b279f6b098
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
nvptx{,64}-apple-darwin as ELF instead of rejecting it outrightly.
30 lines
1.0 KiB
LLVM
30 lines
1.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc < %s -mtriple=nvptx64 -mattr=+ptx73 -mcpu=sm_52 | FileCheck %s
|
|
|
|
target triple = "nvptx64-nvidia-cuda"
|
|
|
|
define void @foo(i64 %a, ptr %p0, ptr %p1) {
|
|
; CHECK-LABEL: foo(
|
|
; CHECK: {
|
|
; CHECK-NEXT: .reg .b64 %rd<8>;
|
|
; CHECK-EMPTY:
|
|
; CHECK-NEXT: // %bb.0:
|
|
; CHECK-NEXT: ld.param.u64 %rd1, [foo_param_0];
|
|
; CHECK-NEXT: add.s64 %rd2, %rd1, 7;
|
|
; CHECK-NEXT: and.b64 %rd3, %rd2, -8;
|
|
; CHECK-NEXT: alloca.u64 %rd4, %rd3, 16;
|
|
; CHECK-NEXT: cvta.local.u64 %rd4, %rd4;
|
|
; CHECK-NEXT: ld.param.u64 %rd5, [foo_param_1];
|
|
; CHECK-NEXT: alloca.u64 %rd6, %rd3, 16;
|
|
; CHECK-NEXT: cvta.local.u64 %rd6, %rd6;
|
|
; CHECK-NEXT: ld.param.u64 %rd7, [foo_param_2];
|
|
; CHECK-NEXT: st.u64 [%rd5], %rd4;
|
|
; CHECK-NEXT: st.u64 [%rd7], %rd6;
|
|
; CHECK-NEXT: ret;
|
|
%b = alloca i8, i64 %a, align 16
|
|
%c = alloca i8, i64 %a, align 16
|
|
store ptr %b, ptr %p0, align 8
|
|
store ptr %c, ptr %p1, align 8
|
|
ret void
|
|
}
|