コード例 #1
0
        public void u8()
        {
            var storage = new u32ArrayMemory(new u32[42]);
            var writer  = new RawBitWriter <u32ArrayMemory>(storage);
            u16 v1      = 42;
            u16 v2      = 666;

            writer.u16(v1, 16);
            Assert.AreEqual(16, writer.BitsWritten);
            writer.u16(v2, 10);
            writer.Align();
            var reader = new RawBitReader <u32ArrayMemory>(storage);
            var rv1    = reader.u16(16);
            var rv2    = reader.u16(10);

            Assert.AreEqual(v1, rv1);
            Assert.AreEqual(v2, rv2);
        }
コード例 #2
0
        /// <summary>
        /// Copies data from array.
        /// </summary>
        public void CopyFrom(ReadOnlySpan <u8> data)
        {
            // may Throw.here as not hot path
            if (data.Length <= 0)
            {
                Throw.Argument("Should be positive", nameof(data.Length));
            }
            if (data.Length > System.UInt16.MaxValue * 4)
            {
                Throw.Argument($"Should be less or equal to {System.UInt16.MaxValue * 4}", nameof(data.Length));
            }


            var length = data.Length;

            Reset();
            var step      = Unsafe.SizeOf <u32>();
            var numChunks = (u16)((length / step) + 1);

            // TODO: move out this expansion from here
            if (chunks.Length < numChunks)
            {
                Chunks = new u32ArrayMemory(new u32[numChunks]); // call it once to stay expanded forever
            }

            // data must be 4 or 8 bytes i64 because 32 and 64 machines https://gafferongames.com/post/reading_and_writing_packets/
            // TODO: possible to optimize to avoid copy? some kind of unsafe cast?
            // TODO: try u64 for performance as most of devices will be 64 bit?
            // https://github.com/nxrighthere/NetStack/issues/1#issuecomment-475212246
            for (u16 i = 0; i < numChunks; i++)
            {
                i32 dataIdx = (i32)i * step;
                u32 chunk   = 0;
                // TODO: ref into data and do block copy of all 4bytes, copy only last 3 bytes by hand
                // may optimize by calculating variable her and doing zero init of remaining blocks
                // may reintepret unsafe as uint, and then if less than 3 then only read last as 1 2 3
                if (dataIdx < length)
                {
                    chunk = (u32)data[dataIdx];
                }

                if (dataIdx + 1 < length)
                {
                    chunk = chunk | (u32)data[dataIdx + 1] << 8;
                }

                if (dataIdx + 2 < length)
                {
                    chunk = chunk | (u32)data[dataIdx + 2] << 16;
                }

                if (dataIdx + 3 < length)
                {
                    chunk = chunk | (u32)data[dataIdx + 3] << 24;
                }

                chunks[i] = chunk;
            }

            // TODO: write sets 1 bit in the end. so we may read all remaining zeroes, minis 1 bit for flag, and make total less
            // TODO: should be do that at all? Consoder multiple buffers writing sequenctially or other length indicator (like prefix)
            // var leadingZeros = BitOperations.LeadingZeroCount(data[length - 1]);
            // totalNumberBits = 8 * length - leadingZeros - 1;
        }