/* All these are in dwords. The maximum is 16K dwords (64KB) of LDS per workgroup. */
const unsigned scratch_lds_size = gfx10_ngg_get_scratch_dw_size(shader);
- /* Scrach is at last of LDS space and 2 dwords aligned, so it may cost more for alignment. */
+ /* Scratch is at last of LDS space and 2 dwords aligned, so it may cost more for alignment. */
const unsigned max_lds_size = 16 * 1024 - ALIGN(scratch_lds_size, 2);
const unsigned target_lds_size = max_lds_size;
unsigned esvert_lds_size = 0;
dpb_size += align(width_in_mb * height_in_mb * 32, alignment);
}
} else {
- // the firmware seems to allways assume a minimum of ref frames
+ // the firmware seems to always assume a minimum of ref frames
max_references = MAX2(NUM_H264_REFS, max_references);
// reference picture buffer
dpb_size = image_size * max_references;
break;
case PIPE_VIDEO_FORMAT_VC1:
- // the firmware seems to allways assume a minimum of ref frames
+ // the firmware seems to always assume a minimum of ref frames
max_references = MAX2(NUM_VC1_REFS, max_references);
// reference picture buffer
#define DRL_MODE_CONTEXTS 3
#define INTER_COMPOUND_MODES (1 + AV1_NEW_NEWMV - AV1_NEAREST_NEARESTMV)
-#define PALATTE_BSIZE_CTXS 9
+#define PALETTE_BSIZE_CTXS 9
#define PALETTE_COLOR_INDEX_CONTEXTS 5
#define PALETTE_Y_MODE_CONTEXTS 3
#define PALETTE_UV_MODE_CONTEXTS 2
uint16_t skip_mode_cdfs[SKIP_CONTEXTS][CDF_SIZE(2)];
uint16_t skip_cdfs[SKIP_CONTEXTS][CDF_SIZE(2)];
uint16_t intra_inter_cdf[AV1_INTRA_INTER_CONTEXTS][CDF_SIZE(2)];
- uint16_t palette_y_mode_cdf[PALATTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)];
+ uint16_t palette_y_mode_cdf[PALETTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)];
uint16_t palette_uv_mode_cdf[PALETTE_UV_MODE_CONTEXTS][CDF_SIZE(2)];
uint16_t filter_intra_cdfs[AV1_BLOCK_SIZES_ALL][CDF_SIZE(2)];
uint16_t wiener_restore_cdf[CDF_SIZE(2)];
uint16_t filter_intra_mode_cdf[CDF_SIZE(FILTER_INTRA_MODES)];
uint16_t eob_flag_cdf32[AV1_PLANE_TYPES][2][CDF_SIZE(6)];
uint16_t eob_flag_cdf64[AV1_PLANE_TYPES][2][CDF_SIZE(7)];
- uint16_t palette_y_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
- uint16_t palette_uv_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
+ uint16_t palette_y_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
+ uint16_t palette_uv_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
uint16_t angle_delta_cdf[DIRECTIONAL_MODES][CDF_SIZE(2 * MAX_ANGLE_DELTA + 1)];
uint16_t reserve3[6];
uint16_t eob_flag_cdf128[AV1_PLANE_TYPES][2][CDF_SIZE(8)];
{ { AOM_CDF2(2235) },{ AOM_CDF2(1423) } }, { { AOM_CDF2(17182) },{ AOM_CDF2(15175) } }, { { AOM_CDF2(30606) },{ AOM_CDF2(30489) } }
};
-static const uint16_t default_palette_y_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)] =
+static const uint16_t default_palette_y_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)] =
{
{ AOM_CDF7(7952, 13000, 18149, 21478, 25527, 29241) }, { AOM_CDF7(7139, 11421, 16195, 19544, 23666, 28073) },
{ AOM_CDF7(7788, 12741, 17325, 20500, 24315, 28530) }, { AOM_CDF7(8271, 14064, 18246, 21564, 25071, 28533) },
{ AOM_CDF7(14940, 20797, 21678, 24186, 27033, 28999) }
};
-static const uint16_t default_palette_uv_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)] =
+static const uint16_t default_palette_uv_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)] =
{
{ AOM_CDF7(8713, 19979, 27128, 29609, 31331, 32272) }, { AOM_CDF7(5839, 15573, 23581, 26947, 29848, 31700) },
{ AOM_CDF7(4426, 11260, 17999, 21483, 25863, 29430) }, { AOM_CDF7(3228, 9464, 14993, 18089, 22523, 27420) },
{ AOM_CDF7(1269, 5435, 10433, 18963, 21700, 25865) }
};
-static const uint16_t default_palette_y_mode_cdf[PALATTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)] =
+static const uint16_t default_palette_y_mode_cdf[PALETTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)] =
{
{ { AOM_CDF2(31676) },{ AOM_CDF2(3419) },{ AOM_CDF2(1261) } }, { { AOM_CDF2(31912) },{ AOM_CDF2(2859) },{ AOM_CDF2(980) } },
{ { AOM_CDF2(31823) },{ AOM_CDF2(3400) },{ AOM_CDF2(781) } }, { { AOM_CDF2(32030) },{ AOM_CDF2(3561) },{ AOM_CDF2(904) } },
uint16_t skip_mode_cdfs[SKIP_CONTEXTS][CDF_SIZE(2)];
uint16_t skip_cdfs[SKIP_CONTEXTS][CDF_SIZE(2)];
uint16_t intra_inter_cdf[AV1_INTRA_INTER_CONTEXTS][CDF_SIZE(2)];
- uint16_t palette_y_mode_cdf[PALATTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)];
+ uint16_t palette_y_mode_cdf[PALETTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][CDF_SIZE(2)];
uint16_t palette_uv_mode_cdf[PALETTE_UV_MODE_CONTEXTS][CDF_SIZE(2)];
uint16_t filter_intra_cdfs[AV1_BLOCK_SIZES_ALL][CDF_SIZE(2)];
uint16_t wiener_restore_cdf[CDF_SIZE(2)];
uint16_t filter_intra_mode_cdf[CDF_SIZE(FILTER_INTRA_MODES)];
uint16_t eob_flag_cdf32[AV1_PLANE_TYPES][2][CDF_SIZE(6)];
uint16_t eob_flag_cdf64[AV1_PLANE_TYPES][2][CDF_SIZE(7)];
- uint16_t palette_y_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
- uint16_t palette_uv_size_cdf[PALATTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
+ uint16_t palette_y_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
+ uint16_t palette_uv_size_cdf[PALETTE_BSIZE_CTXS][CDF_SIZE(PALETTE_SIZES)];
uint16_t angle_delta_cdf[DIRECTIONAL_MODES][CDF_SIZE(2 * MAX_ANGLE_DELTA + 1)];
uint16_t reserve3[6];
uint16_t eob_flag_cdf128[AV1_PLANE_TYPES][2][CDF_SIZE(8)];
break;
case PIPE_VIDEO_FORMAT_VC1:
- // the firmware seems to allways assume a minimum of ref frames
+ // the firmware seems to always assume a minimum of ref frames
max_references = MAX2(NUM_VC1_REFS, max_references);
// reference picture buffer
#define RENCODE_H264_DISABLE_DEBLOCKING_FILTER_IDC_ENABLE 0
#define RENCODE_H264_DISABLE_DEBLOCKING_FILTER_IDC_DISABLE 1
-#define RENCODE_H264_DISABLE_DEBLOCKING_FILTER_IDC_DISALBE_ACROSS_SLICE_BOUNDARY 2
+#define RENCODE_H264_DISABLE_DEBLOCKING_FILTER_IDC_DISABLE_ACROSS_SLICE_BOUNDARY 2
#define RENCODE_INTRA_REFRESH_MODE_NONE 0
#define RENCODE_INTRA_REFRESH_MODE_CTB_MB_ROWS 1
custom_blend == sctx->custom_blend_dcc_decompress)
sctx->flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
- /* When running FMASK decompresion with DCC, we need to run the "eliminate fast clear" pass
+ /* When running FMASK decompression with DCC, we need to run the "eliminate fast clear" pass
* separately because FMASK decompression doesn't eliminate DCC fast clear. This makes
* render->texture transitions more expensive. It can be disabled by
* allow_dcc_msaa_clear_to_reg_for_bpp.
struct si_context *sctx = (struct si_context *)ctx;
struct si_resource *buf = si_resource(resource);
- /* We currently only do anyting here for buffers */
+ /* We currently only do anything here for buffers */
if (resource->target == PIPE_BUFFER)
(void)si_invalidate_buffer(sctx, buf);
}
sctx->b.buffer_subdata(&sctx->b, dst,
PIPE_MAP_WRITE |
- /* TC forbids drivers to invalidate buffers and infer unsychronized mappings,
+ /* TC forbids drivers to invalidate buffers and infer unsynchronized mappings,
* so suppress those optimizations. */
(sctx->tc ? TC_TRANSFER_MAP_NO_INFER_UNSYNCHRONIZED |
TC_TRANSFER_MAP_NO_INVALIDATE : 0),
if (!saved->bo_list)
return;
- /* Sort the list according to VM adddresses first. */
+ /* Sort the list according to VM addresses first. */
qsort(saved->bo_list, saved->bo_count, sizeof(saved->bo_list[0]), (void *)bo_list_compare_va);
fprintf(f, "Buffer list (in units of pages = 4kB):\n" COLOR_YELLOW
#include "vl/vl_video_buffer.h"
#include <sys/utsname.h>
-/* The capabilties reported by the kernel has priority
+/* The capabilities reported by the kernel has priority
over the existing logic in si_get_video_param */
#define QUERYABLE_KERNEL (!!(sscreen->info.drm_minor >= 41))
#define KERNEL_DEC_CAP(codec, attrib) \
* z = [2..8]
* c = [2..8]
*
- * Only MSAA color and depth buffers are overriden.
+ * Only MSAA color and depth buffers are overridden.
*/
if (sscreen->info.has_eqaa_surface_allocator) {
const char *eqaa = debug_get_option("EQAA", NULL);
* - polygon stippling
*
* All preloaded SGPRs and VGPRs are passed through unmodified unless they are
- * overriden by other states. (e.g. per-sample interpolation)
+ * overridden by other states. (e.g. per-sample interpolation)
* Interpolated colors are stored after the preloaded VGPRs.
*/
void si_llvm_build_ps_prolog(struct si_shader_context *ctx, union si_shader_part_key *key,
}
/**
- * Compile the TCS epilog function. This writes tesselation factors to memory
- * based on the output primitive type of the tesselator (determined by TES).
+ * Compile the TCS epilog function. This writes tessellation factors to memory
+ * based on the output primitive type of the tessellator (determined by TES).
*/
void si_llvm_build_tcs_epilog(struct si_shader_context *ctx, union si_shader_part_key *key,
UNUSED bool separate_epilog)
S_028B50_ACCUM_QUAD(11) |
S_028B50_DONUT_SPLIT_GFX81(16);
- /* Testing with Unigine Heaven extreme tesselation yielded best results
+ /* Testing with Unigine Heaven extreme tessellation yielded best results
* with TRAP_SPLIT = 3.
*/
if (sctx->family == CHIP_FIJI || sctx->family >= CHIP_POLARIS10)
shader->ps.db_shader_control |= S_02880C_DUAL_QUAD_DISABLE(1);
/* SPI_BARYC_CNTL.POS_FLOAT_LOCATION
- * Possible vaules:
+ * Possible values:
* 0 -> Position = pixel center
* 1 -> Position = pixel centroid
* 2 -> Position = at sample position
if (util_rast_prim_is_triangles(sel->rast_prim))
sel->rast_prim = PIPE_PRIM_TRIANGLES;
- /* EN_MAX_VERT_OUT_PER_GS_INSTANCE does not work with tesselation so
+ /* EN_MAX_VERT_OUT_PER_GS_INSTANCE does not work with tessellation so
* we can't split workgroups. Disable ngg if any of the following conditions is true:
* - num_invocations * gs.vertices_out > 256
* - LDS usage is too high
if (templ->nr_samples >= 2) {
/* This is hackish (overwriting the const pipe_resource template),
* but should be harmless and gallium frontends can also see
- * the overriden number of samples in the created pipe_resource.
+ * the overridden number of samples in the created pipe_resource.
*/
if (is_zs && sscreen->eqaa_force_z_samples) {
((struct pipe_resource *)templ)->nr_samples =
if (multi_sample && sscreen->info.gfx_level != GFX9)
return 0;
- /* Unsupport formats. */
+ /* Unsupported formats. */
/* TODO: support these formats. */
if (util_format_is_depth_or_stencil(format) ||
util_format_get_num_planes(format) > 1 ||