[client] initial vaapi h264 decode support (unfinished)

This commit is contained in:
Geoffrey McRae 2017-12-31 00:27:26 +11:00
parent a6d2fe73ae
commit fbf08b94aa
5 changed files with 1095 additions and 196 deletions

View File

@ -1,12 +1,12 @@
BINARY = looking-glass-client
CFLAGS = -g -Og -std=gnu99 -march=native -Wall -Werror -I./ -I../common -DDEBUG -DATOMIC_LOCKING
CFLAGS = -g -O3 -std=gnu99 -march=native -Wall -Werror -I./ -I../common -DDEBUG -DATOMIC_LOCKING
LDFLAGS = -lrt
CFLAGS += -ffast-math
CFLAGS += -fdata-sections -ffunction-sections
LDFLAGS += -Wl,--gc-sections
LIBS = sdl2 SDL2_ttf gl glu libssl openssl spice-protocol fontconfig x11 libconfig libva
LIBS = sdl2 SDL2_ttf gl glu libssl openssl spice-protocol fontconfig x11 libconfig libva libva-glx
CFLAGS += $(shell pkg-config --cflags $(LIBS))
LDFLAGS += $(shell pkg-config --libs $(LIBS))
BUILD ?= .build

View File

@ -20,24 +20,107 @@ Place, Suite 330, Boston, MA 02111-1307 USA
#include "lg-decoder.h"
#include "debug.h"
#include "memcpySSE.h"
#include <stdlib.h>
#include <string.h>
#include <SDL2/SDL_syswm.h>
#include <va/va_glx.h>
#define SURFACE_NUM 3
#define NALU_AUD 9
#define SLICE_TYPE_P 0
#define SLICE_TYPE_B 1
#define SLICE_TYPE_I 2
struct Inst
{
bool initialized;
LG_RendererFormat format;
SDL_Window * window;
VADisplay vaDisplay;
int vaMajorVer, vaMinorVer;
VASurfaceID vaSurfaceID[SURFACE_NUM];
VAConfigID vaConfigID;
VAContextID vaContextID;
int lastSID;
int currentSID;
VAPictureH264 curPic;
VAPictureH264 oldPic;
int frameNum;
int fieldCount;
VABufferID picBufferID[SURFACE_NUM];
VABufferID matBufferID[SURFACE_NUM];
VABufferID sliBufferID[SURFACE_NUM];
VABufferID datBufferID[SURFACE_NUM];
bool t2First;
int sliceType;
};
static const unsigned char MatrixBufferH264[] = {
//ScalingList4x4[6][16]
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,
//ScalingList8x8[2][64]
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
};
static bool lgd_h264_create (void ** opaque);
static void lgd_h264_destroy (void * opaque);
static bool lgd_h264_initialize (void * opaque, const LG_RendererFormat format);
static bool lgd_h264_initialize (void * opaque, const LG_RendererFormat format, SDL_Window * window);
static void lgd_h264_deinitialize (void * opaque);
static LG_OutFormat lgd_h264_get_out_format (void * opaque);
static unsigned int lgd_h264_get_frame_pitch(void * opaque);
static bool lgd_h264_decode(void * opaque, uint8_t * dst, size_t dstSize, const uint8_t * src, size_t srcSize);
static unsigned int lgd_h264_get_frame_pitch (void * opaque);
static unsigned int lgd_h264_get_frame_stride(void * opaque);
static bool lgd_h264_decode (void * opaque, const uint8_t * src, size_t srcSize);
static bool lgd_h264_get_buffer (void * opaque, uint8_t * dst, size_t dstSize);
static bool lgd_h264_init_gl_texture (void * opaque, GLenum target, GLuint texture, void ** ref);
static void lgd_h264_free_gl_texture (void * opaque, void * ref);
static bool lgd_h264_update_gl_texture(void * opaque, void * ref);
#define check_surface(x, y, z) _check_surface(__LINE__, x, y, z)
static bool _check_surface(const unsigned int line, struct Inst * this, unsigned int sid, VASurfaceStatus *out)
{
VASurfaceStatus surfStatus;
VAStatus status = vaQuerySurfaceStatus(
this->vaDisplay,
this->vaSurfaceID[sid],
&surfStatus
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaQuerySurfaceStatus: %s", vaErrorStr(status));
return false;
}
#if 0
DEBUG_INFO("L%d: surface %u status: %d", line, sid, surfStatus);
#endif
if (out)
*out = surfStatus;
return true;
}
static bool lgd_h264_create(void ** opaque)
{
@ -49,41 +132,201 @@ static bool lgd_h264_create(void ** opaque)
return false;
}
memset(*opaque, 0, sizeof(struct Inst));
struct Inst * this = (struct Inst *)*opaque;
//struct Inst * this = (struct Inst *)*opaque;
this->vaSurfaceID[0] = VA_INVALID_ID;
this->vaConfigID = VA_INVALID_ID;
this->vaContextID = VA_INVALID_ID;
for(int i = 0; i < SURFACE_NUM; ++i)
this->picBufferID[i] =
this->matBufferID[i] =
this->sliBufferID[i] =
this->datBufferID[i] = VA_INVALID_ID;
lgd_h264_deinitialize(this);
return true;
}
static void lgd_h264_destroy(void * opaque)
{
struct Inst * this = (struct Inst *)opaque;
lgd_h264_deinitialize(opaque);
lgd_h264_deinitialize(this);
free(this);
}
static bool lgd_h264_initialize(void * opaque, const LG_RendererFormat format)
static bool lgd_h264_initialize(void * opaque, const LG_RendererFormat format, SDL_Window * window)
{
struct Inst * this = (struct Inst *)opaque;
if (this->initialized)
lgd_h264_deinitialize(opaque);
lgd_h264_deinitialize(this);
memcpy(&this->format, &format, sizeof(LG_RendererFormat));
this->window = window;
SDL_SysWMinfo wminfo;
SDL_VERSION(&wminfo.version);
if (!SDL_GetWindowWMInfo(window, &wminfo))
{
DEBUG_ERROR("Failed to get SDL window WM Info");
return false;
}
switch(wminfo.subsystem)
{
case SDL_SYSWM_X11:
this->vaDisplay = vaGetDisplayGLX(wminfo.info.x11.display);
break;
default:
DEBUG_ERROR("Unsupported window subsystem");
return false;
}
VAStatus status;
status = vaInitialize(this->vaDisplay, &this->vaMajorVer, &this->vaMinorVer);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaInitialize Failed");
return false;
}
DEBUG_INFO("Vendor: %s", vaQueryVendorString(this->vaDisplay));
VAEntrypoint entryPoints[5];
int entryPointCount;
status = vaQueryConfigEntrypoints(
this->vaDisplay,
VAProfileH264Baseline,
entryPoints,
&entryPointCount
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaQueryConfigEntrypoints Failed");
return false;
}
int ep;
for(ep = 0; ep < entryPointCount; ++ep)
if (entryPoints[ep] == VAEntrypointVLD)
break;
if (ep == entryPointCount)
{
DEBUG_ERROR("Failed to find VAEntrypointVLD index");
return false;
}
VAConfigAttrib attrib;
attrib.type = VAConfigAttribRTFormat;
vaGetConfigAttributes(this->vaDisplay, VAProfileH264Baseline, VAEntrypointVLD, &attrib, 1);
if (!(attrib.value & VA_RT_FORMAT_YUV420))
{
DEBUG_ERROR("Failed to find desired YUV420 RT format");
return false;
}
status = vaCreateConfig(this->vaDisplay, VAProfileH264Baseline, VAEntrypointVLD, &attrib, 1, &this->vaConfigID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaCreateConfig");
return false;
}
status = vaCreateSurfaces(
this->vaDisplay,
VA_RT_FORMAT_YUV420,
this->format.width,
this->format.height,
this->vaSurfaceID,
SURFACE_NUM,
NULL,
0
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaCreateSurfaces");
return false;
}
for(int i = 0; i < SURFACE_NUM; ++i)
if (!check_surface(this, i, NULL))
return false;
status = vaCreateContext(
this->vaDisplay,
this->vaConfigID,
this->format.width,
this->format.height,
VA_PROGRESSIVE,
this->vaSurfaceID,
SURFACE_NUM,
&this->vaContextID
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaCreateContext");
return false;
}
this->currentSID = 0;
this->sliceType = 2;
this->t2First = true;
status = vaBeginPicture(this->vaDisplay, this->vaContextID, this->vaSurfaceID[0]);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaBeginPicture");
return false;
}
return true;
}
static void lgd_h264_deinitialize(void * opaque)
{
struct Inst * this = (struct Inst *)opaque;
if (!this->initialized)
return;
memset(this, 0, sizeof(struct Inst));
for(int i = 0; i < SURFACE_NUM; ++i)
{
if (this->picBufferID[i] != VA_INVALID_ID)
vaDestroyBuffer(this->vaDisplay, this->picBufferID[i]);
if (this->matBufferID[i] != VA_INVALID_ID)
vaDestroyBuffer(this->vaDisplay, this->matBufferID[i]);
if (this->sliBufferID[i] != VA_INVALID_ID)
vaDestroyBuffer(this->vaDisplay, this->sliBufferID[i]);
if (this->datBufferID[i] != VA_INVALID_ID)
vaDestroyBuffer(this->vaDisplay, this->datBufferID[i]);
this->picBufferID[i] =
this->matBufferID[i] =
this->sliBufferID[i] =
this->datBufferID[i] = VA_INVALID_ID;
}
if (this->vaSurfaceID[0] != VA_INVALID_ID)
vaDestroySurfaces(this->vaDisplay, this->vaSurfaceID, SURFACE_NUM);
this->vaSurfaceID[0] = VA_INVALID_ID;
if (this->vaContextID != VA_INVALID_ID)
vaDestroyContext(this->vaDisplay, this->vaContextID);
this->vaContextID = VA_INVALID_ID;
if (this->vaConfigID != VA_INVALID_ID)
vaDestroyConfig(this->vaDisplay, this->vaConfigID);
this->vaConfigID = VA_INVALID_ID;
if (this->vaDisplay)
vaTerminate(this->vaDisplay);
this->vaDisplay = NULL;
}
static LG_OutFormat lgd_h264_get_out_format(void * opaque)
{
return LG_OUTPUT_BGRA;
return LG_OUTPUT_YUV420;
}
static unsigned int lgd_h264_get_frame_pitch(void * opaque)
@ -92,8 +335,546 @@ static unsigned int lgd_h264_get_frame_pitch(void * opaque)
return this->format.width * 4;
}
static bool lgd_h264_decode(void * opaque, uint8_t * dst, size_t dstSize, const uint8_t * src, size_t srcSize)
static unsigned int lgd_h264_get_frame_stride(void * opaque)
{
struct Inst * this = (struct Inst *)opaque;
return this->format.width;
}
static bool get_buffer(struct Inst * this, const VABufferType type, const unsigned int size, VABufferID * buf_id)
{
if (*buf_id != VA_INVALID_ID)
return true;
VAStatus status = vaCreateBuffer(this->vaDisplay, this->vaContextID, type, size, 1, NULL, buf_id);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("Failed to create buffer: %s", vaErrorStr(status));
return false;
}
if (!check_surface(this, this->currentSID, NULL))
return false;
return true;
}
static void set_slice_parameter_buffer(VASliceParameterBufferH264 *p)
{
memset(p, 0, sizeof(VASliceParameterBufferH264));
p->slice_data_size = 0;
p->slice_data_bit_offset = 64;
p->slice_alpha_c0_offset_div2 = 2;
p->slice_beta_offset_div2 = 2;
p->chroma_weight_l0_flag = 1;
p->chroma_weight_l0[0][0] = 1;
p->chroma_offset_l0[0][0] = 0;
p->chroma_weight_l0[0][1] = 1;
p->chroma_offset_l0[0][1] = 0;
p->luma_weight_l1_flag = 1;
p->chroma_weight_l1_flag = 1;
p->luma_weight_l0[0] = 0x01;
for (int i = 0; i < 32; i++)
p->RefPicList0[i].flags =
p->RefPicList1[i].flags = VA_PICTURE_H264_INVALID;
p->RefPicList1[0].picture_id = 0xffffffff;
}
static void set_slice_parameter_buffer_t2(VASliceParameterBufferH264 *p, const bool first)
{
memset(p, 0, sizeof(VASliceParameterBufferH264));
p->slice_data_size = 0;
p->slice_data_bit_offset = 64;
p->slice_alpha_c0_offset_div2 = 2;
p->slice_beta_offset_div2 = 2;
p->slice_type = 2;
if (first)
{
p->luma_weight_l0_flag = 1;
p->chroma_weight_l0_flag = 1;
p->luma_weight_l1_flag = 1;
p->chroma_weight_l1_flag = 1;
}
else
{
p->chroma_weight_l0_flag = 1;
p->chroma_weight_l0[0][0] = 1;
p->chroma_offset_l0[0][0] = 0;
p->chroma_weight_l0[0][1] = 1;
p->chroma_offset_l0[0][1] = 0;
p->luma_weight_l1_flag = 1;
p->chroma_weight_l1_flag = 1;
p->luma_weight_l0[0] = 0x01;
}
for (int i = 0; i < 32; i++)
p->RefPicList0[i].flags =
p->RefPicList1[i].flags = VA_PICTURE_H264_INVALID;
p->RefPicList1[0].picture_id =
p->RefPicList0[0].picture_id = 0xffffffff;
}
static bool setup_pic_buffer(struct Inst * this)
{
VAStatus status;
VABufferID * picBufferID = &this->picBufferID[this->currentSID];
if (!get_buffer(this, VAPictureParameterBufferType, sizeof(VAPictureParameterBufferH264), picBufferID))
{
DEBUG_ERROR("get picBuffer failed");
return false;
}
VAPictureParameterBufferH264 *p;
status = vaMapBuffer(this->vaDisplay, *picBufferID, (void **)&p);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaMapBuffer: %s", vaErrorStr(status));
return false;
}
memset(p, 0, sizeof(VAPictureParameterBufferH264));
p->picture_width_in_mbs_minus1 = (this->format.width + 15) / 16;
p->picture_width_in_mbs_minus1 = (this->format.height + 15) / 16;
p->num_ref_frames = 1;
p->seq_fields.value = 145;
p->pic_fields.value = 0x501;
p->frame_num = this->frameNum % 16;
for(int i = 0; i < 16; ++i)
{
p->ReferenceFrames[i].flags = VA_PICTURE_H264_INVALID;
p->ReferenceFrames[i].picture_id = 0xFFFFFFFF;
}
this->curPic.picture_id = this->vaSurfaceID[this->currentSID];
this->curPic.frame_idx = p->frame_num;
this->curPic.flags = 0;
this->curPic.BottomFieldOrderCnt = this->fieldCount;
this->curPic.TopFieldOrderCnt = this->fieldCount;
memcpy(&p->CurrPic, &this->curPic, sizeof(VAPictureH264));
if (this->sliceType != 2)
{
memcpy(&p->ReferenceFrames[0], &this->oldPic, sizeof(VAPictureH264));
p->ReferenceFrames[0].flags = 0;
}
status = vaUnmapBuffer(this->vaDisplay, *picBufferID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaUnmapBuffer: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool setup_mat_buffer(struct Inst * this)
{
VAStatus status;
VABufferID * matBufferID = &this->matBufferID[this->currentSID];
if (!get_buffer(this, VAIQMatrixBufferType, sizeof(VAIQMatrixBufferH264), matBufferID))
{
DEBUG_ERROR("get matBuffer failed");
return false;
}
VAIQMatrixBufferH264 * m;
status = vaMapBuffer(this->vaDisplay, *matBufferID, (void **)&m);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaMapBuffer: %s", vaErrorStr(status));
return false;
}
memcpy(m, MatrixBufferH264, sizeof(MatrixBufferH264));
status = vaUnmapBuffer(this->vaDisplay, *matBufferID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaUnmapBuffer: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool setup_sli_buffer(struct Inst * this, size_t srcSize)
{
VAStatus status;
VABufferID * sliBufferID = &this->sliBufferID[this->currentSID];
if (!get_buffer(this, VASliceParameterBufferType, sizeof(VASliceParameterBufferH264), sliBufferID))
{
DEBUG_ERROR("get sliBuffer failed");
return false;
}
VASliceParameterBufferH264 * s;
status = vaMapBuffer(this->vaDisplay, *sliBufferID, (void **)&s);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaMapBuffer: %s", vaErrorStr(status));
return false;
}
if (this->sliceType == 2)
{
set_slice_parameter_buffer_t2(s, this->t2First);
this->t2First = false;
}
else
{
set_slice_parameter_buffer(s);
memcpy(&s->RefPicList0[0], &this->oldPic, sizeof(VAPictureH264));
s->RefPicList0[0].flags = 0;
}
s->slice_data_bit_offset = 0;
s->slice_data_size = srcSize;
status = vaUnmapBuffer(this->vaDisplay, *sliBufferID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaUnmapBuffer: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool setup_dat_buffer(struct Inst * this, const uint8_t * src, size_t srcSize)
{
VAStatus status;
VABufferID * datBufferID = &this->datBufferID[this->currentSID];
if (!get_buffer(this, VASliceDataBufferType, srcSize, datBufferID))
{
DEBUG_ERROR("get datBuffer failed");
return false;
}
uint8_t * d;
status = vaMapBuffer(this->vaDisplay, *datBufferID, (void **)&d);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaMapBuffer: %s", vaErrorStr(status));
return false;
}
memcpySSE(d, src, srcSize);
status = vaUnmapBuffer(this->vaDisplay, *datBufferID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaUnmapBuffer: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool parse_nalu(struct Inst * this, const uint8_t * src, size_t size)
{
static const uint8_t startCode[4] = {0x00, 0x00, 0x00, 0x01};
if (memcmp(src, startCode, sizeof(startCode)) != 0)
{
DEBUG_ERROR("Missing start code");
return false;
}
src += 4;
if (*src & 0x80)
{
DEBUG_ERROR("forbidden_zero_bit is set");
return false;
}
// uint8_t nal_ref_idc = (*src & 0x60) >> 5;
uint8_t nal_ref_unit_type = (*src & 0x1F);
++src;
if (nal_ref_unit_type == NALU_AUD)
{
static const int pic_type_to_slice_type[3] =
{
SLICE_TYPE_I,
SLICE_TYPE_P,
SLICE_TYPE_B
};
const uint8_t primary_pic_type = (*src & 0xE0) >> 5;
this->sliceType = pic_type_to_slice_type[primary_pic_type];
return true;
}
return false;
}
static bool lgd_h264_decode(void * opaque, const uint8_t * src, size_t srcSize)
{
VAStatus status;
struct Inst * this = (struct Inst *)opaque;
if (!parse_nalu(this, src, srcSize))
{
DEBUG_ERROR("Failed to parse required information");
return false;
}
// don't start until we have an I-FRAME
if (this->frameNum == 0 && this->sliceType != SLICE_TYPE_I)
return true;
{
if (!setup_pic_buffer(this)) return false;
if (!setup_mat_buffer(this)) return false;
VABufferID bufferIDs[] =
{
this->picBufferID[this->currentSID],
this->matBufferID[this->currentSID]
};
status = vaRenderPicture(this->vaDisplay, this->vaContextID, bufferIDs, 2);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaRenderPicture: %s", vaErrorStr(status));
return false;
}
// intel broke the ABI here, see:
// https://github.com/01org/libva/commit/3eb038aa13bdd785808286c0a4995bd7a1ef07e9
// the buffers are released by vaRenderPicture in old versions
if (this->vaMajorVer == 0 && this->vaMinorVer < 40)
{
this->picBufferID[this->currentSID] =
this->matBufferID[this->currentSID] = VA_INVALID_ID;
}
}
{
if (!setup_sli_buffer(this, srcSize )) return false;
if (!setup_dat_buffer(this, src, srcSize)) return false;
VABufferID bufferIDs[] =
{
this->sliBufferID[this->currentSID],
this->datBufferID[this->currentSID]
};
status = vaRenderPicture(this->vaDisplay, this->vaContextID, bufferIDs, 2);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaRenderPicture: %s", vaErrorStr(status));
return false;
}
// intel broke the ABI here, see:
// https://github.com/01org/libva/commit/3eb038aa13bdd785808286c0a4995bd7a1ef07e9
// the buffers are released by vaRenderPicture in old versions
if (this->vaMajorVer == 0 && this->vaMinorVer < 40)
{
this->sliBufferID[this->currentSID] =
this->datBufferID[this->currentSID] = VA_INVALID_ID;
}
}
status = vaEndPicture(this->vaDisplay, this->vaContextID);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaEndPicture: %s", vaErrorStr(status));
return false;
}
// advance to the next surface and save the old picture info
this->lastSID = this->currentSID;
if (++this->currentSID == SURFACE_NUM)
this->currentSID = 0;
this->frameNum += 1;
this->fieldCount += 2;
memcpy(&this->oldPic, &this->curPic, sizeof(VAPictureH264));
// prepare the next surface
status = vaBeginPicture(this->vaDisplay, this->vaContextID, this->vaSurfaceID[this->currentSID]);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaBeginPicture: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool lgd_h264_get_buffer(void * opaque, uint8_t * dst, size_t dstSize)
{
struct Inst * this = (struct Inst *)opaque;
VAStatus status;
// don't return anything until we have some data
if (this->frameNum == 0)
return true;
// ensure the surface is ready
status = vaSyncSurface(this->vaDisplay, this->vaSurfaceID[this->lastSID]);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaSyncSurface: %s", vaErrorStr(status));
return false;
}
#if 0
// this doesn't work on my system, seems the vdpau va driver is bugged
VASurfaceStatus surfStatus;
if (!check_surface(this, this->lastSID, &surfStatus))
return false;
if (surfStatus != VASurfaceReady)
{
DEBUG_ERROR("vaSyncSurface didn't block, the surface is not ready!");
return false;
}
#endif
// get the decoded data
VAImage decoded =
{
.image_id = VA_INVALID_ID,
.buf = VA_INVALID_ID
};
status = vaDeriveImage(this->vaDisplay, this->vaSurfaceID[this->lastSID], &decoded);
if (status == VA_STATUS_ERROR_OPERATION_FAILED)
{
VAImageFormat format =
{
.fourcc = VA_FOURCC_NV12,
.byte_order = VA_LSB_FIRST,
.bits_per_pixel = 12
};
status = vaCreateImage(
this->vaDisplay,
&format,
this->format.width,
this->format.height,
&decoded
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaCreateImage: %s", vaErrorStr(status));
return false;
}
status = vaPutImage(
this->vaDisplay,
this->vaSurfaceID[this->lastSID],
decoded.image_id,
0 , 0 ,
this->format.width, this->format.height,
0 , 0 ,
this->format.width, this->format.height
);
if (status != VA_STATUS_SUCCESS)
{
vaDestroyImage(this->vaDisplay, decoded.image_id);
DEBUG_ERROR("vaPutImage: %s", vaErrorStr(status));
return false;
}
}
else
{
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaDeriveImage: %s", vaErrorStr(status));
return false;
}
}
uint8_t * d;
status = vaMapBuffer(this->vaDisplay, decoded.buf, (void **)&d);
if (status != VA_STATUS_SUCCESS)
{
vaDestroyImage(this->vaDisplay, decoded.image_id);
DEBUG_ERROR("vaMapBuffer: %s", vaErrorStr(status));
return false;
}
memcpySSE(dst, d, decoded.data_size);
status = vaUnmapBuffer(this->vaDisplay, decoded.buf);
if (status != VA_STATUS_SUCCESS)
{
vaDestroyImage(this->vaDisplay, decoded.image_id);
DEBUG_ERROR("vaUnmapBuffer: %s", vaErrorStr(status));
return false;
}
status = vaDestroyImage(this->vaDisplay, decoded.image_id);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaDestroyImage: %s", vaErrorStr(status));
return false;
}
return true;
}
static bool lgd_h264_init_gl_texture(void * opaque, GLenum target, GLuint texture, void ** ref)
{
struct Inst * this = (struct Inst *)opaque;
VAStatus status;
status = vaCreateSurfaceGLX(this->vaDisplay, target, texture, ref);
if (status != VA_STATUS_SUCCESS)
{
*ref = NULL;
DEBUG_ERROR("vaCreateSurfaceGLX: %s", vaErrorStr(status));
return false;
}
return true;
}
static void lgd_h264_free_gl_texture(void * opaque, void * ref)
{
struct Inst * this = (struct Inst *)opaque;
VAStatus status;
status = vaDestroySurfaceGLX(this->vaDisplay, ref);
if (status != VA_STATUS_SUCCESS)
DEBUG_ERROR("vaDestroySurfaceGLX: %s", vaErrorStr(status));
}
static bool lgd_h264_update_gl_texture(void * opaque, void * ref)
{
struct Inst * this = (struct Inst *)opaque;
VAStatus status;
// don't return anything until we have some data
if (this->frameNum == 0)
return true;
status = vaCopySurfaceGLX(
this->vaDisplay,
ref,
this->vaSurfaceID[this->lastSID],
0
);
if (status != VA_STATUS_SUCCESS)
{
DEBUG_ERROR("vaCopySurfaceGLX: %s", vaErrorStr(status));
return false;
}
return true;
}
@ -106,5 +887,12 @@ const LG_Decoder LGD_H264 =
.deinitialize = lgd_h264_deinitialize,
.get_out_format = lgd_h264_get_out_format,
.get_frame_pitch = lgd_h264_get_frame_pitch,
.decode = lgd_h264_decode
.get_frame_stride = lgd_h264_get_frame_stride,
.decode = lgd_h264_decode,
.get_buffer = lgd_h264_get_buffer,
.has_gl = true,
.init_gl_texture = lgd_h264_init_gl_texture,
.free_gl_texture = lgd_h264_free_gl_texture,
.update_gl_texture = lgd_h264_update_gl_texture
};

View File

@ -28,15 +28,18 @@ Place, Suite 330, Boston, MA 02111-1307 USA
struct Inst
{
LG_RendererFormat format;
const uint8_t * src;
};
static bool lgd_null_create (void ** opaque);
static void lgd_null_destroy (void * opaque);
static bool lgd_null_initialize (void * opaque, const LG_RendererFormat format);
static bool lgd_null_initialize (void * opaque, const LG_RendererFormat format, SDL_Window * window);
static void lgd_null_deinitialize (void * opaque);
static LG_OutFormat lgd_null_get_out_format (void * opaque);
static unsigned int lgd_null_get_frame_pitch(void * opaque);
static bool lgd_null_decode (void * opaque, uint8_t * dst, size_t dstSize, const uint8_t * src, size_t srcSize);
static unsigned int lgd_null_get_frame_pitch (void * opaque);
static unsigned int lgd_null_get_frame_stride(void * opaque);
static bool lgd_null_decode (void * opaque, const uint8_t * src, size_t srcSize);
static bool lgd_null_get_buffer (void * opaque, uint8_t * dst, size_t dstSize);
static bool lgd_null_create(void ** opaque)
{
@ -56,7 +59,7 @@ static void lgd_null_destroy(void * opaque)
free(opaque);
}
static bool lgd_null_initialize(void * opaque, const LG_RendererFormat format)
static bool lgd_null_initialize(void * opaque, const LG_RendererFormat format, SDL_Window * window)
{
struct Inst * this = (struct Inst *)opaque;
memcpy(&this->format, &format, sizeof(LG_RendererFormat));
@ -65,6 +68,8 @@ static bool lgd_null_initialize(void * opaque, const LG_RendererFormat format)
static void lgd_null_deinitialize(void * opaque)
{
struct Inst * this = (struct Inst *)opaque;
memset(this, 0, sizeof(struct Inst));
}
static LG_OutFormat lgd_null_get_out_format(void * opaque)
@ -78,9 +83,26 @@ static unsigned int lgd_null_get_frame_pitch(void * opaque)
return this->format.pitch;
}
static bool lgd_null_decode(void * opaque, uint8_t * dst, size_t dstSize, const uint8_t * src, size_t srcSize)
static unsigned int lgd_null_get_frame_stride(void * opaque)
{
memcpySSE(dst, src, dstSize);
struct Inst * this = (struct Inst *)opaque;
return this->format.stride;
}
static bool lgd_null_decode(void * opaque, const uint8_t * src, size_t srcSize)
{
struct Inst * this = (struct Inst *)opaque;
this->src = src;
return true;
}
static bool lgd_null_get_buffer(void * opaque, uint8_t * dst, size_t dstSize)
{
struct Inst * this = (struct Inst *)opaque;
if (!this->src)
return false;
memcpySSE(dst, this->src, dstSize);
return true;
}
@ -93,5 +115,7 @@ const LG_Decoder LGD_NULL =
.deinitialize = lgd_null_deinitialize,
.get_out_format = lgd_null_get_out_format,
.get_frame_pitch = lgd_null_get_frame_pitch,
.decode = lgd_null_decode
.get_frame_stride = lgd_null_get_frame_stride,
.decode = lgd_null_decode,
.get_buffer = lgd_null_get_buffer
};

View File

@ -19,27 +19,38 @@ Place, Suite 330, Boston, MA 02111-1307 USA
#pragma once
#include "lg-renderer.h"
#include <stdint.h>
#include <stdbool.h>
#include <SDL2/SDL.h>
#include "lg-renderer.h"
#include <GL/gl.h>
typedef enum LG_OutFormat
{
LG_OUTPUT_BGRA
LG_OUTPUT_BGRA,
LG_OUTPUT_YUV420
}
LG_OutFormat;
typedef bool (* LG_DecoderCreate )(void ** opaque);
typedef void (* LG_DecoderDestroy )(void * opaque);
typedef bool (* LG_DecoderInitialize )(void * opaque, const LG_RendererFormat format);
typedef bool (* LG_DecoderInitialize )(void * opaque, const LG_RendererFormat format, SDL_Window * window);
typedef void (* LG_DecoderDeInitialize )(void * opaque);
typedef LG_OutFormat (* LG_DecoderGetOutFormat )(void * opaque);
typedef unsigned int (* LG_DecoderGetFramePitch)(void * opaque);
typedef bool (* LG_DecoderDecode )(void * opaque, uint8_t * dst, size_t dstSize, const uint8_t * src, size_t srcSize);
typedef unsigned int (* LG_DecoderGetFramePitch )(void * opaque);
typedef unsigned int (* LG_DecoderGetFrameStride)(void * opaque);
typedef bool (* LG_DecoderDecode )(void * opaque, const uint8_t * src, size_t srcSize);
typedef bool (* LG_DecoderGetBuffer )(void * opaque, uint8_t * dst, size_t dstSize);
typedef bool (* LG_DecoderInitGLTexture )(void * opaque, GLenum target, GLuint texture, void ** ref);
typedef void (* LG_DecoderFreeGLTexture )(void * opaque, void * ref);
typedef bool (* LG_DecoderUpdateGLTexture)(void * opaque, void * ref);
typedef struct LG_Decoder
{
// mandatory support
const char * name;
LG_DecoderCreate create;
LG_DecoderDestroy destroy;
@ -47,6 +58,14 @@ typedef struct LG_Decoder
LG_DecoderDeInitialize deinitialize;
LG_DecoderGetOutFormat get_out_format;
LG_DecoderGetFramePitch get_frame_pitch;
LG_DecoderGetFrameStride get_frame_stride;
LG_DecoderDecode decode;
LG_DecoderGetBuffer get_buffer;
// optional support
const bool has_gl;
LG_DecoderInitGLTexture init_gl_texture;
LG_DecoderFreeGLTexture free_gl_texture;
LG_DecoderUpdateGLTexture update_gl_texture;
}
LG_Decoder;

View File

@ -36,10 +36,9 @@ Place, Suite 330, Boston, MA 02111-1307 USA
#define BUFFER_COUNT 2
#define FRAME_TEXTURE 0
#define FPS_TEXTURE 1
#define MOUSE_TEXTURE 2
#define TEXTURE_COUNT 3
#define FPS_TEXTURE 0
#define MOUSE_TEXTURE 1
#define TEXTURE_COUNT 2
static PFNGLXGETVIDEOSYNCSGIPROC glXGetVideoSyncSGI = NULL;
static PFNGLXWAITVIDEOSYNCSGIPROC glXWaitVideoSyncSGI = NULL;
@ -82,16 +81,18 @@ struct Inst
uint64_t drawStart;
bool hasBuffers;
GLuint vboID[1];
GLuint vboID[BUFFER_COUNT];
uint8_t * texPixels[BUFFER_COUNT];
LG_Lock syncLock;
int texIndex, wTexIndex;
int texIndex;
int texList;
int fpsList;
int mouseList;
LG_RendererRect destRect;
bool hasTextures;
bool hasTextures, hasFrames;
GLuint frames[BUFFER_COUNT];
void * decoderFrames[BUFFER_COUNT];
GLuint textures[TEXTURE_COUNT];
uint gpuFrameCount;
@ -283,15 +284,8 @@ bool opengl_on_frame_event(void * opaque, const LG_RendererFormat format, const
}
LG_UNLOCK(this->formatLock);
// lock, perform the update, then unlock
LG_LOCK(this->syncLock);
if (!this->decoder->decode(
this->decoderData,
this->texPixels[this->wTexIndex],
this->texSize,
data,
this->format.pitch
))
if (!this->decoder->decode(this->decoderData, data, format.pitch))
{
DEBUG_ERROR("decode returned failure");
LG_UNLOCK(this->syncLock);
@ -430,7 +424,7 @@ bool opengl_render(void * opaque, SDL_Window * window)
glXGetVideoSyncSGI(&before);
SDL_GL_SwapWindow(window);
// wait for the swap to happen to ensure we dont buffer frames //
// wait for the swap to happen to ensure we dont buffer frames
glXGetVideoSyncSGI(&after);
if (before == after)
glXWaitVideoSyncSGI(1, 0, &before);
@ -588,8 +582,8 @@ static bool configure(struct Inst * this, SDL_Window *window)
if (!this->decoder->initialize(
this->decoderData,
this->format
))
this->format,
window))
{
DEBUG_ERROR("Failed to initialize decoder");
return false;
@ -602,6 +596,12 @@ static bool configure(struct Inst * this, SDL_Window *window)
this->vboFormat = GL_BGRA;
break;
case LG_OUTPUT_YUV420:
// fixme
this->intFormat = GL_RGBA8;
this->vboFormat = GL_BGRA;
break;
default:
DEBUG_ERROR("Format not supported");
LG_UNLOCK(this->formatLock);
@ -618,8 +618,10 @@ static bool configure(struct Inst * this, SDL_Window *window)
this->fpsList = glGenLists(1);
this->mouseList = glGenLists(1);
// generate the pixel unpack buffers
glGenBuffers(1, this->vboID);
// generate the pixel unpack buffers if the decoder isn't going to do it for us
if (!this->decoder->has_gl)
{
glGenBuffers(BUFFER_COUNT, this->vboID);
if (check_gl_error("glGenBuffers"))
{
LG_UNLOCK(this->formatLock);
@ -627,7 +629,9 @@ static bool configure(struct Inst * this, SDL_Window *window)
}
this->hasBuffers = true;
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, this->vboID[0]);
for(int i = 0; i < BUFFER_COUNT; ++i)
{
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, this->vboID[i]);
if (check_gl_error("glBindBuffer"))
{
LG_UNLOCK(this->formatLock);
@ -636,7 +640,7 @@ static bool configure(struct Inst * this, SDL_Window *window)
glBufferStorage(
GL_PIXEL_UNPACK_BUFFER,
this->texSize * BUFFER_COUNT,
this->texSize,
NULL,
GL_MAP_WRITE_BIT |
GL_MAP_PERSISTENT_BIT
@ -647,10 +651,10 @@ static bool configure(struct Inst * this, SDL_Window *window)
return false;
}
this->texPixels[0] = glMapBufferRange(
this->texPixels[i] = glMapBufferRange(
GL_PIXEL_UNPACK_BUFFER,
0,
this->texSize * BUFFER_COUNT,
this->texSize,
GL_MAP_WRITE_BIT |
GL_MAP_PERSISTENT_BIT |
GL_MAP_FLUSH_EXPLICIT_BIT
@ -661,11 +665,10 @@ static bool configure(struct Inst * this, SDL_Window *window)
LG_UNLOCK(this->formatLock);
return false;
}
}
}
for(int i = 1; i < BUFFER_COUNT; ++i)
this->texPixels[i] = this->texPixels[i-1] + this->texSize;
// create the textures
// create the overlay textures
glGenTextures(TEXTURE_COUNT, this->textures);
if (check_gl_error("glGenTextures"))
{
@ -674,8 +677,19 @@ static bool configure(struct Inst * this, SDL_Window *window)
}
this->hasTextures = true;
// create the frame texture
glBindTexture(GL_TEXTURE_2D, this->textures[FRAME_TEXTURE]);
// create the frame textures
glGenTextures(BUFFER_COUNT, this->frames);
if (check_gl_error("glGenTextures"))
{
LG_UNLOCK(this->formatLock);
return false;
}
this->hasFrames = true;
for(int i = 0; i < BUFFER_COUNT; ++i)
{
// bind and create the new texture
glBindTexture(GL_TEXTURE_2D, this->frames[i]);
if (check_gl_error("glBindTexture"))
{
LG_UNLOCK(this->formatLock);
@ -687,7 +701,7 @@ static bool configure(struct Inst * this, SDL_Window *window)
0,
this->intFormat,
this->format.width,
this->format.height * BUFFER_COUNT,
this->format.height,
0,
this->vboFormat,
GL_UNSIGNED_BYTE,
@ -699,26 +713,36 @@ static bool configure(struct Inst * this, SDL_Window *window)
return false;
}
if (this->decoder->has_gl)
{
if (!this->decoder->init_gl_texture(
this->decoderData,
GL_TEXTURE_2D,
this->frames[i],
&this->decoderFrames[i]))
{
LG_UNLOCK(this->formatLock);
return false;
}
}
else
{
// configure the texture
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S , GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T , GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
for (int i = 0; i < BUFFER_COUNT; ++i)
{
const float ts = (1.0f / BUFFER_COUNT) * i;
const float te = (1.0f / BUFFER_COUNT) + ts;
}
// create the display lists
glNewList(this->texList + i, GL_COMPILE);
glBindTexture(GL_TEXTURE_2D, this->textures[FRAME_TEXTURE]);
glBindTexture(GL_TEXTURE_2D, this->frames[i]);
glColor4f(1.0f, 1.0f, 1.0f, 1.0f);
glBegin(GL_TRIANGLE_STRIP);
glTexCoord2f(0.0f, ts); glVertex2i(0 , 0 );
glTexCoord2f(1.0f, ts); glVertex2i(this->format.width, 0 );
glTexCoord2f(0.0f, te); glVertex2i(0 , this->format.height);
glTexCoord2f(1.0f, te); glVertex2i(this->format.width, this->format.height);
glTexCoord2f(0.0f, 0.0f); glVertex2i(0 , 0 );
glTexCoord2f(1.0f, 0.0f); glVertex2i(this->format.width, 0 );
glTexCoord2f(0.0f, 1.0f); glVertex2i(0 , this->format.height);
glTexCoord2f(1.0f, 1.0f); glVertex2i(this->format.width, this->format.height);
glEnd();
glEndList();
}
@ -753,9 +777,28 @@ static void deconfigure(struct Inst * this)
this->hasTextures = false;
}
if (this->hasFrames)
{
if (this->decoder->has_gl)
{
for(int i = 0; i < BUFFER_COUNT; ++i)
{
if (this->decoderFrames[i])
this->decoder->free_gl_texture(
this->decoderData,
this->decoderFrames[i]
);
this->decoderFrames[i] = NULL;
}
}
glDeleteTextures(BUFFER_COUNT, this->frames);
this->hasFrames = false;
}
if (this->hasBuffers)
{
glDeleteBuffers(1, this->vboID);
glDeleteBuffers(BUFFER_COUNT, this->vboID);
this->hasBuffers = false;
}
@ -765,7 +808,7 @@ static void deconfigure(struct Inst * this)
this->glContext = NULL;
}
if (this->decoder)
if (this->decoderData)
{
this->decoder->destroy(this->decoderData);
this->decoderData = NULL;
@ -918,51 +961,76 @@ static bool draw_frame(struct Inst * this)
return true;
}
this->texIndex = this->wTexIndex;
if (++this->wTexIndex == BUFFER_COUNT)
this->wTexIndex = 0;
if (++this->texIndex == BUFFER_COUNT)
this->texIndex = 0;
this->frameUpdate = false;
LG_UNLOCK(this->syncLock);
LG_LOCK(this->formatLock);
if (this->decoder->has_gl)
{
if (!this->decoder->update_gl_texture(
this->decoderData,
this->decoderFrames[this->texIndex]
))
{
LG_UNLOCK(this->formatLock);
DEBUG_ERROR("Failed to update the texture from the decoder");
return false;
}
}
else
{
glBindTexture(GL_TEXTURE_2D, this->frames[this->texIndex]);
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, this->vboID[this->texIndex]);
// bind the texture and update it
glBindTexture(GL_TEXTURE_2D , this->textures[FRAME_TEXTURE]);
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, this->vboID[0] );
glPixelStorei(GL_UNPACK_ALIGNMENT , 4 );
glPixelStorei(GL_UNPACK_ROW_LENGTH , this->format.stride );
// copy the buffer to the texture
glFlushMappedBufferRange(
GL_PIXEL_UNPACK_BUFFER,
this->texSize * this->texIndex,
this->texSize
glPixelStorei(GL_UNPACK_ALIGNMENT , 4);
glPixelStorei(GL_UNPACK_ROW_LENGTH ,
this->decoder->get_frame_stride(this->decoderData)
);
if (!this->decoder->get_buffer(
this->decoderData,
this->texPixels[this->texIndex],
this->texSize
))
{
LG_UNLOCK(this->formatLock);
DEBUG_ERROR("Failed to get the buffer from the decoder");
return false;
}
glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, 0, this->texSize);
// update the texture
glTexSubImage2D(
GL_TEXTURE_2D,
0,
0,
this->texIndex * this->format.height,
0,
this->format.width ,
this->format.height,
this->vboFormat,
GL_UNSIGNED_BYTE,
(void*)(this->texIndex * this->texSize)
(void*)0
);
if (check_gl_error("glTexSubImage2D"))
{
DEBUG_ERROR("texIndex: %u, width: %u, height: %u, vboFormat: %x, texSize: %lu",
this->texIndex, this->format.width, this->format.height, this->vboFormat, this->texSize
);
}
// unbind the buffer
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
}
const bool mipmap = this->opt.mipmap && (
(this->format.width > this->destRect.w) ||
(this->format.height > this->destRect.h));
glBindTexture(GL_TEXTURE_2D, this->frames[this->texIndex]);
if (mipmap)
{
glGenerateMipmap(GL_TEXTURE_2D);
@ -974,8 +1042,8 @@ static bool draw_frame(struct Inst * this)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
}
glBindTexture(GL_TEXTURE_2D, 0);
LG_UNLOCK(this->formatLock);
return true;
}