[libclc] Improve __clc_min/max/clamp implementation

[wenju-he]

Replace __clc_max/min with __clc_fmax/fmin in __clc_clamp. FP __clc_min/max/clamp now lowers to @llvm.minimumnum/@llvm.maximumnum, and integer clamp lowers to @llvm.umin/@llvm.umax. This reduce fcmp+select chains and improving codegen. Example change to amdgcn--amdhsa.bc:

in function _Z5clamphhh:
    >   %4 = icmp ugt i8 %0, %2
        %4 = tail call noundef i8 @llvm.umax.i8(i8 %0, i8 %1)
    >   %6 = select i1 %4, i8 %2, i8 %5
    >   ret i8 %6
    <   %5 = tail call noundef i8 @llvm.umin.i8(i8 %2, i8 %4)
    <   ret i8 %5
in function _Z5clampddd:
  in block %3 / %3:
    >   %4 = fcmp ogt double %0, %2
    >   %5 = fcmp olt double %0, %1
    >   %6 = select i1 %5, double %1, double %0
    >   %7 = select i1 %4, double %2, double %6
    >   ret double %7
    <   %4 = tail call noundef double @llvm.maximumnum.f64(double %0, double %1)
    <   %5 = tail call noundef double @llvm.minimumnum.f64(double %4, double %2)
    <   ret double %5

[Copilot]

Pull request overview

This PR optimizes the implementation of __clc_min, __clc_max, and __clc_clamp functions in libclc by leveraging LLVM intrinsics for better code generation. The changes replace floating-point min/max operations with calls to __clc_fmin/__clc_fmax (which lower to @llvm.minimumnum/@llvm.maximumnum), and simplify __clc_clamp to compose min/max operations instead of nested ternary expressions.

Key changes:

• Floating-point __clc_min/__clc_max now delegate to __clc_fmin/__clc_fmax for better LLVM intrinsic lowering
• __clc_clamp implementation simplified from nested ternary to composed __clc_min(__clc_max(x, y), z)
• Integer clamp now benefits from LLVM's @llvm.umin/@llvm.umax intrinsics

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
libclc/clc/lib/generic/shared/clc_min.inc	Adds floating-point path that delegates to __clc_fmin, preserves integer path with ternary
libclc/clc/lib/generic/shared/clc_min.cl	Includes clc_fmin.h header for floating-point delegation
libclc/clc/lib/generic/shared/clc_max.inc	Adds floating-point path that delegates to __clc_fmax, preserves integer path with ternary
libclc/clc/lib/generic/shared/clc_max.cl	Includes clc_fmax.h header for floating-point delegation
libclc/clc/lib/generic/shared/clc_clamp.inc	Replaces nested ternary with __clc_min(__clc_max(...)) composition
libclc/clc/lib/generic/shared/clc_clamp.cl	Includes clc_max.h and clc_min.h headers for composed implementation

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.