JH

I am trying to add density to [InnoDB page compression algorithm list](https://mariadb.com/kb/en/library/compression/). Concatenation of data blocks is impossible. In fact, a lot of scenarios don't allow data concatenation, e.g., real...

I tried the approach of allocating a huge size of memory, using it as an array of memory block with size of each one is 768KB, managing it with a...

@gpnuma As mentioned in my previous post, I am adding density as a candidate algorithm of InnoDB page compression, the [page size](https://dev.mysql.com/doc/refman/5.7/en/innodb-parameters.html#sysvar_innodb_page_size) can be specified when you create the table.

I noticed the gcc/clang speed difference and added Intel C++ Compiler 2018 to the list, then I checked the executables with IDA. The three compilers took completely different approaches for...

@gpnuma Why `uint64_t value = *(uint64_t*)pointer;` is unsafe in comparison to `memcpy(&value, pointer, 8);` ?

If there are millions of 8 bytes to copy, the alignment matters, but I don't think there will be performance impact if there are only 8 bytes. In fact, both...

[density_01263.bin.gz](https://github.com/centaurean/density/files/1871652/density_01263.bin.gz) Please uncompress the attached gz and check the file density_01263.bin with the following test: ``` #include #include #include #include "density/density_api.h" int cc() { uint8_t *dbuf = malloc(100000); uint8_t *cbuf...

@gpnuma Yes, it doesn't crash. The top CPU consumer is still `calloc` which eats up most of the time. I have put the modified version to a proof-of-consistency cluster which...

@gpnuma My dataset is mixed with text and binary data. I will update to your latest source to run the test again.

@gpnuma I reran the Instruments utility. In `density_compress`: ``` +0x1c movq %r8, 8(%rsp) +0x21 movq %rsi, 16(%rsp) +0x26 movq %rdx, 24(%rsp) +0x2b callq "DYLD-STUB$$malloc" +0x30 movl %ebx, %edi +0x32 movl...