RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

blend_vulkan: port for the rewrite

This commit is contained in:
Lynne 2023-02-17 03:11:43 +01:00
parent f29850e98f
commit 40bf1d525e
No known key found for this signature in database
GPG Key ID: A2FEA5F03F034464
1 changed files with 107 additions and 217 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

324
libavfilter/vf_blend_vulkan.c
View File

@ -1,5 +1,7 @@
/*
* copyright (c) 2021-2022 Wu Jianhua <jianhua.wu@intel.com>
* Copyright (c) Lynne
*
* The blend modes are based on the blend.c.
*
* This file is part of FFmpeg.
@ -22,12 +24,11 @@
#include "libavutil/random_seed.h"
#include "libavutil/opt.h"
#include "vulkan_filter.h"
#include "vulkan_spirv.h"
#include "internal.h"
#include "framesync.h"
#include "blend.h"
#define CGS 32
#define IN_TOP 0
#define IN_BOTTOM 1
@ -40,20 +41,18 @@ typedef struct FilterParamsVulkan {
typedef struct BlendVulkanContext {
FFVulkanContext vkctx;
FFVkQueueFamilyCtx qf;
FFVkExecContext *exec;
FFVulkanPipeline *pl;
FFFrameSync fs;
VkDescriptorImageInfo top_images[3];
VkDescriptorImageInfo bottom_images[3];
VkDescriptorImageInfo output_images[3];
int initialized;
FFVulkanPipeline pl;
FFVkExecPool e;
FFVkQueueFamilyCtx qf;
FFVkSPIRVShader shd;
VkSampler sampler;
FilterParamsVulkan params[4];
double all_opacity;
enum BlendMode all_mode;
int initialized;
} BlendVulkanContext;
#define DEFINE_BLEND_MODE(MODE, EXPR) \
@ -125,223 +124,103 @@ static int process_command(AVFilterContext *ctx, const char *cmd, const char *ar
static av_cold int init_filter(AVFilterContext *avctx)
{
int err = 0;
FFVkSampler *sampler;
FFVkSPIRVShader *shd;
uint8_t *spv_data;
size_t spv_len;
void *spv_opaque = NULL;
BlendVulkanContext *s = avctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
FFVkSPIRVShader *shd = &s->shd;
FFVkSPIRVCompiler *spv;
FFVulkanDescriptorSetBinding *desc;
ff_vk_qf_init(vkctx, &s->qf, VK_QUEUE_COMPUTE_BIT, 0);
sampler = ff_vk_init_sampler(vkctx, 1, VK_FILTER_LINEAR);
if (!sampler)
spv = ff_vk_spirv_init();
if (!spv) {
av_log(avctx, AV_LOG_ERROR, "Unable to initialize SPIR-V compiler!\n");
return AVERROR_EXTERNAL;
s->pl = ff_vk_create_pipeline(vkctx, &s->qf);
if (!s->pl)
return AVERROR(ENOMEM);
{
FFVulkanDescriptorSetBinding image_descs[] = {
{
.name = "top_images",
.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
.updater = s->top_images,
.sampler = sampler,
},
{
.name = "bottom_images",
.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
.updater = s->bottom_images,
.sampler = sampler,
},
{
.name = "output_images",
.type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
.mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format),
.mem_quali = "writeonly",
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
.updater = s->output_images,
},
};
shd = ff_vk_init_shader(s->pl, "blend_compute", image_descs[0].stages);
if (!shd)
return AVERROR(ENOMEM);
ff_vk_set_compute_shader_sizes(shd, (int [3]){ CGS, CGS, 1 });
RET(ff_vk_add_descriptor_set(vkctx, s->pl, shd, image_descs, FF_ARRAY_ELEMS(image_descs), 0));
for (int i = 0, j = 0; i < planes; i++) {
for (j = 0; j < i; j++)
if (s->params[i].blend_func == s->params[j].blend_func)
break;
/* note: the bracket is needed, for GLSLD is a macro with multiple statements. */
if (j == i) {
GLSLD(s->params[i].blend_func);
}
}
GLSLC(0, void main() );
GLSLC(0, { );
GLSLC(1, ivec2 size; );
GLSLC(1, const ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
for (int i = 0; i < planes; i++) {
GLSLC(0, );
GLSLF(1, size = imageSize(output_images[%i]); ,i);
GLSLC(1, if (IS_WITHIN(pos, size)) { );
GLSLF(2, const vec4 top = texture(top_images[%i], pos); ,i);
GLSLF(2, const vec4 bottom = texture(bottom_images[%i], pos); ,i);
GLSLF(2, const float opacity = %f; ,s->params[i].opacity);
GLSLF(2, vec4 dst = %s(top, bottom, opacity); ,s->params[i].blend);
GLSLC(0, );
GLSLF(2, imageStore(output_images[%i], pos, dst); ,i);
GLSLC(1, } );
}
GLSLC(0, } );
RET(ff_vk_compile_shader(vkctx, shd, "main"));
RET(ff_vk_init_pipeline_layout(vkctx, s->pl));
RET(ff_vk_init_compute_pipeline(vkctx, s->pl));
}
RET(ff_vk_create_exec_ctx(vkctx, &s->exec, &s->qf));
ff_vk_qf_init(vkctx, &s->qf, VK_QUEUE_COMPUTE_BIT);
RET(ff_vk_exec_pool_init(vkctx, &s->qf, &s->e, s->qf.nb_queues*4, 0, 0, 0, NULL));
RET(ff_vk_init_sampler(vkctx, &s->sampler, 1, VK_FILTER_NEAREST));
RET(ff_vk_shader_init(&s->pl, &s->shd, "blend_compute",
VK_SHADER_STAGE_COMPUTE_BIT, 0));
ff_vk_shader_set_compute_sizes(&s->shd, 32, 32, 1);
desc = (FFVulkanDescriptorSetBinding []) {
{
.name = "top_images",
.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
.samplers = DUP_SAMPLER(s->sampler),
},
{
.name = "bottom_images",
.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
.samplers = DUP_SAMPLER(s->sampler),
},
{
.name = "output_images",
.type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
.mem_layout = ff_vk_shader_rep_fmt(s->vkctx.output_format),
.mem_quali = "writeonly",
.dimensions = 2,
.elems = planes,
.stages = VK_SHADER_STAGE_COMPUTE_BIT,
},
};
RET(ff_vk_pipeline_descriptor_set_add(vkctx, &s->pl, shd, desc, 3, 0, 0));
for (int i = 0, j = 0; i < planes; i++) {
for (j = 0; j < i; j++)
if (s->params[i].blend_func == s->params[j].blend_func)
break;
/* note: the bracket is needed, for GLSLD is a macro with multiple statements. */
if (j == i) {
GLSLD(s->params[i].blend_func);
}
}
GLSLC(0, void main() );
GLSLC(0, { );
GLSLC(1, ivec2 size; );
GLSLC(1, const ivec2 pos = ivec2(gl_GlobalInvocationID.xy); );
for (int i = 0; i < planes; i++) {
GLSLC(0, );
GLSLF(1, size = imageSize(output_images[%i]); ,i);
GLSLC(1, if (IS_WITHIN(pos, size)) { );
GLSLF(2, const vec4 top = texture(top_images[%i], pos); ,i);
GLSLF(2, const vec4 bottom = texture(bottom_images[%i], pos); ,i);
GLSLF(2, const float opacity = %f; ,s->params[i].opacity);
GLSLF(2, vec4 dst = %s(top, bottom, opacity); ,s->params[i].blend);
GLSLC(0, );
GLSLF(2, imageStore(output_images[%i], pos, dst); ,i);
GLSLC(1, } );
}
GLSLC(0, } );
RET(spv->compile_shader(spv, avctx, shd, &spv_data, &spv_len, "main",
&spv_opaque));
RET(ff_vk_shader_create(vkctx, shd, spv_data, spv_len, "main"));
RET(ff_vk_init_compute_pipeline(vkctx, &s->pl, shd));
RET(ff_vk_exec_pipeline_register(vkctx, &s->e, &s->pl));
s->initialized = 1;
fail:
return err;
}
if (spv_opaque)
spv->free_shader(spv, &spv_opaque);
if (spv)
spv->uninit(&spv);
static int process_frames(AVFilterContext *avctx, AVFrame *out_frame, AVFrame *top_frame, AVFrame *bottom_frame)
{
int err = 0;
VkCommandBuffer cmd_buf;
BlendVulkanContext *s = avctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
FFVulkanFunctions *vk = &s->vkctx.vkfn;
const int planes = av_pix_fmt_count_planes(s->vkctx.output_format);
AVVkFrame *out = (AVVkFrame *)out_frame->data[0];
AVVkFrame *top = (AVVkFrame *)top_frame->data[0];
AVVkFrame *bottom = (AVVkFrame *)bottom_frame->data[0];
AVHWFramesContext *top_fc = (AVHWFramesContext*)top_frame->hw_frames_ctx->data;
AVHWFramesContext *bottom_fc = (AVHWFramesContext*)bottom_frame->hw_frames_ctx->data;
const VkFormat *top_formats = av_vkfmt_from_pixfmt(top_fc->sw_format);
const VkFormat *bottom_formats = av_vkfmt_from_pixfmt(bottom_fc->sw_format);
const VkFormat *output_formats = av_vkfmt_from_pixfmt(s->vkctx.output_format);
ff_vk_start_exec_recording(vkctx, s->exec);
cmd_buf = ff_vk_get_exec_buf(s->exec);
for (int i = 0; i < planes; i++) {
RET(ff_vk_create_imageview(vkctx, s->exec,
&s->top_images[i].imageView, top->img[i],
top_formats[i],
ff_comp_identity_map));
RET(ff_vk_create_imageview(vkctx, s->exec,
&s->bottom_images[i].imageView, bottom->img[i],
bottom_formats[i],
ff_comp_identity_map));
RET(ff_vk_create_imageview(vkctx, s->exec,
&s->output_images[i].imageView, out->img[i],
output_formats[i],
ff_comp_identity_map));
s->top_images[i].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
s->bottom_images[i].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
s->output_images[i].imageLayout = VK_IMAGE_LAYOUT_GENERAL;
}
ff_vk_update_descriptor_set(vkctx, s->pl, 0);
for (int i = 0; i < planes; i++) {
VkImageMemoryBarrier barriers[] = {
{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.srcAccessMask = 0,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
.oldLayout = top->layout[i],
.newLayout = s->top_images[i].imageLayout,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = top->img[i],
.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.subresourceRange.levelCount = 1,
.subresourceRange.layerCount = 1,
},
{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.srcAccessMask = 0,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
.oldLayout = bottom->layout[i],
.newLayout = s->bottom_images[i].imageLayout,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = bottom->img[i],
.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.subresourceRange.levelCount = 1,
.subresourceRange.layerCount = 1,
},
{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.srcAccessMask = 0,
.dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.oldLayout = out->layout[i],
.newLayout = s->output_images[i].imageLayout,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = out->img[i],
.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.subresourceRange.levelCount = 1,
.subresourceRange.layerCount = 1,
},
};
vk->CmdPipelineBarrier(cmd_buf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0,
0, NULL, 0, NULL, FF_ARRAY_ELEMS(barriers), barriers);
top->layout[i] = barriers[0].newLayout;
top->access[i] = barriers[0].dstAccessMask;
bottom->layout[i] = barriers[1].newLayout;
bottom->access[i] = barriers[1].dstAccessMask;
out->layout[i] = barriers[2].newLayout;
out->access[i] = barriers[2].dstAccessMask;
}
ff_vk_bind_pipeline_exec(vkctx, s->exec, s->pl);
vk->CmdDispatch(cmd_buf, FFALIGN(s->vkctx.output_width, CGS) / CGS,
FFALIGN(s->vkctx.output_height, CGS) / CGS, 1);
ff_vk_add_exec_dep(vkctx, s->exec, top_frame, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT);
ff_vk_add_exec_dep(vkctx, s->exec, bottom_frame, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT);
ff_vk_add_exec_dep(vkctx, s->exec, out_frame, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT);
err = ff_vk_submit_exec_queue(vkctx, s->exec);
if (err)
return err;
ff_vk_qf_rotate(&s->qf);
return 0;
fail:
ff_vk_discard_exec_deps(s->exec);
return err;
}
@ -375,7 +254,9 @@ static int blend_frame(FFFrameSync *fs)
RET(init_filter(avctx));
}
RET(process_frames(avctx, out, top, bottom));
RET(ff_vk_filter_process_Nin(&s->vkctx, &s->e, &s->pl,
out, (AVFrame *[]){ top, bottom }, 2,
s->sampler, NULL, 0));
return ff_filter_frame(outlink, out);
@ -396,10 +277,19 @@ static av_cold int init(AVFilterContext *avctx)
static av_cold void uninit(AVFilterContext *avctx)
{
BlendVulkanContext *s = avctx->priv;
FFVulkanContext *vkctx = &s->vkctx;
FFVulkanFunctions *vk = &vkctx->vkfn;
ff_framesync_uninit(&s->fs);
ff_vk_exec_pool_free(vkctx, &s->e);
ff_vk_pipeline_free(vkctx, &s->pl);
ff_vk_shader_free(vkctx, &s->shd);
if (s->sampler)
vk->DestroySampler(vkctx->hwctx->act_dev, s->sampler,
vkctx->hwctx->alloc);
ff_vk_uninit(&s->vkctx);
ff_framesync_uninit(&s->fs);
s->initialized = 0;
}