4 ms·

Hi Matt. Thank you for such an excellent project. I'm pretty sure you already know of sharplab.io. A lot of us C# devs prefer it over Compiler Explorer for a b

by Genbox 4y ago

Hi Matt. Thank you for such an excellent project. I'm pretty sure you already know of sharplab.io. A lot of us C# devs prefer it over Compiler Explorer for a bunch of reasons, but the main one is really the speed. My guess is that Sharplab gets it speed because it hooks directly into the JIT compiler itself. There is no overhead from touching disk or the linker.

andyayers 4y ago
IIRC with .NET 7 we will be able to improve this as we can now dump the jit-generated assembly from official builds. .NET 7 will be officially released in a month or so. Currently C#/F#/VB support in Compiler Explorer relies on prejitting which has various issues.
dataflow 4y ago
Confused, how is this related to C++ compilers? Edit: Oh wow I didn't realize the website does other languages, thanks!
- vips7L 4y ago
  Godbolt does more than C++. I often share Java byte code snippets from there
  - pjmlp 4y ago
    I wish it could also do the machine code version like on .NET languages. Maybe there is a way to plug hsdis into the workflow. EDIT: Actually there is already a ticket discussing this, https://github.com/compiler-explorer/compiler-explorer/issues/1364 https://github.com/compiler-explorer/compiler-explorer/issue...
    - vips7L 4y ago
      Have you seen JitWatch? https://github.com/AdoptOpenJDK/jitwatch https://github.com/AdoptOpenJDK/jitwatch
      - pjmlp 4y ago
        Naturally, :)
- mattgodbolt 4y ago
  Why does it have to be related to C++ compilers? Compiler Explorer supports a number of languages, including C#: $ curl -sL https://godbolt.org/api/languages Id | Name csharp | C# fsharp | F# vb | Visual Basic go | Go c | C c++ | C++ fortran | Fortran assembly | Assembly circle | C++ (Circle) circt | CIRCT hlsl | HLSL cppx | Cppx crystal | Crystal dart | Dart erlang | Erlang carbon | Carbon hook | Hook cppx_blue | Cppx-Blue cppx_gold | Cppx-Gold mlir | MLIR cuda | CUDA C++ analysis | Analysis python | Python racket | Racket ruby | Ruby typescript | TypeScript Native d | D ada | Ada cpp_for_opencl | C++ for OpenCL openclc | OpenCL C llvm | LLVM IR cpp2_cppfront | Cpp2-cppfront rust | Rust ispc | ispc java | Java kotlin | Kotlin nim | Nim pony | Pony scala | Scala solidity | Solidity clean | Clean pascal | Pascal haskell | Haskell ocaml | OCaml swift | Swift zig | Zig $
  - jlarocco 4y ago
    Since it's on topic and pretty cool, Common Lisp has built-in support for this with the disassemble function, (disassemble <function-name>): CL-USER> (defun do-something (a b) (format t "a: ~a~%b: ~a~%~%" a b)) DO-SOMETHING CL-USER> (disassemble #'do-something) ; disassembly for DO-SOMETHING ; Size: 216 bytes. Origin: #x537D0F66 ; DO-SOMETHING ; 0F66: 498B4510 MOV RAX, [R13+16] ; thread.binding-stack-pointer ; 0F6A: 488945F8 MOV [RBP-8], RAX ; 0F6E: 840425F87F0050 TEST AL, [#x50007FF8] ; safepoint ; 0F75: 498BB5E8050000 MOV RSI, [R13+1512] ; tls: *STANDARD-OUTPUT* ; 0F7C: 4883FEFF CMP RSI, -1 ; 0F80: 480F443425B05B1450 CMOVEQ RSI, [#x50145BB0] ; *STANDARD-OUTPUT* ; 0F89: 488975E0 MOV [RBP-32], RSI ; 0F8D: 4883EC10 SUB RSP, 16 ; 0F91: 488B1570FFFFFF MOV RDX, [RIP-144] ; "a: " ; 0F98: 488B7DE0 MOV RDI, [RBP-32] ; ... And of course it works for the functions that are defined by the standard, too: CL-USER> (disassemble #'format) ; disassembly for FORMAT ; Size: 581 bytes. Origin: #x52A12CCD ; FORMAT ; CCD: 840425F87F0050 TEST AL, [#x50007FF8] ; safepoint ; CD4: 48817DF817810050 CMP QWORD PTR [RBP-8], #x50008117 ; NIL ; CDC: 0F85CB000000 JNE L4 ; CE2: 4881EC90000000 SUB RSP, 144 ; CE9: 4883E4F0 AND RSP, -16 ; CED: 488D7C240F LEA RDI, [RSP+15] ; CF2: 48C747F1E5000000 MOV QWORD PTR [RDI-15], 229 ; CFA: 48C747F93E000000 MOV QWORD PTR [RDI-7], 62 ; D02: 4881ECB0000000 SUB RSP, 176 ; D09: 4883E4F0 AND RSP, -16 ; D0D: 488BC4 MOV RAX, RSP ; ... CL-USER> (disassemble #'first) CL-USER> (disassemble #'+)
mattgodbolt 4y ago
Hi Genbox, thanks for the kind words. I've never heard of sharplab.io - the speed of C# compilation is not very high on the priorities of CE right now (though we have some issues filed, and some (stale) PRs to help). We welcome help improving it! If you're talking about the speed of other languages than C#, then most of the time is in the compilers itself, not the fairly limited pre- and post-processing we do. We have a pretty sophisticated setup (at least, I think it is...) for handling the 1,500 compiler/language combinations, hundreds of libraries etc (around 2TB of data), but there's always tradeoffs between fast start-up time, manageability and simplicity of adding new compilers, and the run-time performance of running the compilers etc.

by Genbox 4y ago

andyayers 4y ago

IIRC with .NET 7 we will be able to improve this as we can now dump the jit-generated assembly from official builds. .NET 7 will be officially released in a month or so. Currently C#/F#/VB support in Compiler Explorer relies on prejitting which has various issues.

dataflow 4y ago

Confused, how is this related to C++ compilers? Edit: Oh wow I didn't realize the website does other languages, thanks!

vips7L 4y ago
Godbolt does more than C++. I often share Java byte code snippets from there
- pjmlp 4y ago
  I wish it could also do the machine code version like on .NET languages. Maybe there is a way to plug hsdis into the workflow. EDIT: Actually there is already a ticket discussing this, https://github.com/compiler-explorer/compiler-explorer/issues/1364 https://github.com/compiler-explorer/compiler-explorer/issue...
  - vips7L 4y ago
    Have you seen JitWatch? https://github.com/AdoptOpenJDK/jitwatch https://github.com/AdoptOpenJDK/jitwatch
    - pjmlp 4y ago
      Naturally, :)
mattgodbolt 4y ago
Why does it have to be related to C++ compilers? Compiler Explorer supports a number of languages, including C#: $ curl -sL https://godbolt.org/api/languages Id | Name csharp | C# fsharp | F# vb | Visual Basic go | Go c | C c++ | C++ fortran | Fortran assembly | Assembly circle | C++ (Circle) circt | CIRCT hlsl | HLSL cppx | Cppx crystal | Crystal dart | Dart erlang | Erlang carbon | Carbon hook | Hook cppx_blue | Cppx-Blue cppx_gold | Cppx-Gold mlir | MLIR cuda | CUDA C++ analysis | Analysis python | Python racket | Racket ruby | Ruby typescript | TypeScript Native d | D ada | Ada cpp_for_opencl | C++ for OpenCL openclc | OpenCL C llvm | LLVM IR cpp2_cppfront | Cpp2-cppfront rust | Rust ispc | ispc java | Java kotlin | Kotlin nim | Nim pony | Pony scala | Scala solidity | Solidity clean | Clean pascal | Pascal haskell | Haskell ocaml | OCaml swift | Swift zig | Zig $
- jlarocco 4y ago
  Since it's on topic and pretty cool, Common Lisp has built-in support for this with the disassemble function, (disassemble <function-name>): CL-USER> (defun do-something (a b) (format t "a: ~a~%b: ~a~%~%" a b)) DO-SOMETHING CL-USER> (disassemble #'do-something) ; disassembly for DO-SOMETHING ; Size: 216 bytes. Origin: #x537D0F66 ; DO-SOMETHING ; 0F66: 498B4510 MOV RAX, [R13+16] ; thread.binding-stack-pointer ; 0F6A: 488945F8 MOV [RBP-8], RAX ; 0F6E: 840425F87F0050 TEST AL, [#x50007FF8] ; safepoint ; 0F75: 498BB5E8050000 MOV RSI, [R13+1512] ; tls: *STANDARD-OUTPUT* ; 0F7C: 4883FEFF CMP RSI, -1 ; 0F80: 480F443425B05B1450 CMOVEQ RSI, [#x50145BB0] ; *STANDARD-OUTPUT* ; 0F89: 488975E0 MOV [RBP-32], RSI ; 0F8D: 4883EC10 SUB RSP, 16 ; 0F91: 488B1570FFFFFF MOV RDX, [RIP-144] ; "a: " ; 0F98: 488B7DE0 MOV RDI, [RBP-32] ; ... And of course it works for the functions that are defined by the standard, too: CL-USER> (disassemble #'format) ; disassembly for FORMAT ; Size: 581 bytes. Origin: #x52A12CCD ; FORMAT ; CCD: 840425F87F0050 TEST AL, [#x50007FF8] ; safepoint ; CD4: 48817DF817810050 CMP QWORD PTR [RBP-8], #x50008117 ; NIL ; CDC: 0F85CB000000 JNE L4 ; CE2: 4881EC90000000 SUB RSP, 144 ; CE9: 4883E4F0 AND RSP, -16 ; CED: 488D7C240F LEA RDI, [RSP+15] ; CF2: 48C747F1E5000000 MOV QWORD PTR [RDI-15], 229 ; CFA: 48C747F93E000000 MOV QWORD PTR [RDI-7], 62 ; D02: 4881ECB0000000 SUB RSP, 176 ; D09: 4883E4F0 AND RSP, -16 ; D0D: 488BC4 MOV RAX, RSP ; ... CL-USER> (disassemble #'first) CL-USER> (disassemble #'+)

mattgodbolt 4y ago

Hi Genbox, thanks for the kind words. I've never heard of sharplab.io - the speed of C# compilation is not very high on the priorities of CE right now (though we have some issues filed, and some (stale) PRs to help). We welcome help improving it! If you're talking about the speed of other languages than C#, then most of the time is in the compilers itself, not the fairly limited pre- and post-processing we do. We have a pretty sophisticated setup (at least, I think it is...) for handling the 1,500 compiler/language combinations, hundreds of libraries etc (around 2TB of data), but there's always tradeoffs between fast start-up time, manageability and simplicity of adding new compilers, and the run-time performance of running the compilers etc.