This commit refactors free space analysis (and VM analysis) somewhat.
This is still far from perfect, but at least it's not using any
hardcoded values.
This commit also reorganizes different flushing bits here and there,
so that bit common to Exe and Obj are put in the common path, while
Exe-specific are behind an appropriate switch/if statement.
* Convert draft to generate all relevant segments and sections in right places
* Do not prealloc space in text blocks until we can NOP
* Write out LC_LOAD_DYLINKER command
* Add LC_LOAD_DYLIB command in order to specify to load libSystem
* Redo update decl exports (similar to Elf globals, globals need to be contiguous in
memory)
Decouples SQE queueing and SQE prepping methods to allow for non-sequential
SQE allocation schemes as suggested by @daurnimator.
Adds essential SQE prepping methods from liburing to reduce boilerplate.
Removes non-essential .link_with_next_sqe() and .use_registered_fd().
Implemented following Knuth's "Evaluation of Powers" chapter in TAOCP,
some extra complexity is needed to make sure there's no aliasing and
avoid allocating too many limbs.
A brief example to illustrate why the last point is important:
consider 10^123, since 10 is well within the limits of a single limb we
can safely say that the result will surely fit in:
⌈log2(10)⌉ bit * 123 = 492 bits = 7 limbs
A naive calculation using only the number of limbs yields:
1 limb * 123 = 123 limbs
The space savings are noticeable.
This reverts commit c8b4cc2ff9cf15a42f95b2302e02431a0004f5c8.
This includes many C++ standard library headers:
#include <algorithm>
#include <chrono>
#include <ctime>
#include <iomanip>
#include <ios>
#include <iostream>
#include <memory>
#include <sstream>
#include <string>
#include <tuple>
#include <type_traits>
#include <vector>
which adds more than a second of compile time for each file that
includes the header:
ir.cpp before: 8.041s
ir.cpp after: 6.847s
Ensures that the wakeup flag is read after the tail pointer has been
written. It's important to use memory load acquire semantics for the
flags read, otherwise the application and the kernel might not agree on
the consistency of the wakeup flag, leading to I/O starvation.
Refs: 6768ddcc56
Refs: https://github.com/axboe/liburing/issues/219
