3 ms·

I see this argument a lot, but it's silly. I don't have to care whether gcc changed the settled semantics without notice or documentation out of sincere belief

I see this argument a lot, but it's silly. I don't have to care whether gcc changed the settled semantics without notice or documentation out of sincere belief that they were helping or out of a desire to beat a benchmark that doesn't matter to me, or out of habit. It's not the intent of the compiler authors that matters, but their disregard of the needs of application programmers.

Quekid5 5y ago
What are they to do, exactly? Optimization is a very 'generic' process and application programmers want optimization. The only sensible thing to do is to assume UB cannot occur and optimize accordingly. What else is there? I can already predict that whatever you suggest will very shortly end up in Halting Problem territory or will mean: No optimization. There are a lot of UBs that (if defined) would require run-time checking to define. That wouldn't inhibit optimization per se, but it would ultimately mean slower execution.
- vyodaiken 5y ago
  C programmers prefer control to "optimization". And if you assume UB cannot occur, you should not generate code that makes it happen. Radical UB is not required for optimization: in fact it appears to mostly do nothing positive. There is not a single paper or study showing significant better performance for substantial C code that depends on assuming UB can't happen - just a bunch of hand waving.
  - saagarjha 5y ago
    I don't mean to be snide, but there's no paper on it because it's pretty much something you can learn in compiler 101. Without being able to assume that UB doesn't happen, useful optimizations become impossible very quickly.
    - vyodaiken 5y ago
      You are being snide and inaccurate. Studies show 80% or more of GCC optimization improvements come from the core simple methods. Trying to compensate for the lack of data to support your argument by claiming (falsely) it's taught in an elementary course is weak.
      - saagarjha 5y ago
        Please link these studies, or describe your "core simple methods", because it is likely that they rely on programs not exhibiting undefined behavior. I mean, even the most simple optimizations like eliminating unused variables or inlining functions (what if you reach into the stack to detect these?) fall apart. These were taught within the first couple weeks of my compiler optimizations class, and I certainly hope that they were part of yours, because I can't imagine what you could have possibly gone over without starting off with this.

Quekid5 5y ago

What are they to do, exactly? Optimization is a very 'generic' process and application programmers want optimization. The only sensible thing to do is to assume UB cannot occur and optimize accordingly. What else is there? I can already predict that whatever you suggest will very shortly end up in Halting Problem territory or will mean: No optimization. There are a lot of UBs that (if defined) would require run-time checking to define. That wouldn't inhibit optimization per se, but it would ultimately mean slower execution.

vyodaiken 5y ago
C programmers prefer control to "optimization". And if you assume UB cannot occur, you should not generate code that makes it happen. Radical UB is not required for optimization: in fact it appears to mostly do nothing positive. There is not a single paper or study showing significant better performance for substantial C code that depends on assuming UB can't happen - just a bunch of hand waving.
- saagarjha 5y ago
  I don't mean to be snide, but there's no paper on it because it's pretty much something you can learn in compiler 101. Without being able to assume that UB doesn't happen, useful optimizations become impossible very quickly.
  - vyodaiken 5y ago
    You are being snide and inaccurate. Studies show 80% or more of GCC optimization improvements come from the core simple methods. Trying to compensate for the lack of data to support your argument by claiming (falsely) it's taught in an elementary course is weak.
    - saagarjha 5y ago
      Please link these studies, or describe your "core simple methods", because it is likely that they rely on programs not exhibiting undefined behavior. I mean, even the most simple optimizations like eliminating unused variables or inlining functions (what if you reach into the stack to detect these?) fall apart. These were taught within the first couple weeks of my compiler optimizations class, and I certainly hope that they were part of yours, because I can't imagine what you could have possibly gone over without starting off with this.