LLVM integrated assembler: Improving expressions and relocations

In my previous post, LLVM integrated assembler: Improving MCExpr and MCValue delved into enhancements made to LLVM's internal MCExpr and MCValue representations. This post covers recent refinements to MC, focusing on expression resolving and relocation generation.

Symbol equating directives

In GNU Assembler, the following directives are called symbol equating. I have re-read its documentation https://sourceware.org/binutils/docs/as.html. Yes, it uses "equating" instead of "assignment" or "definition".

  • symbol = expression (multiple = on the same symbol is allowed)
  • .set symbol, expression (equivalent to =)
  • .equ symbol, expression (equivalent to =)
  • .equiv symbol, expression (redefinition leads to errors)
  • .eqv symbol, expression (lazy evaluation, not implemented in LLVM integrated assembler)

Read More

LLVM integrated assembler: Improving MCExpr and MCValue

In my previous post, Relocation Generation in Assemblers, I explored some key concepts behind LLVM’s integrated assemblers. This post dives into recent improvements I’ve made to refine that system.

The LLVM integrated assembler handles fixups and relocatable expressions as distinct entities. Relocatable expressions, in particular, are encoded using the MCValue class, which originally looked like this:

1
2
3
4
5
class MCValue {
  const MCSymbolRefExpr *SymA = nullptr, *SymB = nullptr;
  int64_t Cst = 0;
  uint32_t RefKind = 0;
};

Read More

Relocation generation in assemblers

This post explores how GNU Assembler and LLVM integrated assembler generate relocations, an important step to generate a relocatable file. Relocations identify parts of instructions or data that cannot be fully determined during assembly because they depend on the final memory layout, which is only established at link time or load time. These are essentially placeholders that will be filled in (typically with absolute addresses or PC-relative offsets) during the linking process.

Relocation generation: the basics

Symbol references are the primary candidates for relocations. For instance, in the x86-64 instruction movl sym(%rip), %eax (GNU syntax), the assembler calculates the displacement between the program counter (PC) and sym. This distance affects the instruction's encoding and typically triggers a R_X86_64_PC32 relocation, unless sym is a local symbol defined within the current section.

Both the GNU assembler and LLVM integrated assembler utilize multiple passes during assembly, with several key phases relevant to relocation generation:

Read More

Migrating comments to giscus

Followed this guide: https://www.patrickthurmond.com/blog/2023/12/11/commenting-is-available-now-thanks-to-giscus

Add the following to layout/_partial/article.ejs

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
<% if (!index && post.comments) { %>
<section class="giscus"></section>
<script src="https://giscus.app/client.js"
 data-repo="MaskRay/maskray.me"
 data-repo-id="FILL IT UP"
 data-category="Blog Post Comments"
 data-category-id="FILL IT UP"
 data-mapping="pathname"
 data-strict="0"
 data-reactions-enabled="1"
 data-emit-metadata="0"
 data-input-position="bottom"
 data-theme="preferred_color_scheme"
 data-lang="en"
 data-loading="lazy"
 crossorigin="anonymous"
 async>
</script>
<% } %>

Unfortunately comments from Disqus have not been migrated yet. If you've left comments in the past, thank you. Apologies they are now gone.

While you can create Github Discussions via GraphQL API, I haven't found a solution that works out of the box. https://www.davidangulo.xyz/posts/dirty-ruby-script-to-migrate-comments-from-disqus-to-giscus/ provides a Ruby solution, which is promising but no longer works.

1
2
3
4
5
6
7
8
9
Failed to define value method for :name, because EnterpriseOrderField already responds to that method. Use `value_method:` to override the method name or `value_method: false` to disable Enum value me
thod generation.
Failed to define value method for :name, because EnvironmentOrderField already responds to that method. Use `value_method:` to override the method name or `value_method: false` to disable Enum value m
ethod generation.
Failed to define value method for :name, because LabelOrderField already responds to that method. Use `value_method:` to override the method name or `value_method: false` to disable Enum value method
generation.
...
.local/share/gem/ruby/3.3.0/gems/graphql-client-0.25.0/lib/graphql/client.rb:338:in `query': wrong number of arguments (given 2, expected 1) (ArgumentError)
        from g.rb:42:in `create_discussion'

Natural loops

A dominator tree can be used to compute natural loops.

  • For every node H in a post-order traversal of the dominator tree (or the original CFG), find all predecessors that are dominated by H. This identifies all back edges.
  • Each back edge T->H identifies a natural loop with H as the header.
    • Perform a flood fill starting from T in the reversed dominator tree (from exiting block to header)
    • All visited nodes reachable from the root belong to the natural loop associated with the back edge. These nodes are guaranteed to be reachable from H due to the dominator property.
    • Visited nodes unreachable from the root should be ignored.
    • Loops associated with visited nodes are considered subloops.

Read More

Understanding and improving Clang -ftime-report

Clang provides a few options to generate timing report. Among them, -ftime-report and -ftime-trace can be used to analyze the performance of Clang's internal passes.

  • -fproc-stat-report records time and memory on spawned processes (ld, and gas if -fno-integrated-as).
  • -ftime-trace, introduced in 2019, generates Clang timing information in the Chrome Trace Event format (JSON). The format supports nested events, providing a rich view of the front end.
  • -ftime-report: The option name is borrowed from GCC.

This post focuses on the traditional -ftime-report, which uses a line-based textual format.

Understanding -ftime-report output

The output consists of information about multiple timer groups. The last group spans the largest interval and encompasses timing data from other groups.

Up to Clang 19, the last group is called "Clang front-end time report". You would see something like the following.

Read More

Skipping boring functions in debuggers

In debuggers, stepping into a function with arguments that involve function calls may step into the nested function calls, even if they are simple and uninteresting, such as those found in the C++ STL.

GDB

Consider the following example:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
#include <cstdio>
#include <memory>
#include <vector>
using namespace std;

void foo(int i, int j) {
  printf("%d %d\n", i, j);
}

int main() {
  auto i = make_unique<int>(3);
  vector v{1,2};
  foo(*i, v.back()); // step into
}

When GDB stops at the foo call, the step (s) command will step into std::vector::back and std::unique_ptr::operator*. While you can execute finish (fin) and then execute s again, it's time-consuming and distracting, especially when dealing with complex argument expressions.

Read More