fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-05 00:58:05 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

omx: add a compute path in enc_LoadImage_common

Browse source 
Acked-by: Leo Liu <leo.liu@amd.com>
This commit is contained in:
Marek Olšák 2019-02-27 17:19:55 -05:00 • committed by Leo Liu
parent 323e7be91c
commit b276e8358a
4 changed files with 189 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
src/gallium/state_trackers/omx/bellagio/vid_enc.c
View file

@ -161,6 +161,8 @@ static OMX_ERRORTYPE vid_enc_Constructor(OMX_COMPONENTTYPE *comp, OMX_STRING nam
if (!priv->s_pipe)
return OMX_ErrorInsufficientResources;
enc_InitCompute_common(priv);
if (!vl_compositor_init(&priv->compositor, priv->s_pipe)) {
priv->s_pipe->destroy(priv->s_pipe);
priv->s_pipe = NULL;
@ -267,6 +269,7 @@ static OMX_ERRORTYPE vid_enc_Destructor(OMX_COMPONENTTYPE *comp)
if (priv->s_pipe) {
vl_compositor_cleanup_state(&priv->cstate);
vl_compositor_cleanup(&priv->compositor);
enc_ReleaseCompute_common(priv);
priv->s_pipe->destroy(priv->s_pipe);
}

3
src/gallium/state_trackers/omx/tizonia/h264eprc.c
View file

@ -407,6 +407,8 @@ static OMX_ERRORTYPE h264e_prc_create_encoder(void *ap_obj)
if (!priv->s_pipe)
return OMX_ErrorInsufficientResources;
enc_InitCompute_common(priv);
if (!vl_compositor_init(&priv->compositor, priv->s_pipe)) {
priv->s_pipe->destroy(priv->s_pipe);
priv->s_pipe = NULL;
@ -451,6 +453,7 @@ static void h264e_prc_destroy_encoder(void *ap_obj)
if (priv->s_pipe) {
vl_compositor_cleanup_state(&priv->cstate);
vl_compositor_cleanup(&priv->compositor);
enc_ReleaseCompute_common(priv);
priv->s_pipe->destroy(priv->s_pipe);
}

205
src/gallium/state_trackers/omx/vid_enc_common.c
View file

@ -28,6 +28,7 @@
#include "vid_enc_common.h"
#include "vl/vl_video_buffer.h"
#include "tgsi/tgsi_text.h"
void enc_ReleaseTasks(struct list_head *head)
{
@ -306,6 +307,92 @@ void enc_ControlPicture_common(vid_enc_PrivateType * priv, struct pipe_h264_enc_
enc_GetPictureParamPreset(picture);
}
static void *create_compute_state(struct pipe_context *pipe,
const char *source)
{
struct tgsi_token tokens[1024];
struct pipe_compute_state state = {0};
if (!tgsi_text_translate(source, tokens, ARRAY_SIZE(tokens))) {
assert(false);
return NULL;
}
state.ir_type = PIPE_SHADER_IR_TGSI;
state.prog = tokens;
return pipe->create_compute_state(pipe, &state);
}
void enc_InitCompute_common(vid_enc_PrivateType *priv)
{
struct pipe_context *pipe = priv->s_pipe;
struct pipe_screen *screen = pipe->screen;
/* We need the partial last block support. */
if (!screen->get_param(screen, PIPE_CAP_COMPUTE_GRID_INFO_LAST_BLOCK))
return;
static const char *copy_y =
"COMP\n"
"PROPERTY CS_FIXED_BLOCK_WIDTH 64\n"
"PROPERTY CS_FIXED_BLOCK_HEIGHT 1\n"
"PROPERTY CS_FIXED_BLOCK_DEPTH 1\n"
"DCL SV[0], THREAD_ID\n"
"DCL SV[1], BLOCK_ID\n"
"DCL IMAGE[0], 2D, PIPE_FORMAT_R8_UINT\n"
"DCL IMAGE[1], 2D, PIPE_FORMAT_R8_UINT, WR\n"
"DCL TEMP[0..1]\n"
"IMM[0] UINT32 {64, 0, 0, 0}\n"
"UMAD TEMP[0].x, SV[1], IMM[0], SV[0]\n"
"MOV TEMP[0].y, SV[1]\n"
"LOAD TEMP[1].x, IMAGE[0], TEMP[0], 2D, PIPE_FORMAT_R8_UINT\n"
"STORE IMAGE[1].x, TEMP[0], TEMP[1], 2D, PIPE_FORMAT_R8_UINT\n"
"END\n";
static const char *copy_uv =
"COMP\n"
"PROPERTY CS_FIXED_BLOCK_WIDTH 64\n"
"PROPERTY CS_FIXED_BLOCK_HEIGHT 1\n"
"PROPERTY CS_FIXED_BLOCK_DEPTH 1\n"
"DCL SV[0], THREAD_ID\n"
"DCL SV[1], BLOCK_ID\n"
"DCL IMAGE[0], 2D, PIPE_FORMAT_R8_UINT\n"
"DCL IMAGE[2], 2D, PIPE_FORMAT_R8G8_UINT, WR\n"
"DCL CONST[0][0]\n" /* .x = offset of the UV portion in the y direction */
"DCL TEMP[0..4]\n"
"IMM[0] UINT32 {64, 0, 2, 1}\n"
/* Destination R8G8 coordinates */
"UMAD TEMP[0].x, SV[1], IMM[0], SV[0]\n"
"MOV TEMP[0].y, SV[1]\n"
/* Source R8 coordinates of U */
"UMUL TEMP[1].x, TEMP[0], IMM[0].zzzz\n"
"UADD TEMP[1].y, TEMP[0], CONST[0].xxxx\n"
/* Source R8 coordinates of V */
"UADD TEMP[2].x, TEMP[1], IMM[0].wwww\n"
"MOV TEMP[2].y, TEMP[1]\n"
"LOAD TEMP[3].x, IMAGE[0], TEMP[1], 2D, PIPE_FORMAT_R8_UINT\n"
"LOAD TEMP[4].x, IMAGE[0], TEMP[2], 2D, PIPE_FORMAT_R8_UINT\n"
"MOV TEMP[3].y, TEMP[4].xxxx\n"
"STORE IMAGE[2], TEMP[0], TEMP[3], 2D, PIPE_FORMAT_R8G8_UINT\n"
"END\n";
priv->copy_y_shader = create_compute_state(pipe, copy_y);
priv->copy_uv_shader = create_compute_state(pipe, copy_uv);
}
void enc_ReleaseCompute_common(vid_enc_PrivateType *priv)
{
struct pipe_context *pipe = priv->s_pipe;
if (priv->copy_y_shader)
pipe->delete_compute_state(pipe, priv->copy_y_shader);
if (priv->copy_uv_shader)
pipe->delete_compute_state(pipe, priv->copy_uv_shader);
}
OMX_ERRORTYPE enc_LoadImage_common(vid_enc_PrivateType * priv, OMX_VIDEO_PORTDEFINITIONTYPE *def,
OMX_BUFFERHEADERTYPE *buf,
struct pipe_video_buffer *vbuf)
@ -337,7 +424,6 @@ OMX_ERRORTYPE enc_LoadImage_common(vid_enc_PrivateType * priv, OMX_VIDEO_PORTDEF
PIPE_TRANSFER_WRITE, &box,
ptr, def->nStride, 0);
} else {
struct pipe_blit_info blit;
struct vl_video_buffer *dst_buf = (struct vl_video_buffer *)vbuf;
pipe_transfer_unmap(pipe, inp->transfer);
@ -358,42 +444,103 @@ OMX_ERRORTYPE enc_LoadImage_common(vid_enc_PrivateType * priv, OMX_VIDEO_PORTDEF
* - Copy Y to dst_buf->resources[0] as R8.
* - Copy UV to dst_buf->resources[1] as R8G8.
*/
if (priv->copy_y_shader && priv->copy_uv_shader) {
/* Compute path */
/* Set shader images for both copies. */
struct pipe_image_view image[3] = {0};
image[0].resource = inp->resource;
image[0].shader_access = image[0].access = PIPE_IMAGE_ACCESS_READ;
image[0].format = PIPE_FORMAT_R8_UINT;
box.width = def->nFrameWidth;
box.height = def->nFrameHeight;
box.depth = 1;
image[1].resource = dst_buf->resources[0];
image[1].shader_access = image[1].access = PIPE_IMAGE_ACCESS_WRITE;
image[1].format = PIPE_FORMAT_R8_UINT;
/* Copy Y */
pipe->resource_copy_region(pipe,
dst_buf->resources[0],
0, 0, 0, 0, inp->resource, 0, &box);
image[2].resource = dst_buf->resources[1];
image[2].shader_access = image[1].access = PIPE_IMAGE_ACCESS_WRITE;
image[2].format = PIPE_FORMAT_R8G8_UINT;
/* Copy U */
memset(&blit, 0, sizeof(blit));
blit.src.resource = inp->resource;
blit.src.format = inp->resource->format;
pipe->set_shader_images(pipe, PIPE_SHADER_COMPUTE, 0, 3, image);
blit.src.box.x = -1;
blit.src.box.y = def->nFrameHeight;
blit.src.box.width = def->nFrameWidth;
blit.src.box.height = def->nFrameHeight / 2 ;
blit.src.box.depth = 1;
/* Set the constant buffer. */
uint32_t constants[4] = {def->nFrameHeight};
struct pipe_constant_buffer cb = {};
blit.dst.resource = dst_buf->resources[1];
blit.dst.format = blit.dst.resource->format;
cb.buffer_size = sizeof(constants);
cb.user_buffer = constants;
pipe->set_constant_buffer(pipe, PIPE_SHADER_COMPUTE, 0, &cb);
blit.dst.box.width = def->nFrameWidth / 2;
blit.dst.box.height = def->nFrameHeight / 2;
blit.dst.box.depth = 1;
blit.filter = PIPE_TEX_FILTER_NEAREST;
/* Use the optimal block size for the linear image layout. */
struct pipe_grid_info info = {};
info.block[0] = 64;
info.block[1] = 1;
info.block[2] = 1;
info.grid[2] = 1;
blit.mask = PIPE_MASK_R;
pipe->blit(pipe, &blit);
/* Copy Y */
pipe->bind_compute_state(pipe, priv->copy_y_shader);
info.grid[0] = DIV_ROUND_UP(def->nFrameWidth, 64);
info.grid[1] = def->nFrameHeight;
info.last_block[0] = def->nFrameWidth % 64;
pipe->launch_grid(pipe, &info);
/* Copy UV */
pipe->bind_compute_state(pipe, priv->copy_uv_shader);
info.grid[0] = DIV_ROUND_UP(def->nFrameWidth / 2, 64);
info.grid[1] = def->nFrameHeight / 2;
info.last_block[0] = (def->nFrameWidth / 2) % 64;
pipe->launch_grid(pipe, &info);
/* Make the result visible to all clients. */
pipe->memory_barrier(pipe, PIPE_BARRIER_ALL);
/* Unbind. */
pipe->set_shader_images(pipe, PIPE_SHADER_COMPUTE, 0, 3, NULL);
pipe->set_constant_buffer(pipe, PIPE_SHADER_COMPUTE, 0, NULL);
pipe->bind_compute_state(pipe, NULL);
} else {
/* Graphics path */
struct pipe_blit_info blit;
box.width = def->nFrameWidth;
box.height = def->nFrameHeight;
box.depth = 1;
/* Copy Y */
pipe->resource_copy_region(pipe,
dst_buf->resources[0],
0, 0, 0, 0, inp->resource, 0, &box);
/* Copy U */
memset(&blit, 0, sizeof(blit));
blit.src.resource = inp->resource;
blit.src.format = inp->resource->format;
blit.src.box.x = -1;
blit.src.box.y = def->nFrameHeight;
blit.src.box.width = def->nFrameWidth;
blit.src.box.height = def->nFrameHeight / 2 ;
blit.src.box.depth = 1;
blit.dst.resource = dst_buf->resources[1];
blit.dst.format = blit.dst.resource->format;
blit.dst.box.width = def->nFrameWidth / 2;
blit.dst.box.height = def->nFrameHeight / 2;
blit.dst.box.depth = 1;
blit.filter = PIPE_TEX_FILTER_NEAREST;
blit.mask = PIPE_MASK_R;
pipe->blit(pipe, &blit);
/* Copy V */
blit.src.box.x = 0;
blit.mask = PIPE_MASK_G;
pipe->blit(pipe, &blit);
}
/* Copy V */
blit.src.box.x = 0;
blit.mask = PIPE_MASK_G;
pipe->blit(pipe, &blit);
pipe->flush(pipe, NULL, 0);
box.width = inp->resource->width0;

8
src/gallium/state_trackers/omx/vid_enc_common.h
View file

@ -100,7 +100,9 @@ DERIVEDCLASS(vid_enc_PrivateType, omx_base_filter_PrivateType)
struct pipe_video_buffer *scale_buffer[OMX_VID_ENC_NUM_SCALING_BUFFERS]; \
OMX_CONFIG_SCALEFACTORTYPE scale; \
OMX_U32 current_scale_buffer; \
OMX_U32 stacked_frames_num;
OMX_U32 stacked_frames_num; \
void *copy_y_shader; \
void *copy_uv_shader;
ENDCLASS(vid_enc_PrivateType)
#else
@ -151,6 +153,8 @@ struct h264e_prc
bool eos_;
bool in_port_disabled_;
bool out_port_disabled_;
void *copy_y_shader;
void *copy_uv_shader;
};
#endif
@ -167,6 +171,8 @@ void enc_ScaleInput_common(vid_enc_PrivateType * priv,
struct pipe_video_buffer **vbuf, unsigned *size);
void enc_ControlPicture_common(vid_enc_PrivateType * priv,
struct pipe_h264_enc_picture_desc *picture);
void enc_InitCompute_common(vid_enc_PrivateType *priv);
void enc_ReleaseCompute_common(vid_enc_PrivateType *priv);
OMX_ERRORTYPE enc_LoadImage_common(vid_enc_PrivateType * priv, OMX_VIDEO_PORTDEFINITIONTYPE *def,
OMX_BUFFERHEADERTYPE *buf,
struct pipe_video_buffer *vbuf);