Upload Kmake

deps/v8/third_party/zlib/contrib/optimizations/chunkcopy.h

@@ -0,0 +1,564 @@
+/* chunkcopy.h -- fast chunk copy and set operations
+ * Copyright (C) 2017 ARM, Inc.
+ * Copyright 2017 The Chromium Authors
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the Chromium source repository LICENSE file.
+ */
+
+#ifndef CHUNKCOPY_H
+#define CHUNKCOPY_H
+
+#include <stdint.h>
+#include "zutil.h"
+
+#define Z_STATIC_ASSERT(name, assert) typedef char name[(assert) ? 1 : -1]
+
+#if __STDC_VERSION__ >= 199901L
+#define Z_RESTRICT restrict
+#else
+#define Z_RESTRICT
+#endif
+
+#if defined(__clang__) || defined(__GNUC__) || defined(__llvm__)
+#define Z_BUILTIN_MEMCPY __builtin_memcpy
+#define Z_BUILTIN_MEMSET __builtin_memset
+#else
+#define Z_BUILTIN_MEMCPY zmemcpy
+#define Z_BUILTIN_MEMSET zmemset
+#endif
+
+#if defined(INFLATE_CHUNK_SIMD_NEON)
+#include <arm_neon.h>
+typedef uint8x16_t z_vec128i_t;
+#elif defined(INFLATE_CHUNK_SIMD_SSE2)
+#include <emmintrin.h>
+typedef __m128i z_vec128i_t;
+#elif defined(INFLATE_CHUNK_GENERIC)
+typedef struct { uint8_t x[16]; } z_vec128i_t;
+#else
+#error chunkcopy.h inflate chunk SIMD is not defined for your build target
+#endif
+
+/*
+ * Suppress MSan errors about copying uninitialized bytes (crbug.com/1376033).
+ */
+#define Z_DISABLE_MSAN
+#if defined(__has_feature)
+  #if __has_feature(memory_sanitizer)
+    #undef Z_DISABLE_MSAN
+    #define Z_DISABLE_MSAN __attribute__((no_sanitize("memory")))
+  #endif
+#endif
+
+/*
+ * chunk copy type: the z_vec128i_t type size should be exactly 128-bits
+ * and equal to CHUNKCOPY_CHUNK_SIZE.
+ */
+#define CHUNKCOPY_CHUNK_SIZE sizeof(z_vec128i_t)
+
+Z_STATIC_ASSERT(vector_128_bits_wide,
+                CHUNKCOPY_CHUNK_SIZE == sizeof(int8_t) * 16);
+
+/*
+ * Ask the compiler to perform a wide, unaligned load with a machine
+ * instruction appropriate for the z_vec128i_t type.
+ */
+static inline z_vec128i_t loadchunk(
+    const unsigned char FAR* s) Z_DISABLE_MSAN {
+  z_vec128i_t v;
+  Z_BUILTIN_MEMCPY(&v, s, sizeof(v));
+  return v;
+}
+
+/*
+ * Ask the compiler to perform a wide, unaligned store with a machine
+ * instruction appropriate for the z_vec128i_t type.
+ */
+static inline void storechunk(
+    unsigned char FAR* d,
+    const z_vec128i_t v) {
+  Z_BUILTIN_MEMCPY(d, &v, sizeof(v));
+}
+
+/*
+ * Perform a memcpy-like operation, assuming that length is non-zero and that
+ * it's OK to overwrite at least CHUNKCOPY_CHUNK_SIZE bytes of output even if
+ * the length is shorter than this.
+ *
+ * It also guarantees that it will properly unroll the data if the distance
+ * between `out` and `from` is at least CHUNKCOPY_CHUNK_SIZE, which we rely on
+ * in chunkcopy_relaxed().
+ *
+ * Aside from better memory bus utilisation, this means that short copies
+ * (CHUNKCOPY_CHUNK_SIZE bytes or fewer) will fall straight through the loop
+ * without iteration, which will hopefully make the branch prediction more
+ * reliable.
+ */
+static inline unsigned char FAR* chunkcopy_core(
+    unsigned char FAR* out,
+    const unsigned char FAR* from,
+    unsigned len) Z_DISABLE_MSAN {
+  const int bump = (--len % CHUNKCOPY_CHUNK_SIZE) + 1;
+  storechunk(out, loadchunk(from));
+  out += bump;
+  from += bump;
+  len /= CHUNKCOPY_CHUNK_SIZE;
+  while (len-- > 0) {
+    storechunk(out, loadchunk(from));
+    out += CHUNKCOPY_CHUNK_SIZE;
+    from += CHUNKCOPY_CHUNK_SIZE;
+  }
+  return out;
+}
+
+/*
+ * Like chunkcopy_core(), but avoid writing beyond of legal output.
+ *
+ * Accepts an additional pointer to the end of safe output.  A generic safe
+ * copy would use (out + len), but it's normally the case that the end of the
+ * output buffer is beyond the end of the current copy, and this can still be
+ * exploited.
+ */
+static inline unsigned char FAR* chunkcopy_core_safe(
+    unsigned char FAR* out,
+    const unsigned char FAR* from,
+    unsigned len,
+    unsigned char FAR* limit) {
+  Assert(out + len <= limit, "chunk copy exceeds safety limit");
+  if ((limit - out) < (ptrdiff_t)CHUNKCOPY_CHUNK_SIZE) {
+    const unsigned char FAR* Z_RESTRICT rfrom = from;
+    Assert((uintptr_t)out - (uintptr_t)from >= len,
+           "invalid restrict in chunkcopy_core_safe");
+    Assert((uintptr_t)from - (uintptr_t)out >= len,
+           "invalid restrict in chunkcopy_core_safe");
+    if (len & 8) {
+      Z_BUILTIN_MEMCPY(out, rfrom, 8);
+      out += 8;
+      rfrom += 8;
+    }
+    if (len & 4) {
+      Z_BUILTIN_MEMCPY(out, rfrom, 4);
+      out += 4;
+      rfrom += 4;
+    }
+    if (len & 2) {
+      Z_BUILTIN_MEMCPY(out, rfrom, 2);
+      out += 2;
+      rfrom += 2;
+    }
+    if (len & 1) {
+      *out++ = *rfrom++;
+    }
+    return out;
+  }
+  return chunkcopy_core(out, from, len);
+}
+
+/*
+ * Perform short copies until distance can be rewritten as being at least
+ * CHUNKCOPY_CHUNK_SIZE.
+ *
+ * Assumes it's OK to overwrite at least the first 2*CHUNKCOPY_CHUNK_SIZE
+ * bytes of output even if the copy is shorter than this.  This assumption
+ * holds within zlib inflate_fast(), which starts every iteration with at
+ * least 258 bytes of output space available (258 being the maximum length
+ * output from a single token; see inffast.c).
+ */
+static inline unsigned char FAR* chunkunroll_relaxed(
+    unsigned char FAR* out,
+    unsigned FAR* dist,
+    unsigned FAR* len) Z_DISABLE_MSAN {
+  const unsigned char FAR* from = out - *dist;
+  while (*dist < *len && *dist < CHUNKCOPY_CHUNK_SIZE) {
+    storechunk(out, loadchunk(from));
+    out += *dist;
+    *len -= *dist;
+    *dist += *dist;
+  }
+  return out;
+}
+
+#if defined(INFLATE_CHUNK_SIMD_NEON)
+/*
+ * v_load64_dup(): load *src as an unaligned 64-bit int and duplicate it in
+ * every 64-bit component of the 128-bit result (64-bit int splat).
+ */
+static inline z_vec128i_t v_load64_dup(const void* src) {
+  return vcombine_u8(vld1_u8(src), vld1_u8(src));
+}
+
+/*
+ * v_load32_dup(): load *src as an unaligned 32-bit int and duplicate it in
+ * every 32-bit component of the 128-bit result (32-bit int splat).
+ */
+static inline z_vec128i_t v_load32_dup(const void* src) {
+  int32_t i32;
+  Z_BUILTIN_MEMCPY(&i32, src, sizeof(i32));
+  return vreinterpretq_u8_s32(vdupq_n_s32(i32));
+}
+
+/*
+ * v_load16_dup(): load *src as an unaligned 16-bit int and duplicate it in
+ * every 16-bit component of the 128-bit result (16-bit int splat).
+ */
+static inline z_vec128i_t v_load16_dup(const void* src) {
+  int16_t i16;
+  Z_BUILTIN_MEMCPY(&i16, src, sizeof(i16));
+  return vreinterpretq_u8_s16(vdupq_n_s16(i16));
+}
+
+/*
+ * v_load8_dup(): load the 8-bit int *src and duplicate it in every 8-bit
+ * component of the 128-bit result (8-bit int splat).
+ */
+static inline z_vec128i_t v_load8_dup(const void* src) {
+  return vld1q_dup_u8((const uint8_t*)src);
+}
+
+/*
+ * v_store_128(): store the 128-bit vec in a memory destination (that might
+ * not be 16-byte aligned) void* out.
+ */
+static inline void v_store_128(void* out, const z_vec128i_t vec) {
+  vst1q_u8(out, vec);
+}
+
+#elif defined(INFLATE_CHUNK_SIMD_SSE2)
+/*
+ * v_load64_dup(): load *src as an unaligned 64-bit int and duplicate it in
+ * every 64-bit component of the 128-bit result (64-bit int splat).
+ */
+static inline z_vec128i_t v_load64_dup(const void* src) {
+  int64_t i64;
+  Z_BUILTIN_MEMCPY(&i64, src, sizeof(i64));
+  return _mm_set1_epi64x(i64);
+}
+
+/*
+ * v_load32_dup(): load *src as an unaligned 32-bit int and duplicate it in
+ * every 32-bit component of the 128-bit result (32-bit int splat).
+ */
+static inline z_vec128i_t v_load32_dup(const void* src) {
+  int32_t i32;
+  Z_BUILTIN_MEMCPY(&i32, src, sizeof(i32));
+  return _mm_set1_epi32(i32);
+}
+
+/*
+ * v_load16_dup(): load *src as an unaligned 16-bit int and duplicate it in
+ * every 16-bit component of the 128-bit result (16-bit int splat).
+ */
+static inline z_vec128i_t v_load16_dup(const void* src) {
+  int16_t i16;
+  Z_BUILTIN_MEMCPY(&i16, src, sizeof(i16));
+  return _mm_set1_epi16(i16);
+}
+
+/*
+ * v_load8_dup(): load the 8-bit int *src and duplicate it in every 8-bit
+ * component of the 128-bit result (8-bit int splat).
+ */
+static inline z_vec128i_t v_load8_dup(const void* src) {
+  return _mm_set1_epi8(*(const char*)src);
+}
+
+/*
+ * v_store_128(): store the 128-bit vec in a memory destination (that might
+ * not be 16-byte aligned) void* out.
+ */
+static inline void v_store_128(void* out, const z_vec128i_t vec) {
+  _mm_storeu_si128((__m128i*)out, vec);
+}
+#elif defined(INFLATE_CHUNK_GENERIC)
+/*
+ * Default implementations for chunk-copy functions rely on memcpy() being
+ * inlined by the compiler for best performance.  This is most likely to work
+ * as expected when the length argument is constant (as is the case here) and
+ * the target supports unaligned loads and stores.  Since that's not always a
+ * safe assumption, this may need extra compiler arguments such as
+ * `-mno-strict-align` or `-munaligned-access`, or the availability of
+ * extensions like SIMD.
+ */
+
+/*
+ * v_load64_dup(): load *src as an unaligned 64-bit int and duplicate it in
+ * every 64-bit component of the 128-bit result (64-bit int splat).
+ */
+static inline z_vec128i_t v_load64_dup(const void* src) {
+  int64_t in;
+  Z_BUILTIN_MEMCPY(&in, src, sizeof(in));
+  z_vec128i_t out;
+  for (int i = 0; i < sizeof(out); i += sizeof(in)) {
+    Z_BUILTIN_MEMCPY((uint8_t*)&out + i, &in, sizeof(in));
+  }
+  return out;
+}
+
+/*
+ * v_load32_dup(): load *src as an unaligned 32-bit int and duplicate it in
+ * every 32-bit component of the 128-bit result (32-bit int splat).
+ */
+static inline z_vec128i_t v_load32_dup(const void* src) {
+  int32_t in;
+  Z_BUILTIN_MEMCPY(&in, src, sizeof(in));
+  z_vec128i_t out;
+  for (int i = 0; i < sizeof(out); i += sizeof(in)) {
+    Z_BUILTIN_MEMCPY((uint8_t*)&out + i, &in, sizeof(in));
+  }
+  return out;
+}
+
+/*
+ * v_load16_dup(): load *src as an unaligned 16-bit int and duplicate it in
+ * every 16-bit component of the 128-bit result (16-bit int splat).
+ */
+static inline z_vec128i_t v_load16_dup(const void* src) {
+  int16_t in;
+  Z_BUILTIN_MEMCPY(&in, src, sizeof(in));
+  z_vec128i_t out;
+  for (int i = 0; i < sizeof(out); i += sizeof(in)) {
+    Z_BUILTIN_MEMCPY((uint8_t*)&out + i, &in, sizeof(in));
+  }
+  return out;
+}
+
+/*
+ * v_load8_dup(): load the 8-bit int *src and duplicate it in every 8-bit
+ * component of the 128-bit result (8-bit int splat).
+ */
+static inline z_vec128i_t v_load8_dup(const void* src) {
+  int8_t in = *(const uint8_t*)src;
+  z_vec128i_t out;
+  Z_BUILTIN_MEMSET(&out, in, sizeof(out));
+  return out;
+}
+
+/*
+ * v_store_128(): store the 128-bit vec in a memory destination (that might
+ * not be 16-byte aligned) void* out.
+ */
+static inline void v_store_128(void* out, const z_vec128i_t vec) {
+  Z_BUILTIN_MEMCPY(out, &vec, sizeof(vec));
+}
+#endif
+
+/*
+ * Perform an overlapping copy which behaves as a memset() operation, but
+ * supporting periods other than one, and assume that length is non-zero and
+ * that it's OK to overwrite at least CHUNKCOPY_CHUNK_SIZE*3 bytes of output
+ * even if the length is shorter than this.
+ */
+static inline unsigned char FAR* chunkset_core(
+    unsigned char FAR* out,
+    unsigned period,
+    unsigned len) {
+  z_vec128i_t v;
+  const int bump = ((len - 1) % sizeof(v)) + 1;
+
+  switch (period) {
+    case 1:
+      v = v_load8_dup(out - 1);
+      v_store_128(out, v);
+      out += bump;
+      len -= bump;
+      while (len > 0) {
+        v_store_128(out, v);
+        out += sizeof(v);
+        len -= sizeof(v);
+      }
+      return out;
+    case 2:
+      v = v_load16_dup(out - 2);
+      v_store_128(out, v);
+      out += bump;
+      len -= bump;
+      if (len > 0) {
+        v = v_load16_dup(out - 2);
+        do {
+          v_store_128(out, v);
+          out += sizeof(v);
+          len -= sizeof(v);
+        } while (len > 0);
+      }
+      return out;
+    case 4:
+      v = v_load32_dup(out - 4);
+      v_store_128(out, v);
+      out += bump;
+      len -= bump;
+      if (len > 0) {
+        v = v_load32_dup(out - 4);
+        do {
+          v_store_128(out, v);
+          out += sizeof(v);
+          len -= sizeof(v);
+        } while (len > 0);
+      }
+      return out;
+    case 8:
+      v = v_load64_dup(out - 8);
+      v_store_128(out, v);
+      out += bump;
+      len -= bump;
+      if (len > 0) {
+        v = v_load64_dup(out - 8);
+        do {
+          v_store_128(out, v);
+          out += sizeof(v);
+          len -= sizeof(v);
+        } while (len > 0);
+      }
+      return out;
+  }
+  out = chunkunroll_relaxed(out, &period, &len);
+  return chunkcopy_core(out, out - period, len);
+}
+
+/*
+ * Perform a memcpy-like operation, but assume that length is non-zero and that
+ * it's OK to overwrite at least CHUNKCOPY_CHUNK_SIZE bytes of output even if
+ * the length is shorter than this.
+ *
+ * Unlike chunkcopy_core() above, no guarantee is made regarding the behaviour
+ * of overlapping buffers, regardless of the distance between the pointers.
+ * This is reflected in the `restrict`-qualified pointers, allowing the
+ * compiler to re-order loads and stores.
+ */
+static inline unsigned char FAR* chunkcopy_relaxed(
+    unsigned char FAR* Z_RESTRICT out,
+    const unsigned char FAR* Z_RESTRICT from,
+    unsigned len) {
+  Assert((uintptr_t)out - (uintptr_t)from >= len,
+         "invalid restrict in chunkcopy_relaxed");
+  Assert((uintptr_t)from - (uintptr_t)out >= len,
+         "invalid restrict in chunkcopy_relaxed");
+  return chunkcopy_core(out, from, len);
+}
+
+/*
+ * Like chunkcopy_relaxed(), but avoid writing beyond of legal output.
+ *
+ * Unlike chunkcopy_core_safe() above, no guarantee is made regarding the
+ * behaviour of overlapping buffers, regardless of the distance between the
+ * pointers.  This is reflected in the `restrict`-qualified pointers, allowing
+ * the compiler to re-order loads and stores.
+ *
+ * Accepts an additional pointer to the end of safe output.  A generic safe
+ * copy would use (out + len), but it's normally the case that the end of the
+ * output buffer is beyond the end of the current copy, and this can still be
+ * exploited.
+ */
+static inline unsigned char FAR* chunkcopy_safe(
+    unsigned char FAR* out,
+    const unsigned char FAR* Z_RESTRICT from,
+    unsigned len,
+    unsigned char FAR* limit) {
+  Assert(out + len <= limit, "chunk copy exceeds safety limit");
+  Assert((uintptr_t)out - (uintptr_t)from >= len,
+         "invalid restrict in chunkcopy_safe");
+  Assert((uintptr_t)from - (uintptr_t)out >= len,
+         "invalid restrict in chunkcopy_safe");
+
+  return chunkcopy_core_safe(out, from, len, limit);
+}
+
+/*
+ * Perform chunky copy within the same buffer, where the source and destination
+ * may potentially overlap.
+ *
+ * Assumes that len > 0 on entry, and that it's safe to write at least
+ * CHUNKCOPY_CHUNK_SIZE*3 bytes to the output.
+ */
+static inline unsigned char FAR* chunkcopy_lapped_relaxed(
+    unsigned char FAR* out,
+    unsigned dist,
+    unsigned len) {
+  if (dist < len && dist < CHUNKCOPY_CHUNK_SIZE) {
+    return chunkset_core(out, dist, len);
+  }
+  return chunkcopy_core(out, out - dist, len);
+}
+
+/*
+ * Behave like chunkcopy_lapped_relaxed(), but avoid writing beyond of legal
+ * output.
+ *
+ * Accepts an additional pointer to the end of safe output.  A generic safe
+ * copy would use (out + len), but it's normally the case that the end of the
+ * output buffer is beyond the end of the current copy, and this can still be
+ * exploited.
+ */
+static inline unsigned char FAR* chunkcopy_lapped_safe(
+    unsigned char FAR* out,
+    unsigned dist,
+    unsigned len,
+    unsigned char FAR* limit) {
+  Assert(out + len <= limit, "chunk copy exceeds safety limit");
+  if ((limit - out) < (ptrdiff_t)(3 * CHUNKCOPY_CHUNK_SIZE)) {
+    /* TODO(cavalcantii): try harder to optimise this */
+    while (len-- > 0) {
+      *out = *(out - dist);
+      out++;
+    }
+    return out;
+  }
+  return chunkcopy_lapped_relaxed(out, dist, len);
+}
+
+/* TODO(cavalcanti): see crbug.com/1110083. */
+static inline unsigned char FAR* chunkcopy_safe_ugly(unsigned char FAR* out,
+                                                     unsigned dist,
+                                                     unsigned len,
+                                                     unsigned char FAR* limit) {
+#if defined(__GNUC__) && !defined(__clang__)
+  /* Speed is the same as using chunkcopy_safe
+     w/ GCC on ARM (tested gcc 6.3 and 7.5) and avoids
+     undefined behavior.
+  */
+  return chunkcopy_core_safe(out, out - dist, len, limit);
+#elif defined(__clang__) && defined(ARMV8_OS_ANDROID) && !defined(__aarch64__)
+  /* Seems to perform better on 32bit (i.e. Android). */
+  return chunkcopy_core_safe(out, out - dist, len, limit);
+#else
+  /* Seems to perform better on 64bit. */
+  return chunkcopy_lapped_safe(out, dist, len, limit);
+#endif
+}
+
+/*
+ * The chunk-copy code above deals with writing the decoded DEFLATE data to
+ * the output with SIMD methods to increase decode speed. Reading the input
+ * to the DEFLATE decoder with a wide, SIMD method can also increase decode
+ * speed. This option is supported on little endian machines, and reads the
+ * input data in 64-bit (8 byte) chunks.
+ */
+
+#ifdef INFLATE_CHUNK_READ_64LE
+/*
+ * Buffer the input in a uint64_t (8 bytes) in the wide input reading case.
+ */
+typedef uint64_t inflate_holder_t;
+
+/*
+ * Ask the compiler to perform a wide, unaligned load of a uint64_t using a
+ * machine instruction appropriate for the uint64_t type.
+ */
+static inline inflate_holder_t read64le(const unsigned char FAR *in) {
+    inflate_holder_t input;
+    Z_BUILTIN_MEMCPY(&input, in, sizeof(input));
+    return input;
+}
+#else
+/*
+ * Otherwise, buffer the input bits using zlib's default input buffer type.
+ */
+typedef unsigned long inflate_holder_t;
+
+#endif /* INFLATE_CHUNK_READ_64LE */
+
+#undef Z_STATIC_ASSERT
+#undef Z_RESTRICT
+#undef Z_BUILTIN_MEMCPY
+#undef Z_DISABLE_MSAN
+
+#endif /* CHUNKCOPY_H */

deps/v8/third_party/zlib/contrib/optimizations/inffast_chunk.c

@@ -0,0 +1,376 @@
+/* inffast_chunk.c -- fast decoding
+ * Copyright (C) 1995-2017 Mark Adler
+ * Copyright 2023 The Chromium Authors
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "contrib/optimizations/inffast_chunk.h"
+#include "contrib/optimizations/chunkcopy.h"
+
+#ifdef ASMINF
+#  pragma message("Assembler code may have bugs -- use at your own risk")
+#else
+
+/*
+   Decode literal, length, and distance codes and write out the resulting
+   literal and match bytes until either not enough input or output is
+   available, an end-of-block is encountered, or a data error is encountered.
+   When large enough input and output buffers are supplied to inflate(), for
+   example, a 16K input buffer and a 64K output buffer, more than 95% of the
+   inflate() execution time is spent in this routine.
+
+   Entry assumptions:
+
+        state->mode == LEN
+        strm->avail_in >= INFLATE_FAST_MIN_INPUT (6 or 8 bytes + 7 bytes)
+        strm->avail_out >= INFLATE_FAST_MIN_OUTPUT (258 bytes + 2 bytes)
+        start >= strm->avail_out
+        state->bits < 8
+        (state->hold >> state->bits) == 0
+        strm->next_out[0..strm->avail_out] does not overlap with
+              strm->next_in[0..strm->avail_in]
+        strm->state->window is allocated with an additional
+              CHUNKCOPY_CHUNK_SIZE-1 bytes of padding beyond strm->state->wsize
+
+   On return, state->mode is one of:
+
+        LEN -- ran out of enough output space or enough available input
+        TYPE -- reached end of block code, inflate() to interpret next block
+        BAD -- error in block data
+
+   Notes:
+
+    INFLATE_FAST_MIN_INPUT: 6 or 8 bytes + 7 bytes
+
+    - The maximum input bits used by a length/distance pair is 15 bits for the
+      length code, 5 bits for the length extra, 15 bits for the distance code,
+      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
+      Therefore if strm->avail_in >= 6, then there is enough input to avoid
+      checking for available input while decoding.
+
+    - The wide input data reading option reads 64 input bits at a time. Thus,
+      if strm->avail_in >= 8, then there is enough input to avoid checking for
+      available input while decoding. Reading consumes the input with:
+
+          hold |= read64le(in) << bits;
+          in += 6;
+          bits += 48;
+
+      reporting 6 bytes of new input because |bits| is 0..15 (2 bytes rounded
+      up, worst case) and 6 bytes is enough to decode as noted above. At exit,
+      hold &= (1U << bits) - 1 drops excess input to keep the invariant:
+
+          (state->hold >> state->bits) == 0
+
+    INFLATE_FAST_MIN_OUTPUT: 258 bytes + 2 bytes for literals = 260 bytes
+
+    - The maximum bytes that a single length/distance pair can output is 258
+      bytes, which is the maximum length that can be coded.  inflate_fast()
+      requires strm->avail_out >= 260 for each loop to avoid checking for
+      available output space while decoding.
+ */
+void ZLIB_INTERNAL inflate_fast_chunk_(z_streamp strm, unsigned start) {
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *in;      /* local strm->next_in */
+    z_const unsigned char FAR *last;    /* have enough input while in < last */
+    unsigned char FAR *out;     /* local strm->next_out */
+    unsigned char FAR *beg;     /* inflate()'s initial strm->next_out */
+    unsigned char FAR *end;     /* while out < end, enough space available */
+    unsigned char FAR *limit;   /* safety limit for chunky copies */
+#ifdef INFLATE_STRICT
+    unsigned dmax;              /* maximum distance from zlib header */
+#endif
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned wnext;             /* window write index */
+    unsigned char FAR *window;  /* allocated sliding window, if wsize != 0 */
+    inflate_holder_t hold;      /* local strm->hold */
+    unsigned bits;              /* local strm->bits */
+    code const FAR *lcode;      /* local strm->lencode */
+    code const FAR *dcode;      /* local strm->distcode */
+    unsigned lmask;             /* mask for first level of length codes */
+    unsigned dmask;             /* mask for first level of distance codes */
+    code const *here;           /* retrieved table entry */
+    unsigned op;                /* code bits, operation, extra bits, or */
+                                /*  window position, window bytes to copy */
+    unsigned len;               /* match length, unused bytes */
+    unsigned dist;              /* match distance */
+    unsigned char FAR *from;    /* where to copy match from */
+
+    /* copy state to local variables */
+    state = (struct inflate_state FAR *)strm->state;
+    in = strm->next_in;
+    last = in + (strm->avail_in - (INFLATE_FAST_MIN_INPUT - 1));
+    out = strm->next_out;
+    beg = out - (start - strm->avail_out);
+    end = out + (strm->avail_out - (INFLATE_FAST_MIN_OUTPUT - 1));
+    limit = out + strm->avail_out;
+#ifdef INFLATE_STRICT
+    dmax = state->dmax;
+#endif
+    wsize = state->wsize;
+    whave = state->whave;
+    wnext = (state->wnext == 0 && whave >= wsize) ? wsize : state->wnext;
+    window = state->window;
+    hold = state->hold;
+    bits = state->bits;
+    lcode = state->lencode;
+    dcode = state->distcode;
+    lmask = (1U << state->lenbits) - 1;
+    dmask = (1U << state->distbits) - 1;
+
+#ifdef INFLATE_CHUNK_READ_64LE
+#define REFILL() do { \
+        Assert(bits < 64, "### Too many bits in inflate_fast."); \
+        hold |= read64le(in) << bits; \
+        in += 7; \
+        in -= bits >> 3; \
+        bits |= 56; \
+    } while (0)
+#endif
+
+    /* decode literals and length/distances until end-of-block or not enough
+       input data or output space */
+    do {
+#ifdef INFLATE_CHUNK_READ_64LE
+        REFILL();
+#else
+        if (bits < 15) {
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+        }
+#endif
+        here = lcode + (hold & lmask);
+#ifdef INFLATE_CHUNK_READ_64LE
+        if (here->op == 0) {                    /* literal */
+            Tracevv((stderr, here->val >= 0x20 && here->val < 0x7f ?
+                    "inflate:         literal '%c'\n" :
+                    "inflate:         literal 0x%02x\n", here->val));
+            *out++ = (unsigned char)(here->val);
+            hold >>= here->bits;
+            bits -= here->bits;
+            here = lcode + (hold & lmask);
+            if (here->op == 0) {                /* literal */
+                Tracevv((stderr, here->val >= 0x20 && here->val < 0x7f ?
+                        "inflate:    2nd  literal '%c'\n" :
+                        "inflate:    2nd  literal 0x%02x\n", here->val));
+                *out++ = (unsigned char)(here->val);
+                hold >>= here->bits;
+                bits -= here->bits;
+                here = lcode + (hold & lmask);
+            }
+        }
+#endif
+      dolen:
+        op = (unsigned)(here->bits);
+        hold >>= op;
+        bits -= op;
+        op = (unsigned)(here->op);
+        if (op == 0) {                          /* literal */
+            Tracevv((stderr, here->val >= 0x20 && here->val < 0x7f ?
+                    "inflate:         literal '%c'\n" :
+                    "inflate:         literal 0x%02x\n", here->val));
+            *out++ = (unsigned char)(here->val);
+        }
+        else if (op & 16) {                     /* length base */
+            len = (unsigned)(here->val);
+            op &= 15;                           /* number of extra bits */
+            if (op) {
+#ifndef INFLATE_CHUNK_READ_64LE
+                if (bits < op) {
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                }
+#endif
+                len += (unsigned)hold & ((1U << op) - 1);
+                hold >>= op;
+                bits -= op;
+            }
+            Tracevv((stderr, "inflate:         length %u\n", len));
+#ifndef INFLATE_CHUNK_READ_64LE
+            if (bits < 15) {
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+            }
+#endif
+            here = dcode + (hold & dmask);
+          dodist:
+            op = (unsigned)(here->bits);
+            hold >>= op;
+            bits -= op;
+            op = (unsigned)(here->op);
+            if (op & 16) {                      /* distance base */
+                dist = (unsigned)(here->val);
+                op &= 15;                       /* number of extra bits */
+                /* we have two fast-path loads: 10+10 + 15+5 + 15 = 55,
+                   but we may need to refill here in the worst case */
+                if (bits < op) {
+#ifdef INFLATE_CHUNK_READ_64LE
+                    REFILL();
+#else
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                    if (bits < op) {
+                        hold += (unsigned long)(*in++) << bits;
+                        bits += 8;
+                    }
+#endif
+                }
+                dist += (unsigned)hold & ((1U << op) - 1);
+#ifdef INFLATE_STRICT
+                if (dist > dmax) {
+                    strm->msg = (char *)"invalid distance too far back";
+                    state->mode = BAD;
+                    break;
+                }
+#endif
+                hold >>= op;
+                bits -= op;
+                Tracevv((stderr, "inflate:         distance %u\n", dist));
+                op = (unsigned)(out - beg);     /* max distance in output */
+                if (dist > op) {                /* see if copy from window */
+                    op = dist - op;             /* distance back in window */
+                    if (op > whave) {
+                        if (state->sane) {
+                            strm->msg =
+                                (char *)"invalid distance too far back";
+                            state->mode = BAD;
+                            break;
+                        }
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                        if (len <= op - whave) {
+                            do {
+                                *out++ = 0;
+                            } while (--len);
+                            continue;
+                        }
+                        len -= op - whave;
+                        do {
+                            *out++ = 0;
+                        } while (--op > whave);
+                        if (op == 0) {
+                            from = out - dist;
+                            do {
+                                *out++ = *from++;
+                            } while (--len);
+                            continue;
+                        }
+#endif
+                    }
+                    from = window;
+                    if (wnext >= op) {          /* contiguous in window */
+                        from += wnext - op;
+                    }
+                    else {                      /* wrap around window */
+                        op -= wnext;
+                        from += wsize - op;
+                        if (op < len) {         /* some from end of window */
+                            len -= op;
+                            out = chunkcopy_safe(out, from, op, limit);
+                            from = window;      /* more from start of window */
+                            op = wnext;
+                            /* This (rare) case can create a situation where
+                               the first chunkcopy below must be checked.
+                             */
+                        }
+                    }
+                    if (op < len) {             /* still need some from output */
+                        out = chunkcopy_safe(out, from, op, limit);
+                        len -= op;
+                        /* When dist is small the amount of data that can be
+                           copied from the window is also small, and progress
+                           towards the dangerous end of the output buffer is
+                           also small.  This means that for trivial memsets and
+                           for chunkunroll_relaxed() a safety check is
+                           unnecessary.  However, these conditions may not be
+                           entered at all, and in that case it's possible that
+                           the main copy is near the end.
+                          */
+                        out = chunkunroll_relaxed(out, &dist, &len);
+                        out = chunkcopy_safe_ugly(out, dist, len, limit);
+                    } else {
+                        /* from points to window, so there is no risk of
+                           overlapping pointers requiring memset-like behaviour
+                         */
+                        out = chunkcopy_safe(out, from, len, limit);
+                    }
+                }
+                else {
+                    /* Whole reference is in range of current output.  No
+                       range checks are necessary because we start with room
+                       for at least 258 bytes of output, so unroll and roundoff
+                       operations can write beyond `out+len` so long as they
+                       stay within 258 bytes of `out`.
+                     */
+                    out = chunkcopy_lapped_relaxed(out, dist, len);
+                }
+            }
+            else if ((op & 64) == 0) {          /* 2nd level distance code */
+                here = dcode + here->val + (hold & ((1U << op) - 1));
+                goto dodist;
+            }
+            else {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+        }
+        else if ((op & 64) == 0) {              /* 2nd level length code */
+            here = lcode + here->val + (hold & ((1U << op) - 1));
+            goto dolen;
+        }
+        else if (op & 32) {                     /* end-of-block */
+            Tracevv((stderr, "inflate:         end of block\n"));
+            state->mode = TYPE;
+            break;
+        }
+        else {
+            strm->msg = (char *)"invalid literal/length code";
+            state->mode = BAD;
+            break;
+        }
+    } while (in < last && out < end);
+
+    /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
+    len = bits >> 3;
+    in -= len;
+    bits -= len << 3;
+    hold &= (1U << bits) - 1;
+
+    /* update state and return */
+    strm->next_in = in;
+    strm->next_out = out;
+    strm->avail_in = (unsigned)(in < last ?
+        (INFLATE_FAST_MIN_INPUT - 1) + (last - in) :
+        (INFLATE_FAST_MIN_INPUT - 1) - (in - last));
+    strm->avail_out = (unsigned)(out < end ?
+        (INFLATE_FAST_MIN_OUTPUT - 1) + (end - out) :
+        (INFLATE_FAST_MIN_OUTPUT - 1) - (out - end));
+    state->hold = hold;
+    state->bits = bits;
+
+    Assert((state->hold >> state->bits) == 0, "invalid input data state");
+}
+
+/*
+   inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe):
+   - Using bit fields for code structure
+   - Different op definition to avoid & for extra bits (do & for table bits)
+   - Three separate decoding do-loops for direct, window, and wnext == 0
+   - Special case for distance > 1 copies to do overlapped load and store copy
+   - Explicit branch predictions (based on measured branch probabilities)
+   - Deferring match copy and interspersed it with decoding subsequent codes
+   - Swapping literal/length else
+   - Swapping window/direct else
+   - Larger unrolled copy loops (three is about right)
+   - Moving len -= 3 statement into middle of loop
+ */
+
+#endif /* !ASMINF */

deps/v8/third_party/zlib/contrib/optimizations/inffast_chunk.h

@@ -0,0 +1,42 @@
+/* inffast_chunk.h -- header to use inffast_chunk.c
+ * Copyright (C) 1995-2003, 2010 Mark Adler
+ * Copyright (C) 2017 ARM, Inc.
+ * Copyright 2023 The Chromium Authors
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+#include "inffast.h"
+
+/* INFLATE_FAST_MIN_INPUT:
+   The minimum number of input bytes needed so that we can safely call
+   inflate_fast() with only one up-front bounds check. One
+   length/distance code pair (15 bits for the length code, 5 bits for length
+   extra, 15 bits for the distance code, 13 bits for distance extra) requires
+   reading up to 48 input bits. Additionally, in the same iteraction, we may
+   decode two literals from the root-table (requiring MIN_OUTPUT = 258 + 2).
+
+   Each root-table entry is up to 10 bits, for a total of 68 input bits each
+   iteraction.
+
+   The refill variant reads 8 bytes from the buffer at a time, and advances
+   the input pointer by up to 7 bytes, ensuring there are at least 56-bits
+   available in the bit-buffer. The technique was documented by Fabian Giesen
+   on his blog as variant 4 in the article 'Reading bits in far too many ways':
+   https://fgiesen.wordpress.com/2018/02/20/
+
+   In the worst case, we may refill twice in the same iteraction, requiring
+   MIN_INPUT = 8 + 7.
+*/
+#ifdef INFLATE_CHUNK_READ_64LE
+#undef INFLATE_FAST_MIN_INPUT
+#define INFLATE_FAST_MIN_INPUT 15
+#undef INFLATE_FAST_MIN_OUTPUT
+#define INFLATE_FAST_MIN_OUTPUT 260
+#endif
+
+void ZLIB_INTERNAL inflate_fast_chunk_(z_streamp strm, unsigned start);

deps/v8/third_party/zlib/contrib/optimizations/insert_string.h

@@ -0,0 +1,80 @@
+/* insert_string.h
+ *
+ * Copyright 2019 The Chromium Authors
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the Chromium source repository LICENSE file.
+ */
+
+#ifndef INSERT_STRING_H
+#define INSERT_STRING_H
+
+#ifndef INLINE
+#if defined(_MSC_VER) && !defined(__clang__)
+#define INLINE __inline
+#else
+#define INLINE inline
+#endif
+#endif
+
+#include <stdint.h>
+
+/**
+ * Some applications need to match zlib DEFLATE output exactly [3]. Use the
+ * canonical zlib Rabin-Karp rolling hash [1,2] in that case.
+ *
+ *  [1] For a description of the Rabin and Karp algorithm, see "Algorithms"
+ *      book by R. Sedgewick, Addison-Wesley, p252.
+ *  [2] https://www.euccas.me/zlib/#zlib_rabin_karp and also "rolling hash"
+ *      https://en.wikipedia.org/wiki/Rolling_hash
+ *  [3] crbug.com/1316541 AOSP incremental client APK package OTA upgrades.
+ */
+#ifdef CHROMIUM_ZLIB_NO_CASTAGNOLI
+#define USE_ZLIB_RABIN_KARP_ROLLING_HASH
+#endif
+
+/* ===========================================================================
+ * Update a hash value with the given input byte (Rabin-Karp rolling hash).
+ * IN  assertion: all calls to UPDATE_HASH are made with consecutive input
+ *    characters, so that a running hash key can be computed from the previous
+ *    key instead of complete recalculation each time.
+ */
+#define UPDATE_HASH(s, h, c) (h = (((h) << s->hash_shift) ^ (c)) & s->hash_mask)
+
+/* ===========================================================================
+ * Insert string str in the dictionary and set match_head to the previous head
+ * of the hash chain (the most recent string with same hash key). Return
+ * the previous length of the hash chain.
+ * If this file is compiled with -DFASTEST, the compression level is forced
+ * to 1, and no hash chains are maintained.
+ * IN  assertion: all calls to INSERT_STRING are made with consecutive input
+ *    characters and the first MIN_MATCH bytes of str are valid (except for
+ *    the last MIN_MATCH-1 bytes of the input file).
+ */
+local INLINE Pos insert_string(deflate_state* const s, const Pos str) {
+  Pos ret;
+/* insert_string dictionary insertion: ANZAC++ hasher
+ * significantly improves data compression speed.
+ *
+ * Note: the generated compressed output is a valid DEFLATE stream, but will
+ * differ from canonical zlib output.
+ */
+#if defined(USE_ZLIB_RABIN_KARP_ROLLING_HASH)
+  UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH - 1)]);
+#else
+  uint32_t value;
+  // Validated for little endian archs (i.e. x86, Arm). YMMV for big endian.
+  zmemcpy(&value, &s->window[str], sizeof(value));
+  s->ins_h = ((value * 66521 + 66521) >> 16) & s->hash_mask;
+#endif
+
+#ifdef FASTEST
+  ret = s->head[s->ins_h];
+#else
+  ret = s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+  s->head[s->ins_h] = str;
+
+  return ret;
+}
+
+#endif /* INSERT_STRING_H */