jitera-google-protobuf 3.21.12.pre.beta.1

data/ext/google/protobuf_c/protobuf.h

@@ -0,0 +1,120 @@
+// Protocol Buffers - Google's data interchange format
+// Copyright 2014 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef __GOOGLE_PROTOBUF_RUBY_PROTOBUF_H__
+#define __GOOGLE_PROTOBUF_RUBY_PROTOBUF_H__
+
+#include <ruby/encoding.h>
+#include <ruby/ruby.h>
+#include <ruby/vm.h>
+
+#include "defs.h"
+#include "ruby-upb.h"
+
+// These operate on a map field (i.e., a repeated field of submessages whose
+// submessage type is a map-entry msgdef).
+const upb_FieldDef* map_field_key(const upb_FieldDef* field);
+const upb_FieldDef* map_field_value(const upb_FieldDef* field);
+
+// -----------------------------------------------------------------------------
+// Arena
+// -----------------------------------------------------------------------------
+
+// A Ruby object that wraps an underlying upb_Arena.  Any objects that are
+// allocated from this arena should reference the Arena in rb_gc_mark(), to
+// ensure that the object's underlying memory outlives any Ruby object that can
+// reach it.
+
+VALUE Arena_new();
+upb_Arena* Arena_get(VALUE arena);
+
+// Fuses this arena to another, throwing a Ruby exception if this is not
+// possible.
+void Arena_fuse(VALUE arena, upb_Arena* other);
+
+// Pins this Ruby object to the lifetime of this arena, so that as long as the
+// arena is alive this object will not be collected.
+//
+// We use this to guarantee that the "frozen" bit on the object will be
+// remembered, even if the user drops their reference to this precise object.
+void Arena_Pin(VALUE arena, VALUE obj);
+
+// -----------------------------------------------------------------------------
+// ObjectCache
+// -----------------------------------------------------------------------------
+
+// Global object cache from upb array/map/message/symtab to wrapper object.
+//
+// This is a conceptually "weak" cache, in that it does not prevent "val" from
+// being collected (though in Ruby <2.7 is it effectively strong, due to
+// implementation limitations).
+
+// Adds an entry to the cache. The "arena" parameter must give the arena that
+// "key" was allocated from.  In Ruby <2.7.0, it will be used to remove the key
+// from the cache when the arena is destroyed.
+void ObjectCache_Add(const void* key, VALUE val);
+
+// Returns the cached object for this key, if any. Otherwise returns Qnil.
+VALUE ObjectCache_Get(const void* key);
+
+// -----------------------------------------------------------------------------
+// StringBuilder, for inspect
+// -----------------------------------------------------------------------------
+
+struct StringBuilder;
+typedef struct StringBuilder StringBuilder;
+
+StringBuilder* StringBuilder_New();
+void StringBuilder_Free(StringBuilder* b);
+void StringBuilder_Printf(StringBuilder* b, const char* fmt, ...);
+VALUE StringBuilder_ToRubyString(StringBuilder* b);
+
+void StringBuilder_PrintMsgval(StringBuilder* b, upb_MessageValue val,
+                               TypeInfo info);
+
+// -----------------------------------------------------------------------------
+// Utilities.
+// -----------------------------------------------------------------------------
+
+extern VALUE cTypeError;
+
+#ifdef NDEBUG
+#define PBRUBY_ASSERT(expr) \
+  do {                      \
+  } while (false && (expr))
+#else
+#define PBRUBY_ASSERT(expr) assert(expr)
+#endif
+
+#define PBRUBY_MAX(x, y) (((x) > (y)) ? (x) : (y))
+
+#define UPB_UNUSED(var) (void)var
+
+#endif  // __GOOGLE_PROTOBUF_RUBY_PROTOBUF_H__

data/ext/google/protobuf_c/range2-neon.c

@@ -0,0 +1,157 @@
+/*
+ * Process 2x16 bytes in each iteration.
+ * Comments removed for brevity. See range-neon.c for details.
+ */
+#ifdef __aarch64__
+
+#include <stdio.h>
+#include <stdint.h>
+#include <arm_neon.h>
+
+int utf8_naive(const unsigned char *data, int len);
+
+static const uint8_t _first_len_tbl[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 2, 3,
+};
+
+static const uint8_t _first_range_tbl[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8,
+};
+
+static const uint8_t _range_min_tbl[] = {
+    0x00, 0x80, 0x80, 0x80, 0xA0, 0x80, 0x90, 0x80,
+    0xC2, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+};
+static const uint8_t _range_max_tbl[] = {
+    0x7F, 0xBF, 0xBF, 0xBF, 0xBF, 0x9F, 0xBF, 0x8F,
+    0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+};
+
+static const uint8_t _range_adjust_tbl[] = {
+    2, 3, 0, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0,
+};
+
+/* Return 0 on success, -1 on error */
+int utf8_range2(const unsigned char *data, int len)
+{
+    if (len >= 32) {
+        uint8x16_t prev_input = vdupq_n_u8(0);
+        uint8x16_t prev_first_len = vdupq_n_u8(0);
+
+        const uint8x16_t first_len_tbl = vld1q_u8(_first_len_tbl);
+        const uint8x16_t first_range_tbl = vld1q_u8(_first_range_tbl);
+        const uint8x16_t range_min_tbl = vld1q_u8(_range_min_tbl);
+        const uint8x16_t range_max_tbl = vld1q_u8(_range_max_tbl);
+        const uint8x16x2_t range_adjust_tbl = vld2q_u8(_range_adjust_tbl);
+
+        const uint8x16_t const_1 = vdupq_n_u8(1);
+        const uint8x16_t const_2 = vdupq_n_u8(2);
+        const uint8x16_t const_e0 = vdupq_n_u8(0xE0);
+
+        uint8x16_t error1 = vdupq_n_u8(0);
+        uint8x16_t error2 = vdupq_n_u8(0);
+        uint8x16_t error3 = vdupq_n_u8(0);
+        uint8x16_t error4 = vdupq_n_u8(0);
+
+        while (len >= 32) {
+            /******************* two blocks interleaved **********************/
+
+#if defined(__GNUC__) && !defined(__clang__) && (__GNUC__ < 8)
+            /* gcc doesn't support vldq1_u8_x2 until version 8 */
+            const uint8x16_t input_a = vld1q_u8(data);
+            const uint8x16_t input_b = vld1q_u8(data + 16);
+#else
+            /* Forces a double load on Clang */
+            const uint8x16x2_t input_pair = vld1q_u8_x2(data);
+            const uint8x16_t input_a = input_pair.val[0];
+            const uint8x16_t input_b = input_pair.val[1];
+#endif
+
+            const uint8x16_t high_nibbles_a = vshrq_n_u8(input_a, 4);
+            const uint8x16_t high_nibbles_b = vshrq_n_u8(input_b, 4);
+
+            const uint8x16_t first_len_a =
+                vqtbl1q_u8(first_len_tbl, high_nibbles_a);
+            const uint8x16_t first_len_b =
+                vqtbl1q_u8(first_len_tbl, high_nibbles_b);
+
+            uint8x16_t range_a = vqtbl1q_u8(first_range_tbl, high_nibbles_a);
+            uint8x16_t range_b = vqtbl1q_u8(first_range_tbl, high_nibbles_b);
+
+            range_a =
+                vorrq_u8(range_a, vextq_u8(prev_first_len, first_len_a, 15));
+            range_b =
+                vorrq_u8(range_b, vextq_u8(first_len_a, first_len_b, 15));
+
+            uint8x16_t tmp1_a, tmp2_a, tmp1_b, tmp2_b;
+            tmp1_a = vextq_u8(prev_first_len, first_len_a, 14);
+            tmp1_a = vqsubq_u8(tmp1_a, const_1);
+            range_a = vorrq_u8(range_a, tmp1_a);
+
+            tmp1_b = vextq_u8(first_len_a, first_len_b, 14);
+            tmp1_b = vqsubq_u8(tmp1_b, const_1);
+            range_b = vorrq_u8(range_b, tmp1_b);
+
+            tmp2_a = vextq_u8(prev_first_len, first_len_a, 13);
+            tmp2_a = vqsubq_u8(tmp2_a, const_2);
+            range_a = vorrq_u8(range_a, tmp2_a);
+
+            tmp2_b = vextq_u8(first_len_a, first_len_b, 13);
+            tmp2_b = vqsubq_u8(tmp2_b, const_2);
+            range_b = vorrq_u8(range_b, tmp2_b);
+
+            uint8x16_t shift1_a = vextq_u8(prev_input, input_a, 15);
+            uint8x16_t pos_a = vsubq_u8(shift1_a, const_e0);
+            range_a = vaddq_u8(range_a, vqtbl2q_u8(range_adjust_tbl, pos_a));
+
+            uint8x16_t shift1_b = vextq_u8(input_a, input_b, 15);
+            uint8x16_t pos_b = vsubq_u8(shift1_b, const_e0);
+            range_b = vaddq_u8(range_b, vqtbl2q_u8(range_adjust_tbl, pos_b));
+
+            uint8x16_t minv_a = vqtbl1q_u8(range_min_tbl, range_a);
+            uint8x16_t maxv_a = vqtbl1q_u8(range_max_tbl, range_a);
+
+            uint8x16_t minv_b = vqtbl1q_u8(range_min_tbl, range_b);
+            uint8x16_t maxv_b = vqtbl1q_u8(range_max_tbl, range_b);
+
+            error1 = vorrq_u8(error1, vcltq_u8(input_a, minv_a));
+            error2 = vorrq_u8(error2, vcgtq_u8(input_a, maxv_a));
+
+            error3 = vorrq_u8(error3, vcltq_u8(input_b, minv_b));
+            error4 = vorrq_u8(error4, vcgtq_u8(input_b, maxv_b));
+
+            /************************ next iteration *************************/
+            prev_input = input_b;
+            prev_first_len = first_len_b;
+
+            data += 32;
+            len -= 32;
+        }
+        error1 = vorrq_u8(error1, error2);
+        error1 = vorrq_u8(error1, error3);
+        error1 = vorrq_u8(error1, error4);
+
+        if (vmaxvq_u8(error1))
+            return -1;
+
+        uint32_t token4;
+        vst1q_lane_u32(&token4, vreinterpretq_u32_u8(prev_input), 3);
+
+        const int8_t *token = (const int8_t *)&token4;
+        int lookahead = 0;
+        if (token[3] > (int8_t)0xBF)
+            lookahead = 1;
+        else if (token[2] > (int8_t)0xBF)
+            lookahead = 2;
+        else if (token[1] > (int8_t)0xBF)
+            lookahead = 3;
+
+        data -= lookahead;
+        len += lookahead;
+    }
+
+    return utf8_naive(data, len);
+}
+
+#endif

data/ext/google/protobuf_c/range2-sse.c

@@ -0,0 +1,170 @@
+/*
+ * Process 2x16 bytes in each iteration.
+ * Comments removed for brevity. See range-sse.c for details.
+ */
+#ifdef __SSE4_1__
+
+#include <stdio.h>
+#include <stdint.h>
+#include <x86intrin.h>
+
+int utf8_naive(const unsigned char *data, int len);
+
+static const int8_t _first_len_tbl[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 2, 3,
+};
+
+static const int8_t _first_range_tbl[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8,
+};
+
+static const int8_t _range_min_tbl[] = {
+    0x00, 0x80, 0x80, 0x80, 0xA0, 0x80, 0x90, 0x80,
+    0xC2, 0x7F, 0x7F, 0x7F, 0x7F, 0x7F, 0x7F, 0x7F,
+};
+static const int8_t _range_max_tbl[] = {
+    0x7F, 0xBF, 0xBF, 0xBF, 0xBF, 0x9F, 0xBF, 0x8F,
+    0xF4, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
+};
+
+static const int8_t _df_ee_tbl[] = {
+    0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0,
+};
+static const int8_t _ef_fe_tbl[] = {
+    0, 3, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+/* Return 0 on success, -1 on error */
+int utf8_range2(const unsigned char *data, int len)
+{
+    if (len >= 32) {
+        __m128i prev_input = _mm_set1_epi8(0);
+        __m128i prev_first_len = _mm_set1_epi8(0);
+
+        const __m128i first_len_tbl =
+            _mm_loadu_si128((const __m128i *)_first_len_tbl);
+        const __m128i first_range_tbl =
+            _mm_loadu_si128((const __m128i *)_first_range_tbl);
+        const __m128i range_min_tbl =
+            _mm_loadu_si128((const __m128i *)_range_min_tbl);
+        const __m128i range_max_tbl =
+            _mm_loadu_si128((const __m128i *)_range_max_tbl);
+        const __m128i df_ee_tbl =
+            _mm_loadu_si128((const __m128i *)_df_ee_tbl);
+        const __m128i ef_fe_tbl =
+            _mm_loadu_si128((const __m128i *)_ef_fe_tbl);
+
+        __m128i error = _mm_set1_epi8(0);
+
+        while (len >= 32) {
+            /***************************** block 1 ****************************/
+            const __m128i input_a = _mm_loadu_si128((const __m128i *)data);
+
+            __m128i high_nibbles =
+                _mm_and_si128(_mm_srli_epi16(input_a, 4), _mm_set1_epi8(0x0F));
+
+            __m128i first_len_a = _mm_shuffle_epi8(first_len_tbl, high_nibbles);
+
+            __m128i range_a = _mm_shuffle_epi8(first_range_tbl, high_nibbles);
+
+            range_a = _mm_or_si128(
+                    range_a, _mm_alignr_epi8(first_len_a, prev_first_len, 15));
+
+            __m128i tmp;
+            tmp = _mm_alignr_epi8(first_len_a, prev_first_len, 14);
+            tmp = _mm_subs_epu8(tmp, _mm_set1_epi8(1));
+            range_a = _mm_or_si128(range_a, tmp);
+
+            tmp = _mm_alignr_epi8(first_len_a, prev_first_len, 13);
+            tmp = _mm_subs_epu8(tmp, _mm_set1_epi8(2));
+            range_a = _mm_or_si128(range_a, tmp);
+
+            __m128i shift1, pos, range2;
+            shift1 = _mm_alignr_epi8(input_a, prev_input, 15);
+            pos = _mm_sub_epi8(shift1, _mm_set1_epi8(0xEF));
+            tmp = _mm_subs_epu8(pos, _mm_set1_epi8(0xF0));
+            range2 = _mm_shuffle_epi8(df_ee_tbl, tmp);
+            tmp = _mm_adds_epu8(pos, _mm_set1_epi8(0x70));
+            range2 = _mm_add_epi8(range2, _mm_shuffle_epi8(ef_fe_tbl, tmp));
+
+            range_a = _mm_add_epi8(range_a, range2);
+
+            __m128i minv = _mm_shuffle_epi8(range_min_tbl, range_a);
+            __m128i maxv = _mm_shuffle_epi8(range_max_tbl, range_a);
+
+            tmp = _mm_or_si128(
+                      _mm_cmplt_epi8(input_a, minv),
+                      _mm_cmpgt_epi8(input_a, maxv)
+                  );
+            error = _mm_or_si128(error, tmp);
+
+            /***************************** block 2 ****************************/
+            const __m128i input_b = _mm_loadu_si128((const __m128i *)(data+16));
+
+            high_nibbles =
+                _mm_and_si128(_mm_srli_epi16(input_b, 4), _mm_set1_epi8(0x0F));
+
+            __m128i first_len_b = _mm_shuffle_epi8(first_len_tbl, high_nibbles);
+
+            __m128i range_b = _mm_shuffle_epi8(first_range_tbl, high_nibbles);
+
+            range_b = _mm_or_si128(
+                    range_b, _mm_alignr_epi8(first_len_b, first_len_a, 15));
+
+
+            tmp = _mm_alignr_epi8(first_len_b, first_len_a, 14);
+            tmp = _mm_subs_epu8(tmp, _mm_set1_epi8(1));
+            range_b = _mm_or_si128(range_b, tmp);
+
+            tmp = _mm_alignr_epi8(first_len_b, first_len_a, 13);
+            tmp = _mm_subs_epu8(tmp, _mm_set1_epi8(2));
+            range_b = _mm_or_si128(range_b, tmp);
+
+            shift1 = _mm_alignr_epi8(input_b, input_a, 15);
+            pos = _mm_sub_epi8(shift1, _mm_set1_epi8(0xEF));
+            tmp = _mm_subs_epu8(pos, _mm_set1_epi8(0xF0));
+            range2 = _mm_shuffle_epi8(df_ee_tbl, tmp);
+            tmp = _mm_adds_epu8(pos, _mm_set1_epi8(0x70));
+            range2 = _mm_add_epi8(range2, _mm_shuffle_epi8(ef_fe_tbl, tmp));
+
+            range_b = _mm_add_epi8(range_b, range2);
+
+            minv = _mm_shuffle_epi8(range_min_tbl, range_b);
+            maxv = _mm_shuffle_epi8(range_max_tbl, range_b);
+
+
+            tmp = _mm_or_si128(
+                      _mm_cmplt_epi8(input_b, minv),
+                      _mm_cmpgt_epi8(input_b, maxv)
+                  );
+            error = _mm_or_si128(error, tmp);
+
+            /************************ next iteration **************************/
+            prev_input = input_b;
+            prev_first_len = first_len_b;
+
+            data += 32;
+            len -= 32;
+        }
+
+        if (!_mm_testz_si128(error, error))
+            return -1;
+
+        int32_t token4 = _mm_extract_epi32(prev_input, 3);
+        const int8_t *token = (const int8_t *)&token4;
+        int lookahead = 0;
+        if (token[3] > (int8_t)0xBF)
+            lookahead = 1;
+        else if (token[2] > (int8_t)0xBF)
+            lookahead = 2;
+        else if (token[1] > (int8_t)0xBF)
+            lookahead = 3;
+
+        data -= lookahead;
+        len += lookahead;
+    }
+
+    return utf8_naive(data, len);
+}
+
+#endif