12 ms·

Optimizations in C++ Compilers

102 points by guodong 6y ago

orangepanda 6y ago
> I went home that evening and created Compiler Explorer. Nice try. You can’t escape being known as a verb now. Everyone knows the tool as godbolt.
- 2bluesc 6y ago
  Interesting tool I've never used, for the uninitiated: https://godbolt.org/ https://godbolt.org/
  - dr_zoidberg 6y ago
    Interesting, but broken in Firefox. Could not get it to run and then I thought "maybe in Chrome..." and then got it to work.
    - Smar 6y ago
      What did you do to get it to work?
      - dr_zoidberg 6y ago
        Opening in Chrome.
- ladberg 6y ago
  Yep, it honestly never occurred to me that the name of the tool wasn't godbolt. I just went to the site and noticed "Compiler Explorer" for the first time, which seems so generic that I never thought of it as a name.
- gpderetta 6y ago
  Yes, it is somehow also an awesome name for a tool.
- guodong 6y ago
  That's my first impression, too, until I'm curious about what 'godbolt' means, and I googled it...
- projektfu 6y ago
  He pulled an ebay. The product was called AuctionWeb but the URL was http://ebay.com/aw/ http://ebay.com/aw/
usefulcat 6y ago
Compiler Explorer (aka godbolt) is awesome; I use it at least weekly. It's amazing how many more code generation questions occur to me now that there's so much less friction in getting the answers.
- beached_whale 6y ago
  I host it locally for myself too and got the CLion plugin, it is amazing. Seeing actual codegen in projets with multiple files is really something and having the same scrolling with ASM or source too.
mrlonglong 6y ago
Godbolt sounds like a top quality brand of quidditch broom :-D
fluffything 6y ago
> Tail call removal. A recursive function that ends in a call to itself can often be rewritten as a loop, reducing call overhead and reducing the chance of stack overflow. Most important: this optimization enables pipelined execution. When people talk about a CPU executing an integer add instruction in ~1 cycle, what they actually mean is that the add has this latency when the CPU pipelines are full. If you have an 11 stage pipeline... the add can often have a latency of ~11 cycles... if you write the _right_ code for it.
- gpderetta 6y ago
  FWIW, at least on intel cpu call instructions are fully pipelined and effectively zero latency (although, like all jump instructions there is a limit of one call every other cycle).
- tomp 6y ago
  Tail call removal is much broader than that. Compilers will remove even indirect tail calls, which is useful when building a fast interpreter. I tried this with a small example (using Godbolt, of course), now let’s see if it works for a full interpreter.
- ladberg 6y ago
  Interestingly, I've seen infinite recursions get optimized to infinite tail calls. This makes debugging harder because instead of a stack overflow you just have an infinite loop and have to manually go kill the process and get a breakpoint. Then, looking at the code it's not obvious where the infinite loop occurs.
jedbrown 6y ago
The floating point comment leaves out that one can use #pragma omp simd reduction(+:res) as a more precise way to achieve vectorization in the reduction (compile with -fopenmp-simd to only use it for SIMD without linking an OpenMP library): https://godbolt.org/z/17oTz1 https://godbolt.org/z/17oTz1 Unfortunately, the pragma is not supported with the new-style class iterators in a released compiler, though it works in clang-trunk: https://godbolt.org/z/hbP11W https://godbolt.org/z/hbP11W Note that Clang disables floating point contraction by default (so no vfmadd instructions), despite them being more accurate. One usually wants this globally (-ffp-contract=fast) except when trying to bitwise reproduce software compiled for pre-Haswell.
amelius 6y ago
That's a cool bag of tricks. But I'm impressed when compilers start optimizing programs in the big-O sense.
- stephenbennyhat 6y ago
  Did you see the sumToX() example in the paper?
manch23 6y ago
> I hope that some of these optimizations are a pleasant surprise and will factor in your decisions to write clear, intention-revealing code and leave it to the compiler to do the right thing. This was my key takeaway from this article. Writing clear code that is easier to maintain will have good enough performance most of the time. I was particularly impressed with the devirtualization optimizations and will be less likely to shy away from using polymorphism in future due to performance concerns.

102 points by guodong 6y ago

> I went home that evening and created Compiler Explorer. Nice try. You can’t escape being known as a verb now. Everyone knows the tool as godbolt.

2bluesc 6y ago
Interesting tool I've never used, for the uninitiated: https://godbolt.org/ https://godbolt.org/
- dr_zoidberg 6y ago
  Interesting, but broken in Firefox. Could not get it to run and then I thought "maybe in Chrome..." and then got it to work.
  - Smar 6y ago
    What did you do to get it to work?
    - dr_zoidberg 6y ago
      Opening in Chrome.
ladberg 6y ago
Yep, it honestly never occurred to me that the name of the tool wasn't godbolt. I just went to the site and noticed "Compiler Explorer" for the first time, which seems so generic that I never thought of it as a name.
gpderetta 6y ago
Yes, it is somehow also an awesome name for a tool.
guodong 6y ago
That's my first impression, too, until I'm curious about what 'godbolt' means, and I googled it...
projektfu 6y ago
He pulled an ebay. The product was called AuctionWeb but the URL was http://ebay.com/aw/ http://ebay.com/aw/

usefulcat 6y ago

Compiler Explorer (aka godbolt) is awesome; I use it at least weekly. It's amazing how many more code generation questions occur to me now that there's so much less friction in getting the answers.

beached_whale 6y ago
I host it locally for myself too and got the CLion plugin, it is amazing. Seeing actual codegen in projets with multiple files is really something and having the same scrolling with ASM or source too.

mrlonglong 6y ago

Godbolt sounds like a top quality brand of quidditch broom :-D

fluffything 6y ago

> Tail call removal. A recursive function that ends in a call to itself can often be rewritten as a loop, reducing call overhead and reducing the chance of stack overflow. Most important: this optimization enables pipelined execution. When people talk about a CPU executing an integer add instruction in ~1 cycle, what they actually mean is that the add has this latency when the CPU pipelines are full. If you have an 11 stage pipeline... the add can often have a latency of ~11 cycles... if you write the _right_ code for it.

gpderetta 6y ago
FWIW, at least on intel cpu call instructions are fully pipelined and effectively zero latency (although, like all jump instructions there is a limit of one call every other cycle).
tomp 6y ago
Tail call removal is much broader than that. Compilers will remove even indirect tail calls, which is useful when building a fast interpreter. I tried this with a small example (using Godbolt, of course), now let’s see if it works for a full interpreter.
ladberg 6y ago
Interestingly, I've seen infinite recursions get optimized to infinite tail calls. This makes debugging harder because instead of a stack overflow you just have an infinite loop and have to manually go kill the process and get a breakpoint. Then, looking at the code it's not obvious where the infinite loop occurs.

jedbrown 6y ago

The floating point comment leaves out that one can use #pragma omp simd reduction(+:res) as a more precise way to achieve vectorization in the reduction (compile with -fopenmp-simd to only use it for SIMD without linking an OpenMP library): https://godbolt.org/z/17oTz1 https://godbolt.org/z/17oTz1 Unfortunately, the pragma is not supported with the new-style class iterators in a released compiler, though it works in clang-trunk: https://godbolt.org/z/hbP11W https://godbolt.org/z/hbP11W Note that Clang disables floating point contraction by default (so no vfmadd instructions), despite them being more accurate. One usually wants this globally (-ffp-contract=fast) except when trying to bitwise reproduce software compiled for pre-Haswell.

amelius 6y ago

That's a cool bag of tricks. But I'm impressed when compilers start optimizing programs in the big-O sense.

stephenbennyhat 6y ago
Did you see the sumToX() example in the paper?

manch23 6y ago

> I hope that some of these optimizations are a pleasant surprise and will factor in your decisions to write clear, intention-revealing code and leave it to the compiler to do the right thing. This was my key takeaway from this article. Writing clear code that is easier to maintain will have good enough performance most of the time. I was particularly impressed with the devirtualization optimizations and will be less likely to shy away from using polymorphism in future due to performance concerns.