sysutils/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2026-10-24 02:09:03 +03:00
Code Activity

swscale/uops: parametrize filter op result type

Browse Source 
The ops.h infrastructure currently hard-codes this as SWS_PIXEL_F32,
but I want to at least properly parametrize this in case we ever
decide to revisit this decision in the future. In particular, it
may become relevant for trivial kernels or kernels whose intermediates
are bounded, exact integers (which could possibly be output directly
as e.g. U16 or U32).

The FATE change is just because the filter op names gained a suffix.

Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
This commit is contained in:
Niklas Haas
2026-06-09 01:35:51 +02:00
parent 4a8a1f5b8b
commit 6057759ffc
5 changed files with 104 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
libswscale/uops.c
View File

@@ -144,6 +144,11 @@ void ff_sws_uop_name(const SwsUOp *op, char buf[SWS_UOP_NAME_MAX])
const SwsUOpParams *par = &op->par;
switch (op->uop) {
case SWS_UOP_READ_PLANAR_FH:
case SWS_UOP_READ_PLANAR_FV:
case SWS_UOP_READ_PLANAR_FV_FMA:
av_bprintf(&bp, "_%s", ff_sws_pixel_type_name(par->filter.type));
break;
case SWS_UOP_LSHIFT:
case SWS_UOP_RSHIFT:
av_bprintf(&bp, "_%u", par->shift.amount);
@@ -219,6 +224,11 @@ static int generate_entry_struct(void *opaque, void *key)
const SwsUOpParams *par = &uop->par;
switch (uop->uop) {
case SWS_UOP_READ_PLANAR_FH:
case SWS_UOP_READ_PLANAR_FV:
case SWS_UOP_READ_PLANAR_FV_FMA:
av_bprintf(bp, ", .par.filter.type = %s", pixel_types[par->filter.type].full);
break;
case SWS_UOP_LSHIFT:
case SWS_UOP_RSHIFT:
av_bprintf(bp, ", .par.shift.amount = %u", par->shift.amount);
@@ -270,6 +280,11 @@ static int generate_entry_args(void *opaque, void *key)
const SwsUOpParams *par = &uop->par;
switch (uop->uop) {
case SWS_UOP_READ_PLANAR_FH:
case SWS_UOP_READ_PLANAR_FV:
case SWS_UOP_READ_PLANAR_FV_FMA:
av_bprintf(bp, ", %s", pixel_types[par->filter.type].full);
break;
case SWS_UOP_LSHIFT:
case SWS_UOP_RSHIFT:
av_bprintf(bp, ", %u", par->shift.amount);
@@ -440,6 +455,7 @@ static int translate_rw_op(SwsContext *ctx, SwsUOpList *ops, SwsUOpFlags flags,
if (op->rw.filter) {
if (op->op == SWS_OP_WRITE || op->rw.frac || op->rw.packed)
return AVERROR(ENOTSUP);
uop.par.filter.type = SWS_PIXEL_F32;
uop.data.kernel = av_refstruct_ref(op->rw.kernel);
if (op->rw.filter == SWS_OP_FILTER_H) {
uop.uop = SWS_UOP_READ_PLANAR_FH;

5
libswscale/uops.h
View File

@@ -135,6 +135,10 @@ typedef enum SwsUOpType {
SWS_UOP_TYPE_NB,
} SwsUOpType;
typedef struct SwsFilterUOp {
SwsPixelType type; /* pixel type to store result as */
} SwsFilterUOp;
typedef struct SwsShiftUOp {
uint8_t amount;
} SwsShiftUOp;
@@ -172,6 +176,7 @@ typedef struct SwsDitherUOp {
int ff_sws_dither_height(const SwsDitherUOp *dither);
typedef union SwsUOpParams {
SwsFilterUOp filter; /* for SWS_UOP_READ_*_FV/FH */
SwsShiftUOp shift;
SwsSwizzleUOp swizzle;
SwsPackUOp pack;

144
libswscale/uops_macros.h
View File

@@ -30,35 +30,35 @@
MACRO(__VA_ARGS__, u8_read_planar_xyz , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR , .mask = 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_xyzw , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR , .mask = 0xf)
#define SWS_FOR_U8_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fh_x , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xy , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyz , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyzw , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fh_x_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xy_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyz_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyzw_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FH , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U8_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fh_x , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xy , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyz , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyzw , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fh_x_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xy_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyz_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fh_xyzw_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U8_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fv_x , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xy , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyz , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyzw , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fv_x_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xy_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyz_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyzw_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U8_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fv_x , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xy , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyz , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyzw , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fv_x_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xy_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyz_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_xyzw_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U8_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_x , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xy , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyz , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyzw , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_x_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xy_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyz_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyzw_f32 , SWS_PIXEL_U8 , SWS_UOP_READ_PLANAR_FV_FMA, 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U8_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_x , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xy , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyz , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyzw , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf)
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_x_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xy_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyz_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u8_read_planar_fv_fma_xyzw_f32 , .type = SWS_PIXEL_U8 , .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U8_READ_PACKED(MACRO, ...) \
MACRO(__VA_ARGS__, u8_read_packed_xy , SWS_PIXEL_U8 , SWS_UOP_READ_PACKED , 0x3) \
MACRO(__VA_ARGS__, u8_read_packed_xyz , SWS_PIXEL_U8 , SWS_UOP_READ_PACKED , 0x7) \
@@ -306,35 +306,35 @@
MACRO(__VA_ARGS__, u16_read_planar_xyz , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR , .mask = 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_xyzw , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR , .mask = 0xf)
#define SWS_FOR_U16_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fh_x , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xy , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyz , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyzw , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fh_x_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xy_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyz_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyzw_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FH , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U16_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fh_x , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xy , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyz , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyzw , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fh_x_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xy_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyz_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fh_xyzw_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U16_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fv_x , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xy , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyz , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyzw , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fv_x_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xy_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyz_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyzw_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U16_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fv_x , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xy , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyz , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyzw , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fv_x_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xy_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyz_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_xyzw_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U16_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_x , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xy , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyz , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyzw , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_x_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xy_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyz_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyzw_f32 , SWS_PIXEL_U16, SWS_UOP_READ_PLANAR_FV_FMA, 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_U16_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_x , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xy , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyz , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyzw , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf)
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_x_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xy_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyz_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, u16_read_planar_fv_fma_xyzw_f32 , .type = SWS_PIXEL_U16, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_U16_READ_PACKED(MACRO, ...) \
MACRO(__VA_ARGS__, u16_read_packed_xy , SWS_PIXEL_U16, SWS_UOP_READ_PACKED , 0x3) \
MACRO(__VA_ARGS__, u16_read_packed_xyz , SWS_PIXEL_U16, SWS_UOP_READ_PACKED , 0x7) \
@@ -790,35 +790,35 @@
#define SWS_FOR_F32_READ_PLANAR(MACRO, ...)
#define SWS_FOR_STRUCT_F32_READ_PLANAR(MACRO, ...)
#define SWS_FOR_F32_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fh_x , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xy , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyz , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyzw , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fh_x_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xy_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyz_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyzw_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FH , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_F32_READ_PLANAR_FH(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fh_x , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xy , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyz , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyzw , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fh_x_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xy_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyz_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fh_xyzw_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FH , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_F32_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fv_x , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xy , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyz , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyzw , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fv_x_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xy_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyz_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyzw_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV , 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_F32_READ_PLANAR_FV(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fv_x , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xy , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyz , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyzw , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fv_x_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xy_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyz_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_xyzw_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV , .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_F32_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_x , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xy , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyz , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyzw , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_x_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x1, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xy_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x3, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyz_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0x7, SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyzw_f32 , SWS_PIXEL_F32, SWS_UOP_READ_PLANAR_FV_FMA, 0xf, SWS_PIXEL_F32)
#define SWS_FOR_STRUCT_F32_READ_PLANAR_FV_FMA(MACRO, ...) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_x , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xy , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyz , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyzw , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf)
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_x_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x1, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xy_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x3, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyz_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0x7, .par.filter.type = SWS_PIXEL_F32) \
MACRO(__VA_ARGS__, f32_read_planar_fv_fma_xyzw_f32 , .type = SWS_PIXEL_F32, .uop = SWS_UOP_READ_PLANAR_FV_FMA, .mask = 0xf, .par.filter.type = SWS_PIXEL_F32)
#define SWS_FOR_F32_READ_PACKED(MACRO, ...)
#define SWS_FOR_STRUCT_F32_READ_PACKED(MACRO, ...)
#define SWS_FOR_F32_READ_NIBBLE(MACRO, ...)

12
libswscale/uops_tmpl.c
View File

@@ -219,6 +219,9 @@ SWS_FOR_STRUCT(PX, WRITE_BIT, DECL_ENTRY)
DECL_SETUP(setup_filter_v, params, out)
{
if (params->uop->par.filter.type != SWS_PIXEL_F32)
return AVERROR(ENOTSUP);
const SwsFilterWeights *filter = params->uop->data.kernel;
static_assert(sizeof(out->priv.ptr) <= sizeof(int32_t[2]),
">8 byte pointers not supported");
@@ -238,8 +241,9 @@ DECL_SETUP(setup_filter_v, params, out)
}
/* Fully general vertical planar filter case */
DECL_READ(read_planar_fv, const SwsCompMask mask)
DECL_READ(read_planar_fv, const SwsCompMask mask, const SwsPixelType type)
{
av_assert2(type == SWS_PIXEL_F32);
const SwsOpExec *exec = iter->exec;
const float *restrict weights = impl->priv.ptr;
const int filter_size = impl->priv.i32[2];
@@ -278,6 +282,9 @@ DECL_READ(read_planar_fv, const SwsCompMask mask)
DECL_SETUP(setup_filter_h, params, out)
{
if (params->uop->par.filter.type != SWS_PIXEL_F32)
return AVERROR(ENOTSUP);
SwsFilterWeights *filter = params->uop->data.kernel;
out->priv.ptr = av_refstruct_ref(filter->weights);
out->priv.i32[2] = filter->filter_size;
@@ -286,8 +293,9 @@ DECL_SETUP(setup_filter_h, params, out)
}
/* Fully general horizontal planar filter case */
DECL_READ(read_planar_fh, const SwsCompMask mask)
DECL_READ(read_planar_fh, const SwsCompMask mask, const SwsPixelType type)
{
av_assert2(type == SWS_PIXEL_F32);
const SwsOpExec *exec = iter->exec;
const int *restrict weights = impl->priv.ptr;
const int filter_size = impl->priv.i32[2];

2
tests/ref/fate/sws-ops-list
View File

@@ -1 +1 @@
900da2b2c6276da01ae3f158d02abf0b
e2f26cb6df5c11015e613016bb1a004a