minhngoc25a
/
freetype2
mirror of git://git.savannah.gnu.org/freetype/freetype2.git
2
1
Fork 2
Code Issues 4 Releases Wiki Activity

Formatting and ChangeLog additions for previous commits.

This commit is contained in:
Werner Lemberg 2021-07-16 07:40:56 +02:00
parent c37c087382
commit 93771d619f
3 changed files with 181 additions and 98 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

80
ChangeLog
View File

@ -1,38 +1,88 @@
2021-07-15 Ben Wagner <bungeman@chromium.org>
* src/smooth/ftgrays.c: Guard inclusion of `emmintrin.h`.
Guard inclusion of `emmintrin.h` with `#ifdef __SSE2__`. The gcc
version of this header, `xmmintrin.h`, and `mmintrin.h` check that
the appropriate defines are set before defining anything (are
internally guarded). However, the clang versions of these includes
are not internally guarded. As a result of this, externally guard
the inclusion of these headers.
2021-07-15 David Turner <david@freetype.org> 2021-07-15 David Turner <david@freetype.org>
[smooth] Implement Bezier quadratic arc flattenning with DDA [smooth] Implement Bézier quadratic arc flattening with DDA.
Benchmarking shows that this provides a very slighty performance Benchmarking shows that this provides a very slighty performance
boost when rendering fonts with lots of quadratic bezier arcs, boost when rendering fonts with lots of quadratic Bézier arcs,
compared to the recursive arc splitting, but only when SSE2 is compared to the recursive arc splitting, but only when SSE2 is
available, or on 64-bit CPUs. available, or on 64-bit CPUs.
On a 2017 Core i5-7300U CPU on Linux/x86_64:
ftbench -p -s10 -t5 -cb DroidSansFallbackFull.ttf
Before: 4.033 us/op (best of 5 runs for all numbers)
After: 3.876 us/op
ftbench -p -s60 -t5 -cb DroidSansFallbackFull.ttf
Before: 13.467 us/op
After: 13.385 us/op
* src/smooth/ftgrays.c (gray_render_conic): New implementation * src/smooth/ftgrays.c (gray_render_conic): New implementation
based on DDA and optionally SSE2. based on DDA and optionally SSE2.
2021-07-15 David Turner <david@freetype.org> 2021-07-15 David Turner <david@freetype.org>
[smooth] Minor speedup to smooth rasterizer [smooth] Minor speedup to smooth rasterizer.
This speeds up the smooth rasterizer by avoiding a conditional This speeds up the smooth rasterizer by avoiding conditional
branches in the hot path. branches in the hot path.
* src/smooth/ftgrays.c: Define a null cell used to both as a - Define a fixed 'null cell', which will be pointed to whenever the
sentinel for all linked-lists, and to accumulate coverage and current cell is outside of the current target region. This avoids
area values for "out-of-bounds" cell positions without a a `ras.cell != NULL` check in the `FT_INTEGRATE` macro.
conditional check.
- Also use the null cell as a sentinel at the end of all `ycells`
linked-lists, by setting its x coordinate to `INT_MAX`. This
avoids a `if (!cell)` check in `gray_set_cell` as well.
- Slightly change the worker struct fields to perform a little less
operations during rendering.
Example results (on a 2013 Corei5-3337U CPU)
out/ftbench -p -s10 -t5 -bc DroidSansFallbackFull.ttf
Before: 5.472 us/op
After: 5.275 us/op
out/ftbench -p -s60 -t5 -bc DroidSansFallbackFull.ttf
Before: 17.988 us/op
After: 17.389 us/op
* src/smooth/ftgrays.c (grat_TWorker): Replace `num_cells` field with
`cell_free` and `cell_limit`.
(NULL_CELL_PTR, CELL_MAX_X_VALUE, CELL_IS_NULL): New macros.
(gray_dump_cells, gray_set_cell, gray_sweep, gray_sweep_direct,
gray_convert_glyph_inner, gray_convert_glyph): Updated.
2021-07-15 David Turner <david@freetype.org> 2021-07-15 David Turner <david@freetype.org>
Replaces download-test-fonts.sh with download-test-fonts.py which [tests] Rewrite download script in Python3.
does the same work, and also avoids downloading anything if the
files are already installed with the right content.
Now uses the first 8 byte of each file's sha256 hash for the digest. This commit replaces the bash script with a Python script that does
the same work, plus avoiding to download anything if the files are
already installed with the right content.
* tests/scripts/download-test-fonts.sh: Removed We now use the first 8 bytes of each file's sha256 hash for the
* tests/scripts/download-test-fonts.py: New script digest.
* tests/README.md: Updated
* tests/scripts/download-test-fonts.sh: Removed.
* tests/scripts/download-test-fonts.py: New script.
* tests/README.md: Updated.
2021-07-15 Alex Richardson <Alexander.Richardson@cl.cam.ac.uk> 2021-07-15 Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>

138
src/smooth/ftgrays.c
View File

@ -487,8 +487,8 @@ typedef ptrdiff_t FT_PtrDist;
PCell cell_free; /* call allocation next free slot */ PCell cell_free; /* call allocation next free slot */
PCell cell_limit; /* cell allocation limit */ PCell cell_limit; /* cell allocation limit */
PCell* ycells; /* array of cell linked-lists, one per */ PCell* ycells; /* array of cell linked-lists; one per */
/* vertical coordinate in the current band. */ /* vertical coordinate in the current band */
PCell cells; /* cell storage area */ PCell cells; /* cell storage area */
FT_PtrDist max_cells; /* cell storage capacity */ FT_PtrDist max_cells; /* cell storage capacity */
@ -513,19 +513,21 @@ typedef ptrdiff_t FT_PtrDist;
static gray_TWorker ras; static gray_TWorker ras;
#endif #endif
/* Return a pointer to the "null cell", used as a sentinel at the end */ /*
/* of all ycells[] linked lists. Its x coordinate should be maximal */ * Return a pointer to the 'null cell', used as a sentinel at the end of
/* to ensure no NULL checks are necessary when looking for an insertion */ * all `ycells` linked lists. Its x coordinate should be maximal to
/* point in gray_set_cell(). Other loops should check the cell pointer */ * ensure no NULL checks are necessary when looking for an insertion point
/* with CELL_IS_NULL() to detect the end of the list. */ * in `gray_set_cell`. Other loops should check the cell pointer with
#define NULL_CELL_PTR(ras) (ras).cells * CELL_IS_NULL() to detect the end of the list.
*/
#define NULL_CELL_PTR( ras ) (ras).cells
/* The |x| value of the null cell. Must be the largest possible */ /* The |x| value of the null cell. Must be the largest possible */
/* integer value stored in a TCell.x field. */ /* integer value stored in a `TCell.x` field. */
#define CELL_MAX_X_VALUE INT_MAX #define CELL_MAX_X_VALUE INT_MAX
/* Return true iff |cell| points to the null cell. */ /* Return true iff |cell| points to the null cell. */
#define CELL_IS_NULL(cell) ((cell)->x == CELL_MAX_X_VALUE) #define CELL_IS_NULL( cell ) ( (cell)->x == CELL_MAX_X_VALUE )
#define FT_INTEGRATE( ras, a, b ) \ #define FT_INTEGRATE( ras, a, b ) \
@ -556,7 +558,7 @@ typedef ptrdiff_t FT_PtrDist;
printf( "%3d:", y ); printf( "%3d:", y );
for ( ; !CELL_IS_NULL(cell); cell = cell->next ) for ( ; !CELL_IS_NULL( cell ); cell = cell->next )
printf( " (%3d, c:%4d, a:%6d)", printf( " (%3d, c:%4d, a:%6d)",
cell->x, cell->cover, cell->area ); cell->x, cell->cover, cell->area );
printf( "\n" ); printf( "\n" );
@ -584,9 +586,11 @@ typedef ptrdiff_t FT_PtrDist;
/* Note that if a cell is to the left of the clipping region, it is */ /* Note that if a cell is to the left of the clipping region, it is */
/* actually set to the (min_ex-1) horizontal position. */ /* actually set to the (min_ex-1) horizontal position. */
TCoord ey_index = ey - ras.min_ey; TCoord ey_index = ey - ras.min_ey;
if ( ey_index < 0 || ey_index >= ras.count_ey || ex >= ras.max_ex ) if ( ey_index < 0 || ey_index >= ras.count_ey || ex >= ras.max_ex )
ras.cell = NULL_CELL_PTR(ras); ras.cell = NULL_CELL_PTR( ras );
else else
{ {
PCell* pcell = ras.ycells + ey_index; PCell* pcell = ras.ycells + ey_index;
@ -610,7 +614,7 @@ typedef ptrdiff_t FT_PtrDist;
/* insert new cell */ /* insert new cell */
cell = ras.cell_free++; cell = ras.cell_free++;
if (cell >= ras.cell_limit) if ( cell >= ras.cell_limit )
ft_longjmp( ras.jump_buffer, 1 ); ft_longjmp( ras.jump_buffer, 1 );
cell->x = ex; cell->x = ex;
@ -978,6 +982,7 @@ typedef ptrdiff_t FT_PtrDist;
} }
gray_set_cell( RAS_VAR_ ex1, ey1 ); gray_set_cell( RAS_VAR_ ex1, ey1 );
} while ( ex1 != ex2 || ey1 != ey2 ); } while ( ex1 != ex2 || ey1 != ey2 );
} }
@ -987,30 +992,37 @@ typedef ptrdiff_t FT_PtrDist;
FT_INTEGRATE( ras, fy2 - fy1, fx1 + fx2 ); FT_INTEGRATE( ras, fy2 - fy1, fx1 + fx2 );
End: End:
ras.x = to_x; ras.x = to_x;
ras.y = to_y; ras.y = to_y;
} }
#endif #endif
/* Benchmarking shows that using DDA to flatten the quadratic bezier /*
* arcs is slightly faster in the following cases: * Benchmarking shows that using DDA to flatten the quadratic Bézier arcs
* * is slightly faster in the following cases:
* - When the host CPU is 64-bit. *
* - When SSE2 SIMD registers and instructions are available (even on x86). * - When the host CPU is 64-bit.
* * - When SSE2 SIMD registers and instructions are available (even on
* For other cases, using binary splits is actually slightly faster. * x86).
*/ *
#if defined(__SSE2__) || defined(__x86_64__) || defined(__aarch64__) || defined(_M_AMD64) || defined(_M_ARM64) * For other cases, using binary splits is actually slightly faster.
#define BEZIER_USE_DDA 1 */
#if defined( __SSE2__ ) || \
defined( __x86_64__ ) || \
defined( __aarch64__ ) || \
defined( _M_AMD64 ) || \
defined( _M_ARM64 )
# define BEZIER_USE_DDA 1
#else #else
#define BEZIER_USE_DDA 0 # define BEZIER_USE_DDA 0
#endif #endif
#if BEZIER_USE_DDA #if BEZIER_USE_DDA
#ifdef __SSE2__ #ifdef __SSE2__
#include <emmintrin.h> # include <emmintrin.h>
#endif #endif
static void static void
@ -1058,8 +1070,8 @@ typedef ptrdiff_t FT_PtrDist;
{ {
dx >>= 2; dx >>= 2;
shift += 1; shift += 1;
}
while (dx > ONE_PIXEL / 4); } while ( dx > ONE_PIXEL / 4 );
/* /*
* The (P0,P1,P2) arc equation, for t in [0,1] range: * The (P0,P1,P2) arc equation, for t in [0,1] range:
@ -1102,12 +1114,17 @@ typedef ptrdiff_t FT_PtrDist;
* Q << 32 = (2 * B << (32 - N)) + (A << (32 - N - N)) * Q << 32 = (2 * B << (32 - N)) + (A << (32 - N - N))
* = (B << (33 - N)) + (A << (32 - N - N)) * = (B << (33 - N)) + (A << (32 - N - N))
*/ */
#ifdef __SSE2__ #ifdef __SSE2__
/* Experience shows that for small shift values, SSE2 is actually slower. */ /* Experience shows that for small shift values, */
if (shift > 2) { /* SSE2 is actually slower. */
union { if ( shift > 2 )
struct { FT_Int64 ax, ay, bx, by; } i; {
struct { __m128i a, b; } vec; union
{
struct { FT_Int64 ax, ay, bx, by; } i;
struct { __m128i a, b; } vec;
} u; } u;
u.i.ax = p0.x + p2.x - 2 * p1.x; u.i.ax = p0.x + p2.x - 2 * p1.x;
@ -1138,10 +1155,11 @@ typedef ptrdiff_t FT_PtrDist;
p = _mm_add_epi64(p, q); p = _mm_add_epi64(p, q);
q = _mm_add_epi64(q, r); q = _mm_add_epi64(q, r);
_mm_store_si128(&v.vec, p); _mm_store_si128( &v.vec, p );
gray_render_line( RAS_VAR_ v.i.px_hi, v.i.py_hi); gray_render_line( RAS_VAR_ v.i.px_hi, v.i.py_hi );
} }
return; return;
} }
#endif /* !__SSE2__ */ #endif /* !__SSE2__ */
@ -1167,13 +1185,15 @@ typedef ptrdiff_t FT_PtrDist;
qx += rx; qx += rx;
qy += ry; qy += ry;
gray_render_line( RAS_VAR_ (FT_Pos)(px >> 32), (FT_Pos)(py >> 32)); gray_render_line( RAS_VAR_ (FT_Pos)( px >> 32 ),
(FT_Pos)( py >> 32 ) );
} }
} }
#else /* !BEZIER_USE_DDA */ #else /* !BEZIER_USE_DDA */
/* Note that multiple attempts to speed up the function below /*
* Note that multiple attempts to speed up the function below
* with SSE2 intrinsics, using various data layouts, have turned * with SSE2 intrinsics, using various data layouts, have turned
* out to be slower than the non-SIMD code below. * out to be slower than the non-SIMD code below.
*/ */
@ -1264,12 +1284,14 @@ typedef ptrdiff_t FT_PtrDist;
#endif /* !BEZIER_USE_DDA */ #endif /* !BEZIER_USE_DDA */
/* For cubic bezier, binary splits are still faster than DDA
/*
* For cubic Bézier, binary splits are still faster than DDA
* because the splits are adaptive to how quickly each sub-arc * because the splits are adaptive to how quickly each sub-arc
* approaches their chord trisection points. * approaches their chord trisection points.
* *
* It might be useful to experiment with SSE2 to speed up * It might be useful to experiment with SSE2 to speed up
* gray_split_cubic() though. * `gray_split_cubic`, though.
*/ */
static void static void
gray_split_cubic( FT_Vector* base ) gray_split_cubic( FT_Vector* base )
@ -1361,6 +1383,7 @@ typedef ptrdiff_t FT_PtrDist;
} }
} }
static int static int
gray_move_to( const FT_Vector* to, gray_move_to( const FT_Vector* to,
gray_PWorker worker ) gray_PWorker worker )
@ -1428,7 +1451,7 @@ typedef ptrdiff_t FT_PtrDist;
unsigned char* line = ras.target.origin - ras.target.pitch * y; unsigned char* line = ras.target.origin - ras.target.pitch * y;
for ( ; !CELL_IS_NULL(cell); cell = cell->next ) for ( ; !CELL_IS_NULL( cell ); cell = cell->next )
{ {
if ( cover != 0 && cell->x > x ) if ( cover != 0 && cell->x > x )
{ {
@ -1476,7 +1499,7 @@ typedef ptrdiff_t FT_PtrDist;
TArea area; TArea area;
for ( ; !CELL_IS_NULL(cell); cell = cell->next ) for ( ; !CELL_IS_NULL( cell ); cell = cell->next )
{ {
if ( cover != 0 && cell->x > x ) if ( cover != 0 && cell->x > x )
{ {
@ -1898,19 +1921,19 @@ typedef ptrdiff_t FT_PtrDist;
/* memory management */ /* memory management */
n = ( height * sizeof ( PCell ) + sizeof ( TCell ) - 1 ) / sizeof ( TCell ); n = ( height * sizeof ( PCell ) + sizeof ( TCell ) - 1 ) / sizeof ( TCell );
ras.cells = buffer + n; ras.cells = buffer + n;
ras.max_cells = (FT_PtrDist)( FT_MAX_GRAY_POOL - n ); ras.max_cells = (FT_PtrDist)( FT_MAX_GRAY_POOL - n );
ras.cell_limit = ras.cells + ras.max_cells; ras.cell_limit = ras.cells + ras.max_cells;
ras.ycells = (PCell*)buffer; ras.ycells = (PCell*)buffer;
/* Initialize the null cell is at the start of the 'cells' array. */ /* Initialize the null cell at the start of the `cells` array. */
/* Note that this requires ras.cell_free initialization to skip */ /* Note that this requires `ras.cell_free` initialization to skip */
/* over the first entry in the array. */ /* over the first entry in the array. */
PCell null_cell = NULL_CELL_PTR(ras); PCell null_cell = NULL_CELL_PTR( ras );
null_cell->x = CELL_MAX_X_VALUE; null_cell->x = CELL_MAX_X_VALUE;
null_cell->area = 0; null_cell->area = 0;
null_cell->cover = 0; null_cell->cover = 0;
null_cell->next = NULL;; null_cell->next = NULL;;
for ( y = yMin; y < yMax; ) for ( y = yMin; y < yMax; )
{ {
@ -1928,7 +1951,8 @@ typedef ptrdiff_t FT_PtrDist;
TCoord w; TCoord w;
int error; int error;
for (w = 0; w < width; ++w)
for ( w = 0; w < width; ++w )
ras.ycells[w] = null_cell; ras.ycells[w] = null_cell;
ras.cell_free = ras.cells + 1; /* NOTE: Skip over the null cell. */ ras.cell_free = ras.cells + 1; /* NOTE: Skip over the null cell. */

61
tests/scripts/download-test-fonts.py
View File

@ -1,8 +1,7 @@
#!/usr/bin/env python3 #!/usr/bin/env python3
"""Download test fonts used by the FreeType regression test programs. """Download test fonts used by the FreeType regression test programs. These
These will be copied to $FREETYPE/tests/data/ by default. will be copied to $FREETYPE/tests/data/ by default."""
"""
import argparse import argparse
import collections import collections
@ -15,8 +14,8 @@ import zipfile
from typing import Callable, List, Optional, Tuple from typing import Callable, List, Optional, Tuple
# The list of download items describing the font files to install. # The list of download items describing the font files to install. Each
# Each download item is a dictionary with one of the following schemas: # download item is a dictionary with one of the following schemas:
# #
# - File item: # - File item:
# #
@ -28,8 +27,8 @@ from typing import Callable, List, Optional, Tuple
# install_name # install_name
# Type: file name string # Type: file name string
# Required: No # Required: No
# Description: Installation name for the font file, only provided if it # Description: Installation name for the font file, only provided if
# must be different from the original URL's basename. # it must be different from the original URL's basename.
# #
# hex_digest # hex_digest
# Type: hexadecimal string # Type: hexadecimal string
@ -39,7 +38,7 @@ from typing import Callable, List, Optional, Tuple
# - Zip items: # - Zip items:
# #
# These items correspond to one or more font files that are embedded in a # These items correspond to one or more font files that are embedded in a
# remote zip archive. Each entry has the following fields: # remote zip archive. Each entry has the following fields:
# #
# zip_url # zip_url
# Type: URL string. # Type: URL string.
@ -52,23 +51,25 @@ from typing import Callable, List, Optional, Tuple
# Description: A list of entries describing a single font file to be # Description: A list of entries describing a single font file to be
# extracted from the archive # extracted from the archive
# #
# Apart from that, some schemas are used for dictionaries used inside download # Apart from that, some schemas are used for dictionaries used inside
# items: # download items:
# #
# - File entries: # - File entries:
# #
# These are dictionaries describing a single font file to extract from an archive. # These are dictionaries describing a single font file to extract from an
# archive.
# #
# filename # filename
# Type: file path string # Type: file path string
# Required: Yes # Required: Yes
# Description: Path of source file, relative to the archive's top-level directory. # Description: Path of source file, relative to the archive's
# top-level directory.
# #
# install_name # install_name
# Type: file name string # Type: file name string
# Required: No # Required: No
# Description: Installation name for the font file, only provided if it must be # Description: Installation name for the font file; only provided if
# different from the original filename value. # it must be different from the original filename value.
# #
# hex_digest # hex_digest
# Type: hexadecimal string # Type: hexadecimal string
@ -90,7 +91,8 @@ _DOWNLOAD_ITEMS = [
def digest_data(data: bytes): def digest_data(data: bytes):
"""Compute the digest of a given input byte string, which are the first 8 bytes of its sha256 hash.""" """Compute the digest of a given input byte string, which are the first
8 bytes of its sha256 hash."""
m = hashlib.sha256() m = hashlib.sha256()
m.update(data) m.update(data)
return m.digest()[:8] return m.digest()[:8]
@ -155,14 +157,16 @@ def extract_file_from_zip_archive(
Args: Args:
archive: Input ZipFile objec. archive: Input ZipFile objec.
archive_name: Archive name or URL, only used to generate a human-readable error archive_name: Archive name or URL, only used to generate a
message. human-readable error message.
filepath: Input filepath in archive. filepath: Input filepath in archive.
expected_digest: Optional digest for the file. expected_digest: Optional digest for the file.
Returns: Returns:
A new File instance corresponding to the extract file. A new File instance corresponding to the extract file.
Raises: Raises:
ValueError if expected_digest is not None and does not match the extracted file. ValueError if expected_digest is not None and does not match the
extracted file.
""" """
file = archive.open(filepath) file = archive.open(filepath)
if expected_digest is not None: if expected_digest is not None:
@ -181,7 +185,8 @@ def _get_and_install_file(
force_download: bool, force_download: bool,
get_content: Callable[[], bytes], get_content: Callable[[], bytes],
) -> bool: ) -> bool:
if not force_download and hex_digest is not None and os.path.exists(install_path): if not force_download and hex_digest is not None \
and os.path.exists(install_path):
with open(install_path, "rb") as f: with open(install_path, "rb") as f:
content: bytes = f.read() content: bytes = f.read()
if bytes.fromhex(hex_digest) == digest_data(content): if bytes.fromhex(hex_digest) == digest_data(content):
@ -200,14 +205,15 @@ def download_and_install_item(
Args: Args:
item: Download item as a dictionary, see above for schema. item: Download item as a dictionary, see above for schema.
install_dir: Installation directory. install_dir: Installation directory.
force_download: Set to True to force download and installation, even if force_download: Set to True to force download and installation, even
the font file is already installed with the right content. if the font file is already installed with the right content.
Returns: Returns:
A list of (install_name, status) tuples, where 'install_name' is the file's A list of (install_name, status) tuples, where 'install_name' is the
installation name under 'install_dir', and 'status' is a boolean that is True file's installation name under 'install_dir', and 'status' is a
to indicate that the file was downloaded and installed, or False to indicate that boolean that is True to indicate that the file was downloaded and
the file is already installed with the right content. installed, or False to indicate that the file is already installed
with the right content.
""" """
if "file_url" in item: if "file_url" in item:
file_url = item["file_url"] file_url = item["file_url"]
@ -284,10 +290,13 @@ def main():
for install_name, status in download_and_install_item( for install_name, status in download_and_install_item(
item, args.install_dir, args.force item, args.install_dir, args.force
): ):
print("%s %s" % (install_name, "INSTALLED" if status else "UP-TO-DATE")) print("%s %s" % (install_name,
"INSTALLED" if status else "UP-TO-DATE"))
return 0 return 0
if __name__ == "__main__": if __name__ == "__main__":
sys.exit(main()) sys.exit(main())
# EOF