Files
clang-p2996/llvm/test/CodeGen/XCore/inline-asm.ll
Fangrui Song f1987c74ee [XCore,test] Change llc -march= to -mtriple=
Similar to 806761a762

-mtriple= specifies the full target triple while -march= merely sets the
architecture part of the default target triple (e.g. Windows, macOS),
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 xcore-apple-darwin as ELF instead
of rejecting it outrightly.
2024-12-15 10:31:38 -08:00

54 lines
1.1 KiB
LLVM

; RUN: llc < %s -mtriple=xcore | FileCheck %s
; CHECK-LABEL: f1:
; CHECK: foo r0
define i32 @f1() nounwind {
entry:
%asmtmp = tail call i32 asm sideeffect "foo $0", "=r"() nounwind
ret i32 %asmtmp
}
; CHECK-LABEL: f2:
; CHECK: foo 5
define void @f2() nounwind {
entry:
tail call void asm sideeffect "foo $0", "i"(i32 5) nounwind
ret void
}
; CHECK-LABEL: f3:
; CHECK: foo 42
define void @f3() nounwind {
entry:
tail call void asm sideeffect "foo ${0:c}", "i"(i32 42) nounwind
ret void
}
; CHECK-LABEL: f4:
; CHECK: foo -99
define void @f4() nounwind {
entry:
tail call void asm sideeffect "foo ${0:n}", "i"(i32 99) nounwind
ret void
}
@x = external global i32
@y = external global i32, section ".cp.rodata"
; CHECK-LABEL: f5:
; CHECK: ldw r0, dp[x]
; CHECK: retsp 0
define i32 @f5() nounwind {
entry:
%asmtmp = call i32 asm "ldw $0, $1", "=r,*m"(ptr elementtype(i32) @x) nounwind
ret i32 %asmtmp
}
; CHECK-LABEL: f6:
; CHECK: ldw r0, cp[y]
; CHECK: retsp 0
define i32 @f6() nounwind {
entry:
%asmtmp = call i32 asm "ldw $0, $1", "=r,*m"(ptr elementtype(i32) @y) nounwind
ret i32 %asmtmp
}