[Git][freetype/freetype][multiply-shift] 4 commits: [smooth] Detect SSE2

Alexei Podtelezhnikov pushed to branch multiply-shift at FreeType / FreeType

Commits:

a842a098

by Ben Wagner at 2021-08-27T09:20:26+02:00

[smooth] Detect SSE2 with MSVC for x86

MSVC does not set `__SSE2__`. Instead one must check whether `_M_IX86_FP` is
defined and greater than or equal to 2.

* src/smooth/ftgrays.c (FT_SSE2): New macro.
Use it where appropriate.

6804ce29

by Alexei Podtelezhnikov at 2021-08-27T09:46:40-04:00

[builds/windows] Revise SSE2 settings.

* builds/windows/vc2010/freetype.vcxproj [x64]: Remove explicit SSE2.
* builds/windows/visualc/freetype.vcproj [Win32]: Add explicit SSE2.

d1c20005

by Werner Lemberg at 2021-08-28T07:36:08+02:00

[truetype] Fix compilation if !TT_CONFIG_OPTION_BYTECODE_INTERPRETER.

* src/truetype/ttgxvar.c (tt_cvt_ready_iterator): Compile function
conditionally.
(tt_face_vary_cvt) [!TT_CONFIG_OPTION_BYTECODE_INTERPRETER]: Add code.

Fixes #1091.

930140e9

by Alexei Podtelezhnikov at 2021-08-29T21:55:19+00:00

[smooth] Reduce shift in multiply-shift optimization.

Smaller shifts that keep the division operands of FT_UDIVPREP within
32 bits result in slightly faster divisions, which is noticeable in
the overall performance.  The loss of precision is tolerable until the
divisors (the components dx and dy) approach 32 - PIXEL_BITS. With
PIXEL_BITS = 8, this corresponds to 65,000 pixels or the bitmap size
that we refuse to render anyway.

Using `ftbench -p -s60 -t5 -bc timesi.ttf`,

Before: 8.52 us/op
After:  8.32 us/op

Changes:

builds/windows/vc2010/freetype.vcxproj

@@ -242,7 +242,6 @@
        <CompileAs>Default</CompileAs>
        <DisableSpecificWarnings>4001</DisableSpecificWarnings>
        <MultiProcessorCompilation>true</MultiProcessorCompilation>
 -      <EnableEnhancedInstructionSet>StreamingSIMDExtensions2</EnableEnhancedInstructionSet>
        <IntrinsicFunctions>true</IntrinsicFunctions>
      </ClCompile>
      <ResourceCompile>
@@ -296,7 +295,6 @@
        <CompileAs>Default</CompileAs>
        <DisableSpecificWarnings>4001</DisableSpecificWarnings>
        <MultiProcessorCompilation>true</MultiProcessorCompilation>
 -      <EnableEnhancedInstructionSet>StreamingSIMDExtensions2</EnableEnhancedInstructionSet>
        <IntrinsicFunctions>true</IntrinsicFunctions>
      </ClCompile>
      <ResourceCompile>

builds/windows/visualc/freetype.vcproj

@@ -45,6 +45,7 @@
  				PreprocessorDefinitions="NDEBUG;WIN32;_LIB;_CRT_SECURE_NO_WARNINGS;FT2_BUILD_LIBRARY;DLL_EXPORT"
  				StringPooling="true"
  				RuntimeLibrary="2"
 +				EnableEnhancedInstructionSet="2"
  				EnableFunctionLevelLinking="true"
  				DisableLanguageExtensions="true"
  				WarningLevel="4"
@@ -121,6 +122,7 @@
  				PreprocessorDefinitions="NDEBUG;WIN32;_LIB;_CRT_SECURE_NO_WARNINGS;FT2_BUILD_LIBRARY"
  				StringPooling="true"
  				RuntimeLibrary="0"
 +				EnableEnhancedInstructionSet="2"
  				EnableFunctionLevelLinking="true"
  				DisableLanguageExtensions="true"
  				WarningLevel="4"

src/smooth/ftgrays.c

@@ -386,12 +386,11 @@ typedef ptrdiff_t  FT_PtrDist;
    /* divisors to provide sufficient accuracy of the multiply-shift.    */
    /* It should not exceed (64 - PIXEL_BITS) to prevent overflowing and */
    /* leave enough room for 64-bit unsigned multiplication however.     */
 -#define FT_UDIVPREP( c, b )                                                 \
 -  FT_Int64  b ## _r = c ? (FT_Int64)( ~(FT_UInt64)0 >> PIXEL_BITS ) / ( b ) \
 +#define FT_UDIVPREP( c, b )                            \
 +  FT_Int64  b ## _r = c ? (FT_Int64)0xFFFFFFFF / ( b ) \
                      : 0
 -#define FT_UDIV( a, b )                                         \
 -  (TCoord)( ( (FT_UInt64)( a ) * (FT_UInt64)( b ## _r ) ) >>    \
 -            ( sizeof( FT_UInt64 ) * FT_CHAR_BIT - PIXEL_BITS ) )
 +#define FT_UDIV( a, b )                                           \
 +  (TCoord)( ( (FT_UInt64)( a ) * (FT_UInt64)( b ## _r ) ) >> 32 )
    /* Scale area and apply fill rule to calculate the coverage byte. */
@@ -999,10 +998,17 @@ typedef ptrdiff_t  FT_PtrDist;
+    *
     * For other cases, using binary splits is actually slightly faster.
     */
 -#if defined( __SSE2__ )    || \
 -    defined( __x86_64__ )  || \
 +#if defined( __SSE2__ )                          || \
 +    defined( __x86_64__ )                        || \
 +    defined( _M_AMD64 )                          || \
 +    ( defined( _M_IX86_FP ) && _M_IX86_FP >= 2 )
 +#  define FT_SSE2 1
 +#else
 +#  define FT_SSE2 0
 +#endif
++
 +#if FT_SSE2                || \
      defined( __aarch64__ ) || \
 -    defined( _M_AMD64 )    || \
      defined( _M_ARM64 )
  #  define BEZIER_USE_DDA  1
  #else
@@ -1022,7 +1028,7 @@ typedef ptrdiff_t  FT_PtrDist;
  #if BEZIER_USE_DDA
 -#ifdef __SSE2__
 +#if FT_SSE2
  #  include <emmintrin.h>
  #endif
@@ -1135,7 +1141,7 @@ typedef ptrdiff_t  FT_PtrDist;
       *             = (B << (33 - N)) + (A << (32 - 2*N))
       */
 -#ifdef __SSE2__
 +#if FT_SSE2
      /* Experience shows that for small shift values, */
      /* SSE2 is actually slower.                      */
      if ( shift > 2 )
@@ -1192,7 +1198,7 @@ typedef ptrdiff_t  FT_PtrDist;
        return;
+     }
 -#endif  /* __SSE2__ */
 +#endif  /* FT_SSE2 */
      rx = LEFT_SHIFT( ax, 33 - 2 * shift );
      ry = LEFT_SHIFT( ay, 33 - 2 * shift );

src/truetype/ttgxvar.c

@@ -3164,6 +3164,8 @@
    /*************************************************************************/
 +#ifdef TT_CONFIG_GPTION_BYTECODE_INTERPRETER
++
    static FT_Error
    tt_cvt_ready_iterator( FT_ListNode  node,
                           void*        user )
@@ -3178,6 +3180,9 @@
      return FT_Err_Ok;
+   }
 +#endif /* TT_CONFIG_OPTION_BYTECODE_INTERPRETER */
++
++
    /**************************************************************************
+    *
@@ -3206,6 +3211,8 @@
    tt_face_vary_cvt( TT_Face    face,
                      FT_Stream  stream )
+   {
 +#ifdef TT_CONFIG_GPTION_BYTECODE_INTERPRETER
++
      FT_Error   error;
      FT_Memory  memory = stream->memory;
@@ -3526,6 +3533,16 @@
                       NULL );
      return error;
++
 +#else /* !TT_CONFIG_OPTION_BYTECODE_INTERPRETER */
++
 +    FT_UNUSED( face );
 +    FT_UNUSED( stream );
++
 +    return FT_Err_Ok;
++
 +#endif /* !TT_CONFIG_OPTION_BYTECODE_INTERPRETER */
++
+   }

From:	Alexei Podtelezhnikov (@apodtele)
Subject:	[Git][freetype/freetype][multiply-shift] 4 commits: [smooth] Detect SSE2 with MSVC for x86
Date:	Sun, 29 Aug 2021 21:55:23 +0000