QString: insert a number of 8-character SIMD loops

We don't have _mm_cvtsi64_si128() (the REX.W expansion of MOVD [0F 6E]),
but we do have _mm_loadl_epi64(), the SSE2 expansion of the MMX MOVQ at
opcode 0F 7E. Ditto for _mm_cvtsi128_si64() and _mm_storel_epi64(). And
those work even in 32-bit mode. By doing this, we can reduce the tail
unrolled loops by half, reducing code size.

I'm not adding these new SIMD sections to -Os builds.

Change-Id: Ib48364abee9f464c96c6fffd152e405310ef67be
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>

1 files changed, 39 insertions, 0 deletions

diff --git a/src/corelib/codecs/qutfcodec.cpp b/src/corelib/codecs/qutfcodec.cpp
index 08cc99f4dc..96be45ff4e 100644
--- a/src/corelib/codecs/qutfcodec.cpp
+++ b/src/corelib/codecs/qutfcodec.cpp
@@ -102,6 +102,26 @@ static inline bool simdEncodeAscii(uchar *&dst, const ushort *&nextAscii, const
             return false;
         }
     }
+
+    if (end - src >= 8) {
+        // do eight characters at a time
+        __m128i data = _mm_loadu_si128(reinterpret_cast<const __m128i *>(src));
+        __m128i packed = _mm_packus_epi16(data, data);
+        __m128i nonAscii = _mm_cmpgt_epi8(packed, _mm_setzero_si128());
+
+        // store even non-ASCII
+        _mm_storel_epi64(reinterpret_cast<__m128i *>(dst), packed);
+
+        uchar n = ~_mm_movemask_epi8(nonAscii);
+        if (n) {
+            nextAscii = src + qBitScanReverse(n) + 1;
+            n = qCountTrailingZeroBits(n);
+            dst += n;
+            src += n;
+            return false;
+        }
+    }
+
     return src == end;
 }
 
@@ -150,6 +170,25 @@ static inline bool simdDecodeAscii(ushort *&dst, const uchar *&nextAscii, const
         return false;
 
     }
+
+    if (end - src >= 8) {
+        __m128i data = _mm_loadl_epi64(reinterpret_cast<const __m128i *>(src));
+        uint n = _mm_movemask_epi8(data) & 0xff;
+        if (!n) {
+            // unpack and store
+            _mm_storeu_si128(reinterpret_cast<__m128i *>(dst), _mm_unpacklo_epi8(data, _mm_setzero_si128()));
+        } else {
+            while (!(n & 1)) {
+                *dst++ = *src++;
+                n >>= 1;
+            }
+
+            n = qBitScanReverse(n);
+            nextAscii = src + n + 1;
+            return false;
+        }
+    }
+
     return src == end;
 }