Fix murmur64a's tail handling.

Also, split up the murmur64 proc into murmur64a and murmur64b as they're distinct hashes with their own output.

core/hash/hash.odin

@@ -172,108 +172,114 @@ murmur32 :: proc(data: []byte, seed := u32(0)) -> u32 {
 	return h1
 }
 
+// See https://github.com/aappleby/smhasher/blob/master/src/MurmurHash2.cpp#L96
 @(optimization_mode="speed")
-murmur64 :: proc(data: []byte, seed := u64(0x9747b28c)) -> u64 {
-	when size_of(int) == 8 {
-		m :: 0xc6a4a7935bd1e995
-		r :: 47
+murmur64a :: proc(data: []byte, seed := u64(0x9747b28c)) -> u64 {
+	m :: 0xc6a4a7935bd1e995
+	r :: 47
 
-		h: u64 = seed ~ (u64(len(data)) * m)
-		data64 := mem.slice_ptr(cast(^u64)raw_data(data), len(data)/size_of(u64))
+	h: u64 = seed ~ (u64(len(data)) * m)
+	data64 := mem.slice_data_cast([]u64, data)
 
-		for _, i in data64 {
-			k := data64[i]
+	for _, i in data64 {
+		k := data64[i]
 
-			k *= m
-			k ~= k>>r
-			k *= m
+		k *= m
+		k ~= k>>r
+		k *= m
 
-			h ~= k
-			h *= m
-		}
+		h ~= k
+		h *= m
+	}
 
-		switch len(data)&7 {
-		case 7: h ~= u64(data[6]) << 48; fallthrough
-		case 6: h ~= u64(data[5]) << 40; fallthrough
-		case 5: h ~= u64(data[4]) << 32; fallthrough
-		case 4: h ~= u64(data[3]) << 24; fallthrough
-		case 3: h ~= u64(data[2]) << 16; fallthrough
-		case 2: h ~= u64(data[1]) << 8;  fallthrough
-		case 1:
-			h ~= u64(data[0])
-			h *= m
-		}
+	offset := len(data64) * size_of(u64)
 
-		h ~= h>>r
+	switch len(data)&7 {
+	case 7: h ~= u64(data[offset + 6]) << 48; fallthrough
+	case 6: h ~= u64(data[offset + 5]) << 40; fallthrough
+	case 5: h ~= u64(data[offset + 4]) << 32; fallthrough
+	case 4: h ~= u64(data[offset + 3]) << 24; fallthrough
+	case 3: h ~= u64(data[offset + 2]) << 16; fallthrough
+	case 2: h ~= u64(data[offset + 1]) << 8;  fallthrough
+	case 1:
+		h ~= u64(data[offset + 0])
 		h *= m
-		h ~= h>>r
-
-		return h
-	} else {
-		m :: 0x5bd1e995
-		r :: 24
-
-		h1 := u32(seed) ~ u32(len(data))
-		h2 := u32(seed) >> 32
-		data32 := mem.slice_ptr(cast(^u32)raw_data(data), len(data)/size_of(u32))
-		len := len(data)
-		i := 0
-
-		for len >= 8 {
-			k1, k2: u32
-			k1 = data32[i]; i += 1
-			k1 *= m
-			k1 ~= k1>>r
-			k1 *= m
-			h1 *= m
-			h1 ~= k1
-			len -= 4
-
-			k2 = data32[i]; i += 1
-			k2 *= m
-			k2 ~= k2>>r
-			k2 *= m
-			h2 *= m
-			h2 ~= k2
-			len -= 4
-		}
+	}
 
-		if len >= 4 {
-			k1: u32
-			k1 = data32[i]; i += 1
-			k1 *= m
-			k1 ~= k1>>r
-			k1 *= m
-			h1 *= m
-			h1 ~= k1
-			len -= 4
-		}
+	h ~= h>>r
+	h *= m
+	h ~= h>>r
 
-		// TODO(bill): Fix this
-		#no_bounds_check data8 := mem.slice_to_bytes(data32[i:])[:3]
-		switch len {
-		case 3:
-			h2 ~= u32(data8[2]) << 16
-			fallthrough
-		case 2:
-			h2 ~= u32(data8[1]) << 8
-			fallthrough
-		case 1:
-			h2 ~= u32(data8[0])
-			h2 *= m
-		}
+	return h
+}
 
-		h1 ~= h2>>18
+// See https://github.com/aappleby/smhasher/blob/master/src/MurmurHash2.cpp#L140
+@(optimization_mode="speed")
+murmur64b :: proc(data: []byte, seed := u64(0x9747b28c)) -> u64 {
+	m :: 0x5bd1e995
+	r :: 24
+
+	h1 := u32(seed) ~ u32(len(data))
+	h2 := u32(seed) >> 32
+
+	data32 := mem.slice_ptr(cast(^u32)raw_data(data), len(data)/size_of(u32))
+	len := len(data)
+	i := 0
+
+	for len >= 8 {
+		k1, k2: u32
+		k1 = data32[i]; i += 1
+		k1 *= m
+		k1 ~= k1>>r
+		k1 *= m
 		h1 *= m
-		h2 ~= h1>>22
+		h1 ~= k1
+		len -= 4
+
+		k2 = data32[i]; i += 1
+		k2 *= m
+		k2 ~= k2>>r
+		k2 *= m
 		h2 *= m
-		h1 ~= h2>>17
+		h2 ~= k2
+		len -= 4
+	}
+
+	if len >= 4 {
+		k1: u32
+		k1 = data32[i]; i += 1
+		k1 *= m
+		k1 ~= k1>>r
+		k1 *= m
 		h1 *= m
-		h2 ~= h1>>19
-		h2 *= m
+		h1 ~= k1
+		len -= 4
+	}
 
-		return u64(h1)<<32 | u64(h2)
+	// TODO(bill): Fix this
+	#no_bounds_check data8 := mem.slice_to_bytes(data32[i:])[:3]
+	switch len {
+	case 3:
+		h2 ~= u32(data8[2]) << 16
+		fallthrough
+	case 2:
+		h2 ~= u32(data8[1]) << 8
+		fallthrough
+	case 1:
+		h2 ~= u32(data8[0])
+		h2 *= m
 	}
+
+	h1 ~= h2>>18
+	h1 *= m
+	h2 ~= h1>>22
+	h2 *= m
+	h1 ~= h2>>17
+	h1 *= m
+	h2 ~= h1>>19
+	h2 *= m
+
+	return u64(h1)<<32 | u64(h2)
 }
 
 @(optimization_mode="speed")