (own units)
"tolerance" selection layer (own units)
pixel-perfect selection (everything else)

Vtec234 · 2018-06-21T22:28:06Z

I've now pushed a commit fixing all code compliance issues (copyright dates, mostly). And since Kevin will be down for a bit, I've tested what I could. At least make checkall passes with no errors on my x86_64 Arch Linux and the renderer.tests.renderer_demo works, so this should be mergeable as the first step towards the new renderer. @TheJJ could you please take a look at the code?

TheJJ

that was about half the lines, more to come :)
if you want, i can fix some of the things myself.

TheJJ · 2018-06-19T11:35:58Z

libopenage/renderer/README.md

@@ -0,0 +1,149 @@
+# Openage graphics


can we move this file to doc/code/renderer/README.md? we didn't mix code and documentation so far.

TheJJ · 2018-06-19T11:37:31Z

libopenage/renderer/README.md

+
+__TODO name__
+`openage::graphics` - the level 2 system
+


And level 3 is the "presenter", which dispatches events to gui, sound, graphics and receives input.

It's uncertain yet whether there would be level 2 and 3 or whether we can just have the game rendering code and general presentation code in one unit, which would just be level 2. I think it depends on how much code there will be, so that if it's little we can have one unit, but if there's a lot two levels would be better.

TheJJ · 2018-06-19T11:42:43Z

libopenage/renderer/opengl/CMakeLists.txt

+  texture_array.cpp
+  vertex_array.cpp
+  window.cpp
+)


cmake should be indented with tabs

TheJJ · 2018-06-19T12:25:19Z

libopenage/renderer/opengl/lookup.h

+  std::make_pair(resources::index_t::U8, GL_UNSIGNED_BYTE),
+  std::make_pair(resources::index_t::U16, GL_UNSIGNED_SHORT),
+  std::make_pair(resources::index_t::U32, GL_UNSIGNED_INT)
+);


indent with tabs pls, all the above indents are 2 spaces.

TheJJ · 2018-06-19T12:25:38Z

libopenage/renderer/opengl/render_target.h

+	gl_render_target_t type;
+
+	/// For textures target type, the framebuffer.
+	std::experimental::optional<GlFramebuffer> framebuffer;


no need for experimental any more

TheJJ · 2018-06-26T16:31:18Z

libopenage/renderer/opengl/vertex_array.h

+	///
+	/// A shader input mapping is only allowed when there is a single element in `buffers`. In such a case,
+	/// the vertex inputs are paired with VAO attributes according to the mapping instead of in ascending order.
+	GlVertexArray(std::vector<std::pair<GlBuffer const&, resources::VertexInputInfo const&>> buffers);


const & for the buffers-vector?

It's just a vector of pointers, so I think passing by value is fine.

TheJJ · 2018-06-26T16:33:28Z

libopenage/renderer/opengl/window.cpp

+
+	// We need HIGHDPI for eventual support of GUI scaling.
+	// TODO HIGHDPI fails (requires newer SDL2?)
+	int32_t window_flags = SDL_WINDOW_OPENGL | SDL_WINDOW_RESIZABLE | SDL_WINDOW_MAXIMIZED;


SDL_WINDOW_ALLOW_HIGHDPI | window should be created in high-DPI mode if supported (>= SDL 2.0.1)

but I don't see what the highdpi mode would do. we'll scale the UI with qt, the window resolution should remain unaffected? (https://wiki.libsdl.org/SDL_GL_GetDrawableSize)
I think we can safely ignore highdpi stuff from SDL.

Unsure about this as I've never done high-DPI stuff. The docs seem to say that one can have a larger rendering space than the actual window dimensions with ALLOW_HIGHDPI.

TheJJ · 2018-06-26T16:49:23Z

libopenage/renderer/opengl/window.h

+
+	/// The window's OpenGL context. It's optional because it can't be constructed immediately,
+	/// but after the constructor runs it's guaranteed to be available.
+	std::experimental::optional<opengl::GlContext> context;


no more experimental::

TheJJ · 2018-06-29T12:10:29Z

libopenage/renderer/resources/CMakeLists.txt

+  mesh_data.cpp
+  shader_source.cpp
+  texture_data.cpp
+  texture_info.cpp


TheJJ · 2018-06-29T12:12:12Z

libopenage/renderer/resources/mesh_data.cpp

+	return this->data;
+}
+
+std::experimental::optional<std::vector<uint8_t>> const &MeshData::get_ids() const {


no more std::experimental

Vtec234 · 2018-06-30T14:27:08Z

The builds are failing because of missing Vulkan on the test machines.

TheJJ · 2018-07-08T18:46:10Z

libopenage/renderer/resources/texture_data.cpp

+	log::log(MSG(info) << "Saving texture data to " << file);
+
+	if (this->info.get_format() != pixel_format::rgba8) {
+		throw "unimplemented";


pls use our Error

TheJJ · 2018-07-08T18:47:50Z

libopenage/renderer/resources/texture_data.cpp

+	amask = 0xff000000;
+#endif
+
+	SDL_Surface *surf = SDL_CreateRGBSurfaceFrom(


nit: use a uniqueptr here so that if the write fails/exceptions occur before freesurface, the surface is still free'd

TheJJ · 2018-07-08T18:48:54Z

libopenage/renderer/resources/texture_info.cpp

+	std::make_pair(pixel_format::depth24, 3)
+);
+
+size_t pixel_size(pixel_format fmt) {


can be a constexpr function, that should allow more optimizations then

TheJJ · 2018-07-08T20:19:37Z

libopenage/renderer/tests.cpp

+			pmat << xScale, 0, 0, 0,
+							0, 1, 0, 0,
+							0, 0, 1, 0,
+							0, 0, 0, 1;


TheJJ · 2018-07-08T20:20:30Z

libopenage/renderer/tests.cpp

+			}
+			id--; //real id is id-1
+			log::log(INFO << "Object number " << id << " clicked.");
+			renderer->display_into_data().store(path / "/assets/screen.png");


we should store this to /tmp or somewhere else at a disposable location? otherwise the assets folder will be filled up.
I suspect this is a debugging leftover, though.

TheJJ · 2018-07-10T12:51:51Z

libopenage/renderer/tests.cpp

+#define TEST_DBG(txt)                  \
+	printf("before %s\n", txt);          \
+	window.get_context()->check_error(); \
+	printf("after %s\n", txt);


i think that macro is a leftover.

It's for debugging yes, but I'd just leave it because it comes in handy often when debugging the tests.

TheJJ · 2018-07-10T12:54:36Z

libopenage/renderer/texture_array.h

+  Texture2dArray(resources::Texture2dInfo);
+
+  /// Information about the size, format, etc. of every layer in this array.
+  resources::Texture2dInfo layer_info;


TheJJ · 2018-07-10T12:54:44Z

libopenage/renderer/vulkan/CMakeLists.txt

+  graphics_device.cpp
+  loader.cpp
+  renderer.cpp
+  windowvk.cpp


TheJJ · 2018-07-10T12:55:28Z

libopenage/renderer/vulkan/graphics_device.cpp

+			return {};
+		}
+
+		return details;


indent 1 tab too much

TheJJ · 2018-07-10T13:00:05Z

libopenage/renderer/vulkan/renderer.cpp

+
+	// TODO reinit swapchain on window resize
+
+	auto dir = std::make_shared<util::fslike::Directory>("/home/wojtek/Programming/C++/openage/");


😃
you already create the root directory correctly, you can simply pass it here for access to the shaders.

TheJJ · 2018-07-12T15:59:07Z

To make Vulkan optional, we can decide between a macro that disables the content of each file, or we don't even add the TU's to libopenage. We'd need the macro anyway in order to disable the codepaths that would select vulkan.

Co-authored-by: Erik Griese <erik.griese@student.hpi.uni-potsdam.de> Co-authored-by: surajhanchinal <surajhanchinal@gmail.com> Co-authored-by: Jonas Jelten <jj@sft.mx>

TheJJ added lang: c++ Done in C++ code nice new thing ☺ A new feature that was not there before area: renderer Concerns our graphics renderer labels May 22, 2017

TheJJ mentioned this pull request May 22, 2017

[WIP] New renderer #287

Closed

42 tasks

Vtec234 mentioned this pull request Jun 11, 2017

GUI subsystem refactor #855

Open

TheJJ mentioned this pull request Oct 1, 2017

Some HUD textures missing (red cross) when converting from AoE 2 HD #912

Open

Vtec234 mentioned this pull request Nov 13, 2017

rebase of new renderer #933

Closed

Vtec234 force-pushed the new-renderer branch 2 times, most recently from ae41129 to 23650fc Compare November 13, 2017 17:36

Vtec234 force-pushed the new-renderer branch from 16a2d34 to 69fc6ea Compare January 19, 2018 22:25

zuntrax added the big stuff High-impact changes, mainly foundation work label Jan 21, 2018

TheJJ mentioned this pull request Feb 16, 2018

Coord rewrite #975

Merged

zuntrax mentioned this pull request May 6, 2018

Render manager #34

Closed

Vtec234 force-pushed the new-renderer branch from ca15b1b to fff4b9b Compare June 14, 2018 22:08

Vtec234 changed the title ~~[WIP] New renderer, cont.~~ New renderer: initial abstraction implementation Jun 15, 2018

TheJJ mentioned this pull request Jun 22, 2018

This is the new WIP renderer #1022

Closed

TheJJ requested changes Jun 29, 2018

View reviewed changes

TheJJ mentioned this pull request Jul 2, 2018

Units (Villagers) being hidden by a shadow #712

Closed

TheJJ requested changes Jul 12, 2018

View reviewed changes

Vtec234 force-pushed the new-renderer branch 3 times, most recently from e05a6cb to ef622b7 Compare July 21, 2018 11:35

Vtec234 force-pushed the new-renderer branch from 0528c26 to cec2431 Compare July 22, 2018 13:24

Vtec234 and others added 3 commits July 22, 2018 19:03

renderer: level 1 implementation

e54cf15

Co-authored-by: Erik Griese <erik.griese@student.hpi.uni-potsdam.de> Co-authored-by: surajhanchinal <surajhanchinal@gmail.com> Co-authored-by: Jonas Jelten <jj@sft.mx>

renderer: code compliance

7fba029

renderer: remove experimental namespace, misc fixes

49d46c2

Vtec234 added 4 commits July 22, 2018 19:03

renderer: move doc, misc parameter fix

802558b

renderer: factor out 2D subtexture struct, misc fixes

b72316b

renderer: clang-tidy

5d280f7

cmake: fix renderer Vulkan-less build, check for Eigen

f0f6993

Vtec234 force-pushed the new-renderer branch from cec2431 to f0f6993 Compare July 22, 2018 17:14

TheJJ approved these changes Jul 22, 2018

View reviewed changes

TheJJ merged commit 93a17ce into SFTtech:master Jul 22, 2018

heinezen mentioned this pull request Jun 24, 2023

Roadmap to release 0.5.0 #1518

Closed

13 tasks


		// TODO reinit swapchain on window resize

		auto dir = std::make_shared<util::fslike::Directory>("/home/wojtek/Programming/C++/openage/");

Uh oh!

New renderer: initial abstraction implementation #850

New renderer: initial abstraction implementation #850

Uh oh!

Conversation

Vtec234 commented May 22, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

heinezen commented Feb 2, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Vtec234 commented Jun 21, 2018

Uh oh!

TheJJ left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Vtec234 commented Jun 30, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TheJJ commented Jul 12, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Vtec234 commented May 22, 2017 •

edited

Loading

heinezen commented Feb 2, 2018 •

edited

Loading