(a > b) | (a < b) is not simplified only for the case b=0 #62586

k-arrows · 2023-05-07T05:36:23Z

int foo(int a)
{
    return (a > 0) | (a < 0);
}

int bar(int a)
{
    return (a != 0);
}

foo(int):                                # @foo(int)
        xor     eax, eax
        test    edi, edi
        setg    al
        shr     edi, 31
        or      eax, edi
        ret
bar(int):                                # @bar(int)
        xor     eax, eax
        test    edi, edi
        setne   al
        ret

RKSimon · 2023-05-07T09:54:44Z

define i32 @foo(i32 %0) {
  %2 = icmp sgt i32 %0, 0
  %3 = zext i1 %2 to i32
  %4 = lshr i32 %0, 31
  %5 = or i32 %4, %3
  ret i32 %5
}

k-arrows · 2023-05-07T10:24:36Z

https://alive2.llvm.org/ce/z/33HzjE

define i32 @src(i32 %0) {
%1:
  %2 = icmp sgt i32 %0, 0
  %3 = zext i1 %2 to i32
  %4 = lshr i32 %0, 31
  %5 = or i32 %4, %3
  ret i32 %5
}
=>
define i32 @tgt(i32 %0) {
%1:
  %2 = icmp ne i32 %0, 0
  %3 = zext i1 %2 to i32
  ret i32 %3
}
Transformation seems to be correct!

XChy · 2023-07-03T05:42:33Z

Patch for the simplification:patch

…NFC) Tests for an upcoming (A > 0) | (A < 0) -> zext (A != 0) fold. Related issue: [[ #62586 | (a > b) | (a < b) is not simplified only for the case b=0 ]] Differential Revision: https://reviews.llvm.org/D154089

[InstCombine] Transform (A > 0) | (A < 0) -> zext (A != 0) fold This extends **foldCastedBitwiseLogic** to handle the similar cases. Actually, for `(A > B) | (A < B)`, when B != 0, it can be optimized to `zext( A != B )` by **foldAndOrOfICmpsUsingRanges**. However, when B = 0, **transformZExtICmp** will transform `zext(A < 0) to i32` into `A << 31`, which cannot be optimized by **foldAndOrOfICmpsUsingRanges**. Because I'm new to LLVM and has no concise knowledge about how LLVM decides the order of optimization, I choose to extend **foldCastedBitwiseLogic** to fold `( A << (X - 1) ) | ((A > 0) zext to iX) -> (A != 0) zext to iX`. And the equivalent fold follows: ``` A << (X - 1) ) | ((A > 0) zext to iX -> A < 0 | A > 0 -> (A != 0) zext to iX ``` It's proved by [[https://alive2.llvm.org/ce/z/33HzjE|alive-tv]] Related issue: [[#62586 | (a > b) | (a < b) is not simplified only for the case b=0 ]] Reviewed By: goldstein.w.n Differential Revision: https://reviews.llvm.org/D154126

k-arrows · 2023-07-08T01:35:09Z

I will close this issue since the fix has been merged.

github-actions bot added the new issue label May 7, 2023

RKSimon added llvm:instcombine and removed new issue labels May 7, 2023

k-arrows closed this as completed Jul 8, 2023

k-arrows mentioned this issue Jul 8, 2023

(a >= b-1) & (a < b) is not simplified for the case b=0 #63751

Closed

k-arrows mentioned this issue Jan 31, 2024

[InstCombine] Fold A pred C ? (A >> BW - 1) : 1 --> ZExt(A pred C ? A < 0 : 1) #69961

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

(a > b) | (a < b) is not simplified only for the case b=0 #62586

(a > b) | (a < b) is not simplified only for the case b=0 #62586

k-arrows commented May 7, 2023 •

edited by VoltrexKeyva

RKSimon commented May 7, 2023

k-arrows commented May 7, 2023

XChy commented Jul 3, 2023

k-arrows commented Jul 8, 2023

(a > b) | (a < b) is not simplified only for the case b=0 #62586

(a > b) | (a < b) is not simplified only for the case b=0 #62586

Comments

k-arrows commented May 7, 2023 • edited by VoltrexKeyva

RKSimon commented May 7, 2023

k-arrows commented May 7, 2023

XChy commented Jul 3, 2023

k-arrows commented Jul 8, 2023

k-arrows commented May 7, 2023 •

edited by VoltrexKeyva