Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improve zigzag scan #32

Merged
merged 1 commit into from
Oct 6, 2023
Merged

Improve zigzag scan #32

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 16 additions & 8 deletions lib/block_coding_128.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,10 +89,14 @@ auto row4_ne_0 = VecFromMask(s16, Eq(row4, zero));
auto row5_ne_0 = VecFromMask(s16, Eq(row5, zero));
auto row6_ne_0 = VecFromMask(s16, Eq(row6, zero));
auto row7_ne_0 = VecFromMask(s16, Eq(row7, zero));
auto row10_ne_0 = ConcatEven(u8, BitCast(u8, row0_ne_0), BitCast(u8, row1_ne_0));
auto row32_ne_0 = ConcatEven(u8, BitCast(u8, row2_ne_0), BitCast(u8, row3_ne_0));
auto row54_ne_0 = ConcatEven(u8, BitCast(u8, row4_ne_0), BitCast(u8, row5_ne_0));
auto row76_ne_0 = ConcatEven(u8, BitCast(u8, row6_ne_0), BitCast(u8, row7_ne_0));
auto row10_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row1_ne_0), BitCast(u16, row0_ne_0));
auto row32_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row3_ne_0), BitCast(u16, row2_ne_0));
auto row54_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row5_ne_0), BitCast(u16, row4_ne_0));
auto row76_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row7_ne_0), BitCast(u16, row6_ne_0));
// auto row10_ne_0 = ConcatEven(u8, BitCast(u8, row0_ne_0), BitCast(u8, row1_ne_0));
// auto row32_ne_0 = ConcatEven(u8, BitCast(u8, row2_ne_0), BitCast(u8, row3_ne_0));
// auto row54_ne_0 = ConcatEven(u8, BitCast(u8, row4_ne_0), BitCast(u8, row5_ne_0));
// auto row76_ne_0 = ConcatEven(u8, BitCast(u8, row6_ne_0), BitCast(u8, row7_ne_0));

/* { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 } */
HWY_ALIGN constexpr uint64_t bm[] = {0x0102040810204080, 0x0102040810204080};
Expand Down Expand Up @@ -129,10 +133,14 @@ auto row6_lz = LeadingZeroCount(abs_row6);
auto row7_lz = LeadingZeroCount(abs_row7);

/* Narrow leading zero count to 8 bits. */
auto row01_lz = ConcatEven(u8, BitCast(u8, row1_lz), BitCast(u8, row0_lz));
auto row23_lz = ConcatEven(u8, BitCast(u8, row3_lz), BitCast(u8, row2_lz));
auto row45_lz = ConcatEven(u8, BitCast(u8, row5_lz), BitCast(u8, row4_lz));
auto row67_lz = ConcatEven(u8, BitCast(u8, row7_lz), BitCast(u8, row6_lz));
auto row01_lz = OrderedTruncate2To(u8, BitCast(u16, row0_lz), BitCast(u16, row1_lz));
auto row23_lz = OrderedTruncate2To(u8, BitCast(u16, row2_lz), BitCast(u16, row3_lz));
auto row45_lz = OrderedTruncate2To(u8, BitCast(u16, row4_lz), BitCast(u16, row5_lz));
auto row67_lz = OrderedTruncate2To(u8, BitCast(u16, row6_lz), BitCast(u16, row7_lz));
// auto row01_lz = ConcatEven(u8, BitCast(u8, row1_lz), BitCast(u8, row0_lz));
// auto row23_lz = ConcatEven(u8, BitCast(u8, row3_lz), BitCast(u8, row2_lz));
// auto row45_lz = ConcatEven(u8, BitCast(u8, row5_lz), BitCast(u8, row4_lz));
// auto row67_lz = ConcatEven(u8, BitCast(u8, row7_lz), BitCast(u8, row6_lz));
/* Compute nbits needed to specify magnitude of each coefficient. */
const auto sixteen = Set(u8, 16);
auto row01_nbits = Sub(sixteen, row01_lz);
Expand Down
8 changes: 4 additions & 4 deletions lib/block_coding_256.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,8 +67,8 @@ auto row01_ne_0 = VecFromMask(s16, Eq(row01, zero));
auto row23_ne_0 = VecFromMask(s16, Eq(row23, zero));
auto row45_ne_0 = VecFromMask(s16, Eq(row45, zero));
auto row67_ne_0 = VecFromMask(s16, Eq(row67, zero));
auto row3210_ne_0 = ConcatEven(u8, BitCast(u8, row23_ne_0), BitCast(u8, row01_ne_0));
auto row7654_ne_0 = ConcatEven(u8, BitCast(u8, row67_ne_0), BitCast(u8, row45_ne_0));
auto row3210_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row01_ne_0), BitCast(u16, row23_ne_0));
auto row7654_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row45_ne_0), BitCast(u16, row67_ne_0));

/* { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 } */
HWY_ALIGN constexpr uint64_t bm[] = {0x0102040810204080, 0x0102040810204080, 0x0102040810204080,
Expand Down Expand Up @@ -96,8 +96,8 @@ auto row23_lz = LeadingZeroCount(abs_row23);
auto row45_lz = LeadingZeroCount(abs_row45);
auto row67_lz = LeadingZeroCount(abs_row67);
/* Narrow leading zero count to 8 bits. */
auto row0123_lz = ConcatEven(u8, BitCast(u8, row23_lz), BitCast(u8, row01_lz));
auto row4567_lz = ConcatEven(u8, BitCast(u8, row67_lz), BitCast(u8, row45_lz));
auto row0123_lz = OrderedTruncate2To(u8, BitCast(u16, row01_lz), BitCast(u16, row23_lz));
auto row4567_lz = OrderedTruncate2To(u8, BitCast(u16, row45_lz), BitCast(u16, row67_lz));
/* Compute nbits needed to specify magnitude of each coefficient. */
auto row0123_nbits = Sub(Set(u8, 16), row0123_lz);
auto row4567_nbits = Sub(Set(u8, 16), row4567_lz);
Expand Down
4 changes: 2 additions & 2 deletions lib/block_coding_512.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,7 @@ auto row4567 = TwoTablesLookupLanes(s16, v0, v1, SetTableIndices(s16, &indices[1
auto zero = Zero(s16);
auto row0123_ne_0 = VecFromMask(s16, Eq(row0123, zero));
auto row4567_ne_0 = VecFromMask(s16, Eq(row4567, zero));
auto row76543210_ne_0 = ConcatEven(u8, BitCast(u8, row4567_ne_0), BitCast(u8, row0123_ne_0));
auto row76543210_ne_0 = OrderedTruncate2To(u8, BitCast(u16, row0123_ne_0), BitCast(u16, row4567_ne_0));

/* { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 } */
HWY_ALIGN constexpr uint64_t bm[] = {0x0102040810204080, 0x0102040810204080, 0x0102040810204080,
Expand All @@ -50,7 +50,7 @@ auto abs_row4567 = Abs(row4567);
auto row0123_lz = LeadingZeroCount(abs_row0123);
auto row4567_lz = LeadingZeroCount(abs_row4567);
/* Narrow leading zero count to 8 bits. */
auto row01234567_lz = ConcatEven(u8, BitCast(u8, row4567_lz), BitCast(u8, row0123_lz));
auto row01234567_lz = OrderedTruncate2To(u8, BitCast(u16, row0123_lz), BitCast(u16, row4567_lz));
/* Compute nbits needed to specify magnitude of each coefficient. */
auto row01234567_nbits = Sub(Set(u8, 16), row01234567_lz);
/* Store nbits. */
Expand Down