2208c97c1b
Similar to 806761a7629df268c8aed49657aeccffa6bca449 -mtriple= specifies the full target triple while -march= merely sets the architecture part of the default target triple, leaving a target triple which may not make sense. Therefore, -march= is error-prone and not recommended for tests without a target triple. The issue has been benign as we recognize $unknown-apple-darwin as ELF instead of rejecting it outrightly.
39 lines
1.2 KiB
LLVM
39 lines
1.2 KiB
LLVM
; RUN: llc -mtriple=hexagon -verify-machineinstrs < %s | FileCheck %s
|
|
|
|
; Check for sane output.
|
|
; CHECK: vmpyweh
|
|
|
|
target triple = "hexagon"
|
|
|
|
declare i32 @llvm.hexagon.S2.clb(i32) #0
|
|
declare i32 @llvm.hexagon.S2.asl.r.r(i32, i32) #0
|
|
declare i32 @llvm.hexagon.S2.vrndpackwh(i64) #0
|
|
declare i64 @llvm.hexagon.M2.mmpyl.s1(i64, i64) #0
|
|
|
|
define i64 @fred(i32 %a0, i32 %a1) local_unnamed_addr #1 {
|
|
b2:
|
|
br i1 undef, label %b15, label %b3
|
|
|
|
b3: ; preds = %b2
|
|
%v4 = tail call i32 @llvm.hexagon.S2.clb(i32 %a1) #0
|
|
%v5 = add nsw i32 %v4, -32
|
|
%v6 = zext i32 %v5 to i64
|
|
%v7 = shl nuw i64 %v6, 32
|
|
%v8 = or i64 %v7, 0
|
|
%v9 = tail call i32 @llvm.hexagon.S2.asl.r.r(i32 %a0, i32 0)
|
|
%v10 = tail call i32 @llvm.hexagon.S2.vrndpackwh(i64 %v8)
|
|
%v11 = sext i32 %v9 to i64
|
|
%v12 = sext i32 %v10 to i64
|
|
%v13 = tail call i64 @llvm.hexagon.M2.mmpyl.s1(i64 %v11, i64 %v12)
|
|
%v14 = and i64 %v13, 4294967295
|
|
br label %b15
|
|
|
|
b15: ; preds = %b3, %b2
|
|
%v16 = phi i64 [ %v14, %b3 ], [ 0, %b2 ]
|
|
%v17 = or i64 0, %v16
|
|
ret i64 %v17
|
|
}
|
|
|
|
attributes #0 = { nounwind readnone }
|
|
attributes #1 = { nounwind "target-cpu"="hexagonv55" }
|