Skip to content

[AArch64][SME] Precommit tests for LUT4I Chain issues (NFC) #161505

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
MacDue merged 1 commit into from
Oct 1, 2025
Merged

[AArch64][SME] Precommit tests for LUT4I Chain issues (NFC) #161505

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 51 additions & 0 deletions llvm/test/CodeGen/AArch64/pr161420.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,51 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s | FileCheck %s

target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128-Fn32"
target triple = "arm64-apple-macosx15.0.0"

; From: https://github.com/llvm/llvm-project/issues/161420. This test checks that
; two `luti4` instructions are emitted. FIXME: This is currently broken!
define void @pluto(ptr %arg, ptr %arg1, ptr %arg2, ptr %arg3) #0 {
; CHECK-LABEL: pluto:
; CHECK: ; %bb.0: ; %bb
; CHECK-NEXT: mov w8, #0 ; =0x0
; CHECK-NEXT: ldr zt0, [x1]
; CHECK-NEXT: ldr z0, [x3]
; CHECK-NEXT: ptrue pn8.h
; CHECK-NEXT: ld1h { z4.h - z7.h }, pn8/z, [x0]
; CHECK-NEXT: luti4 { z0.h - z3.h }, zt0, z0[0]
; CHECK-NEXT: fmla za.h[w8, 2, vgx4], { z4.h - z7.h }, { z0.h - z3.h }
; CHECK-NEXT: ret
bb:
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %arg1)
%load = load <vscale x 16 x i8>, ptr %arg3, align 16
%call = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c16()
%call4 = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x4.nxv8f16(target("aarch64.svcount") %call, ptr %arg)
%extractvalue = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call4, 0
%extractvalue5 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call4, 1
%extractvalue6 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call4, 2
%extractvalue7 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call4, 3
%call8 = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> %load, i32 0)
%extractvalue9 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call8, 0
%extractvalue10 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call8, 1
%extractvalue11 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call8, 2
%extractvalue12 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call8, 3
tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 0, <vscale x 8 x half> %extractvalue, <vscale x 8 x half> %extractvalue5, <vscale x 8 x half> %extractvalue6, <vscale x 8 x half> %extractvalue7, <vscale x 8 x half> %extractvalue9, <vscale x 8 x half> %extractvalue10, <vscale x 8 x half> %extractvalue11, <vscale x 8 x half> %extractvalue12)
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %arg2)
%call13 = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> %load, i32 0)
%extractvalue14 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call13, 0
%extractvalue15 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call13, 1
%extractvalue16 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call13, 2
%extractvalue17 = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %call13, 3
tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 2, <vscale x 8 x half> %extractvalue, <vscale x 8 x half> %extractvalue5, <vscale x 8 x half> %extractvalue6, <vscale x 8 x half> %extractvalue7, <vscale x 8 x half> %extractvalue14, <vscale x 8 x half> %extractvalue15, <vscale x 8 x half> %extractvalue16, <vscale x 8 x half> %extractvalue17)
ret void
}

declare void @llvm.aarch64.sme.ldr.zt(i32, ptr)
declare target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c16()
declare { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x4.nxv8f16(target("aarch64.svcount"), ptr)
declare { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 immarg, <vscale x 16 x i8>, i32 immarg)
declare void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)

attributes #0 = { mustprogress nofree noinline norecurse nosync nounwind ssp willreturn uwtable(sync) "aarch64_inout_za" "aarch64_inout_zt0" "aarch64_pstate_sm_enabled" "target-cpu"="apple-m1" "target-features"="+fp-armv8,+lse,+neon,+sme,+sme-f16f16,+sme2,+v8.1a,+v8.2a,+v8.3a,+v8.4a,+v8.5a,+v8a" }
18 changes: 18 additions & 0 deletions llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane-x4.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,24 @@ define {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscal
ret {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} %res
}

; Tests multiple identical luti4 intrinsics with ZT0 loads interspersed, are not CSD'd.
; FIXME: This is currently broken!
define void @test_multiple_luti4_zt_i8(ptr %ptrA, ptr %ptrB, <vscale x 16 x i8> %x) {
; CHECK-LABEL: test_multiple_luti4_zt_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: luti4 { z0.s - z3.s }, zt0, z0[1]
; CHECK-NEXT: // fake_use: $z0 $z0_z1_z2_z3
; CHECK-NEXT: ret
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %ptrA)
%res1 = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 1)
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %ptrB)
%res2 = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 1)

call void (...) @llvm.fake.use({<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} %res1)
call void (...) @llvm.fake.use({<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} %res2)
ret void
}

declare {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32, <vscale x 16 x i8>, i32)
declare {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32, <vscale x 16 x i8>, i32)
declare {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32, <vscale x 16 x i8>, i32)
Expand Down
20 changes: 20 additions & 0 deletions llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,4 +14,24 @@ define {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16
ret {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>} %res
}

; Tests multiple identical luti4 intrinsics with ZT0 loads interspersed, are not CSD'd.
; FIXME: This is currently broken!
define void @test_multiple_luti4_zt_i8(ptr %ptrA, ptr %ptrB, <vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1) #0 {
; CHECK-LABEL: test_multiple_luti4_zt_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT: // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT: luti4 { z0.b - z3.b }, zt0, { z0, z1 }
; CHECK-NEXT: // fake_use: $z0 $z0_z1_z2_z3
; CHECK-NEXT: ret
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %ptrA)
%res1 = call {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>} @llvm.aarch64.sme.luti4.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1)
tail call void @llvm.aarch64.sme.ldr.zt(i32 0, ptr %ptrB)
%res2 = call {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>} @llvm.aarch64.sme.luti4.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> %v0, <vscale x 16 x i8> %v1)

call void (...) @llvm.fake.use({ <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res1)
call void (...) @llvm.fake.use({ <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res2)
ret void
}

attributes #0 = { "target-features"="+sme2,+sme-lutv2"}