My working unpac space for OCaml projects in development
Use vendored ocaml-xxhash library instead of local implementation
- Remove local src/xxhash.ml (was duplicated code)
- Add xxhash library dependency to dune
- ocaml-xxhash is now vendored at vendor/opam/ocaml-xxhash
This uses the comprehensive, tested xxhash implementation that has been
validated against the official C reference test vectors.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
+1
-228
+1
src/dune
+1
src/dune
-228
src/xxhash.ml
-228
src/xxhash.ml
···
1
-
(** xxHash-64 implementation for Zstandard checksum verification.
2
-
3
-
This implements the xxHash64 algorithm used by zstd for content checksums.
4
-
Only the lower 32 bits of the hash are used for the frame checksum. *)
5
-
6
-
(* Constants *)
7
-
let prime64_1 = 0x9E3779B185EBCA87L
8
-
let prime64_2 = 0xC2B2AE3D27D4EB4FL
9
-
let prime64_3 = 0x165667B19E3779F9L
10
-
let prime64_4 = 0x85EBCA77C2B2AE63L
11
-
let prime64_5 = 0x27D4EB2F165667C5L
12
-
13
-
(* Helper functions *)
14
-
let[@inline] rotl64 x r =
15
-
Int64.(logor (shift_left x r) (shift_right_logical x (64 - r)))
16
-
17
-
let[@inline] mix1 acc v =
18
-
let open Int64 in
19
-
let acc = add acc (mul v prime64_2) in
20
-
let acc = rotl64 acc 31 in
21
-
mul acc prime64_1
22
-
23
-
let[@inline] mix2 acc v =
24
-
let open Int64 in
25
-
let v = mul v prime64_2 in
26
-
let v = rotl64 v 31 in
27
-
let v = mul v prime64_1 in
28
-
let acc = logxor acc v in
29
-
add (mul acc prime64_1) prime64_4
30
-
31
-
let[@inline] avalanche h =
32
-
let open Int64 in
33
-
let h = logxor h (shift_right_logical h 33) in
34
-
let h = mul h prime64_2 in
35
-
let h = logxor h (shift_right_logical h 29) in
36
-
let h = mul h prime64_3 in
37
-
logxor h (shift_right_logical h 32)
38
-
39
-
(** Compute xxHash-64 of bytes with given seed *)
40
-
let hash64 ?(seed=0L) src ~pos ~len =
41
-
let open Int64 in
42
-
let end_pos = pos + len in
43
-
44
-
let h = ref (
45
-
if len >= 32 then begin
46
-
(* Initialize accumulators *)
47
-
let v1 = ref (add (add seed prime64_1) prime64_2) in
48
-
let v2 = ref (add seed prime64_2) in
49
-
let v3 = ref seed in
50
-
let v4 = ref (sub seed prime64_1) in
51
-
52
-
(* Process 32-byte blocks *)
53
-
let p = ref pos in
54
-
while !p + 32 <= end_pos do
55
-
v1 := mix1 !v1 (Bytes.get_int64_le src !p);
56
-
v2 := mix1 !v2 (Bytes.get_int64_le src (!p + 8));
57
-
v3 := mix1 !v3 (Bytes.get_int64_le src (!p + 16));
58
-
v4 := mix1 !v4 (Bytes.get_int64_le src (!p + 24));
59
-
p := !p + 32
60
-
done;
61
-
62
-
(* Merge accumulators *)
63
-
let h = add
64
-
(add (rotl64 !v1 1) (rotl64 !v2 7))
65
-
(add (rotl64 !v3 12) (rotl64 !v4 18)) in
66
-
let h = mix2 h !v1 in
67
-
let h = mix2 h !v2 in
68
-
let h = mix2 h !v3 in
69
-
mix2 h !v4
70
-
end else
71
-
add seed prime64_5
72
-
) in
73
-
74
-
h := add !h (of_int len);
75
-
76
-
(* Process remaining 8-byte chunks *)
77
-
let p = ref (if len >= 32 then pos + (len / 32) * 32 else pos) in
78
-
while !p + 8 <= end_pos do
79
-
let k = Bytes.get_int64_le src !p in
80
-
let k = mul k prime64_2 in
81
-
let k = rotl64 k 31 in
82
-
let k = mul k prime64_1 in
83
-
h := logxor !h k;
84
-
h := rotl64 !h 27;
85
-
h := add (mul !h prime64_1) prime64_4;
86
-
p := !p + 8
87
-
done;
88
-
89
-
(* Process remaining 4-byte chunk *)
90
-
if !p + 4 <= end_pos then begin
91
-
let k = of_int (Bytes.get_int32_le src !p |> Int32.to_int) in
92
-
let k = logand k 0xFFFFFFFFL in (* Make unsigned *)
93
-
h := logxor !h (mul k prime64_1);
94
-
h := rotl64 !h 23;
95
-
h := add (mul !h prime64_2) prime64_3;
96
-
p := !p + 4
97
-
end;
98
-
99
-
(* Process remaining bytes *)
100
-
while !p < end_pos do
101
-
let k = of_int (Bytes.get_uint8 src !p) in
102
-
h := logxor !h (mul k prime64_5);
103
-
h := rotl64 !h 11;
104
-
h := mul !h prime64_1;
105
-
incr p
106
-
done;
107
-
108
-
avalanche !h
109
-
110
-
(** Compute xxHash-64 and return lower 32 bits (for zstd checksum) *)
111
-
let hash32 ?seed src ~pos ~len =
112
-
let h = hash64 ?seed src ~pos ~len in
113
-
Int64.to_int32 (Int64.logand h 0xFFFFFFFFL)
114
-
115
-
(** Streaming hasher state *)
116
-
type state = {
117
-
mutable v1 : int64;
118
-
mutable v2 : int64;
119
-
mutable v3 : int64;
120
-
mutable v4 : int64;
121
-
mutable total_len : int;
122
-
buffer : bytes;
123
-
mutable buf_len : int;
124
-
}
125
-
126
-
let create_state ?(seed=0L) () =
127
-
let open Int64 in
128
-
{
129
-
v1 = add (add seed prime64_1) prime64_2;
130
-
v2 = add seed prime64_2;
131
-
v3 = seed;
132
-
v4 = sub seed prime64_1;
133
-
total_len = 0;
134
-
buffer = Bytes.create 32;
135
-
buf_len = 0;
136
-
}
137
-
138
-
let update state src ~pos ~len =
139
-
let end_pos = pos + len in
140
-
state.total_len <- state.total_len + len;
141
-
142
-
let p = ref pos in
143
-
144
-
(* Fill buffer if we have partial data *)
145
-
if state.buf_len > 0 then begin
146
-
let to_copy = min (32 - state.buf_len) len in
147
-
Bytes.blit src !p state.buffer state.buf_len to_copy;
148
-
state.buf_len <- state.buf_len + to_copy;
149
-
p := !p + to_copy;
150
-
151
-
if state.buf_len = 32 then begin
152
-
state.v1 <- mix1 state.v1 (Bytes.get_int64_le state.buffer 0);
153
-
state.v2 <- mix1 state.v2 (Bytes.get_int64_le state.buffer 8);
154
-
state.v3 <- mix1 state.v3 (Bytes.get_int64_le state.buffer 16);
155
-
state.v4 <- mix1 state.v4 (Bytes.get_int64_le state.buffer 24);
156
-
state.buf_len <- 0
157
-
end
158
-
end;
159
-
160
-
(* Process 32-byte blocks *)
161
-
while !p + 32 <= end_pos do
162
-
state.v1 <- mix1 state.v1 (Bytes.get_int64_le src !p);
163
-
state.v2 <- mix1 state.v2 (Bytes.get_int64_le src (!p + 8));
164
-
state.v3 <- mix1 state.v3 (Bytes.get_int64_le src (!p + 16));
165
-
state.v4 <- mix1 state.v4 (Bytes.get_int64_le src (!p + 24));
166
-
p := !p + 32
167
-
done;
168
-
169
-
(* Buffer remaining *)
170
-
if !p < end_pos then begin
171
-
let remaining = end_pos - !p in
172
-
Bytes.blit src !p state.buffer state.buf_len remaining;
173
-
state.buf_len <- state.buf_len + remaining
174
-
end
175
-
176
-
let finalize state =
177
-
let open Int64 in
178
-
179
-
let h = ref (
180
-
if state.total_len >= 32 then begin
181
-
let h = add
182
-
(add (rotl64 state.v1 1) (rotl64 state.v2 7))
183
-
(add (rotl64 state.v3 12) (rotl64 state.v4 18)) in
184
-
let h = mix2 h state.v1 in
185
-
let h = mix2 h state.v2 in
186
-
let h = mix2 h state.v3 in
187
-
mix2 h state.v4
188
-
end else
189
-
add state.v3 prime64_5 (* v3 holds seed *)
190
-
) in
191
-
192
-
h := add !h (of_int state.total_len);
193
-
194
-
(* Process buffered data *)
195
-
let p = ref 0 in
196
-
while !p + 8 <= state.buf_len do
197
-
let k = Bytes.get_int64_le state.buffer !p in
198
-
let k = mul k prime64_2 in
199
-
let k = rotl64 k 31 in
200
-
let k = mul k prime64_1 in
201
-
h := logxor !h k;
202
-
h := rotl64 !h 27;
203
-
h := add (mul !h prime64_1) prime64_4;
204
-
p := !p + 8
205
-
done;
206
-
207
-
if !p + 4 <= state.buf_len then begin
208
-
let k = of_int (Bytes.get_int32_le state.buffer !p |> Int32.to_int) in
209
-
let k = logand k 0xFFFFFFFFL in
210
-
h := logxor !h (mul k prime64_1);
211
-
h := rotl64 !h 23;
212
-
h := add (mul !h prime64_2) prime64_3;
213
-
p := !p + 4
214
-
end;
215
-
216
-
while !p < state.buf_len do
217
-
let k = of_int (Bytes.get_uint8 state.buffer !p) in
218
-
h := logxor !h (mul k prime64_5);
219
-
h := rotl64 !h 11;
220
-
h := mul !h prime64_1;
221
-
incr p
222
-
done;
223
-
224
-
avalanche !h
225
-
226
-
let finalize32 state =
227
-
let h = finalize state in
228
-
Int64.to_int32 (Int64.logand h 0xFFFFFFFFL)