*
**************************************************************************/
-/*
- * Authors:
- * Christian König <christian.koenig@amd.com>
- *
- */
-
#include <sys/types.h>
#include <assert.h>
#include <errno.h>
#include "vl/vl_defines.h"
#include "vl/vl_mpeg12_decoder.h"
-#include "r600_pipe_common.h"
+#include "radeonsi/si_pipe.h"
#include "radeon_video.h"
#include "radeon_uvd.h"
#define FB_BUFFER_OFFSET 0x1000
#define FB_BUFFER_SIZE 2048
+#define FB_BUFFER_SIZE_TONGA (2048 * 64)
#define IT_SCALING_TABLE_SIZE 992
+#define UVD_SESSION_CONTEXT_SIZE (128 * 1024)
/* UVD decoder representation */
struct ruvd_decoder {
struct pipe_screen *screen;
struct radeon_winsys* ws;
- struct radeon_winsys_cs* cs;
+ struct radeon_cmdbuf* cs;
unsigned cur_buffer;
struct rvid_buffer msg_fb_it_buffers[NUM_BUFFERS];
struct ruvd_msg *msg;
uint32_t *fb;
+ unsigned fb_size;
uint8_t *it;
struct rvid_buffer bs_buffers[NUM_BUFFERS];
struct rvid_buffer dpb;
bool use_legacy;
struct rvid_buffer ctx;
+ struct rvid_buffer sessionctx;
+ struct {
+ unsigned data0;
+ unsigned data1;
+ unsigned cmd;
+ unsigned cntl;
+ } reg;
+
+ void *render_pic_list[16];
};
/* flush IB to the hardware */
-static void flush(struct ruvd_decoder *dec)
+static int flush(struct ruvd_decoder *dec, unsigned flags)
{
- dec->ws->cs_flush(dec->cs, RADEON_FLUSH_ASYNC, NULL, 0);
+ return dec->ws->cs_flush(dec->cs, flags, NULL);
}
/* add a new set register command to the IB */
static void set_reg(struct ruvd_decoder *dec, unsigned reg, uint32_t val)
{
- uint32_t *pm4 = dec->cs->buf;
- pm4[dec->cs->cdw++] = RUVD_PKT0(reg >> 2, 0);
- pm4[dec->cs->cdw++] = val;
+ radeon_emit(dec->cs, RUVD_PKT0(reg >> 2, 0));
+ radeon_emit(dec->cs, val);
}
/* send a command to the VCPU through the GPCOM registers */
static void send_cmd(struct ruvd_decoder *dec, unsigned cmd,
- struct radeon_winsys_cs_handle* cs_buf, uint32_t off,
+ struct pb_buffer* buf, uint32_t off,
enum radeon_bo_usage usage, enum radeon_bo_domain domain)
{
int reloc_idx;
- reloc_idx = dec->ws->cs_add_buffer(dec->cs, cs_buf, usage, domain,
+ reloc_idx = dec->ws->cs_add_buffer(dec->cs, buf, usage | RADEON_USAGE_SYNCHRONIZED,
+ domain,
RADEON_PRIO_UVD);
if (!dec->use_legacy) {
uint64_t addr;
- addr = dec->ws->buffer_get_virtual_address(cs_buf);
+ addr = dec->ws->buffer_get_virtual_address(buf);
addr = addr + off;
- set_reg(dec, RUVD_GPCOM_VCPU_DATA0, addr);
- set_reg(dec, RUVD_GPCOM_VCPU_DATA1, addr >> 32);
+ set_reg(dec, dec->reg.data0, addr);
+ set_reg(dec, dec->reg.data1, addr >> 32);
} else {
+ off += dec->ws->buffer_get_reloc_offset(buf);
set_reg(dec, RUVD_GPCOM_VCPU_DATA0, off);
set_reg(dec, RUVD_GPCOM_VCPU_DATA1, reloc_idx * 4);
}
- set_reg(dec, RUVD_GPCOM_VCPU_CMD, cmd << 1);
+ set_reg(dec, dec->reg.cmd, cmd << 1);
}
/* do the codec needs an IT buffer ?*/
buf = &dec->msg_fb_it_buffers[dec->cur_buffer];
/* and map it for CPU access */
- ptr = dec->ws->buffer_map(buf->res->cs_buf, dec->cs, PIPE_TRANSFER_WRITE);
+ ptr = dec->ws->buffer_map(buf->res->buf, dec->cs, PIPE_TRANSFER_WRITE);
/* calc buffer offsets */
dec->msg = (struct ruvd_msg *)ptr;
+ memset(dec->msg, 0, sizeof(*dec->msg));
+
dec->fb = (uint32_t *)(ptr + FB_BUFFER_OFFSET);
if (have_it(dec))
- dec->it = (uint8_t *)(ptr + FB_BUFFER_OFFSET + FB_BUFFER_SIZE);
+ dec->it = (uint8_t *)(ptr + FB_BUFFER_OFFSET + dec->fb_size);
}
/* unmap and send a message command to the VCPU */
buf = &dec->msg_fb_it_buffers[dec->cur_buffer];
/* unmap the buffer */
- dec->ws->buffer_unmap(buf->res->cs_buf);
+ dec->ws->buffer_unmap(buf->res->buf);
dec->msg = NULL;
dec->fb = NULL;
dec->it = NULL;
+
+ if (dec->sessionctx.res)
+ send_cmd(dec, RUVD_CMD_SESSION_CONTEXT_BUFFER,
+ dec->sessionctx.res->buf, 0, RADEON_USAGE_READWRITE,
+ RADEON_DOMAIN_VRAM);
+
/* and send it to the hardware */
- send_cmd(dec, RUVD_CMD_MSG_BUFFER, buf->res->cs_buf, 0,
+ send_cmd(dec, RUVD_CMD_MSG_BUFFER, buf->res->buf, 0,
RADEON_USAGE_READ, RADEON_DOMAIN_GTT);
}
case PIPE_VIDEO_FORMAT_HEVC:
return RUVD_CODEC_H265;
+ case PIPE_VIDEO_FORMAT_JPEG:
+ return RUVD_CODEC_MJPEG;
+
default:
assert(0);
return 0;
}
}
-static unsigned calc_ctx_size(struct ruvd_decoder *dec)
+static unsigned calc_ctx_size_h264_perf(struct ruvd_decoder *dec)
+{
+ unsigned width_in_mb, height_in_mb, ctx_size;
+ unsigned width = align(dec->base.width, VL_MACROBLOCK_WIDTH);
+ unsigned height = align(dec->base.height, VL_MACROBLOCK_HEIGHT);
+
+ unsigned max_references = dec->base.max_references + 1;
+
+ // picture width & height in 16 pixel units
+ width_in_mb = width / VL_MACROBLOCK_WIDTH;
+ height_in_mb = align(height / VL_MACROBLOCK_HEIGHT, 2);
+
+ if (!dec->use_legacy) {
+ unsigned fs_in_mb = width_in_mb * height_in_mb;
+ unsigned num_dpb_buffer;
+ switch(dec->base.level) {
+ case 30:
+ num_dpb_buffer = 8100 / fs_in_mb;
+ break;
+ case 31:
+ num_dpb_buffer = 18000 / fs_in_mb;
+ break;
+ case 32:
+ num_dpb_buffer = 20480 / fs_in_mb;
+ break;
+ case 41:
+ num_dpb_buffer = 32768 / fs_in_mb;
+ break;
+ case 42:
+ num_dpb_buffer = 34816 / fs_in_mb;
+ break;
+ case 50:
+ num_dpb_buffer = 110400 / fs_in_mb;
+ break;
+ case 51:
+ num_dpb_buffer = 184320 / fs_in_mb;
+ break;
+ default:
+ num_dpb_buffer = 184320 / fs_in_mb;
+ break;
+ }
+ num_dpb_buffer++;
+ max_references = MAX2(MIN2(NUM_H264_REFS, num_dpb_buffer), max_references);
+ ctx_size = max_references * align(width_in_mb * height_in_mb * 192, 256);
+ } else {
+ // the firmware seems to always assume a minimum of ref frames
+ max_references = MAX2(NUM_H264_REFS, max_references);
+ // macroblock context buffer
+ ctx_size = align(width_in_mb * height_in_mb * max_references * 192, 256);
+ }
+
+ return ctx_size;
+}
+
+static unsigned calc_ctx_size_h265_main(struct ruvd_decoder *dec)
{
unsigned width = align(dec->base.width, VL_MACROBLOCK_WIDTH);
unsigned height = align(dec->base.height, VL_MACROBLOCK_HEIGHT);
return ((width + 255) / 16) * ((height + 255) / 16) * 16 * max_references + 52 * 1024;
}
+static unsigned calc_ctx_size_h265_main10(struct ruvd_decoder *dec, struct pipe_h265_picture_desc *pic)
+{
+ unsigned block_size, log2_ctb_size, width_in_ctb, height_in_ctb, num_16x16_block_per_ctb;
+ unsigned context_buffer_size_per_ctb_row, cm_buffer_size, max_mb_address, db_left_tile_pxl_size;
+ unsigned db_left_tile_ctx_size = 4096 / 16 * (32 + 16 * 4);
+
+ unsigned width = align(dec->base.width, VL_MACROBLOCK_WIDTH);
+ unsigned height = align(dec->base.height, VL_MACROBLOCK_HEIGHT);
+ unsigned coeff_10bit = (pic->pps->sps->bit_depth_luma_minus8 || pic->pps->sps->bit_depth_chroma_minus8) ? 2 : 1;
+
+ unsigned max_references = dec->base.max_references + 1;
+
+ if (dec->base.width * dec->base.height >= 4096*2000)
+ max_references = MAX2(max_references, 8);
+ else
+ max_references = MAX2(max_references, 17);
+
+ block_size = (1 << (pic->pps->sps->log2_min_luma_coding_block_size_minus3 + 3));
+ log2_ctb_size = block_size + pic->pps->sps->log2_diff_max_min_luma_coding_block_size;
+
+ width_in_ctb = (width + ((1 << log2_ctb_size) - 1)) >> log2_ctb_size;
+ height_in_ctb = (height + ((1 << log2_ctb_size) - 1)) >> log2_ctb_size;
+
+ num_16x16_block_per_ctb = ((1 << log2_ctb_size) >> 4) * ((1 << log2_ctb_size) >> 4);
+ context_buffer_size_per_ctb_row = align(width_in_ctb * num_16x16_block_per_ctb * 16, 256);
+ max_mb_address = (unsigned) ceil(height * 8 / 2048.0);
+
+ cm_buffer_size = max_references * context_buffer_size_per_ctb_row * height_in_ctb;
+ db_left_tile_pxl_size = coeff_10bit * (max_mb_address * 2 * 2048 + 1024);
+
+ return cm_buffer_size + db_left_tile_ctx_size + db_left_tile_pxl_size;
+}
+
+static unsigned get_db_pitch_alignment(struct ruvd_decoder *dec)
+{
+ if (((struct si_screen*)dec->screen)->info.family < CHIP_VEGA10)
+ return 16;
+ else
+ return 32;
+}
+
/* calculate size of reference picture buffer */
static unsigned calc_dpb_size(struct ruvd_decoder *dec)
{
unsigned max_references = dec->base.max_references + 1;
// aligned size of a single frame
- image_size = width * height;
+ image_size = align(width, get_db_pitch_alignment(dec)) * height;
image_size += image_size / 2;
image_size = align(image_size, 1024);
num_dpb_buffer++;
max_references = MAX2(MIN2(NUM_H264_REFS, num_dpb_buffer), max_references);
dpb_size = image_size * max_references;
- dpb_size += max_references * align(width_in_mb * height_in_mb * 192, alignment);
- dpb_size += align(width_in_mb * height_in_mb * 32, alignment);
+ if ((dec->stream_type != RUVD_CODEC_H264_PERF) ||
+ (((struct si_screen*)dec->screen)->info.family < CHIP_POLARIS10)) {
+ dpb_size += max_references * align(width_in_mb * height_in_mb * 192, alignment);
+ dpb_size += align(width_in_mb * height_in_mb * 32, alignment);
+ }
} else {
// the firmware seems to allways assume a minimum of ref frames
max_references = MAX2(NUM_H264_REFS, max_references);
// reference picture buffer
dpb_size = image_size * max_references;
- // macroblock context buffer
- dpb_size += width_in_mb * height_in_mb * max_references * 192;
- // IT surface buffer
- dpb_size += width_in_mb * height_in_mb * 32;
+ if ((dec->stream_type != RUVD_CODEC_H264_PERF) ||
+ (((struct si_screen*)dec->screen)->info.family < CHIP_POLARIS10)) {
+ // macroblock context buffer
+ dpb_size += width_in_mb * height_in_mb * max_references * 192;
+ // IT surface buffer
+ dpb_size += width_in_mb * height_in_mb * 32;
+ }
}
break;
}
width = align (width, 16);
height = align (height, 16);
- dpb_size = align((width * height * 3) / 2, 256) * max_references;
+ if (dec->base.profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10)
+ dpb_size = align((align(width, get_db_pitch_alignment(dec)) * height * 9) / 4, 256) * max_references;
+ else
+ dpb_size = align((align(width, get_db_pitch_alignment(dec)) * height * 3) / 2, 256) * max_references;
break;
case PIPE_VIDEO_FORMAT_VC1:
dpb_size = MAX2(dpb_size, 30 * 1024 * 1024);
break;
+ case PIPE_VIDEO_FORMAT_JPEG:
+ dpb_size = 0;
+ break;
+
default:
// something is missing here
assert(0);
memset(&result, 0, sizeof(result));
switch (pic->base.profile) {
case PIPE_VIDEO_PROFILE_MPEG4_AVC_BASELINE:
+ case PIPE_VIDEO_PROFILE_MPEG4_AVC_CONSTRAINED_BASELINE:
result.profile = RUVD_H264_PROFILE_BASELINE;
break;
result.log2_max_pic_order_cnt_lsb_minus4 = pic->pps->sps->log2_max_pic_order_cnt_lsb_minus4;
switch (dec->base.chroma_format) {
+ case PIPE_VIDEO_CHROMA_FORMAT_NONE:
+ /* TODO: assert? */
+ break;
case PIPE_VIDEO_CHROMA_FORMAT_400:
result.chroma_format = 0;
break;
struct pipe_h265_picture_desc *pic)
{
struct ruvd_h265 result;
- unsigned i;
+ unsigned i, j;
memset(&result, 0, sizeof(result));
result.sps_info_flags |= pic->pps->sps->sps_temporal_mvp_enabled_flag << 6;
result.sps_info_flags |= pic->pps->sps->strong_intra_smoothing_enabled_flag << 7;
result.sps_info_flags |= pic->pps->sps->separate_colour_plane_flag << 8;
- if (((struct r600_common_screen*)dec->screen)->family == CHIP_CARRIZO)
+ if (((struct si_screen*)dec->screen)->info.family == CHIP_CARRIZO)
result.sps_info_flags |= 1 << 9;
+ if (pic->UseRefPicList == true)
+ result.sps_info_flags |= 1 << 10;
result.chroma_format = pic->pps->sps->chroma_format_idc;
result.bit_depth_luma_minus8 = pic->pps->sps->bit_depth_luma_minus8;
result.row_height_minus1[i] = pic->pps->row_height_minus1[i];
result.num_delta_pocs_ref_rps_idx = pic->NumDeltaPocsOfRefRpsIdx;
- result.curr_idx = pic->CurrPicOrderCntVal;
result.curr_poc = pic->CurrPicOrderCntVal;
+ for (i = 0 ; i < 16 ; i++) {
+ for (j = 0; (pic->ref[j] != NULL) && (j < 16) ; j++) {
+ if (dec->render_pic_list[i] == pic->ref[j])
+ break;
+ if (j == 15)
+ dec->render_pic_list[i] = NULL;
+ else if (pic->ref[j+1] == NULL)
+ dec->render_pic_list[i] = NULL;
+ }
+ }
+ for (i = 0 ; i < 16 ; i++) {
+ if (dec->render_pic_list[i] == NULL) {
+ dec->render_pic_list[i] = target;
+ result.curr_idx = i;
+ break;
+ }
+ }
+
vl_video_buffer_set_associated_data(target, &dec->base,
- (void *)(uintptr_t)pic->CurrPicOrderCntVal,
+ (void *)(uintptr_t)result.curr_idx,
&ruvd_destroy_associated_data);
for (i = 0; i < 16; ++i) {
memcpy(dec->it + 480, pic->pps->sps->ScalingList16x16, 6 * 64);
memcpy(dec->it + 864, pic->pps->sps->ScalingList32x32, 2 * 64);
+ for (i = 0 ; i < 2 ; i++) {
+ for (j = 0 ; j < 15 ; j++)
+ result.direct_reflist[i][j] = pic->RefPicList[i][j];
+ }
+
+ if (pic->base.profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10) {
+ if (target->buffer_format == PIPE_FORMAT_P016) {
+ result.p010_mode = 1;
+ result.msb_mode = 1;
+ } else {
+ result.luma_10to8 = 5;
+ result.chroma_10to8 = 5;
+ result.sclr_luma10to8 = 4;
+ result.sclr_chroma10to8 = 4;
+ }
+ }
+
/* TODO
result.highestTid;
result.isNonRef;
for (i = 0; i < 2; ++i)
result.ref_pic_idx[i] = get_ref_pic_idx(dec, pic->ref[i]);
- result.load_intra_quantiser_matrix = 1;
- result.load_nonintra_quantiser_matrix = 1;
-
- for (i = 0; i < 64; ++i) {
- result.intra_quantiser_matrix[i] = pic->intra_matrix[zscan[i]];
- result.nonintra_quantiser_matrix[i] = pic->non_intra_matrix[zscan[i]];
+ if(pic->intra_matrix) {
+ result.load_intra_quantiser_matrix = 1;
+ for (i = 0; i < 64; ++i) {
+ result.intra_quantiser_matrix[i] = pic->intra_matrix[zscan[i]];
+ }
+ }
+ if(pic->non_intra_matrix) {
+ result.load_nonintra_quantiser_matrix = 1;
+ for (i = 0; i < 64; ++i) {
+ result.nonintra_quantiser_matrix[i] = pic->non_intra_matrix[zscan[i]];
+ }
}
result.profile_and_level_indication = 0;
return result;
}
+static void get_mjpeg_slice_header(struct ruvd_decoder *dec, struct pipe_mjpeg_picture_desc *pic)
+{
+ int size = 0, saved_size, len_pos, i;
+ uint16_t *bs;
+ uint8_t *buf = dec->bs_ptr;
+
+ /* SOI */
+ buf[size++] = 0xff;
+ buf[size++] = 0xd8;
+
+ /* DQT */
+ buf[size++] = 0xff;
+ buf[size++] = 0xdb;
+
+ len_pos = size++;
+ size++;
+
+ for (i = 0; i < 4; ++i) {
+ if (pic->quantization_table.load_quantiser_table[i] == 0)
+ continue;
+
+ buf[size++] = i;
+ memcpy((buf + size), &pic->quantization_table.quantiser_table[i], 64);
+ size += 64;
+ }
+
+ bs = (uint16_t*)&buf[len_pos];
+ *bs = util_bswap16(size - 4);
+
+ saved_size = size;
+
+ /* DHT */
+ buf[size++] = 0xff;
+ buf[size++] = 0xc4;
+
+ len_pos = size++;
+ size++;
+
+ for (i = 0; i < 2; ++i) {
+ if (pic->huffman_table.load_huffman_table[i] == 0)
+ continue;
+
+ buf[size++] = 0x00 | i;
+ memcpy((buf + size), &pic->huffman_table.table[i].num_dc_codes, 16);
+ size += 16;
+ memcpy((buf + size), &pic->huffman_table.table[i].dc_values, 12);
+ size += 12;
+ }
+
+ for (i = 0; i < 2; ++i) {
+ if (pic->huffman_table.load_huffman_table[i] == 0)
+ continue;
+
+ buf[size++] = 0x10 | i;
+ memcpy((buf + size), &pic->huffman_table.table[i].num_ac_codes, 16);
+ size += 16;
+ memcpy((buf + size), &pic->huffman_table.table[i].ac_values, 162);
+ size += 162;
+ }
+
+ bs = (uint16_t*)&buf[len_pos];
+ *bs = util_bswap16(size - saved_size - 2);
+
+ saved_size = size;
+
+ /* DRI */
+ if (pic->slice_parameter.restart_interval) {
+ buf[size++] = 0xff;
+ buf[size++] = 0xdd;
+ buf[size++] = 0x00;
+ buf[size++] = 0x04;
+ bs = (uint16_t*)&buf[size++];
+ *bs = util_bswap16(pic->slice_parameter.restart_interval);
+ saved_size = ++size;
+ }
+
+ /* SOF */
+ buf[size++] = 0xff;
+ buf[size++] = 0xc0;
+
+ len_pos = size++;
+ size++;
+
+ buf[size++] = 0x08;
+
+ bs = (uint16_t*)&buf[size++];
+ *bs = util_bswap16(pic->picture_parameter.picture_height);
+ size++;
+
+ bs = (uint16_t*)&buf[size++];
+ *bs = util_bswap16(pic->picture_parameter.picture_width);
+ size++;
+
+ buf[size++] = pic->picture_parameter.num_components;
+
+ for (i = 0; i < pic->picture_parameter.num_components; ++i) {
+ buf[size++] = pic->picture_parameter.components[i].component_id;
+ buf[size++] = pic->picture_parameter.components[i].h_sampling_factor << 4 |
+ pic->picture_parameter.components[i].v_sampling_factor;
+ buf[size++] = pic->picture_parameter.components[i].quantiser_table_selector;
+ }
+
+ bs = (uint16_t*)&buf[len_pos];
+ *bs = util_bswap16(size - saved_size - 2);
+
+ saved_size = size;
+
+ /* SOS */
+ buf[size++] = 0xff;
+ buf[size++] = 0xda;
+
+ len_pos = size++;
+ size++;
+
+ buf[size++] = pic->slice_parameter.num_components;
+
+ for (i = 0; i < pic->slice_parameter.num_components; ++i) {
+ buf[size++] = pic->slice_parameter.components[i].component_selector;
+ buf[size++] = pic->slice_parameter.components[i].dc_table_selector << 4 |
+ pic->slice_parameter.components[i].ac_table_selector;
+ }
+
+ buf[size++] = 0x00;
+ buf[size++] = 0x3f;
+ buf[size++] = 0x00;
+
+ bs = (uint16_t*)&buf[len_pos];
+ *bs = util_bswap16(size - saved_size - 2);
+
+ dec->bs_ptr += size;
+ dec->bs_size += size;
+}
+
/**
* destroy this video decoder
*/
assert(decoder);
map_msg_fb_it_buf(dec);
- memset(dec->msg, 0, sizeof(*dec->msg));
dec->msg->size = sizeof(*dec->msg);
dec->msg->msg_type = RUVD_MSG_DESTROY;
dec->msg->stream_handle = dec->stream_handle;
send_msg_buf(dec);
- flush(dec);
+ flush(dec, 0);
dec->ws->cs_destroy(dec->cs);
for (i = 0; i < NUM_BUFFERS; ++i) {
- rvid_destroy_buffer(&dec->msg_fb_it_buffers[i]);
- rvid_destroy_buffer(&dec->bs_buffers[i]);
+ si_vid_destroy_buffer(&dec->msg_fb_it_buffers[i]);
+ si_vid_destroy_buffer(&dec->bs_buffers[i]);
}
- rvid_destroy_buffer(&dec->dpb);
- if (u_reduce_video_profile(dec->base.profile) == PIPE_VIDEO_FORMAT_HEVC)
- rvid_destroy_buffer(&dec->ctx);
+ si_vid_destroy_buffer(&dec->dpb);
+ si_vid_destroy_buffer(&dec->ctx);
+ si_vid_destroy_buffer(&dec->sessionctx);
FREE(dec);
}
dec->bs_size = 0;
dec->bs_ptr = dec->ws->buffer_map(
- dec->bs_buffers[dec->cur_buffer].res->cs_buf,
+ dec->bs_buffers[dec->cur_buffer].res->buf,
dec->cs, PIPE_TRANSFER_WRITE);
}
const unsigned *sizes)
{
struct ruvd_decoder *dec = (struct ruvd_decoder*)decoder;
+ enum pipe_video_format format = u_reduce_video_profile(picture->profile);
unsigned i;
assert(decoder);
if (!dec->bs_ptr)
return;
+ if (format == PIPE_VIDEO_FORMAT_JPEG)
+ get_mjpeg_slice_header(dec, (struct pipe_mjpeg_picture_desc*)picture);
+
for (i = 0; i < num_buffers; ++i) {
struct rvid_buffer *buf = &dec->bs_buffers[dec->cur_buffer];
unsigned new_size = dec->bs_size + sizes[i];
+ if (format == PIPE_VIDEO_FORMAT_JPEG)
+ new_size += 2; /* save for EOI */
+
if (new_size > buf->res->buf->size) {
- dec->ws->buffer_unmap(buf->res->cs_buf);
- if (!rvid_resize_buffer(dec->screen, dec->cs, buf, new_size)) {
+ dec->ws->buffer_unmap(buf->res->buf);
+ if (!si_vid_resize_buffer(dec->screen, dec->cs, buf, new_size)) {
RVID_ERR("Can't resize bitstream buffer!");
return;
}
- dec->bs_ptr = dec->ws->buffer_map(buf->res->cs_buf, dec->cs,
+ dec->bs_ptr = dec->ws->buffer_map(buf->res->buf, dec->cs,
PIPE_TRANSFER_WRITE);
if (!dec->bs_ptr)
return;
dec->bs_size += sizes[i];
dec->bs_ptr += sizes[i];
}
+
+ if (format == PIPE_VIDEO_FORMAT_JPEG) {
+ ((uint8_t *)dec->bs_ptr)[0] = 0xff; /* EOI */
+ ((uint8_t *)dec->bs_ptr)[1] = 0xd9;
+ dec->bs_size += 2;
+ dec->bs_ptr += 2;
+ }
}
/**
struct pipe_picture_desc *picture)
{
struct ruvd_decoder *dec = (struct ruvd_decoder*)decoder;
- struct radeon_winsys_cs_handle *dt;
+ struct pb_buffer *dt;
struct rvid_buffer *msg_fb_it_buf, *bs_buf;
unsigned bs_size;
bs_size = align(dec->bs_size, 128);
memset(dec->bs_ptr, 0, bs_size - dec->bs_size);
- dec->ws->buffer_unmap(bs_buf->res->cs_buf);
+ dec->ws->buffer_unmap(bs_buf->res->buf);
map_msg_fb_it_buf(dec);
dec->msg->size = sizeof(*dec->msg);
dec->msg->body.decode.width_in_samples = dec->base.width;
dec->msg->body.decode.height_in_samples = dec->base.height;
- dec->msg->body.decode.dpb_size = dec->dpb.res->buf->size;
+ if ((picture->profile == PIPE_VIDEO_PROFILE_VC1_SIMPLE) ||
+ (picture->profile == PIPE_VIDEO_PROFILE_VC1_MAIN)) {
+ dec->msg->body.decode.width_in_samples = align(dec->msg->body.decode.width_in_samples, 16) / 16;
+ dec->msg->body.decode.height_in_samples = align(dec->msg->body.decode.height_in_samples, 16) / 16;
+ }
+
+ if (dec->dpb.res)
+ dec->msg->body.decode.dpb_size = dec->dpb.res->buf->size;
dec->msg->body.decode.bsd_size = bs_size;
- dec->msg->body.decode.db_pitch = dec->base.width;
+ dec->msg->body.decode.db_pitch = align(dec->base.width, get_db_pitch_alignment(dec));
+
+ if (dec->stream_type == RUVD_CODEC_H264_PERF &&
+ ((struct si_screen*)dec->screen)->info.family >= CHIP_POLARIS10)
+ dec->msg->body.decode.dpb_reserved = dec->ctx.res->buf->size;
dt = dec->set_dtb(dec->msg, (struct vl_video_buffer *)target);
+ if (((struct si_screen*)dec->screen)->info.family >= CHIP_STONEY)
+ dec->msg->body.decode.dt_wa_chroma_top_offset = dec->msg->body.decode.dt_pitch / 2;
switch (u_reduce_video_profile(picture->profile)) {
case PIPE_VIDEO_FORMAT_MPEG4_AVC:
case PIPE_VIDEO_FORMAT_HEVC:
dec->msg->body.decode.codec.h265 = get_h265_msg(dec, target, (struct pipe_h265_picture_desc*)picture);
+ if (dec->ctx.res == NULL) {
+ unsigned ctx_size;
+ if (dec->base.profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10)
+ ctx_size = calc_ctx_size_h265_main10(dec, (struct pipe_h265_picture_desc*)picture);
+ else
+ ctx_size = calc_ctx_size_h265_main(dec);
+ if (!si_vid_create_buffer(dec->screen, &dec->ctx, ctx_size, PIPE_USAGE_DEFAULT)) {
+ RVID_ERR("Can't allocated context buffer.\n");
+ }
+ si_vid_clear_buffer(decoder->context, &dec->ctx);
+ }
+
+ if (dec->ctx.res)
+ dec->msg->body.decode.dpb_reserved = dec->ctx.res->buf->size;
break;
case PIPE_VIDEO_FORMAT_VC1:
dec->msg->body.decode.codec.mpeg4 = get_mpeg4_msg(dec, (struct pipe_mpeg4_picture_desc*)picture);
break;
+ case PIPE_VIDEO_FORMAT_JPEG:
+ break;
+
default:
assert(0);
return;
dec->msg->body.decode.extension_support = 0x1;
/* set at least the feedback buffer size */
- dec->fb[0] = FB_BUFFER_SIZE;
+ dec->fb[0] = dec->fb_size;
send_msg_buf(dec);
- send_cmd(dec, RUVD_CMD_DPB_BUFFER, dec->dpb.res->cs_buf, 0,
- RADEON_USAGE_READWRITE, RADEON_DOMAIN_VRAM);
- if (u_reduce_video_profile(picture->profile) == PIPE_VIDEO_FORMAT_HEVC) {
- send_cmd(dec, RUVD_CMD_CONTEXT_BUFFER, dec->ctx.res->cs_buf, 0,
+ if (dec->dpb.res)
+ send_cmd(dec, RUVD_CMD_DPB_BUFFER, dec->dpb.res->buf, 0,
RADEON_USAGE_READWRITE, RADEON_DOMAIN_VRAM);
- }
- send_cmd(dec, RUVD_CMD_BITSTREAM_BUFFER, bs_buf->res->cs_buf,
+
+ if (dec->ctx.res)
+ send_cmd(dec, RUVD_CMD_CONTEXT_BUFFER, dec->ctx.res->buf, 0,
+ RADEON_USAGE_READWRITE, RADEON_DOMAIN_VRAM);
+ send_cmd(dec, RUVD_CMD_BITSTREAM_BUFFER, bs_buf->res->buf,
0, RADEON_USAGE_READ, RADEON_DOMAIN_GTT);
send_cmd(dec, RUVD_CMD_DECODING_TARGET_BUFFER, dt, 0,
RADEON_USAGE_WRITE, RADEON_DOMAIN_VRAM);
- send_cmd(dec, RUVD_CMD_FEEDBACK_BUFFER, msg_fb_it_buf->res->cs_buf,
+ send_cmd(dec, RUVD_CMD_FEEDBACK_BUFFER, msg_fb_it_buf->res->buf,
FB_BUFFER_OFFSET, RADEON_USAGE_WRITE, RADEON_DOMAIN_GTT);
if (have_it(dec))
- send_cmd(dec, RUVD_CMD_ITSCALING_TABLE_BUFFER, msg_fb_it_buf->res->cs_buf,
- FB_BUFFER_OFFSET + FB_BUFFER_SIZE, RADEON_USAGE_READ, RADEON_DOMAIN_GTT);
- set_reg(dec, RUVD_ENGINE_CNTL, 1);
+ send_cmd(dec, RUVD_CMD_ITSCALING_TABLE_BUFFER, msg_fb_it_buf->res->buf,
+ FB_BUFFER_OFFSET + dec->fb_size, RADEON_USAGE_READ, RADEON_DOMAIN_GTT);
+ set_reg(dec, dec->reg.cntl, 1);
- flush(dec);
+ flush(dec, PIPE_FLUSH_ASYNC);
next_buffer(dec);
}
/**
* create and UVD decoder
*/
-struct pipe_video_codec *ruvd_create_decoder(struct pipe_context *context,
- const struct pipe_video_codec *templ,
- ruvd_set_dtb set_dtb)
+struct pipe_video_codec *si_common_uvd_create_decoder(struct pipe_context *context,
+ const struct pipe_video_codec *templ,
+ ruvd_set_dtb set_dtb)
{
- struct radeon_winsys* ws = ((struct r600_common_context *)context)->ws;
- struct r600_common_context *rctx = (struct r600_common_context*)context;
+ struct si_context *sctx = (struct si_context*)context;
+ struct radeon_winsys *ws = sctx->ws;
unsigned dpb_size;
unsigned width = templ->width, height = templ->height;
unsigned bs_buf_size;
- struct radeon_info info;
struct ruvd_decoder *dec;
- int i;
-
- ws->query_info(ws, &info);
+ int r, i;
switch(u_reduce_video_profile(templ->profile)) {
case PIPE_VIDEO_FORMAT_MPEG12:
- if (templ->entrypoint > PIPE_VIDEO_ENTRYPOINT_BITSTREAM || info.family < CHIP_PALM)
+ if (templ->entrypoint > PIPE_VIDEO_ENTRYPOINT_BITSTREAM)
return vl_create_mpeg12_decoder(context, templ);
/* fall through */
case PIPE_VIDEO_FORMAT_MPEG4:
+ width = align(width, VL_MACROBLOCK_WIDTH);
+ height = align(height, VL_MACROBLOCK_HEIGHT);
+ break;
case PIPE_VIDEO_FORMAT_MPEG4_AVC:
width = align(width, VL_MACROBLOCK_WIDTH);
height = align(height, VL_MACROBLOCK_HEIGHT);
if (!dec)
return NULL;
- if (info.drm_major < 3)
- dec->use_legacy = TRUE;
+ if (sctx->screen->info.drm_major < 3)
+ dec->use_legacy = true;
dec->base = *templ;
dec->base.context = context;
dec->base.end_frame = ruvd_end_frame;
dec->base.flush = ruvd_flush;
- dec->stream_type = profile2stream_type(dec, info.family);
+ dec->stream_type = profile2stream_type(dec, sctx->family);
dec->set_dtb = set_dtb;
- dec->stream_handle = rvid_alloc_stream_handle();
+ dec->stream_handle = si_vid_alloc_stream_handle();
dec->screen = context->screen;
dec->ws = ws;
- dec->cs = ws->cs_create(rctx->ctx, RING_UVD, NULL, NULL, NULL);
+ dec->cs = ws->cs_create(sctx->ctx, RING_UVD, NULL, NULL);
if (!dec->cs) {
RVID_ERR("Can't get command submission context.\n");
goto error;
}
- bs_buf_size = width * height * 512 / (16 * 16);
+ for (i = 0; i < 16; i++)
+ dec->render_pic_list[i] = NULL;
+ dec->fb_size = (sctx->family == CHIP_TONGA) ? FB_BUFFER_SIZE_TONGA :
+ FB_BUFFER_SIZE;
+ bs_buf_size = width * height * (512 / (16 * 16));
for (i = 0; i < NUM_BUFFERS; ++i) {
- unsigned msg_fb_it_size = FB_BUFFER_OFFSET + FB_BUFFER_SIZE;
+ unsigned msg_fb_it_size = FB_BUFFER_OFFSET + dec->fb_size;
STATIC_ASSERT(sizeof(struct ruvd_msg) <= FB_BUFFER_OFFSET);
if (have_it(dec))
msg_fb_it_size += IT_SCALING_TABLE_SIZE;
- if (!rvid_create_buffer(dec->screen, &dec->msg_fb_it_buffers[i],
+ if (!si_vid_create_buffer(dec->screen, &dec->msg_fb_it_buffers[i],
msg_fb_it_size, PIPE_USAGE_STAGING)) {
RVID_ERR("Can't allocated message buffers.\n");
goto error;
}
- if (!rvid_create_buffer(dec->screen, &dec->bs_buffers[i],
+ if (!si_vid_create_buffer(dec->screen, &dec->bs_buffers[i],
bs_buf_size, PIPE_USAGE_STAGING)) {
RVID_ERR("Can't allocated bitstream buffers.\n");
goto error;
}
- rvid_clear_buffer(context, &dec->msg_fb_it_buffers[i]);
- rvid_clear_buffer(context, &dec->bs_buffers[i]);
+ si_vid_clear_buffer(context, &dec->msg_fb_it_buffers[i]);
+ si_vid_clear_buffer(context, &dec->bs_buffers[i]);
}
dpb_size = calc_dpb_size(dec);
-
- if (!rvid_create_buffer(dec->screen, &dec->dpb, dpb_size, PIPE_USAGE_DEFAULT)) {
- RVID_ERR("Can't allocated dpb.\n");
- goto error;
+ if (dpb_size) {
+ if (!si_vid_create_buffer(dec->screen, &dec->dpb, dpb_size, PIPE_USAGE_DEFAULT)) {
+ RVID_ERR("Can't allocated dpb.\n");
+ goto error;
+ }
+ si_vid_clear_buffer(context, &dec->dpb);
}
- rvid_clear_buffer(context, &dec->dpb);
-
- if (u_reduce_video_profile(dec->base.profile) == PIPE_VIDEO_FORMAT_HEVC) {
- unsigned ctx_size = calc_ctx_size(dec);
- if (!rvid_create_buffer(dec->screen, &dec->ctx, ctx_size, PIPE_USAGE_DEFAULT)) {
+ if (dec->stream_type == RUVD_CODEC_H264_PERF && sctx->family >= CHIP_POLARIS10) {
+ unsigned ctx_size = calc_ctx_size_h264_perf(dec);
+ if (!si_vid_create_buffer(dec->screen, &dec->ctx, ctx_size, PIPE_USAGE_DEFAULT)) {
RVID_ERR("Can't allocated context buffer.\n");
goto error;
}
- rvid_clear_buffer(context, &dec->ctx);
+ si_vid_clear_buffer(context, &dec->ctx);
+ }
+
+ if (sctx->family >= CHIP_POLARIS10 && sctx->screen->info.drm_minor >= 3) {
+ if (!si_vid_create_buffer(dec->screen, &dec->sessionctx,
+ UVD_SESSION_CONTEXT_SIZE,
+ PIPE_USAGE_DEFAULT)) {
+ RVID_ERR("Can't allocated session ctx.\n");
+ goto error;
+ }
+ si_vid_clear_buffer(context, &dec->sessionctx);
+ }
+
+ if (sctx->family >= CHIP_VEGA10) {
+ dec->reg.data0 = RUVD_GPCOM_VCPU_DATA0_SOC15;
+ dec->reg.data1 = RUVD_GPCOM_VCPU_DATA1_SOC15;
+ dec->reg.cmd = RUVD_GPCOM_VCPU_CMD_SOC15;
+ dec->reg.cntl = RUVD_ENGINE_CNTL_SOC15;
+ } else {
+ dec->reg.data0 = RUVD_GPCOM_VCPU_DATA0;
+ dec->reg.data1 = RUVD_GPCOM_VCPU_DATA1;
+ dec->reg.cmd = RUVD_GPCOM_VCPU_CMD;
+ dec->reg.cntl = RUVD_ENGINE_CNTL;
}
map_msg_fb_it_buf(dec);
dec->msg->body.create.height_in_samples = dec->base.height;
dec->msg->body.create.dpb_size = dpb_size;
send_msg_buf(dec);
- flush(dec);
+ r = flush(dec, 0);
+ if (r)
+ goto error;
+
next_buffer(dec);
return &dec->base;
if (dec->cs) dec->ws->cs_destroy(dec->cs);
for (i = 0; i < NUM_BUFFERS; ++i) {
- rvid_destroy_buffer(&dec->msg_fb_it_buffers[i]);
- rvid_destroy_buffer(&dec->bs_buffers[i]);
+ si_vid_destroy_buffer(&dec->msg_fb_it_buffers[i]);
+ si_vid_destroy_buffer(&dec->bs_buffers[i]);
}
- rvid_destroy_buffer(&dec->dpb);
- if (u_reduce_video_profile(dec->base.profile) == PIPE_VIDEO_FORMAT_HEVC)
- rvid_destroy_buffer(&dec->ctx);
+ si_vid_destroy_buffer(&dec->dpb);
+ si_vid_destroy_buffer(&dec->ctx);
+ si_vid_destroy_buffer(&dec->sessionctx);
FREE(dec);
}
/* calculate top/bottom offset */
-static unsigned texture_offset(struct radeon_surf *surface, unsigned layer)
+static unsigned texture_offset(struct radeon_surf *surface, unsigned layer,
+ enum ruvd_surface_type type)
{
- return surface->level[0].offset +
- layer * surface->level[0].slice_size;
+ switch (type) {
+ default:
+ case RUVD_SURFACE_TYPE_LEGACY:
+ return surface->u.legacy.level[0].offset +
+ layer * (uint64_t)surface->u.legacy.level[0].slice_size_dw * 4;
+ break;
+ case RUVD_SURFACE_TYPE_GFX9:
+ return surface->u.gfx9.surf_offset +
+ layer * surface->u.gfx9.surf_slice_size;
+ break;
+ }
}
/* hw encode the aspect of macro tiles */
/**
* fill decoding target field from the luma and chroma surfaces
*/
-void ruvd_set_dt_surfaces(struct ruvd_msg *msg, struct radeon_surf *luma,
- struct radeon_surf *chroma)
+void si_uvd_set_dt_surfaces(struct ruvd_msg *msg, struct radeon_surf *luma,
+ struct radeon_surf *chroma, enum ruvd_surface_type type)
{
- msg->body.decode.dt_pitch = luma->level[0].pitch_bytes;
- switch (luma->level[0].mode) {
- case RADEON_SURF_MODE_LINEAR_ALIGNED:
- msg->body.decode.dt_tiling_mode = RUVD_TILE_LINEAR;
- msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_LINEAR;
- break;
- case RADEON_SURF_MODE_1D:
- msg->body.decode.dt_tiling_mode = RUVD_TILE_8X8;
- msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_1D_THIN;
- break;
- case RADEON_SURF_MODE_2D:
- msg->body.decode.dt_tiling_mode = RUVD_TILE_8X8;
- msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_2D_THIN;
- break;
+ switch (type) {
default:
- assert(0);
- break;
- }
+ case RUVD_SURFACE_TYPE_LEGACY:
+ msg->body.decode.dt_pitch = luma->u.legacy.level[0].nblk_x * luma->blk_w;
+ switch (luma->u.legacy.level[0].mode) {
+ case RADEON_SURF_MODE_LINEAR_ALIGNED:
+ msg->body.decode.dt_tiling_mode = RUVD_TILE_LINEAR;
+ msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_LINEAR;
+ break;
+ case RADEON_SURF_MODE_1D:
+ msg->body.decode.dt_tiling_mode = RUVD_TILE_8X8;
+ msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_1D_THIN;
+ break;
+ case RADEON_SURF_MODE_2D:
+ msg->body.decode.dt_tiling_mode = RUVD_TILE_8X8;
+ msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_2D_THIN;
+ break;
+ default:
+ assert(0);
+ break;
+ }
- msg->body.decode.dt_luma_top_offset = texture_offset(luma, 0);
- msg->body.decode.dt_chroma_top_offset = texture_offset(chroma, 0);
- if (msg->body.decode.dt_field_mode) {
- msg->body.decode.dt_luma_bottom_offset = texture_offset(luma, 1);
- msg->body.decode.dt_chroma_bottom_offset = texture_offset(chroma, 1);
- } else {
- msg->body.decode.dt_luma_bottom_offset = msg->body.decode.dt_luma_top_offset;
- msg->body.decode.dt_chroma_bottom_offset = msg->body.decode.dt_chroma_top_offset;
- }
+ msg->body.decode.dt_luma_top_offset = texture_offset(luma, 0, type);
+ if (chroma)
+ msg->body.decode.dt_chroma_top_offset = texture_offset(chroma, 0, type);
+ if (msg->body.decode.dt_field_mode) {
+ msg->body.decode.dt_luma_bottom_offset = texture_offset(luma, 1, type);
+ if (chroma)
+ msg->body.decode.dt_chroma_bottom_offset = texture_offset(chroma, 1, type);
+ } else {
+ msg->body.decode.dt_luma_bottom_offset = msg->body.decode.dt_luma_top_offset;
+ msg->body.decode.dt_chroma_bottom_offset = msg->body.decode.dt_chroma_top_offset;
+ }
- assert(luma->bankw == chroma->bankw);
- assert(luma->bankh == chroma->bankh);
- assert(luma->mtilea == chroma->mtilea);
+ if (chroma) {
+ assert(luma->u.legacy.bankw == chroma->u.legacy.bankw);
+ assert(luma->u.legacy.bankh == chroma->u.legacy.bankh);
+ assert(luma->u.legacy.mtilea == chroma->u.legacy.mtilea);
+ }
- msg->body.decode.dt_surf_tile_config |= RUVD_BANK_WIDTH(bank_wh(luma->bankw));
- msg->body.decode.dt_surf_tile_config |= RUVD_BANK_HEIGHT(bank_wh(luma->bankh));
- msg->body.decode.dt_surf_tile_config |= RUVD_MACRO_TILE_ASPECT_RATIO(macro_tile_aspect(luma->mtilea));
+ msg->body.decode.dt_surf_tile_config |= RUVD_BANK_WIDTH(bank_wh(luma->u.legacy.bankw));
+ msg->body.decode.dt_surf_tile_config |= RUVD_BANK_HEIGHT(bank_wh(luma->u.legacy.bankh));
+ msg->body.decode.dt_surf_tile_config |= RUVD_MACRO_TILE_ASPECT_RATIO(macro_tile_aspect(luma->u.legacy.mtilea));
+ break;
+ case RUVD_SURFACE_TYPE_GFX9:
+ msg->body.decode.dt_pitch = luma->u.gfx9.surf_pitch * luma->blk_w;
+ /* SWIZZLE LINEAR MODE */
+ msg->body.decode.dt_tiling_mode = RUVD_TILE_LINEAR;
+ msg->body.decode.dt_array_mode = RUVD_ARRAY_MODE_LINEAR;
+ msg->body.decode.dt_luma_top_offset = texture_offset(luma, 0, type);
+ msg->body.decode.dt_chroma_top_offset = texture_offset(chroma, 0, type);
+ if (msg->body.decode.dt_field_mode) {
+ msg->body.decode.dt_luma_bottom_offset = texture_offset(luma, 1, type);
+ msg->body.decode.dt_chroma_bottom_offset = texture_offset(chroma, 1, type);
+ } else {
+ msg->body.decode.dt_luma_bottom_offset = msg->body.decode.dt_luma_top_offset;
+ msg->body.decode.dt_chroma_bottom_offset = msg->body.decode.dt_chroma_top_offset;
+ }
+ msg->body.decode.dt_surf_tile_config = 0;
+ break;
+ }
}