lavfi/glslang: migrate to the C API and robustify library detection
Finally, this is as close to usable as it gets for glslang. Much faster to compile as well, and eliminates the need for a C++ compiler, which is great. Also, changes to the resource limits won't break users, as we can use designated initializers in C90.
This commit is contained in:
Lynne
parent
fef85c376a
commit
246f841b53
|
|
@ -3564,7 +3564,7 @@ ass_filter_deps="libass"
|
|||
atempo_filter_deps="avcodec"
|
||||
atempo_filter_select="rdft"
|
||||
avgblur_opencl_filter_deps="opencl"
|
||||
avgblur_vulkan_filter_deps="libglslang"
|
||||
avgblur_vulkan_filter_deps="vulkan libglslang"
|
||||
azmq_filter_deps="libzmq"
|
||||
blackframe_filter_deps="gpl"
|
||||
bm3d_filter_deps="avcodec"
|
||||
|
|
@ -3572,7 +3572,7 @@ bm3d_filter_select="dct"
|
|||
boxblur_filter_deps="gpl"
|
||||
boxblur_opencl_filter_deps="opencl gpl"
|
||||
bs2b_filter_deps="libbs2b"
|
||||
chromaber_vulkan_filter_deps="libglslang"
|
||||
chromaber_vulkan_filter_deps="vulkan libglslang"
|
||||
colorkey_opencl_filter_deps="opencl"
|
||||
colormatrix_filter_deps="gpl"
|
||||
convolution_opencl_filter_deps="opencl"
|
||||
|
|
@ -3635,7 +3635,7 @@ openclsrc_filter_deps="opencl"
|
|||
overlay_opencl_filter_deps="opencl"
|
||||
overlay_qsv_filter_deps="libmfx"
|
||||
overlay_qsv_filter_select="qsvvpp"
|
||||
overlay_vulkan_filter_deps="libglslang"
|
||||
overlay_vulkan_filter_deps="vulkan libglslang"
|
||||
owdenoise_filter_deps="gpl"
|
||||
pad_opencl_filter_deps="opencl"
|
||||
pan_filter_deps="swresample"
|
||||
|
|
@ -3698,7 +3698,7 @@ zmq_filter_deps="libzmq"
|
|||
zoompan_filter_deps="swscale"
|
||||
zscale_filter_deps="libzimg const_nan"
|
||||
scale_vaapi_filter_deps="vaapi"
|
||||
scale_vulkan_filter_deps="libglslang"
|
||||
scale_vulkan_filter_deps="vulkan libglslang"
|
||||
vpp_qsv_filter_deps="libmfx"
|
||||
vpp_qsv_filter_select="qsvvpp"
|
||||
xfade_opencl_filter_deps="opencl"
|
||||
|
|
@ -6440,7 +6440,12 @@ enabled fontconfig && enable libfontconfig
|
|||
enabled libfontconfig && require_pkg_config libfontconfig fontconfig "fontconfig/fontconfig.h" FcInit
|
||||
enabled libfreetype && require_pkg_config libfreetype freetype2 "ft2build.h FT_FREETYPE_H" FT_Init_FreeType
|
||||
enabled libfribidi && require_pkg_config libfribidi fribidi fribidi.h fribidi_version_info
|
||||
enabled libglslang && require_cpp libglslang glslang/SPIRV/GlslangToSpv.h "glslang::TIntermediate*" -lglslang -lMachineIndependent -lOSDependent -lHLSL -lOGLCompiler -lGenericCodeGen -lSPVRemapper -lSPIRV -lSPIRV-Tools-opt -lSPIRV-Tools -lpthread -lstdc++
|
||||
enabled libglslang && { check_lib libglslang glslang/Include/glslang_c_interface.h glslang_initialize_process \
|
||||
-lglslang -lMachineIndependent -lOSDependent -lHLSL -lOGLCompiler -lGenericCodeGen \
|
||||
-lSPVRemapper -lSPIRV -lSPIRV-Tools-opt -lSPIRV-Tools -lpthread -lstdc++ -lm ||
|
||||
require libglslang glslang/Include/glslang_c_interface.h glslang_initialize_process \
|
||||
-lglslang -lOSDependent -lHLSL -lOGLCompiler \
|
||||
-lSPVRemapper -lSPIRV -lSPIRV-Tools-opt -lSPIRV-Tools -lpthread -lstdc++ -lm; }
|
||||
enabled libgme && { check_pkg_config libgme libgme gme/gme.h gme_new_emu ||
|
||||
require libgme gme/gme.h gme_new_emu -lgme -lstdc++; }
|
||||
enabled libgsm && { for gsm_hdr in "gsm.h" "gsm/gsm.h"; do
|
||||
|
|
|
|||
|
|
@ -0,0 +1,256 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* modify it under the terms of the GNU Lesser General Public
|
||||
* License as published by the Free Software Foundation; either
|
||||
* version 2.1 of the License, or (at your option) any later version.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
* Lesser General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU Lesser General Public
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include <pthread.h>
|
||||
|
||||
#include <glslang/build_info.h>
|
||||
#include <glslang/Include/glslang_c_interface.h>
|
||||
|
||||
#include "libavutil/mem.h"
|
||||
#include "libavutil/avassert.h"
|
||||
|
||||
#include "glslang.h"
|
||||
|
||||
static pthread_mutex_t glslang_mutex = PTHREAD_MUTEX_INITIALIZER;
|
||||
static int glslang_refcount = 0;
|
||||
|
||||
static const glslang_resource_t glslc_resource_limits = {
|
||||
.max_lights = 32,
|
||||
.max_clip_planes = 6,
|
||||
.max_texture_units = 32,
|
||||
.max_texture_coords = 32,
|
||||
.max_vertex_attribs = 64,
|
||||
.max_vertex_uniform_components = 4096,
|
||||
.max_varying_floats = 64,
|
||||
.max_vertex_texture_image_units = 32,
|
||||
.max_combined_texture_image_units = 80,
|
||||
.max_texture_image_units = 32,
|
||||
.max_fragment_uniform_components = 4096,
|
||||
.max_draw_buffers = 32,
|
||||
.max_vertex_uniform_vectors = 128,
|
||||
.max_varying_vectors = 8,
|
||||
.max_fragment_uniform_vectors = 16,
|
||||
.max_vertex_output_vectors = 16,
|
||||
.max_fragment_input_vectors = 15,
|
||||
.min_program_texel_offset = -8,
|
||||
.max_program_texel_offset = 7,
|
||||
.max_clip_distances = 8,
|
||||
.max_compute_work_group_count_x = 65535,
|
||||
.max_compute_work_group_count_y = 65535,
|
||||
.max_compute_work_group_count_z = 65535,
|
||||
.max_compute_work_group_size_x = 1024,
|
||||
.max_compute_work_group_size_y = 1024,
|
||||
.max_compute_work_group_size_z = 64,
|
||||
.max_compute_uniform_components = 1024,
|
||||
.max_compute_texture_image_units = 16,
|
||||
.max_compute_image_uniforms = 8,
|
||||
.max_compute_atomic_counters = 8,
|
||||
.max_compute_atomic_counter_buffers = 1,
|
||||
.max_varying_components = 60,
|
||||
.max_vertex_output_components = 64,
|
||||
.max_geometry_input_components = 64,
|
||||
.max_geometry_output_components = 128,
|
||||
.max_fragment_input_components = 128,
|
||||
.max_image_units = 8,
|
||||
.max_combined_image_units_and_fragment_outputs = 8,
|
||||
.max_combined_shader_output_resources = 8,
|
||||
.max_image_samples = 0,
|
||||
.max_vertex_image_uniforms = 0,
|
||||
.max_tess_control_image_uniforms = 0,
|
||||
.max_tess_evaluation_image_uniforms = 0,
|
||||
.max_geometry_image_uniforms = 0,
|
||||
.max_fragment_image_uniforms = 8,
|
||||
.max_combined_image_uniforms = 8,
|
||||
.max_geometry_texture_image_units = 16,
|
||||
.max_geometry_output_vertices = 256,
|
||||
.max_geometry_total_output_components = 1024,
|
||||
.max_geometry_uniform_components = 1024,
|
||||
.max_geometry_varying_components = 64,
|
||||
.max_tess_control_input_components = 128,
|
||||
.max_tess_control_output_components = 128,
|
||||
.max_tess_control_texture_image_units = 16,
|
||||
.max_tess_control_uniform_components = 1024,
|
||||
.max_tess_control_total_output_components = 4096,
|
||||
.max_tess_evaluation_input_components = 128,
|
||||
.max_tess_evaluation_output_components = 128,
|
||||
.max_tess_evaluation_texture_image_units = 16,
|
||||
.max_tess_evaluation_uniform_components = 1024,
|
||||
.max_tess_patch_components = 120,
|
||||
.max_patch_vertices = 32,
|
||||
.max_tess_gen_level = 64,
|
||||
.max_viewports = 16,
|
||||
.max_vertex_atomic_counters = 0,
|
||||
.max_tess_control_atomic_counters = 0,
|
||||
.max_tess_evaluation_atomic_counters = 0,
|
||||
.max_geometry_atomic_counters = 0,
|
||||
.max_fragment_atomic_counters = 8,
|
||||
.max_combined_atomic_counters = 8,
|
||||
.max_atomic_counter_bindings = 1,
|
||||
.max_vertex_atomic_counter_buffers = 0,
|
||||
.max_tess_control_atomic_counter_buffers = 0,
|
||||
.max_tess_evaluation_atomic_counter_buffers = 0,
|
||||
.max_geometry_atomic_counter_buffers = 0,
|
||||
.max_fragment_atomic_counter_buffers = 1,
|
||||
.max_combined_atomic_counter_buffers = 1,
|
||||
.max_atomic_counter_buffer_size = 16384,
|
||||
.max_transform_feedback_buffers = 4,
|
||||
.max_transform_feedback_interleaved_components = 64,
|
||||
.max_cull_distances = 8,
|
||||
.max_combined_clip_and_cull_distances = 8,
|
||||
.max_samples = 4,
|
||||
.max_mesh_output_vertices_nv = 256,
|
||||
.max_mesh_output_primitives_nv = 512,
|
||||
.max_mesh_work_group_size_x_nv = 32,
|
||||
.max_mesh_work_group_size_y_nv = 1,
|
||||
.max_mesh_work_group_size_z_nv = 1,
|
||||
.max_task_work_group_size_x_nv = 32,
|
||||
.max_task_work_group_size_y_nv = 1,
|
||||
.max_task_work_group_size_z_nv = 1,
|
||||
.max_mesh_view_count_nv = 4,
|
||||
.maxDualSourceDrawBuffersEXT = 1,
|
||||
|
||||
.limits = {
|
||||
.non_inductive_for_loops = 1,
|
||||
.while_loops = 1,
|
||||
.do_while_loops = 1,
|
||||
.general_uniform_indexing = 1,
|
||||
.general_attribute_matrix_vector_indexing = 1,
|
||||
.general_varying_indexing = 1,
|
||||
.general_sampler_indexing = 1,
|
||||
.general_variable_indexing = 1,
|
||||
.general_constant_matrix_vector_indexing = 1,
|
||||
}
|
||||
};
|
||||
|
||||
int ff_vk_glslang_shader_compile(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
uint8_t **data, size_t *size, void **opaque)
|
||||
{
|
||||
const char *messages;
|
||||
glslang_shader_t *glslc_shader;
|
||||
glslang_program_t *glslc_program;
|
||||
|
||||
static const glslang_stage_t glslc_stage[] = {
|
||||
[VK_SHADER_STAGE_VERTEX_BIT] = GLSLANG_STAGE_VERTEX,
|
||||
[VK_SHADER_STAGE_FRAGMENT_BIT] = GLSLANG_STAGE_FRAGMENT,
|
||||
[VK_SHADER_STAGE_COMPUTE_BIT] = GLSLANG_STAGE_COMPUTE,
|
||||
};
|
||||
|
||||
const glslang_input_t glslc_input = {
|
||||
.language = GLSLANG_SOURCE_GLSL,
|
||||
.stage = glslc_stage[shd->shader.stage],
|
||||
.client = GLSLANG_CLIENT_VULKAN,
|
||||
/* GLSLANG_TARGET_VULKAN_1_2 before 11.6 resulted in targeting 1.0 */
|
||||
#if (((GLSLANG_VERSION_MAJOR) > 11) || ((GLSLANG_VERSION_MAJOR) == 11 && \
|
||||
(((GLSLANG_VERSION_MINOR) > 6) || ((GLSLANG_VERSION_MINOR) == 6 && \
|
||||
((GLSLANG_VERSION_PATCH) > 0)))))
|
||||
.client_version = GLSLANG_TARGET_VULKAN_1_2,
|
||||
.target_language_version = GLSLANG_TARGET_SPV_1_5,
|
||||
#else
|
||||
.client_version = GLSLANG_TARGET_VULKAN_1_1,
|
||||
.target_language_version = GLSLANG_TARGET_SPV_1_3,
|
||||
#endif
|
||||
.target_language = GLSLANG_TARGET_SPV,
|
||||
.code = shd->src.str,
|
||||
.default_version = 460,
|
||||
.default_profile = GLSLANG_NO_PROFILE,
|
||||
.force_default_version_and_profile = false,
|
||||
.forward_compatible = false,
|
||||
.messages = GLSLANG_MSG_DEFAULT_BIT,
|
||||
.resource = &glslc_resource_limits,
|
||||
};
|
||||
|
||||
av_assert0(glslang_refcount);
|
||||
|
||||
if (!(glslc_shader = glslang_shader_create(&glslc_input)))
|
||||
return AVERROR(ENOMEM);
|
||||
|
||||
if (!glslang_shader_preprocess(glslc_shader, &glslc_input)) {
|
||||
ff_vk_print_shader(avctx, shd, AV_LOG_WARNING);
|
||||
av_log(avctx, AV_LOG_ERROR, "Unable to preprocess shader: %s (%s)!\n",
|
||||
glslang_shader_get_info_log(glslc_shader),
|
||||
glslang_shader_get_info_debug_log(glslc_shader));
|
||||
glslang_shader_delete(glslc_shader);
|
||||
return AVERROR(EINVAL);
|
||||
}
|
||||
|
||||
if (!glslang_shader_parse(glslc_shader, &glslc_input)) {
|
||||
ff_vk_print_shader(avctx, shd, AV_LOG_WARNING);
|
||||
av_log(avctx, AV_LOG_ERROR, "Unable to parse shader: %s (%s)!\n",
|
||||
glslang_shader_get_info_log(glslc_shader),
|
||||
glslang_shader_get_info_debug_log(glslc_shader));
|
||||
glslang_shader_delete(glslc_shader);
|
||||
return AVERROR(EINVAL);
|
||||
}
|
||||
|
||||
if (!(glslc_program = glslang_program_create())) {
|
||||
glslang_shader_delete(glslc_shader);
|
||||
return AVERROR(EINVAL);
|
||||
}
|
||||
|
||||
glslang_program_add_shader(glslc_program, glslc_shader);
|
||||
|
||||
if (!glslang_program_link(glslc_program, GLSLANG_MSG_SPV_RULES_BIT |
|
||||
GLSLANG_MSG_VULKAN_RULES_BIT)) {
|
||||
ff_vk_print_shader(avctx, shd, AV_LOG_WARNING);
|
||||
av_log(avctx, AV_LOG_ERROR, "Unable to link shader: %s (%s)!\n",
|
||||
glslang_program_get_info_log(glslc_program),
|
||||
glslang_program_get_info_debug_log(glslc_program));
|
||||
glslang_program_delete(glslc_program);
|
||||
glslang_shader_delete(glslc_shader);
|
||||
return AVERROR(EINVAL);
|
||||
}
|
||||
|
||||
glslang_program_SPIRV_generate(glslc_program, glslc_input.stage);
|
||||
|
||||
messages = glslang_program_SPIRV_get_messages(glslc_program);
|
||||
if (messages)
|
||||
av_log(avctx, AV_LOG_WARNING, "%s\n", messages);
|
||||
|
||||
glslang_shader_delete(glslc_shader);
|
||||
|
||||
*size = glslang_program_SPIRV_get_size(glslc_program) * sizeof(unsigned int);
|
||||
*data = (void *)glslang_program_SPIRV_get_ptr(glslc_program);
|
||||
*opaque = glslc_program;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
void ff_vk_glslang_shader_free(void *opaque)
|
||||
{
|
||||
glslang_program_delete(opaque);
|
||||
}
|
||||
|
||||
int ff_vk_glslang_init(void)
|
||||
{
|
||||
int ret = 0;
|
||||
|
||||
pthread_mutex_lock(&glslang_mutex);
|
||||
if (glslang_refcount++ == 0)
|
||||
ret = !glslang_initialize_process();
|
||||
pthread_mutex_unlock(&glslang_mutex);
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
void ff_vk_glslang_uninit(void)
|
||||
{
|
||||
pthread_mutex_lock(&glslang_mutex);
|
||||
if (glslang_refcount && (--glslang_refcount == 0))
|
||||
glslang_finalize_process();
|
||||
pthread_mutex_unlock(&glslang_mutex);
|
||||
}
|
||||
|
|
@ -1,241 +0,0 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* modify it under the terms of the GNU Lesser General Public
|
||||
* License as published by the Free Software Foundation; either
|
||||
* version 2.1 of the License, or (at your option) any later version.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
* Lesser General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU Lesser General Public
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include <pthread.h>
|
||||
|
||||
extern "C" {
|
||||
#include "libavutil/mem.h"
|
||||
#include "libavutil/avassert.h"
|
||||
}
|
||||
|
||||
#include <glslang/Include/ResourceLimits.h>
|
||||
#include <glslang/Public/ShaderLang.h>
|
||||
#include <glslang/SPIRV/GlslangToSpv.h>
|
||||
|
||||
#include "glslang.h"
|
||||
|
||||
using namespace glslang;
|
||||
|
||||
static pthread_mutex_t glslang_mutex = PTHREAD_MUTEX_INITIALIZER;
|
||||
static int glslang_refcount = 0;
|
||||
|
||||
/* We require Vulkan 1.1 */
|
||||
#define GLSL_VERSION EShTargetVulkan_1_1
|
||||
|
||||
/* Vulkan 1.1 implementations require SPIR-V 1.3 to be implemented */
|
||||
#define SPIRV_VERSION EShTargetSpv_1_3
|
||||
|
||||
// Taken from glslang's examples, which apparently generally bases the choices
|
||||
// on OpenGL specification limits
|
||||
static const TBuiltInResource DefaultTBuiltInResource = {
|
||||
/* .MaxLights = */ 32,
|
||||
/* .MaxClipPlanes = */ 6,
|
||||
/* .MaxTextureUnits = */ 32,
|
||||
/* .MaxTextureCoords = */ 32,
|
||||
/* .MaxVertexAttribs = */ 64,
|
||||
/* .MaxVertexUniformComponents = */ 4096,
|
||||
/* .MaxVaryingFloats = */ 64,
|
||||
/* .MaxVertexTextureImageUnits = */ 32,
|
||||
/* .MaxCombinedTextureImageUnits = */ 80,
|
||||
/* .MaxTextureImageUnits = */ 32,
|
||||
/* .MaxFragmentUniformComponents = */ 4096,
|
||||
/* .MaxDrawBuffers = */ 32,
|
||||
/* .MaxVertexUniformVectors = */ 128,
|
||||
/* .MaxVaryingVectors = */ 8,
|
||||
/* .MaxFragmentUniformVectors = */ 16,
|
||||
/* .MaxVertexOutputVectors = */ 16,
|
||||
/* .MaxFragmentInputVectors = */ 15,
|
||||
/* .MinProgramTexelOffset = */ -8,
|
||||
/* .MaxProgramTexelOffset = */ 7,
|
||||
/* .MaxClipDistances = */ 8,
|
||||
/* .MaxComputeWorkGroupCountX = */ 65535,
|
||||
/* .MaxComputeWorkGroupCountY = */ 65535,
|
||||
/* .MaxComputeWorkGroupCountZ = */ 65535,
|
||||
/* .MaxComputeWorkGroupSizeX = */ 1024,
|
||||
/* .MaxComputeWorkGroupSizeY = */ 1024,
|
||||
/* .MaxComputeWorkGroupSizeZ = */ 64,
|
||||
/* .MaxComputeUniformComponents = */ 1024,
|
||||
/* .MaxComputeTextureImageUnits = */ 16,
|
||||
/* .MaxComputeImageUniforms = */ 8,
|
||||
/* .MaxComputeAtomicCounters = */ 8,
|
||||
/* .MaxComputeAtomicCounterBuffers = */ 1,
|
||||
/* .MaxVaryingComponents = */ 60,
|
||||
/* .MaxVertexOutputComponents = */ 64,
|
||||
/* .MaxGeometryInputComponents = */ 64,
|
||||
/* .MaxGeometryOutputComponents = */ 128,
|
||||
/* .MaxFragmentInputComponents = */ 128,
|
||||
/* .MaxImageUnits = */ 8,
|
||||
/* .MaxCombinedImageUnitsAndFragmentOutputs = */ 8,
|
||||
/* .MaxCombinedShaderOutputResources = */ 8,
|
||||
/* .MaxImageSamples = */ 0,
|
||||
/* .MaxVertexImageUniforms = */ 0,
|
||||
/* .MaxTessControlImageUniforms = */ 0,
|
||||
/* .MaxTessEvaluationImageUniforms = */ 0,
|
||||
/* .MaxGeometryImageUniforms = */ 0,
|
||||
/* .MaxFragmentImageUniforms = */ 8,
|
||||
/* .MaxCombinedImageUniforms = */ 8,
|
||||
/* .MaxGeometryTextureImageUnits = */ 16,
|
||||
/* .MaxGeometryOutputVertices = */ 256,
|
||||
/* .MaxGeometryTotalOutputComponents = */ 1024,
|
||||
/* .MaxGeometryUniformComponents = */ 1024,
|
||||
/* .MaxGeometryVaryingComponents = */ 64,
|
||||
/* .MaxTessControlInputComponents = */ 128,
|
||||
/* .MaxTessControlOutputComponents = */ 128,
|
||||
/* .MaxTessControlTextureImageUnits = */ 16,
|
||||
/* .MaxTessControlUniformComponents = */ 1024,
|
||||
/* .MaxTessControlTotalOutputComponents = */ 4096,
|
||||
/* .MaxTessEvaluationInputComponents = */ 128,
|
||||
/* .MaxTessEvaluationOutputComponents = */ 128,
|
||||
/* .MaxTessEvaluationTextureImageUnits = */ 16,
|
||||
/* .MaxTessEvaluationUniformComponents = */ 1024,
|
||||
/* .MaxTessPatchComponents = */ 120,
|
||||
/* .MaxPatchVertices = */ 32,
|
||||
/* .MaxTessGenLevel = */ 64,
|
||||
/* .MaxViewports = */ 16,
|
||||
/* .MaxVertexAtomicCounters = */ 0,
|
||||
/* .MaxTessControlAtomicCounters = */ 0,
|
||||
/* .MaxTessEvaluationAtomicCounters = */ 0,
|
||||
/* .MaxGeometryAtomicCounters = */ 0,
|
||||
/* .MaxFragmentAtomicCounters = */ 8,
|
||||
/* .MaxCombinedAtomicCounters = */ 8,
|
||||
/* .MaxAtomicCounterBindings = */ 1,
|
||||
/* .MaxVertexAtomicCounterBuffers = */ 0,
|
||||
/* .MaxTessControlAtomicCounterBuffers = */ 0,
|
||||
/* .MaxTessEvaluationAtomicCounterBuffers = */ 0,
|
||||
/* .MaxGeometryAtomicCounterBuffers = */ 0,
|
||||
/* .MaxFragmentAtomicCounterBuffers = */ 1,
|
||||
/* .MaxCombinedAtomicCounterBuffers = */ 1,
|
||||
/* .MaxAtomicCounterBufferSize = */ 16384,
|
||||
/* .MaxTransformFeedbackBuffers = */ 4,
|
||||
/* .MaxTransformFeedbackInterleavedComponents = */ 64,
|
||||
/* .MaxCullDistances = */ 8,
|
||||
/* .MaxCombinedClipAndCullDistances = */ 8,
|
||||
/* .MaxSamples = */ 4,
|
||||
/* .maxMeshOutputVerticesNV = */ 256,
|
||||
/* .maxMeshOutputPrimitivesNV = */ 512,
|
||||
/* .maxMeshWorkGroupSizeX_NV = */ 32,
|
||||
/* .maxMeshWorkGroupSizeY_NV = */ 1,
|
||||
/* .maxMeshWorkGroupSizeZ_NV = */ 1,
|
||||
/* .maxTaskWorkGroupSizeX_NV = */ 32,
|
||||
/* .maxTaskWorkGroupSizeY_NV = */ 1,
|
||||
/* .maxTaskWorkGroupSizeZ_NV = */ 1,
|
||||
/* .maxMeshViewCountNV = */ 4,
|
||||
|
||||
.limits = {
|
||||
/* .nonInductiveForLoops = */ 1,
|
||||
/* .whileLoops = */ 1,
|
||||
/* .doWhileLoops = */ 1,
|
||||
/* .generalUniformIndexing = */ 1,
|
||||
/* .generalAttributeMatrixVectorIndexing = */ 1,
|
||||
/* .generalVaryingIndexing = */ 1,
|
||||
/* .generalSamplerIndexing = */ 1,
|
||||
/* .generalVariableIndexing = */ 1,
|
||||
/* .generalConstantMatrixVectorIndexing = */ 1,
|
||||
}
|
||||
};
|
||||
|
||||
GLSlangResult *glslang_compile(const char *glsl, enum GLSlangStage stage)
|
||||
{
|
||||
GLSlangResult *res = (GLSlangResult *)av_mallocz(sizeof(*res));
|
||||
if (!res)
|
||||
return NULL;
|
||||
|
||||
static const EShLanguage lang[] = {
|
||||
[GLSLANG_VERTEX] = EShLangVertex,
|
||||
[GLSLANG_FRAGMENT] = EShLangFragment,
|
||||
[GLSLANG_COMPUTE] = EShLangCompute,
|
||||
};
|
||||
|
||||
assert(glslang_refcount);
|
||||
TShader *shader = new TShader(lang[stage]);
|
||||
if (!shader) {
|
||||
res->rval = AVERROR(ENOMEM);
|
||||
return res;
|
||||
}
|
||||
|
||||
shader->setEnvClient(EShClientVulkan, GLSL_VERSION);
|
||||
shader->setEnvTarget(EShTargetSpv, SPIRV_VERSION);
|
||||
shader->setStrings(&glsl, 1);
|
||||
if (!shader->parse(&DefaultTBuiltInResource, GLSL_VERSION, true, EShMsgDefault)) {
|
||||
res->error_msg = av_strdup(shader->getInfoLog());
|
||||
res->rval = AVERROR_EXTERNAL;
|
||||
delete shader;
|
||||
return res;
|
||||
}
|
||||
|
||||
TProgram *prog = new TProgram();
|
||||
if (!prog) {
|
||||
res->rval = AVERROR(ENOMEM);
|
||||
delete shader;
|
||||
return res;
|
||||
}
|
||||
|
||||
prog->addShader(shader);
|
||||
if (!prog->link(EShMsgDefault)) {
|
||||
res->error_msg = av_strdup(prog->getInfoLog());
|
||||
res->rval = AVERROR_EXTERNAL;
|
||||
delete shader;
|
||||
delete prog;
|
||||
return res;
|
||||
}
|
||||
|
||||
std::vector<unsigned int> spirv; /* Result */
|
||||
|
||||
SpvOptions options; /* Options - by default all optimizations are off */
|
||||
options.generateDebugInfo = false; /* Makes sense for files but not here */
|
||||
options.disassemble = false; /* Will print disassembly on compilation */
|
||||
options.validate = false; /* Validates the generated SPIRV, unneeded */
|
||||
options.disableOptimizer = false; /* For debugging */
|
||||
options.optimizeSize = true; /* Its faster */
|
||||
|
||||
GlslangToSpv(*prog->getIntermediate(lang[stage]), spirv, NULL, &options);
|
||||
|
||||
res->size = spirv.size()*sizeof(unsigned int);
|
||||
res->data = av_memdup(spirv.data(), res->size);
|
||||
if (!res->data) {
|
||||
res->rval = AVERROR(ENOMEM);
|
||||
delete shader;
|
||||
delete prog;
|
||||
return res;
|
||||
}
|
||||
|
||||
delete shader;
|
||||
delete prog;
|
||||
|
||||
return res;
|
||||
}
|
||||
|
||||
int glslang_init(void)
|
||||
{
|
||||
int ret = 0;
|
||||
|
||||
pthread_mutex_lock(&glslang_mutex);
|
||||
if (glslang_refcount++ == 0)
|
||||
ret = !InitializeProcess();
|
||||
pthread_mutex_unlock(&glslang_mutex);
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
void glslang_uninit(void)
|
||||
{
|
||||
pthread_mutex_lock(&glslang_mutex);
|
||||
if (glslang_refcount && (--glslang_refcount == 0))
|
||||
FinalizeProcess();
|
||||
pthread_mutex_unlock(&glslang_mutex);
|
||||
}
|
||||
|
|
@ -19,34 +19,23 @@
|
|||
#ifndef AVFILTER_GLSLANG_H
|
||||
#define AVFILTER_GLSLANG_H
|
||||
|
||||
#include <stdlib.h>
|
||||
#include "vulkan.h"
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C" {
|
||||
#endif
|
||||
/**
|
||||
* Un/initialize glslang's global state. Thread-safe and reference counted.
|
||||
*/
|
||||
int ff_vk_glslang_init(void);
|
||||
void ff_vk_glslang_uninit(void);
|
||||
|
||||
int glslang_init(void);
|
||||
void glslang_uninit(void);
|
||||
/**
|
||||
* Compile GLSL into SPIR-V using glslang.
|
||||
*/
|
||||
int ff_vk_glslang_shader_compile(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
uint8_t **data, size_t *size, void **opaque);
|
||||
|
||||
typedef struct GLSlangResult {
|
||||
int rval;
|
||||
char *error_msg;
|
||||
|
||||
void *data; /* Shader data or NULL */
|
||||
size_t size;
|
||||
} GLSlangResult;
|
||||
|
||||
enum GLSlangStage {
|
||||
GLSLANG_VERTEX,
|
||||
GLSLANG_FRAGMENT,
|
||||
GLSLANG_COMPUTE,
|
||||
};
|
||||
|
||||
/* Compile GLSL into a SPIRV stream, if possible */
|
||||
GLSlangResult *glslang_compile(const char *glsl, enum GLSlangStage stage);
|
||||
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
/**
|
||||
* Frees the shader-specific context.
|
||||
*/
|
||||
void ff_vk_glslang_shader_free(void *opaque);
|
||||
|
||||
#endif /* AVFILTER_GLSLANG_H */
|
||||
|
|
|
|||
|
|
@ -70,7 +70,7 @@ static const char blur_kernel[] = {
|
|||
static av_cold int init_filter(AVFilterContext *ctx, AVFrame *in)
|
||||
{
|
||||
int err;
|
||||
SPIRVShader *shd;
|
||||
FFSPIRVShader *shd;
|
||||
AvgBlurVulkanContext *s = ctx->priv;
|
||||
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
|
||||
VkSampler *sampler = ff_vk_init_sampler(ctx, 1, VK_FILTER_LINEAR);
|
||||
|
|
|
|||
|
|
@ -109,8 +109,8 @@ static av_cold int init_filter(AVFilterContext *ctx, AVFrame *in)
|
|||
},
|
||||
};
|
||||
|
||||
SPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "chromaber_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
FFSPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "chromaber_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
if (!shd)
|
||||
return AVERROR(ENOMEM);
|
||||
|
||||
|
|
|
|||
|
|
@ -136,8 +136,8 @@ static av_cold int init_filter(AVFilterContext *ctx)
|
|||
.buf_content = "ivec2 o_offset[3], o_size[3];",
|
||||
};
|
||||
|
||||
SPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "overlay_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
FFSPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "overlay_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
if (!shd)
|
||||
return AVERROR(ENOMEM);
|
||||
|
||||
|
|
|
|||
|
|
@ -170,8 +170,8 @@ static av_cold int init_filter(AVFilterContext *ctx, AVFrame *in)
|
|||
.buf_content = "mat4 yuv_matrix;",
|
||||
};
|
||||
|
||||
SPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "scale_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
FFSPIRVShader *shd = ff_vk_init_shader(ctx, s->pl, "scale_compute",
|
||||
VK_SHADER_STAGE_COMPUTE_BIT);
|
||||
if (!shd)
|
||||
return AVERROR(ENOMEM);
|
||||
|
||||
|
|
|
|||
|
|
@ -794,7 +794,7 @@ int ff_vk_filter_init(AVFilterContext *avctx)
|
|||
|
||||
s->output_format = AV_PIX_FMT_NONE;
|
||||
|
||||
if (glslang_init())
|
||||
if (ff_vk_glslang_init())
|
||||
return AVERROR_EXTERNAL;
|
||||
|
||||
return 0;
|
||||
|
|
@ -924,11 +924,11 @@ int ff_vk_create_imageview(AVFilterContext *avctx, FFVkExecContext *e,
|
|||
return 0;
|
||||
}
|
||||
|
||||
FN_CREATING(VulkanPipeline, SPIRVShader, shader, shaders, shaders_num)
|
||||
SPIRVShader *ff_vk_init_shader(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
const char *name, VkShaderStageFlags stage)
|
||||
FN_CREATING(VulkanPipeline, FFSPIRVShader, shader, shaders, shaders_num)
|
||||
FFSPIRVShader *ff_vk_init_shader(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
const char *name, VkShaderStageFlags stage)
|
||||
{
|
||||
SPIRVShader *shd = create_shader(pl);
|
||||
FFSPIRVShader *shd = create_shader(pl);
|
||||
if (!shd)
|
||||
return NULL;
|
||||
|
||||
|
|
@ -946,7 +946,7 @@ SPIRVShader *ff_vk_init_shader(AVFilterContext *avctx, VulkanPipeline *pl,
|
|||
return shd;
|
||||
}
|
||||
|
||||
void ff_vk_set_compute_shader_sizes(AVFilterContext *avctx, SPIRVShader *shd,
|
||||
void ff_vk_set_compute_shader_sizes(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
int local_size[3])
|
||||
{
|
||||
shd->local_size[0] = local_size[0];
|
||||
|
|
@ -958,7 +958,7 @@ void ff_vk_set_compute_shader_sizes(AVFilterContext *avctx, SPIRVShader *shd,
|
|||
shd->local_size[0], shd->local_size[1], shd->local_size[2]);
|
||||
}
|
||||
|
||||
static void print_shader(AVFilterContext *avctx, SPIRVShader *shd, int prio)
|
||||
void ff_vk_print_shader(AVFilterContext *avctx, FFSPIRVShader *shd, int prio)
|
||||
{
|
||||
int line = 0;
|
||||
const char *p = shd->src.str;
|
||||
|
|
@ -979,51 +979,39 @@ static void print_shader(AVFilterContext *avctx, SPIRVShader *shd, int prio)
|
|||
av_bprint_finalize(&buf, NULL);
|
||||
}
|
||||
|
||||
int ff_vk_compile_shader(AVFilterContext *avctx, SPIRVShader *shd,
|
||||
int ff_vk_compile_shader(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
const char *entrypoint)
|
||||
{
|
||||
int err;
|
||||
VkResult ret;
|
||||
VulkanFilterContext *s = avctx->priv;
|
||||
FFVulkanFunctions *vk = &s->vkfn;
|
||||
VkShaderModuleCreateInfo shader_create;
|
||||
GLSlangResult *res;
|
||||
|
||||
static const enum GLSlangStage emap[] = {
|
||||
[VK_SHADER_STAGE_VERTEX_BIT] = GLSLANG_VERTEX,
|
||||
[VK_SHADER_STAGE_FRAGMENT_BIT] = GLSLANG_FRAGMENT,
|
||||
[VK_SHADER_STAGE_COMPUTE_BIT] = GLSLANG_COMPUTE,
|
||||
};
|
||||
uint8_t *spirv;
|
||||
size_t spirv_size;
|
||||
void *priv;
|
||||
|
||||
shd->shader.pName = entrypoint;
|
||||
|
||||
res = glslang_compile(shd->src.str, emap[shd->shader.stage]);
|
||||
if (!res)
|
||||
return AVERROR(ENOMEM);
|
||||
err = ff_vk_glslang_shader_compile(avctx, shd, &spirv, &spirv_size, &priv);
|
||||
if (err < 0)
|
||||
return err;
|
||||
|
||||
if (res->rval) {
|
||||
av_log(avctx, AV_LOG_ERROR, "Error compiling shader %s: %s!\n",
|
||||
shd->name, av_err2str(res->rval));
|
||||
print_shader(avctx, shd, AV_LOG_ERROR);
|
||||
if (res->error_msg)
|
||||
av_log(avctx, AV_LOG_ERROR, "%s", res->error_msg);
|
||||
av_free(res->error_msg);
|
||||
return res->rval;
|
||||
}
|
||||
ff_vk_print_shader(avctx, shd, AV_LOG_VERBOSE);
|
||||
|
||||
print_shader(avctx, shd, AV_LOG_VERBOSE);
|
||||
av_log(avctx, AV_LOG_VERBOSE, "Shader %s compiled! Size: %zu bytes\n",
|
||||
shd->name, spirv_size);
|
||||
|
||||
shader_create.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
|
||||
shader_create.pNext = NULL;
|
||||
shader_create.codeSize = res->size;
|
||||
shader_create.codeSize = spirv_size;
|
||||
shader_create.flags = 0;
|
||||
shader_create.pCode = res->data;
|
||||
shader_create.pCode = (void *)spirv;
|
||||
|
||||
ret = vk->CreateShaderModule(s->hwctx->act_dev, &shader_create, NULL,
|
||||
&shd->shader.module);
|
||||
|
||||
/* Free the GLSlangResult struct */
|
||||
av_free(res->data);
|
||||
av_free(res);
|
||||
ff_vk_glslang_shader_free(priv);
|
||||
|
||||
if (ret != VK_SUCCESS) {
|
||||
av_log(avctx, AV_LOG_ERROR, "Unable to create shader module: %s\n",
|
||||
|
|
@ -1031,9 +1019,6 @@ int ff_vk_compile_shader(AVFilterContext *avctx, SPIRVShader *shd,
|
|||
return AVERROR_EXTERNAL;
|
||||
}
|
||||
|
||||
av_log(avctx, AV_LOG_VERBOSE, "Shader %s linked! Size: %zu bytes\n",
|
||||
shd->name, shader_create.codeSize);
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
|
@ -1059,7 +1044,7 @@ static const struct descriptor_props {
|
|||
};
|
||||
|
||||
int ff_vk_add_descriptor_set(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
SPIRVShader *shd, VulkanDescriptorSetBinding *desc,
|
||||
FFSPIRVShader *shd, VulkanDescriptorSetBinding *desc,
|
||||
int num, int only_print_to_shader)
|
||||
{
|
||||
VkResult ret;
|
||||
|
|
@ -1423,7 +1408,7 @@ static void free_pipeline(VulkanFilterContext *s, VulkanPipeline *pl)
|
|||
FFVulkanFunctions *vk = &s->vkfn;
|
||||
|
||||
for (int i = 0; i < pl->shaders_num; i++) {
|
||||
SPIRVShader *shd = pl->shaders[i];
|
||||
FFSPIRVShader *shd = pl->shaders[i];
|
||||
av_bprint_finalize(&shd->src, NULL);
|
||||
vk->DestroyShaderModule(s->hwctx->act_dev, shd->shader.module,
|
||||
s->hwctx->alloc);
|
||||
|
|
@ -1471,7 +1456,7 @@ void ff_vk_filter_uninit(AVFilterContext *avctx)
|
|||
VulkanFilterContext *s = avctx->priv;
|
||||
FFVulkanFunctions *vk = &s->vkfn;
|
||||
|
||||
glslang_uninit();
|
||||
ff_vk_glslang_uninit();
|
||||
|
||||
for (int i = 0; i < s->exec_ctx_num; i++)
|
||||
free_exec_ctx(s, s->exec_ctx[i]);
|
||||
|
|
|
|||
|
|
@ -55,12 +55,12 @@
|
|||
/* Useful for attaching immutable samplers to arrays */
|
||||
#define DUP_SAMPLER_ARRAY4(x) (VkSampler []){ x, x, x, x, }
|
||||
|
||||
typedef struct SPIRVShader {
|
||||
typedef struct FFSPIRVShader {
|
||||
const char *name; /* Name for id/debugging purposes */
|
||||
AVBPrint src;
|
||||
int local_size[3]; /* Compute shader workgroup sizes */
|
||||
VkPipelineShaderStageCreateInfo shader;
|
||||
} SPIRVShader;
|
||||
} FFSPIRVShader;
|
||||
|
||||
typedef struct VulkanDescriptorSetBinding {
|
||||
const char *name;
|
||||
|
|
@ -89,7 +89,7 @@ typedef struct VulkanPipeline {
|
|||
VkPipeline pipeline;
|
||||
|
||||
/* Shaders */
|
||||
SPIRVShader **shaders;
|
||||
FFSPIRVShader **shaders;
|
||||
int shaders_num;
|
||||
|
||||
/* Push consts */
|
||||
|
|
@ -249,28 +249,33 @@ VulkanPipeline *ff_vk_create_pipeline(AVFilterContext *avctx);
|
|||
/**
|
||||
* Inits a shader for a specific pipeline. Will be auto-freed on uninit.
|
||||
*/
|
||||
SPIRVShader *ff_vk_init_shader(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
const char *name, VkShaderStageFlags stage);
|
||||
FFSPIRVShader *ff_vk_init_shader(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
const char *name, VkShaderStageFlags stage);
|
||||
|
||||
/**
|
||||
* Writes the workgroup size for a shader.
|
||||
*/
|
||||
void ff_vk_set_compute_shader_sizes(AVFilterContext *avctx, SPIRVShader *shd,
|
||||
void ff_vk_set_compute_shader_sizes(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
int local_size[3]);
|
||||
|
||||
/**
|
||||
* Adds a descriptor set to the shader and registers them in the pipeline.
|
||||
*/
|
||||
int ff_vk_add_descriptor_set(AVFilterContext *avctx, VulkanPipeline *pl,
|
||||
SPIRVShader *shd, VulkanDescriptorSetBinding *desc,
|
||||
FFSPIRVShader *shd, VulkanDescriptorSetBinding *desc,
|
||||
int num, int only_print_to_shader);
|
||||
|
||||
/**
|
||||
* Compiles the shader, entrypoint must be set to "main".
|
||||
*/
|
||||
int ff_vk_compile_shader(AVFilterContext *avctx, SPIRVShader *shd,
|
||||
int ff_vk_compile_shader(AVFilterContext *avctx, FFSPIRVShader *shd,
|
||||
const char *entrypoint);
|
||||
|
||||
/**
|
||||
* Pretty print shader, mainly used by shader compilers.
|
||||
*/
|
||||
void ff_vk_print_shader(AVFilterContext *avctx, FFSPIRVShader *shd, int prio);
|
||||
|
||||
/**
|
||||
* Initializes the pipeline layout after all shaders and descriptor sets have
|
||||
* been finished.
|
||||
|
|
|
|||
Loading…
Reference in New Issue