freetype-commit
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[freetype2] gsoc-anurag-2022-final bdeb70bc1: [dense] Add SIMD support t


From: Werner Lemberg
Subject: [freetype2] gsoc-anurag-2022-final bdeb70bc1: [dense] Add SIMD support to rasterizer
Date: Sat, 19 Nov 2022 02:07:19 -0500 (EST)

branch: gsoc-anurag-2022-final
commit bdeb70bc18d5537e45185f3195d2f055f1bbee87
Author: Anurag Thakur <anurag105csec21@bpitindia.edu.in>
Commit: Anurag Thakur <anurag105csec21@bpitindia.edu.in>

    [dense] Add SIMD support to rasterizer
    
    * src/dense/ftdense: Use SSE4.1 for final accumulation step
    (FT_SSE4_1): Macro which checks if SSE4.1 is available
    
    * src/dense/rules.mk: Enable linking for SSE4.1
---
 src/dense/ftdense.c | 41 ++++++++++++++++++++++++++++++++++++++++-
 src/dense/rules.mk  |  6 +++---
 2 files changed, 43 insertions(+), 4 deletions(-)

diff --git a/src/dense/ftdense.c b/src/dense/ftdense.c
index 70c610885..a6b69f53e 100644
--- a/src/dense/ftdense.c
+++ b/src/dense/ftdense.c
@@ -12,6 +12,22 @@
 #include "ftdense.h"
 #include "ftdenseerrs.h"
 
+#if defined( __SSE4_1__ )                          || \
+    defined( __x86_64__ )                        || \
+    defined( _M_AMD64 )                          || \
+    ( defined( _M_IX86_FP ) && _M_IX86_FP >= 2 )
+#  define FT_SSE4_1 1
+#else
+#  define FT_SSE4_1 0
+#endif
+
+
+#if FT_SSE4_1
+
+    #include <tmmintrin.h>
+
+#endif
+
 #define PIXEL_BITS 8
 
 #define ONE_PIXEL  ( 1 << PIXEL_BITS )
@@ -349,9 +365,30 @@ dense_render_glyph( dense_worker* worker, const FT_Bitmap* 
target )
                                          &dense_decompose_funcs, worker );
   // Render into bitmap
   const float* source = worker->m_a;
-
   unsigned char* dest     = target->buffer;
   unsigned char* dest_end = target->buffer + worker->m_w * worker->m_h;
+
+#if FT_SSE4_1
+
+  __m128 offset = _mm_setzero_ps();
+  __m128i mask = _mm_set1_epi32(0x0c080400);
+  __m128 sign_mask = _mm_set1_ps(-0.f);
+  for (int i = 0; i < worker->m_h*worker->m_w; i += 4) {
+    __m128 x = _mm_load_ps(&source[i]);
+    x = _mm_add_ps(x, _mm_castsi128_ps(_mm_slli_si128(_mm_castps_si128(x), 
4)));
+    x = _mm_add_ps(x, _mm_shuffle_ps(_mm_setzero_ps(), x, 0x40));
+    x = _mm_add_ps(x, offset);
+    __m128 y = _mm_andnot_ps(sign_mask, x);  // fabs(x)
+    y = _mm_min_ps(y, _mm_set1_ps(1.0f));
+    y = _mm_mul_ps(y, _mm_set1_ps(255.0f));
+    __m128i z = _mm_cvtps_epi32(y);
+    z = _mm_shuffle_epi8(z, mask);
+    _mm_store_ss((float *)&dest[i], (__m128)z);
+    offset = _mm_shuffle_ps(x, x, _MM_SHUFFLE(3, 3, 3, 3));
+  }
+
+#else /* FT_SSE4_1 */
+
   float          value    = 0.0f;
   while ( dest < dest_end )
   {
@@ -368,6 +405,8 @@ dense_render_glyph( dense_worker* worker, const FT_Bitmap* 
target )
     dest++;
   }
 
+#endif /* FT_SSE4_1 */
+
   free(worker->m_a);
   return error;
 }
diff --git a/src/dense/rules.mk b/src/dense/rules.mk
index 005116873..306a068f6 100644
--- a/src/dense/rules.mk
+++ b/src/dense/rules.mk
@@ -22,9 +22,9 @@ DENSE_DIR := $(SRC_DIR)/dense
 #
 DENSE_COMPILE := $(CC) $(ANSIFLAGS)                               \
                         $I$(subst /,$(COMPILER_SEP),$(DENSE_DIR)) \
-                        $(INCLUDE_FLAGS)                           \
-                        $(FT_CFLAGS)
-
+                        $(INCLUDE_FLAGS)                          \
+                        $(FT_CFLAGS)                              \
+                        "-msse4.1"
 
 # DENSE driver sources (i.e., C files)
 #



reply via email to

[Prev in Thread] Current Thread [Next in Thread]