[AArch64][GlobalISel] Fail legalization for unknown libcalls.

[davemgreen]

If, like powi on windows, the libcall is unavailable we should fall back to SDAG. Currently we try and generate a call to "".

[llvmbot]

@llvm/pr-subscribers-backend-aarch64

@llvm/pr-subscribers-llvm-globalisel

Author: David Green (davemgreen)

Changes

If, like powi on windows, the libcall is unavailable we should fall back to SDAG. Currently we try and generate a call to "".

---

Full diff: https://github.com/llvm/llvm-project/pull/81873.diff

2 Files Affected:

• (modified) llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp (+2)
• (added) llvm/test/CodeGen/AArch64/win64-fpowi.ll (+136)

diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index e39fdae1ccbedb..bd2dae32484c69 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -596,6 +596,8 @@ llvm::createLibcall(MachineIRBuilder &MIRBuilder, RTLIB::Libcall Libcall,
                     LostDebugLocObserver &LocObserver, MachineInstr *MI) {
   auto &TLI = *MIRBuilder.getMF().getSubtarget().getTargetLowering();
   const char *Name = TLI.getLibcallName(Libcall);
+  if (!Name)
+    return LegalizerHelper::UnableToLegalize;
   const CallingConv::ID CC = TLI.getLibcallCallingConv(Libcall);
   return createLibcall(MIRBuilder, Name, Result, Args, CC, LocObserver, MI);
 }
diff --git a/llvm/test/CodeGen/AArch64/win64-fpowi.ll b/llvm/test/CodeGen/AArch64/win64-fpowi.ll
new file mode 100644
index 00000000000000..3eb74f8394ec4e
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/win64-fpowi.ll
@@ -0,0 +1,136 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=aarch64-pc-windows-msvc19 -verify-machineinstrs %s -o - | FileCheck %s
+; RUN: llc -mtriple=aarch64-pc-windows-msvc19 -global-isel -global-isel-abort=2 -verify-machineinstrs %s -o - | FileCheck %s
+
+define double @powi_f64(double %a, i32 %b) {
+; CHECK-LABEL: powi_f64:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    scvtf d1, w0
+; CHECK-NEXT:    b pow
+entry:
+  %c = call double @llvm.powi.f64.i32(double %a, i32 %b)
+  ret double %c
+}
+
+define float @powi_f32(float %a, i32 %b) {
+; CHECK-LABEL: powi_f32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    scvtf s1, w0
+; CHECK-NEXT:    b powf
+entry:
+  %c = call float @llvm.powi.f32.i32(float %a, i32 %b)
+  ret float %c
+}
+
+define half @powi_f16(half %a, i32 %b) {
+; CHECK-LABEL: powi_f16:
+; CHECK:       .seh_proc powi_f16
+; CHECK-NEXT:  // %bb.0: // %entry
+; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_reg_x x30, 16
+; CHECK-NEXT:    .seh_endprologue
+; CHECK-NEXT:    fcvt s0, h0
+; CHECK-NEXT:    scvtf s1, w0
+; CHECK-NEXT:    bl powf
+; CHECK-NEXT:    fcvt h0, s0
+; CHECK-NEXT:    .seh_startepilogue
+; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_reg_x x30, 16
+; CHECK-NEXT:    .seh_endepilogue
+; CHECK-NEXT:    ret
+; CHECK-NEXT:    .seh_endfunclet
+; CHECK-NEXT:    .seh_endproc
+entry:
+  %c = call half @llvm.powi.f16.i32(half %a, i32 %b)
+  ret half %c
+}
+
+define <2 x double> @powi_v2f64(<2 x double> %a, i32 %b) {
+; CHECK-LABEL: powi_v2f64:
+; CHECK:       .seh_proc powi_v2f64
+; CHECK-NEXT:  // %bb.0: // %entry
+; CHECK-NEXT:    sub sp, sp, #48
+; CHECK-NEXT:    .seh_stackalloc 48
+; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_reg x30, 32
+; CHECK-NEXT:    str d8, [sp, #40] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_freg d8, 40
+; CHECK-NEXT:    .seh_endprologue
+; CHECK-NEXT:    scvtf d8, w0
+; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
+; CHECK-NEXT:    mov d0, v0.d[1]
+; CHECK-NEXT:    fmov d1, d8
+; CHECK-NEXT:    bl pow
+; CHECK-NEXT:    fmov d1, d8
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT:    bl pow
+; CHECK-NEXT:    ldr q1, [sp, #16] // 16-byte Folded Reload
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT:    mov v0.d[1], v1.d[0]
+; CHECK-NEXT:    .seh_startepilogue
+; CHECK-NEXT:    ldr d8, [sp, #40] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_freg d8, 40
+; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_reg x30, 32
+; CHECK-NEXT:    add sp, sp, #48
+; CHECK-NEXT:    .seh_stackalloc 48
+; CHECK-NEXT:    .seh_endepilogue
+; CHECK-NEXT:    ret
+; CHECK-NEXT:    .seh_endfunclet
+; CHECK-NEXT:    .seh_endproc
+entry:
+  %c = call <2 x double> @llvm.powi.v2f64.i32(<2 x double> %a, i32 %b)
+  ret <2 x double> %c
+}
+
+define <2 x float> @powi_v2f32(<2 x float> %a, i32 %b) {
+; CHECK-LABEL: powi_v2f32:
+; CHECK:       .seh_proc powi_v2f32
+; CHECK-NEXT:  // %bb.0: // %entry
+; CHECK-NEXT:    sub sp, sp, #48
+; CHECK-NEXT:    .seh_stackalloc 48
+; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_reg x30, 32
+; CHECK-NEXT:    str d8, [sp, #40] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_freg d8, 40
+; CHECK-NEXT:    .seh_endprologue
+; CHECK-NEXT:    scvtf s8, w0
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
+; CHECK-NEXT:    mov s0, v0.s[1]
+; CHECK-NEXT:    fmov s1, s8
+; CHECK-NEXT:    bl powf
+; CHECK-NEXT:    fmov s1, s8
+; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
+; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
+; CHECK-NEXT:    bl powf
+; CHECK-NEXT:    ldr q1, [sp, #16] // 16-byte Folded Reload
+; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
+; CHECK-NEXT:    mov v0.s[1], v1.s[0]
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT:    .seh_startepilogue
+; CHECK-NEXT:    ldr d8, [sp, #40] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_freg d8, 40
+; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_reg x30, 32
+; CHECK-NEXT:    add sp, sp, #48
+; CHECK-NEXT:    .seh_stackalloc 48
+; CHECK-NEXT:    .seh_endepilogue
+; CHECK-NEXT:    ret
+; CHECK-NEXT:    .seh_endfunclet
+; CHECK-NEXT:    .seh_endproc
+entry:
+  %c = call <2 x float> @llvm.powi.v2f32.i32(<2 x float> %a, i32 %b)
+  ret <2 x float> %c
+}
+
+declare <2 x double> @llvm.powi.v2f64.i32(<2 x double>, i32)
+declare <2 x float> @llvm.powi.v2f32.i32(<2 x float>, i32)
+declare double @llvm.powi.f64.i32(double, i32)
+declare float @llvm.powi.f32.i32(float, i32)
+declare half @llvm.powi.f16.i32(half, i32)

[kiranchandramohan]

Thanks @davemgreen for the fix. Could we have this in the release branch too? We have an internal team that requires this for their tests to pass.

[davemgreen]

Yep that sounds like a good idea. Hopefully it is a simple one to get ported over.