From a42eb01c87675698ae5972421f8896f85f048f2b Mon Sep 17 00:00:00 2001
From: Masahiko Sawada <sawada.mshk@gmail.com>
Date: Mon, 24 Oct 2022 14:07:09 +0900
Subject: [PATCH v24 1/9] introduce vector8_min and vector8_highbit_mask

TODO: commit message
TODO: Remove uint64 case.

separate-commit TODO: move non-SIMD fallbacks to own header
to clean up the #ifdef maze.
---
 src/include/port/simd.h | 47 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 47 insertions(+)

diff --git a/src/include/port/simd.h b/src/include/port/simd.h
index c836360d4b..f0bba33c53 100644
--- a/src/include/port/simd.h
+++ b/src/include/port/simd.h
@@ -77,6 +77,7 @@ static inline bool vector8_has(const Vector8 v, const uint8 c);
 static inline bool vector8_has_zero(const Vector8 v);
 static inline bool vector8_has_le(const Vector8 v, const uint8 c);
 static inline bool vector8_is_highbit_set(const Vector8 v);
+static inline uint32 vector8_highbit_mask(const Vector8 v);
 #ifndef USE_NO_SIMD
 static inline bool vector32_is_highbit_set(const Vector32 v);
 #endif
@@ -96,6 +97,7 @@ static inline Vector8 vector8_ssub(const Vector8 v1, const Vector8 v2);
  */
 #ifndef USE_NO_SIMD
 static inline Vector8 vector8_eq(const Vector8 v1, const Vector8 v2);
+static inline Vector8 vector8_min(const Vector8 v1, const Vector8 v2);
 static inline Vector32 vector32_eq(const Vector32 v1, const Vector32 v2);
 #endif
 
@@ -277,6 +279,36 @@ vector8_is_highbit_set(const Vector8 v)
 #endif
 }
 
+/*
+ * Return the bitmask of the high-bit of each element.
+ */
+static inline uint32
+vector8_highbit_mask(const Vector8 v)
+{
+#ifdef USE_SSE2
+	return (uint32) _mm_movemask_epi8(v);
+#elif defined(USE_NEON)
+	static const uint8 mask[16] = {
+		1 << 0, 1 << 1, 1 << 2, 1 << 3,
+		1 << 4, 1 << 5, 1 << 6, 1 << 7,
+		1 << 0, 1 << 1, 1 << 2, 1 << 3,
+		1 << 4, 1 << 5, 1 << 6, 1 << 7,
+	  };
+
+	uint8x16_t masked = vandq_u8(vld1q_u8(mask), (uint8x16_t) vshrq_n_s8(v, 7));
+	uint8x16_t maskedhi = vextq_u8(masked, masked, 8);
+
+	return (uint32) vaddvq_u16((uint16x8_t) vzip1q_u8(masked, maskedhi));
+#else
+	uint32 mask = 0;
+
+	for (Size i = 0; i < sizeof(Vector8); i++)
+		mask |= (((const uint8 *) &v)[i] >> 7) << i;
+
+	return mask;
+#endif
+}
+
 /*
  * Exactly like vector8_is_highbit_set except for the input type, so it
  * looks at each byte separately.
@@ -372,4 +404,19 @@ vector32_eq(const Vector32 v1, const Vector32 v2)
 }
 #endif							/* ! USE_NO_SIMD */
 
+/*
+ * Compare the given vectors and return the vector of minimum elements.
+ */
+#ifndef USE_NO_SIMD
+static inline Vector8
+vector8_min(const Vector8 v1, const Vector8 v2)
+{
+#ifdef USE_SSE2
+	return _mm_min_epu8(v1, v2);
+#elif defined(USE_NEON)
+	return vminq_u8(v1, v2);
+#endif
+}
+#endif							/* ! USE_NO_SIMD */
+
 #endif							/* SIMD_H */
-- 
2.31.1