mirror/godot
2
0
Fork 0
mirror of https://github.com/godotengine/godot.git synced 2024-12-21 10:25:24 +08:00
Code Issues Packages Projects Releases Wiki Activity
7c015a768f
godot/drivers/metal/metal_objects.h
Matias N. Goldberg c77cbf096b Improvements from TheForge (see description) 
The work was performed by collaboration of TheForge and Google. I am
merely splitting it up into smaller PRs and cleaning it up.

This is the most "risky" PR so far because the previous ones have been
miscellaneous stuff aimed at either [improve
debugging](https://github.com/godotengine/godot/pull/90993) (e.g. device
lost), [improve Android
experience](https://github.com/godotengine/godot/pull/96439) (add Swappy
for better Frame Pacing + Pre-Transformed Swapchains for slightly better
performance), or harmless [ASTC
improvements](https://github.com/godotengine/godot/pull/96045) (better
performance by simply toggling a feature when available).

However this PR contains larger modifications aimed at improving
performance or reducing memory fragmentation. With greater
modifications, come greater risks of bugs or breakage.

Changes introduced by this PR:

TBDR GPUs (e.g. most of Android + iOS + M1 Apple) support rendering to
Render Targets that are not backed by actual GPU memory (everything
stays in cache). This works as long as load action isn't `LOAD`, and
store action must be `DONT_CARE`. This saves VRAM (it also makes
painfully obvious when a mistake introduces a performance regression).
Of particular usefulness is when doing MSAA and keeping the raw MSAA
content is not necessary.

Some GPUs get faster when the sampler settings are hard-coded into the
GLSL shaders (instead of being dynamically bound at runtime). This
required changes to the GLSL shaders, PSO creation routines, Descriptor
creation routines, and Descriptor binding routines.

 - `bool immutable_samplers_enabled = true`

Setting it to false enforces the old behavior. Useful for debugging bugs
and regressions.

Immutable samplers requires that the samplers stay... immutable, hence
this boolean is useful if the promise gets broken. We might want to turn
this into a `GLOBAL_DEF` setting.

Instead of creating dozen/hundreds/thousands of `VkDescriptorSet` every
frame that need to be freed individually when they are no longer needed,
they all get freed at once by resetting the whole pool. Once the whole
pool is no longer in use by the GPU, it gets reset and its memory
recycled. Descriptor sets that are created to be kept around for longer
or forever (i.e. not created and freed within the same frame) **must
not** use linear pools. There may be more than one pool per frame. How
many pools per frame Godot ends up with depends on its capacity, and
that is controlled by
`rendering/rendering_device/vulkan/max_descriptors_per_pool`.

- **Possible improvement for later:** It should be possible for Godot
to adapt to how many descriptors per pool are needed on a per-key basis
(i.e. grow their capacity like `std::vector` does) after rendering a few
frames; which would be better than the current solution of having a
single global value for all pools (`max_descriptors_per_pool`) that the
user needs to tweak.

 - `bool linear_descriptor_pools_enabled = true`

Setting it to false enforces the old behavior. Useful for debugging bugs
and regressions.
Setting it to false is required when workarounding driver bugs (e.g.
Adreno 730).

A ridiculous optimization. Ridiculous because the original code
should've done this in the first place. Previously Godot was doing the
following:

  1. Create a command buffer **pool**. One per frame.
  2. Create multiple command buffers from the pool in point 1.
3. Call `vkBeginCommandBuffer` on the cmd buffer in point 2. This
resets the cmd buffer because Godot requests the
`VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT` flag.
  4. Add commands to the cmd buffers from point 2.
  5. Submit those commands.
6. On frame N + 2, recycle the buffer pool and cmd buffers from pt 1 &
2, and repeat from step 3.

The problem here is that step 3 resets each command buffer individually.
Initially Godot used to have 1 cmd buffer per pool, thus the impact is
very low.

But not anymore (specially with Adreno workarounds to force splitting
compute dispatches into a new cmd buffer, more on this later). However
Godot keeps around a very low amount of command buffers per frame.

The recommended method is to reset the whole pool, to reset all cmd
buffers at once. Hence the new steps would be:

  1. Create a command buffer **pool**. One per frame.
  2. Create multiple command buffers from the pool in point 1.
3. Call `vkBeginCommandBuffer` on the cmd buffer in point 2, which is
already reset/empty (see step 6).
  4. Add commands to the cmd buffers from point 2.
  5. Submit those commands.
6. On frame N + 2, recycle the buffer pool and cmd buffers from pt 1 &
2, call `vkResetCommandPool` and repeat from step 3.

**Possible issues:** @dariosamo added `transfer_worker` which creates a
command buffer pool:

```cpp
transfer_worker->command_pool =
driver->command_pool_create(transfer_queue_family,
RDD::COMMAND_BUFFER_TYPE_PRIMARY);
```

As expected, validation was complaining that command buffers were being
reused without being reset (that's good, we now know Validation Layers
will warn us of wrong use).
I fixed it by adding:

```cpp
void RenderingDevice::_wait_for_transfer_worker(TransferWorker
*p_transfer_worker) {
	driver->fence_wait(p_transfer_worker->command_fence);
	driver->command_pool_reset(p_transfer_worker->command_pool); //
! New line !
```

**Secondary cmd buffers are subject to the same issue but I didn't alter
them. I talked this with Dario and he is aware of this.**
Secondary cmd buffers are currently disabled due to other issues (it's
disabled on master).

 - `bool RenderingDeviceCommons::command_pool_reset_enabled`

Setting it to false enforces the old behavior. Useful for debugging bugs
and regressions.

There's no other reason for this boolean. Possibly once it becomes well
tested, the boolean could be removed entirely.

Adds `command_bind_render_uniform_sets` and
`add_draw_list_bind_uniform_sets` (+ compute variants).

It performs the same as `add_draw_list_bind_uniform_set` (notice
singular vs plural), but on multiple consecutive uniform sets, thus
reducing graph and draw call overhead.

 - `bool descriptor_set_batching = true;`

Setting it to false enforces the old behavior. Useful for debugging bugs
and regressions.

There's no other reason for this boolean. Possibly once it becomes well
tested, the boolean could be removed entirely.

Godot currently does the following:

 1. Fill the entire cmd buffer with commands.
 2. `submit()`
    - Wait with a semaphore for the swapchain.
- Trigger a semaphore to indicate when we're done (so the swapchain
can submit).
 3. `present()`

The optimization opportunity here is that 95% of Godot's rendering is
done offscreen.
Then a fullscreen pass copies everything to the swapchain. Godot doesn't
practically render directly to the swapchain.

The problem with this is that the GPU has to wait for the swapchain to
be released **to start anything**, when we could start *much earlier*.
Only the final blit pass must wait for the swapchain.

TheForge changed it to the following (more complicated, I'm simplifying
the idea):

 1. Fill the entire cmd buffer with commands.
 2. In `screen_prepare_for_drawing` do `submit()`
    - There are no semaphore waits for the swapchain.
    - Trigger a semaphore to indicate when we're done.
3. Fill a new cmd buffer that only does the final blit to the
swapchain.
 4. `submit()`
    - Wait with a semaphore for the submit() from step 2.
- Wait with a semaphore for the swapchain (so the swapchain can
submit).
- Trigger a semaphore to indicate when we're done (so the swapchain
can submit).
 5. `present()`

Dario discovered this problem independently while working on a different
platform.

**However TheForge's solution had to be rewritten from scratch:** The
complexity to achieve the solution was high and quite difficult to
maintain with the way Godot works now (after Übershaders PR).
But on the other hand, re-implementing the solution became much simpler
because Dario already had to do something similar: To fix an Adreno 730
driver bug, he had to implement splitting command buffers. **This is
exactly what we need!**. Thus it was re-written using this existing
functionality for a new purpose.

To achieve this, I added a new argument, `bool p_split_cmd_buffer`, to
`RenderingDeviceGraph::add_draw_list_begin`, which is only set to true
by `RenderingDevice::draw_list_begin_for_screen`.

The graph will split the draw list into its own command buffer.

 - `bool split_swapchain_into_its_own_cmd_buffer = true;`

Setting it to false enforces the old behavior. This might be necessary
for consoles which follow an alternate solution to the same problem.
If not, then we should consider removing it.

PR #90993 added `shader_destroy_modules()` but it was not actually in
use.

This PR adds several places where `shader_destroy_modules()` is called
after initialization to free up memory of SPIR-V structures that are no
longer needed.
2024-12-09 11:49:28 -03:00

965 lines
32 KiB
C++
Raw Blame History

/**************************************************************************/
/* metal_objects.h */
/**************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/**************************************************************************/
/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/**************************************************************************/
/**************************************************************************/
/* */
/* Portions of this code were derived from MoltenVK. */
/* */
/* Copyright (c) 2015-2023 The Brenwill Workshop Ltd. */
/* (http://www.brenwill.com) */
/* */
/* Licensed under the Apache License, Version 2.0 (the "License"); */
/* you may not use this file except in compliance with the License. */
/* You may obtain a copy of the License at */
/* */
/* http://www.apache.org/licenses/LICENSE-2.0 */
/* */
/* Unless required by applicable law or agreed to in writing, software */
/* distributed under the License is distributed on an "AS IS" BASIS, */
/* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or */
/* implied. See the License for the specific language governing */
/* permissions and limitations under the License. */
/**************************************************************************/
#ifndef METAL_OBJECTS_H
#define METAL_OBJECTS_H
#import "metal_device_properties.h"
#import "metal_utils.h"
#import "pixel_formats.h"
#import "servers/rendering/rendering_device_driver.h"
#import <CommonCrypto/CommonDigest.h>
#import <Foundation/Foundation.h>
#import <Metal/Metal.h>
#import <QuartzCore/CAMetalLayer.h>
#import <simd/simd.h>
#import <zlib.h>
#import <initializer_list>
#import <optional>
#import <spirv.hpp>
// These types can be used in Vector and other containers that use
// pointer operations not supported by ARC.
namespace MTL {
#define MTL_CLASS(name) \
class name { \
public: \
name(id<MTL##name> obj = nil) : m_obj(obj) {} \
operator id<MTL##name>() const { return m_obj; } \
id<MTL##name> m_obj; \
};
MTL_CLASS(Texture)
} //namespace MTL
/// Metal buffer index for the view mask when rendering multi-view.
const uint32_t VIEW_MASK_BUFFER_INDEX = 24;
enum ShaderStageUsage : uint32_t {
None = 0,
Vertex = RDD::SHADER_STAGE_VERTEX_BIT,
Fragment = RDD::SHADER_STAGE_FRAGMENT_BIT,
TesselationControl = RDD::SHADER_STAGE_TESSELATION_CONTROL_BIT,
TesselationEvaluation = RDD::SHADER_STAGE_TESSELATION_EVALUATION_BIT,
Compute = RDD::SHADER_STAGE_COMPUTE_BIT,
};
_FORCE_INLINE_ ShaderStageUsage &operator|=(ShaderStageUsage &p_a, int p_b) {
p_a = ShaderStageUsage(uint32_t(p_a) | uint32_t(p_b));
return p_a;
}
enum StageResourceUsage : uint32_t {
VertexRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_VERTEX * 2),
VertexWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_VERTEX * 2),
FragmentRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_FRAGMENT * 2),
FragmentWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_FRAGMENT * 2),
TesselationControlRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2),
TesselationControlWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_CONTROL * 2),
TesselationEvaluationRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2),
TesselationEvaluationWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_TESSELATION_EVALUATION * 2),
ComputeRead = (MTLResourceUsageRead << RDD::SHADER_STAGE_COMPUTE * 2),
ComputeWrite = (MTLResourceUsageWrite << RDD::SHADER_STAGE_COMPUTE * 2),
};
typedef LocalVector<__unsafe_unretained id<MTLResource>> ResourceVector;
typedef HashMap<StageResourceUsage, ResourceVector> ResourceUsageMap;
enum class MDCommandBufferStateType {
None,
Render,
Compute,
Blit,
};
enum class MDPipelineType {
None,
Render,
Compute,
};
class MDRenderPass;
class MDPipeline;
class MDRenderPipeline;
class MDComputePipeline;
class MDFrameBuffer;
class RenderingDeviceDriverMetal;
class MDUniformSet;
class MDShader;
#pragma mark - Resource Factory
struct ClearAttKey {
const static uint32_t COLOR_COUNT = MAX_COLOR_ATTACHMENT_COUNT;
const static uint32_t DEPTH_INDEX = COLOR_COUNT;
const static uint32_t STENCIL_INDEX = DEPTH_INDEX + 1;
const static uint32_t ATTACHMENT_COUNT = STENCIL_INDEX + 1;
enum Flags : uint16_t {
CLEAR_FLAGS_NONE = 0,
CLEAR_FLAGS_LAYERED = 1 << 0,
};
Flags flags = CLEAR_FLAGS_NONE;
uint16_t sample_count = 0;
uint16_t pixel_formats[ATTACHMENT_COUNT] = { 0 };
_FORCE_INLINE_ void set_color_format(uint32_t p_idx, MTLPixelFormat p_fmt) { pixel_formats[p_idx] = p_fmt; }
_FORCE_INLINE_ void set_depth_format(MTLPixelFormat p_fmt) { pixel_formats[DEPTH_INDEX] = p_fmt; }
_FORCE_INLINE_ void set_stencil_format(MTLPixelFormat p_fmt) { pixel_formats[STENCIL_INDEX] = p_fmt; }
_FORCE_INLINE_ MTLPixelFormat depth_format() const { return (MTLPixelFormat)pixel_formats[DEPTH_INDEX]; }
_FORCE_INLINE_ MTLPixelFormat stencil_format() const { return (MTLPixelFormat)pixel_formats[STENCIL_INDEX]; }
_FORCE_INLINE_ void enable_layered_rendering() { flags::set(flags, CLEAR_FLAGS_LAYERED); }
_FORCE_INLINE_ bool is_enabled(uint32_t p_idx) const { return pixel_formats[p_idx] != 0; }
_FORCE_INLINE_ bool is_depth_enabled() const { return pixel_formats[DEPTH_INDEX] != 0; }
_FORCE_INLINE_ bool is_stencil_enabled() const { return pixel_formats[STENCIL_INDEX] != 0; }
_FORCE_INLINE_ bool is_layered_rendering_enabled() const { return flags::any(flags, CLEAR_FLAGS_LAYERED); }
_FORCE_INLINE_ bool operator==(const ClearAttKey &p_rhs) const {
return memcmp(this, &p_rhs, sizeof(ClearAttKey)) == 0;
}
uint32_t hash() const {
uint32_t h = hash_murmur3_one_32(flags);
h = hash_murmur3_one_32(sample_count, h);
h = hash_murmur3_buffer(pixel_formats, ATTACHMENT_COUNT * sizeof(pixel_formats[0]), h);
return hash_fmix32(h);
}
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDResourceFactory {
private:
RenderingDeviceDriverMetal *device_driver;
id<MTLFunction> new_func(NSString *p_source, NSString *p_name, NSError **p_error);
id<MTLFunction> new_clear_vert_func(ClearAttKey &p_key);
id<MTLFunction> new_clear_frag_func(ClearAttKey &p_key);
NSString *get_format_type_string(MTLPixelFormat p_fmt);
public:
id<MTLRenderPipelineState> new_clear_pipeline_state(ClearAttKey &p_key, NSError **p_error);
id<MTLDepthStencilState> new_depth_stencil_state(bool p_use_depth, bool p_use_stencil);
MDResourceFactory(RenderingDeviceDriverMetal *p_device_driver) :
device_driver(p_device_driver) {}
~MDResourceFactory() = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDResourceCache {
private:
typedef HashMap<ClearAttKey, id<MTLRenderPipelineState>, HashableHasher<ClearAttKey>> HashMap;
std::unique_ptr<MDResourceFactory> resource_factory;
HashMap clear_states;
struct {
id<MTLDepthStencilState> all;
id<MTLDepthStencilState> depth_only;
id<MTLDepthStencilState> stencil_only;
id<MTLDepthStencilState> none;
} clear_depth_stencil_state;
public:
id<MTLRenderPipelineState> get_clear_render_pipeline_state(ClearAttKey &p_key, NSError **p_error);
id<MTLDepthStencilState> get_depth_stencil_state(bool p_use_depth, bool p_use_stencil);
explicit MDResourceCache(RenderingDeviceDriverMetal *p_device_driver) :
resource_factory(new MDResourceFactory(p_device_driver)) {}
~MDResourceCache() = default;
};
enum class MDAttachmentType : uint8_t {
None = 0,
Color = 1 << 0,
Depth = 1 << 1,
Stencil = 1 << 2,
};
_FORCE_INLINE_ MDAttachmentType &operator|=(MDAttachmentType &p_a, MDAttachmentType p_b) {
flags::set(p_a, p_b);
return p_a;
}
_FORCE_INLINE_ bool operator&(MDAttachmentType p_a, MDAttachmentType p_b) {
return uint8_t(p_a) & uint8_t(p_b);
}
struct MDSubpass {
uint32_t subpass_index = 0;
uint32_t view_count = 0;
LocalVector<RDD::AttachmentReference> input_references;
LocalVector<RDD::AttachmentReference> color_references;
RDD::AttachmentReference depth_stencil_reference;
LocalVector<RDD::AttachmentReference> resolve_references;
MTLFmtCaps getRequiredFmtCapsForAttachmentAt(uint32_t p_index) const;
};
struct API_AVAILABLE(macos(11.0), ios(14.0)) MDAttachment {
private:
uint32_t index = 0;
uint32_t firstUseSubpassIndex = 0;
uint32_t lastUseSubpassIndex = 0;
public:
MTLPixelFormat format = MTLPixelFormatInvalid;
MDAttachmentType type = MDAttachmentType::None;
MTLLoadAction loadAction = MTLLoadActionDontCare;
MTLStoreAction storeAction = MTLStoreActionDontCare;
MTLLoadAction stencilLoadAction = MTLLoadActionDontCare;
MTLStoreAction stencilStoreAction = MTLStoreActionDontCare;
uint32_t samples = 1;
/*!
* @brief Returns true if this attachment is first used in the given subpass.
* @param p_subpass
* @return
*/
_FORCE_INLINE_ bool isFirstUseOf(MDSubpass const &p_subpass) const {
return p_subpass.subpass_index == firstUseSubpassIndex;
}
/*!
* @brief Returns true if this attachment is last used in the given subpass.
* @param p_subpass
* @return
*/
_FORCE_INLINE_ bool isLastUseOf(MDSubpass const &p_subpass) const {
return p_subpass.subpass_index == lastUseSubpassIndex;
}
void linkToSubpass(MDRenderPass const &p_pass);
MTLStoreAction getMTLStoreAction(MDSubpass const &p_subpass,
bool p_is_rendering_entire_area,
bool p_has_resolve,
bool p_can_resolve,
bool p_is_stencil) const;
bool configureDescriptor(MTLRenderPassAttachmentDescriptor *p_desc,
PixelFormats &p_pf,
MDSubpass const &p_subpass,
id<MTLTexture> p_attachment,
bool p_is_rendering_entire_area,
bool p_has_resolve,
bool p_can_resolve,
bool p_is_stencil) const;
/** Returns whether this attachment should be cleared in the subpass. */
bool shouldClear(MDSubpass const &p_subpass, bool p_is_stencil) const;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDRenderPass {
public:
Vector<MDAttachment> attachments;
Vector<MDSubpass> subpasses;
uint32_t get_sample_count() const {
return attachments.is_empty() ? 1 : attachments[0].samples;
}
MDRenderPass(Vector<MDAttachment> &p_attachments, Vector<MDSubpass> &p_subpasses);
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDCommandBuffer {
private:
RenderingDeviceDriverMetal *device_driver = nullptr;
id<MTLCommandQueue> queue = nil;
id<MTLCommandBuffer> commandBuffer = nil;
void _end_compute_dispatch();
void _end_blit();
#pragma mark - Render
void _render_set_dirty_state();
void _render_bind_uniform_sets();
void _populate_vertices(simd::float4 *p_vertices, Size2i p_fb_size, VectorView<Rect2i> p_rects);
uint32_t _populate_vertices(simd::float4 *p_vertices, uint32_t p_index, Rect2i const &p_rect, Size2i p_fb_size);
void _end_render_pass();
void _render_clear_render_area();
public:
MDCommandBufferStateType type = MDCommandBufferStateType::None;
struct RenderState {
MDRenderPass *pass = nullptr;
MDFrameBuffer *frameBuffer = nullptr;
MDRenderPipeline *pipeline = nullptr;
LocalVector<RDD::RenderPassClearValue> clear_values;
LocalVector<MTLViewport> viewports;
LocalVector<MTLScissorRect> scissors;
std::optional<Color> blend_constants;
uint32_t current_subpass = UINT32_MAX;
Rect2i render_area = {};
bool is_rendering_entire_area = false;
MTLRenderPassDescriptor *desc = nil;
id<MTLRenderCommandEncoder> encoder = nil;
id<MTLBuffer> __unsafe_unretained index_buffer = nil; // Buffer is owned by RDD.
MTLIndexType index_type = MTLIndexTypeUInt16;
uint32_t index_offset = 0;
LocalVector<id<MTLBuffer> __unsafe_unretained> vertex_buffers;
LocalVector<NSUInteger> vertex_offsets;
ResourceUsageMap resource_usage;
// clang-format off
enum DirtyFlag: uint8_t {
DIRTY_NONE = 0b0000'0000,
DIRTY_PIPELINE = 0b0000'0001, //! pipeline state
DIRTY_UNIFORMS = 0b0000'0010, //! uniform sets
DIRTY_DEPTH = 0b0000'0100, //! depth / stenci state
DIRTY_VERTEX = 0b0000'1000, //! vertex buffers
DIRTY_VIEWPORT = 0b0001'0000, //! viewport rectangles
DIRTY_SCISSOR = 0b0010'0000, //! scissor rectangles
DIRTY_BLEND = 0b0100'0000, //! blend state
DIRTY_RASTER = 0b1000'0000, //! encoder state like cull mode
DIRTY_ALL = 0xff,
};
// clang-format on
BitField<DirtyFlag> dirty = DIRTY_NONE;
LocalVector<MDUniformSet *> uniform_sets;
// Bit mask of the uniform sets that are dirty, to prevent redundant binding.
uint64_t uniform_set_mask = 0;
_FORCE_INLINE_ void reset();
void end_encoding();
_ALWAYS_INLINE_ const MDSubpass &get_subpass() const {
DEV_ASSERT(pass != nullptr);
return pass->subpasses[current_subpass];
}
_FORCE_INLINE_ void mark_viewport_dirty() {
if (viewports.is_empty()) {
return;
}
dirty.set_flag(DirtyFlag::DIRTY_VIEWPORT);
}
_FORCE_INLINE_ void mark_scissors_dirty() {
if (scissors.is_empty()) {
return;
}
dirty.set_flag(DirtyFlag::DIRTY_SCISSOR);
}
_FORCE_INLINE_ void mark_vertex_dirty() {
if (vertex_buffers.is_empty()) {
return;
}
dirty.set_flag(DirtyFlag::DIRTY_VERTEX);
}
_FORCE_INLINE_ void mark_uniforms_dirty(std::initializer_list<uint32_t> l) {
if (uniform_sets.is_empty()) {
return;
}
for (uint32_t i : l) {
if (i < uniform_sets.size() && uniform_sets[i] != nullptr) {
uniform_set_mask |= 1 << i;
}
}
dirty.set_flag(DirtyFlag::DIRTY_UNIFORMS);
}
_FORCE_INLINE_ void mark_uniforms_dirty(void) {
if (uniform_sets.is_empty()) {
return;
}
for (uint32_t i = 0; i < uniform_sets.size(); i++) {
if (uniform_sets[i] != nullptr) {
uniform_set_mask |= 1 << i;
}
}
dirty.set_flag(DirtyFlag::DIRTY_UNIFORMS);
}
_FORCE_INLINE_ void mark_blend_dirty() {
if (!blend_constants.has_value()) {
return;
}
dirty.set_flag(DirtyFlag::DIRTY_BLEND);
}
MTLScissorRect clip_to_render_area(MTLScissorRect p_rect) const {
uint32_t raLeft = render_area.position.x;
uint32_t raRight = raLeft + render_area.size.width;
uint32_t raBottom = render_area.position.y;
uint32_t raTop = raBottom + render_area.size.height;
p_rect.x = CLAMP(p_rect.x, raLeft, MAX(raRight - 1, raLeft));
p_rect.y = CLAMP(p_rect.y, raBottom, MAX(raTop - 1, raBottom));
p_rect.width = MIN(p_rect.width, raRight - p_rect.x);
p_rect.height = MIN(p_rect.height, raTop - p_rect.y);
return p_rect;
}
Rect2i clip_to_render_area(Rect2i p_rect) const {
int32_t raLeft = render_area.position.x;
int32_t raRight = raLeft + render_area.size.width;
int32_t raBottom = render_area.position.y;
int32_t raTop = raBottom + render_area.size.height;
p_rect.position.x = CLAMP(p_rect.position.x, raLeft, MAX(raRight - 1, raLeft));
p_rect.position.y = CLAMP(p_rect.position.y, raBottom, MAX(raTop - 1, raBottom));
p_rect.size.width = MIN(p_rect.size.width, raRight - p_rect.position.x);
p_rect.size.height = MIN(p_rect.size.height, raTop - p_rect.position.y);
return p_rect;
}
} render;
// State specific for a compute pass.
struct ComputeState {
MDComputePipeline *pipeline = nullptr;
id<MTLComputeCommandEncoder> encoder = nil;
ResourceUsageMap resource_usage;
_FORCE_INLINE_ void reset() {
pipeline = nil;
encoder = nil;
// Keep the keys, as they are likely to be used again.
for (KeyValue<StageResourceUsage, LocalVector<__unsafe_unretained id<MTLResource>>> &kv : resource_usage) {
kv.value.clear();
}
}
void end_encoding();
} compute;
// State specific to a blit pass.
struct {
id<MTLBlitCommandEncoder> encoder = nil;
_FORCE_INLINE_ void reset() {
encoder = nil;
}
} blit;
_FORCE_INLINE_ id<MTLCommandBuffer> get_command_buffer() const {
return commandBuffer;
}
void begin();
void commit();
void end();
id<MTLBlitCommandEncoder> blit_command_encoder();
void encodeRenderCommandEncoderWithDescriptor(MTLRenderPassDescriptor *p_desc, NSString *p_label);
void bind_pipeline(RDD::PipelineID p_pipeline);
#pragma mark - Render Commands
void render_bind_uniform_set(RDD::UniformSetID p_uniform_set, RDD::ShaderID p_shader, uint32_t p_set_index);
void render_bind_uniform_sets(VectorView<RDD::UniformSetID> p_uniform_sets, RDD::ShaderID p_shader, uint32_t p_first_set_index, uint32_t p_set_count);
void render_clear_attachments(VectorView<RDD::AttachmentClear> p_attachment_clears, VectorView<Rect2i> p_rects);
void render_set_viewport(VectorView<Rect2i> p_viewports);
void render_set_scissor(VectorView<Rect2i> p_scissors);
void render_set_blend_constants(const Color &p_constants);
void render_begin_pass(RDD::RenderPassID p_render_pass,
RDD::FramebufferID p_frameBuffer,
RDD::CommandBufferType p_cmd_buffer_type,
const Rect2i &p_rect,
VectorView<RDD::RenderPassClearValue> p_clear_values);
void render_next_subpass();
void render_draw(uint32_t p_vertex_count,
uint32_t p_instance_count,
uint32_t p_base_vertex,
uint32_t p_first_instance);
void render_bind_vertex_buffers(uint32_t p_binding_count, const RDD::BufferID *p_buffers, const uint64_t *p_offsets);
void render_bind_index_buffer(RDD::BufferID p_buffer, RDD::IndexBufferFormat p_format, uint64_t p_offset);
void render_draw_indexed(uint32_t p_index_count,
uint32_t p_instance_count,
uint32_t p_first_index,
int32_t p_vertex_offset,
uint32_t p_first_instance);
void render_draw_indexed_indirect(RDD::BufferID p_indirect_buffer, uint64_t p_offset, uint32_t p_draw_count, uint32_t p_stride);
void render_draw_indexed_indirect_count(RDD::BufferID p_indirect_buffer, uint64_t p_offset, RDD::BufferID p_count_buffer, uint64_t p_count_buffer_offset, uint32_t p_max_draw_count, uint32_t p_stride);
void render_draw_indirect(RDD::BufferID p_indirect_buffer, uint64_t p_offset, uint32_t p_draw_count, uint32_t p_stride);
void render_draw_indirect_count(RDD::BufferID p_indirect_buffer, uint64_t p_offset, RDD::BufferID p_count_buffer, uint64_t p_count_buffer_offset, uint32_t p_max_draw_count, uint32_t p_stride);
void render_end_pass();
#pragma mark - Compute Commands
void compute_bind_uniform_set(RDD::UniformSetID p_uniform_set, RDD::ShaderID p_shader, uint32_t p_set_index);
void compute_bind_uniform_sets(VectorView<RDD::UniformSetID> p_uniform_sets, RDD::ShaderID p_shader, uint32_t p_first_set_index, uint32_t p_set_count);
void compute_dispatch(uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups);
void compute_dispatch_indirect(RDD::BufferID p_indirect_buffer, uint64_t p_offset);
MDCommandBuffer(id<MTLCommandQueue> p_queue, RenderingDeviceDriverMetal *p_device_driver) :
device_driver(p_device_driver), queue(p_queue) {
type = MDCommandBufferStateType::None;
}
MDCommandBuffer() = default;
};
#if (TARGET_OS_OSX && __MAC_OS_X_VERSION_MAX_ALLOWED < 140000) || (TARGET_OS_IOS && __IPHONE_OS_VERSION_MAX_ALLOWED < 170000)
#define MTLBindingAccess MTLArgumentAccess
#define MTLBindingAccessReadOnly MTLArgumentAccessReadOnly
#define MTLBindingAccessReadWrite MTLArgumentAccessReadWrite
#define MTLBindingAccessWriteOnly MTLArgumentAccessWriteOnly
#endif
struct API_AVAILABLE(macos(11.0), ios(14.0)) BindingInfo {
MTLDataType dataType = MTLDataTypeNone;
uint32_t index = 0;
MTLBindingAccess access = MTLBindingAccessReadOnly;
MTLResourceUsage usage = 0;
MTLTextureType textureType = MTLTextureType2D;
spv::ImageFormat imageFormat = spv::ImageFormatUnknown;
uint32_t arrayLength = 0;
bool isMultisampled = false;
inline MTLArgumentDescriptor *new_argument_descriptor() const {
MTLArgumentDescriptor *desc = MTLArgumentDescriptor.argumentDescriptor;
desc.dataType = dataType;
desc.index = index;
desc.access = access;
desc.textureType = textureType;
desc.arrayLength = arrayLength;
return desc;
}
size_t serialize_size() const {
return sizeof(uint32_t) * 8 /* 8 uint32_t fields */;
}
template <typename W>
void serialize(W &p_writer) const {
p_writer.write((uint32_t)dataType);
p_writer.write(index);
p_writer.write((uint32_t)access);
p_writer.write((uint32_t)usage);
p_writer.write((uint32_t)textureType);
p_writer.write(imageFormat);
p_writer.write(arrayLength);
p_writer.write(isMultisampled);
}
template <typename R>
void deserialize(R &p_reader) {
p_reader.read((uint32_t &)dataType);
p_reader.read(index);
p_reader.read((uint32_t &)access);
p_reader.read((uint32_t &)usage);
p_reader.read((uint32_t &)textureType);
p_reader.read((uint32_t &)imageFormat);
p_reader.read(arrayLength);
p_reader.read(isMultisampled);
}
};
using RDC = RenderingDeviceCommons;
typedef API_AVAILABLE(macos(11.0), ios(14.0)) HashMap<RDC::ShaderStage, BindingInfo> BindingInfoMap;
struct API_AVAILABLE(macos(11.0), ios(14.0)) UniformInfo {
uint32_t binding;
ShaderStageUsage active_stages = None;
BindingInfoMap bindings;
BindingInfoMap bindings_secondary;
};
struct API_AVAILABLE(macos(11.0), ios(14.0)) UniformSet {
LocalVector<UniformInfo> uniforms;
uint32_t buffer_size = 0;
HashMap<RDC::ShaderStage, uint32_t> offsets;
HashMap<RDC::ShaderStage, id<MTLArgumentEncoder>> encoders;
};
struct ShaderCacheEntry;
enum class ShaderLoadStrategy {
DEFAULT,
LAZY,
};
/// A Metal shader library.
@interface MDLibrary : NSObject {
ShaderCacheEntry *_entry;
};
- (id<MTLLibrary>)library;
- (NSError *)error;
- (void)setLabel:(NSString *)label;
+ (instancetype)newLibraryWithCacheEntry:(ShaderCacheEntry *)entry
device:(id<MTLDevice>)device
source:(NSString *)source
options:(MTLCompileOptions *)options
strategy:(ShaderLoadStrategy)strategy;
@end
struct SHA256Digest {
unsigned char data[CC_SHA256_DIGEST_LENGTH];
uint32_t hash() const {
uint32_t c = crc32(0, data, CC_SHA256_DIGEST_LENGTH);
return c;
}
SHA256Digest() {
bzero(data, CC_SHA256_DIGEST_LENGTH);
}
SHA256Digest(const char *p_data, size_t p_length) {
CC_SHA256(p_data, (CC_LONG)p_length, data);
}
_FORCE_INLINE_ uint32_t short_sha() const {
return __builtin_bswap32(*(uint32_t *)&data[0]);
}
};
template <>
struct HashMapComparatorDefault<SHA256Digest> {
static bool compare(const SHA256Digest &p_lhs, const SHA256Digest &p_rhs) {
return memcmp(p_lhs.data, p_rhs.data, CC_SHA256_DIGEST_LENGTH) == 0;
}
};
/// A cache entry for a Metal shader library.
struct ShaderCacheEntry {
RenderingDeviceDriverMetal &owner;
/// A hash of the Metal shader source code.
SHA256Digest key;
CharString name;
RD::ShaderStage stage = RD::SHADER_STAGE_VERTEX;
/// This reference must be weak, to ensure that when the last strong reference to the library
/// is released, the cache entry is freed.
MDLibrary *__weak library = nil;
/// Notify the cache that this entry is no longer needed.
void notify_free() const;
ShaderCacheEntry(RenderingDeviceDriverMetal &p_owner, SHA256Digest p_key) :
owner(p_owner), key(p_key) {
}
~ShaderCacheEntry() = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDShader {
public:
CharString name;
Vector<UniformSet> sets;
virtual void encode_push_constant_data(VectorView<uint32_t> p_data, MDCommandBuffer *p_cb) = 0;
MDShader(CharString p_name, Vector<UniformSet> p_sets) :
name(p_name), sets(p_sets) {}
virtual ~MDShader() = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDComputeShader final : public MDShader {
public:
struct {
uint32_t binding = -1;
uint32_t size = 0;
} push_constants;
MTLSize local = {};
MDLibrary *kernel;
#if DEV_ENABLED
CharString kernel_source;
#endif
void encode_push_constant_data(VectorView<uint32_t> p_data, MDCommandBuffer *p_cb) final;
MDComputeShader(CharString p_name, Vector<UniformSet> p_sets, MDLibrary *p_kernel);
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDRenderShader final : public MDShader {
public:
struct {
struct {
int32_t binding = -1;
uint32_t size = 0;
} vert;
struct {
int32_t binding = -1;
uint32_t size = 0;
} frag;
} push_constants;
bool needs_view_mask_buffer = false;
MDLibrary *vert;
MDLibrary *frag;
#if DEV_ENABLED
CharString vert_source;
CharString frag_source;
#endif
void encode_push_constant_data(VectorView<uint32_t> p_data, MDCommandBuffer *p_cb) final;
MDRenderShader(CharString p_name,
bool p_needs_view_mask_buffer,
Vector<UniformSet> p_sets,
MDLibrary *p_vert, MDLibrary *p_frag);
};
_FORCE_INLINE_ StageResourceUsage &operator|=(StageResourceUsage &p_a, uint32_t p_b) {
p_a = StageResourceUsage(uint32_t(p_a) | p_b);
return p_a;
}
_FORCE_INLINE_ StageResourceUsage stage_resource_usage(RDC::ShaderStage p_stage, MTLResourceUsage p_usage) {
return StageResourceUsage(p_usage << (p_stage * 2));
}
_FORCE_INLINE_ MTLResourceUsage resource_usage_for_stage(StageResourceUsage p_usage, RDC::ShaderStage p_stage) {
return MTLResourceUsage((p_usage >> (p_stage * 2)) & 0b11);
}
template <>
struct HashMapComparatorDefault<RDD::ShaderID> {
static bool compare(const RDD::ShaderID &p_lhs, const RDD::ShaderID &p_rhs) {
return p_lhs.id == p_rhs.id;
}
};
struct BoundUniformSet {
id<MTLBuffer> buffer;
ResourceUsageMap usage_to_resources;
/// Perform a 2-way merge each key of `ResourceVector` resources from this set into the
/// destination set.
///
/// Assumes the vectors of resources are sorted.
void merge_into(ResourceUsageMap &p_dst) const;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDUniformSet {
public:
uint32_t index;
LocalVector<RDD::BoundUniform> uniforms;
HashMap<MDShader *, BoundUniformSet> bound_uniforms;
BoundUniformSet &boundUniformSetForShader(MDShader *p_shader, id<MTLDevice> p_device);
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDPipeline {
public:
MDPipelineType type;
explicit MDPipeline(MDPipelineType p_type) :
type(p_type) {}
virtual ~MDPipeline() = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDRenderPipeline final : public MDPipeline {
public:
id<MTLRenderPipelineState> state = nil;
id<MTLDepthStencilState> depth_stencil = nil;
uint32_t push_constant_size = 0;
uint32_t push_constant_stages_mask = 0;
SampleCount sample_count = SampleCount1;
struct {
MTLCullMode cull_mode = MTLCullModeNone;
MTLTriangleFillMode fill_mode = MTLTriangleFillModeFill;
MTLDepthClipMode clip_mode = MTLDepthClipModeClip;
MTLWinding winding = MTLWindingClockwise;
MTLPrimitiveType render_primitive = MTLPrimitiveTypePoint;
struct {
bool enabled = false;
} depth_test;
struct {
bool enabled = false;
float depth_bias = 0.0;
float slope_scale = 0.0;
float clamp = 0.0;
_FORCE_INLINE_ void apply(id<MTLRenderCommandEncoder> __unsafe_unretained p_enc) const {
if (!enabled) {
return;
}
[p_enc setDepthBias:depth_bias slopeScale:slope_scale clamp:clamp];
}
} depth_bias;
struct {
bool enabled = false;
uint32_t front_reference = 0;
uint32_t back_reference = 0;
_FORCE_INLINE_ void apply(id<MTLRenderCommandEncoder> __unsafe_unretained p_enc) const {
if (!enabled)
return;
[p_enc setStencilFrontReferenceValue:front_reference backReferenceValue:back_reference];
}
} stencil;
struct {
bool enabled = false;
float r = 0.0;
float g = 0.0;
float b = 0.0;
float a = 0.0;
_FORCE_INLINE_ void apply(id<MTLRenderCommandEncoder> __unsafe_unretained p_enc) const {
//if (!enabled)
// return;
[p_enc setBlendColorRed:r green:g blue:b alpha:a];
}
} blend;
_FORCE_INLINE_ void apply(id<MTLRenderCommandEncoder> __unsafe_unretained p_enc) const {
[p_enc setCullMode:cull_mode];
[p_enc setTriangleFillMode:fill_mode];
[p_enc setDepthClipMode:clip_mode];
[p_enc setFrontFacingWinding:winding];
depth_bias.apply(p_enc);
stencil.apply(p_enc);
blend.apply(p_enc);
}
} raster_state;
MDRenderShader *shader = nil;
MDRenderPipeline() :
MDPipeline(MDPipelineType::Render) {}
~MDRenderPipeline() final = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDComputePipeline final : public MDPipeline {
public:
id<MTLComputePipelineState> state = nil;
struct {
MTLSize local = {};
} compute_state;
MDComputeShader *shader = nil;
explicit MDComputePipeline(id<MTLComputePipelineState> p_state) :
MDPipeline(MDPipelineType::Compute), state(p_state) {}
~MDComputePipeline() final = default;
};
class API_AVAILABLE(macos(11.0), ios(14.0)) MDFrameBuffer {
Vector<MTL::Texture> textures;
public:
Size2i size;
MDFrameBuffer(Vector<MTL::Texture> p_textures, Size2i p_size) :
textures(p_textures), size(p_size) {}
MDFrameBuffer() {}
/// Returns the texture at the given index.
_ALWAYS_INLINE_ MTL::Texture get_texture(uint32_t p_idx) const {
return textures[p_idx];
}
/// Returns true if the texture at the given index is not nil.
_ALWAYS_INLINE_ bool has_texture(uint32_t p_idx) const {
return textures[p_idx] != nil;
}
/// Set the texture at the given index.
_ALWAYS_INLINE_ void set_texture(uint32_t p_idx, MTL::Texture p_texture) {
textures.write[p_idx] = p_texture;
}
/// Unset or nil the texture at the given index.
_ALWAYS_INLINE_ void unset_texture(uint32_t p_idx) {
textures.write[p_idx] = nil;
}
/// Resizes buffers to the specified size.
_ALWAYS_INLINE_ void set_texture_count(uint32_t p_size) {
textures.resize(p_size);
}
virtual ~MDFrameBuffer() = default;
};
// These functions are used to convert between Objective-C objects and
// the RIDs used by Godot, respecting automatic reference counting.
namespace rid {
// Converts an Objective-C object to a pointer, and incrementing the
// reference count.
_FORCE_INLINE_
void *owned(id p_id) {
return (__bridge_retained void *)p_id;
}
#define MAKE_ID(FROM, TO) \
_FORCE_INLINE_ TO make(FROM p_obj) { return TO(owned(p_obj)); }
MAKE_ID(id<MTLTexture>, RDD::TextureID)
MAKE_ID(id<MTLBuffer>, RDD::BufferID)
MAKE_ID(id<MTLSamplerState>, RDD::SamplerID)
MAKE_ID(MTLVertexDescriptor *, RDD::VertexFormatID)
MAKE_ID(id<MTLCommandQueue>, RDD::CommandPoolID)
// Converts a pointer to an Objective-C object without changing the reference count.
_FORCE_INLINE_
auto get(RDD::ID p_id) {
return (p_id.id) ? (__bridge ::id)(void *)p_id.id : nil;
}
// Converts a pointer to an Objective-C object, and decrements the reference count.
_FORCE_INLINE_
auto release(RDD::ID p_id) {
return (__bridge_transfer ::id)(void *)p_id.id;
}
} // namespace rid
#endif // METAL_OBJECTS_H
Reference in New Issue View Git Blame Copy Permalink