[Encoder] Add an SSE4.2 implementation of WelsGetNonZeroCount

Avoid touching some cache lines by using popcnt instead of table lookups. Also gives a speedup of ~1.4x on Haswell as compared with SSE2.

Sindre Aamås committed 10 years ago

Commit 4645bd21 parent d906dda4 files · +22 −0Browse files

Showing the changes to test/encoder/EncUT_EncoderMbAux.cpp (1 of 4 files). Show all

Unified · Split

 TEST (EncodeMbAuxTest, WelsGetNoneZeroCount_sse2) {
   TestGetNoneZeroCount (WelsGetNoneZeroCount_sse2);
 }
+TEST (EncodeMbAuxTest, WelsGetNoneZeroCount_sse42) {
+  if (WelsCPUFeatureDetect (0) & WELS_CPU_SSE42)
+    TestGetNoneZeroCount (WelsGetNoneZeroCount_sse42);
+}
 #endif
 #define WELS_ABS_LC(a) ((sign ^ (int32_t)(a)) - sign)
 #define NEW_QUANT(pDct, ff, mf) (((ff)+ WELS_ABS_LC(pDct))*(mf)) >>16

Keyboard shortcuts