static const uint32_t pp_rgbx_load_save_nv12_gen7[][4] = {
};
static const uint32_t pp_nv12_load_save_rgbx_gen7[][4] = {
+#include "shaders/post_processing/gen7/pl2_to_rgbx.g7b"
};
static VAStatus gen7_pp_plx_avs_initialize(VADriverContextP ctx, struct i965_post_processing_context *pp_context,
NULL,
},
+ gen7_pp_plx_avs_initialize,
+ },
+
+};
+
+static const uint32_t pp_null_gen75[][4] = {
+};
+
+static const uint32_t pp_nv12_load_save_nv12_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl2_to_pl2.g75b"
+};
+
+static const uint32_t pp_nv12_load_save_pl3_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl2_to_pl3.g75b"
+};
+
+static const uint32_t pp_pl3_load_save_nv12_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl3_to_pl2.g75b"
+};
+
+static const uint32_t pp_pl3_load_save_pl3_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl3_to_pl3.g75b"
+};
+
+static const uint32_t pp_nv12_scaling_gen75[][4] = {
+#include "shaders/post_processing/gen7/avs.g75b"
+};
+
+static const uint32_t pp_nv12_avs_gen75[][4] = {
+#include "shaders/post_processing/gen7/avs.g75b"
+};
+
+static const uint32_t pp_nv12_dndi_gen75[][4] = {
+// #include "shaders/post_processing/gen7/dndi.g75b"
+};
+
+static const uint32_t pp_nv12_dn_gen75[][4] = {
+// #include "shaders/post_processing/gen7/nv12_dn_nv12.g75b"
+};
+static const uint32_t pp_nv12_load_save_pa_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl2_to_pa.g75b"
+};
+static const uint32_t pp_pl3_load_save_pa_gen75[][4] = {
+#include "shaders/post_processing/gen7/pl3_to_pa.g75b"
+};
+static const uint32_t pp_pa_load_save_nv12_gen75[][4] = {
+#include "shaders/post_processing/gen7/pa_to_pl2.g75b"
+};
+static const uint32_t pp_pa_load_save_pl3_gen75[][4] = {
+#include "shaders/post_processing/gen7/pa_to_pl3.g75b"
+};
+static const uint32_t pp_rgbx_load_save_nv12_gen75[][4] = {
+};
+static const uint32_t pp_nv12_load_save_rgbx_gen75[][4] = {
+};
+
+static struct pp_module pp_modules_gen75[] = {
+ {
+ {
+ "NULL module (for testing)",
+ PP_NULL,
+ pp_null_gen75,
+ sizeof(pp_null_gen75),
+ NULL,
+ },
+
+ pp_null_initialize,
+ },
+
+ {
+ {
+ "NV12_NV12",
+ PP_NV12_LOAD_SAVE_N12,
+ pp_nv12_load_save_nv12_gen75,
+ sizeof(pp_nv12_load_save_nv12_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "NV12_PL3",
+ PP_NV12_LOAD_SAVE_PL3,
+ pp_nv12_load_save_pl3_gen75,
+ sizeof(pp_nv12_load_save_pl3_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "PL3_NV12",
+ PP_PL3_LOAD_SAVE_N12,
+ pp_pl3_load_save_nv12_gen75,
+ sizeof(pp_pl3_load_save_nv12_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "PL3_PL3",
+ PP_PL3_LOAD_SAVE_N12,
+ pp_pl3_load_save_pl3_gen75,
+ sizeof(pp_pl3_load_save_pl3_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "NV12 Scaling module",
+ PP_NV12_SCALING,
+ pp_nv12_scaling_gen75,
+ sizeof(pp_nv12_scaling_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "NV12 AVS module",
+ PP_NV12_AVS,
+ pp_nv12_avs_gen75,
+ sizeof(pp_nv12_avs_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "NV12 DNDI module",
+ PP_NV12_DNDI,
+ pp_nv12_dndi_gen75,
+ sizeof(pp_nv12_dndi_gen75),
+ NULL,
+ },
+
+ gen7_pp_nv12_dndi_initialize,
+ },
+
+ {
+ {
+ "NV12 DN module",
+ PP_NV12_DN,
+ pp_nv12_dn_gen75,
+ sizeof(pp_nv12_dn_gen75),
+ NULL,
+ },
+
+ gen7_pp_nv12_dn_initialize,
+ },
+ {
+ {
+ "NV12_PA module",
+ PP_NV12_LOAD_SAVE_PA,
+ pp_nv12_load_save_pa_gen75,
+ sizeof(pp_nv12_load_save_pa_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "PL3_PA module",
+ PP_PL3_LOAD_SAVE_PA,
+ pp_pl3_load_save_pa_gen75,
+ sizeof(pp_pl3_load_save_pa_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "PA_NV12 module",
+ PP_PA_LOAD_SAVE_NV12,
+ pp_pa_load_save_nv12_gen75,
+ sizeof(pp_pa_load_save_nv12_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "PA_PL3 module",
+ PP_PA_LOAD_SAVE_PL3,
+ pp_pa_load_save_pl3_gen75,
+ sizeof(pp_pa_load_save_pl3_gen75),
+ NULL,
+ },
+
+ gen7_pp_plx_avs_initialize,
+ },
+
+ {
+ {
+ "RGBX_NV12 module",
+ PP_RGBX_LOAD_SAVE_NV12,
+ pp_rgbx_load_save_nv12_gen75,
+ sizeof(pp_rgbx_load_save_nv12_gen75),
+ NULL,
+ },
+
+ pp_plx_load_save_plx_initialize,
+ },
+
+ {
+ {
+ "NV12_RGBX module",
+ PP_NV12_LOAD_SAVE_RGBX,
+ pp_nv12_load_save_rgbx_gen75,
+ sizeof(pp_nv12_load_save_rgbx_gen75),
+ NULL,
+ },
+
pp_plx_load_save_plx_initialize,
},
int width, int height, int pitch, int format,
int index, int is_target)
{
+ struct i965_driver_data * const i965 = i965_driver_data(ctx);
struct gen7_surface_state *ss;
dri_bo *ss_bo;
unsigned int tiling;
ss->ss2.height = height - 1;
ss->ss3.pitch = pitch - 1;
gen7_pp_set_surface_tiling(ss, tiling);
+ if (IS_HASWELL(i965->intel.device_id))
+ gen7_render_set_surface_scs(ss);
dri_bo_emit_reloc(ss_bo,
I915_GEM_DOMAIN_RENDER, is_target ? I915_GEM_DOMAIN_RENDER : 0,
surf_bo_offset,
fourcc == VA_FOURCC('I', 'M', 'C', '1')) ? 1 : 2;
int interleaved_uv = fourcc == VA_FOURCC('N', 'V', '1', '2');
int packed_yuv = (fourcc == VA_FOURCC('Y', 'U', 'Y', '2') || fourcc == VA_FOURCC('U', 'Y', 'V', 'Y'));
+ int rgbx_format = (fourcc == VA_FOURCC('R', 'G', 'B', 'A') ||
+ fourcc == VA_FOURCC('R', 'G', 'B', 'X') ||
+ fourcc == VA_FOURCC('B', 'G', 'R', 'A') ||
+ fourcc == VA_FOURCC('B', 'G', 'R', 'X'));
if (surface->type == I965_SURFACE_TYPE_SURFACE) {
obj_surface = SURFACE(surface->id);
width[0] = obj_surface->orig_width; /* surface foramt is YCBCR, width is specified in units of pixels */
pitch[0] = obj_surface->width * 2;
- }
+ } else if (rgbx_format) {
+ if (is_target)
+ width[0] = obj_surface->orig_width * 4; /* surface format is R8, so quad the width */
+ }
width[1] = obj_surface->cb_cr_width;
height[1] = obj_surface->cb_cr_height;
pitch[0] = obj_image->image.pitches[0];
offset[0] = obj_image->image.offsets[0];
- if (packed_yuv) {
+ if (rgbx_format) {
+ if (is_target)
+ width[0] = obj_image->image.width * 4; /* surface format is R8, so quad the width */
+ } else if (packed_yuv) {
if (is_target)
width[0] = obj_image->image.width * 2; /* surface format is R8, so double the width */
else
gen7_pp_set_surface_state(ctx, pp_context,
bo, 0,
width[0] / 4, height[0], pitch[0],
- I965_SURFACEFORMAT_R8_SINT,
+ I965_SURFACEFORMAT_R8_UINT,
base_index, 1);
-
- if (!packed_yuv) {
+ if (rgbx_format) {
+ struct gen7_pp_static_parameter *pp_static_parameter = pp_context->pp_static_parameter;
+ /* the format is MSB: X-B-G-R */
+ pp_static_parameter->grf2.save_avs_rgb_swap = 0;
+ if ((fourcc == VA_FOURCC('B', 'G', 'R', 'A')) ||
+ (fourcc == VA_FOURCC('B', 'G', 'R', 'X'))) {
+ /* It is stored as MSB: X-R-G-B */
+ pp_static_parameter->grf2.save_avs_rgb_swap = 1;
+ }
+ }
+ if (!packed_yuv && !rgbx_format) {
if (interleaved_uv) {
gen7_pp_set_surface_state(ctx, pp_context,
bo, offset[1],
pp_load_save_set_block_parameter(struct i965_post_processing_context *pp_context, int x, int y)
{
struct pp_inline_parameter *pp_inline_parameter = pp_context->pp_inline_parameter;
+ struct pp_load_save_context *pp_load_save_context = (struct pp_load_save_context *)&pp_context->private_context;
- pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16;
- pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8;
+ pp_inline_parameter->grf5.destination_block_horizontal_origin = x * 16 + pp_load_save_context->dest_x;
+ pp_inline_parameter->grf5.destination_block_vertical_origin = y * 8 + pp_load_save_context->dest_y;
return 0;
}
pp_context->pp_x_steps = pp_load_save_x_steps;
pp_context->pp_y_steps = pp_load_save_y_steps;
pp_context->pp_set_block_parameter = pp_load_save_set_block_parameter;
- pp_load_save_context->dest_h = ALIGN(height[Y], 8);
- pp_load_save_context->dest_w = ALIGN(width[Y], 16);
- pp_inline_parameter->grf5.block_count_x = ALIGN(width[Y], 16) / 16; /* 1 x N */
- pp_inline_parameter->grf5.number_blocks = ALIGN(width[Y], 16) / 16;
+ int dst_left_edge_extend = dst_rect->x%GPU_ASM_X_OFFSET_ALIGNMENT;;
+ pp_load_save_context->dest_x = dst_rect->x - dst_left_edge_extend;
+ pp_load_save_context->dest_y = dst_rect->y;
+ pp_load_save_context->dest_h = ALIGN(dst_rect->height, 8);
+ pp_load_save_context->dest_w = ALIGN(dst_rect->width+dst_left_edge_extend, 16);
+
+ pp_inline_parameter->grf5.block_count_x = pp_load_save_context->dest_w / 16; /* 1 x N */
+ pp_inline_parameter->grf5.number_blocks = pp_load_save_context->dest_w / 16;
pp_static_parameter->grf3.horizontal_origin_offset = src_rect->x;
pp_static_parameter->grf3.vertical_origin_offset = src_rect->y;
pp_context->pp_y_steps = pp_scaling_y_steps;
pp_context->pp_set_block_parameter = pp_scaling_set_block_parameter;
- pp_scaling_context->dest_x = dst_rect->x;
+ int dst_left_edge_extend = dst_rect->x%GPU_ASM_X_OFFSET_ALIGNMENT;
+ float src_left_edge_extend = (float)dst_left_edge_extend*src_rect->width/dst_rect->width;
+ pp_scaling_context->dest_x = dst_rect->x - dst_left_edge_extend;
pp_scaling_context->dest_y = dst_rect->y;
- pp_scaling_context->dest_w = ALIGN(dst_rect->width, 16);
- pp_scaling_context->dest_h = ALIGN(dst_rect->height, 16);
- pp_scaling_context->src_normalized_x = (float)src_rect->x / in_w;
+ pp_scaling_context->dest_w = ALIGN(dst_rect->width + dst_left_edge_extend, 16);
+ pp_scaling_context->dest_h = ALIGN(dst_rect->height, 8);
+ pp_scaling_context->src_normalized_x = (float)(src_rect->x - src_left_edge_extend)/ in_w;
pp_scaling_context->src_normalized_y = (float)src_rect->y / in_h;
pp_static_parameter->grf1.r1_6.normalized_video_y_scaling_step = (float) src_rect->height / in_h / dst_rect->height;
- pp_inline_parameter->grf5.normalized_video_x_scaling_step = (float) src_rect->width / in_w / dst_rect->width;
+ pp_inline_parameter->grf5.normalized_video_x_scaling_step = (float) (src_rect->width + src_left_edge_extend)/ in_w / (dst_rect->width + dst_left_edge_extend);
pp_inline_parameter->grf5.block_count_x = pp_scaling_context->dest_w / 16; /* 1 x N */
pp_inline_parameter->grf5.number_blocks = pp_scaling_context->dest_w / 16;
for (x = 0; x < x_steps; x++) {
if (!pp_context->pp_set_block_parameter(pp_context, x, y)) {
// some common block parameter update goes here, apply to all pp functions
- update_block_mask_parameter (pp_context, x, y, x_steps, y_steps);
+ if (IS_GEN6(i965->intel.device_id))
+ update_block_mask_parameter (pp_context, x, y, x_steps, y_steps);
*command_ptr++ = (CMD_MEDIA_OBJECT | (command_length_in_dws - 2));
*command_ptr++ = 0;
intel_batchbuffer_end_atomic(batch);
}
+VAStatus
+i965_scaling_processing(
+ VADriverContextP ctx,
+ VASurfaceID src_surface_id,
+ const VARectangle *src_rect,
+ VASurfaceID dst_surface_id,
+ const VARectangle *dst_rect,
+ unsigned int flags)
+{
+ VAStatus va_status = VA_STATUS_SUCCESS;
+ struct i965_driver_data *i965 = i965_driver_data(ctx);
+ struct object_surface *src_surface_obj = SURFACE(src_surface_id);
+ struct object_surface *dst_surface_obj = SURFACE(dst_surface_id);
+
+ assert(src_surface_obj->fourcc == VA_FOURCC('N', 'V', '1', '2'));
+ assert(dst_surface_obj->fourcc == VA_FOURCC('N', 'V', '1', '2'));
+
+ if (HAS_PP(i965) && (flags & I965_PP_FLAG_AVS)) {
+ struct i965_surface src_surface;
+ struct i965_surface dst_surface;
+
+ _i965LockMutex(&i965->pp_mutex);
+
+ src_surface.id = src_surface_id;
+ src_surface.type = I965_SURFACE_TYPE_SURFACE;
+ src_surface.flags = I965_SURFACE_FLAG_FRAME;
+ dst_surface.id = dst_surface_id;
+ dst_surface.type = I965_SURFACE_TYPE_SURFACE;
+ dst_surface.flags = I965_SURFACE_FLAG_FRAME;
+
+ va_status = i965_post_processing_internal(ctx, i965->pp_context,
+ &src_surface,
+ src_rect,
+ &dst_surface,
+ dst_rect,
+ PP_NV12_AVS,
+ NULL);
+
+ _i965UnlockMutex(&i965->pp_mutex);
+ }
+
+ return va_status;
+}
+
VASurfaceID
i965_post_processing(
VADriverContextP ctx,
assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen5));
assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen6));
assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen7));
+ assert(NUM_PP_MODULES == ARRAY_ELEMS(pp_modules_gen75));
- if (IS_GEN7(i965->intel.device_id))
+ if (IS_HASWELL(i965->intel.device_id))
+ memcpy(pp_context->pp_modules, pp_modules_gen75, sizeof(pp_context->pp_modules));
+ else if (IS_GEN7(i965->intel.device_id))
memcpy(pp_context->pp_modules, pp_modules_gen7, sizeof(pp_context->pp_modules));
else if (IS_GEN6(i965->intel.device_id))
memcpy(pp_context->pp_modules, pp_modules_gen6, sizeof(pp_context->pp_modules));
I965_SURFACE_FLAG_BOTTOME_FIELD_FIRST
};
-static void
+void
i965_proc_picture(VADriverContextP ctx,
VAProfile profile,
union codec_state *codec_state,
dst_surface.type = I965_SURFACE_TYPE_SURFACE;
i965_vpp_clear_surface(ctx, &proc_context->pp_context, proc_state->current_render_target, pipeline_param->output_background_color);
+
+ // load/save doesn't support different origin offset for src and dst surface
if (src_rect.width == dst_rect.width &&
- src_rect.height == dst_rect.height) {
+ src_rect.height == dst_rect.height &&
+ src_rect.x == dst_rect.x &&
+ src_rect.y == dst_rect.y) {
i965_post_processing_internal(ctx, &proc_context->pp_context,
&src_surface,
&src_rect,
proc_context->base.destroy = i965_proc_context_destroy;
proc_context->base.run = i965_proc_picture;
- proc_context->base.batch = intel_batchbuffer_new(intel, I915_EXEC_RENDER);
+ proc_context->base.batch = intel_batchbuffer_new(intel, I915_EXEC_RENDER, 0);
i965_post_processing_context_init(ctx, &proc_context->pp_context, proc_context->base.batch);
return (struct hw_context *)proc_context;