Fangrui Song b279f6b098 [NVPTX,test] Change llc -march= to -mtriple=
Similar to 806761a7629df268c8aed49657aeccffa6bca449

-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.
2024-12-15 10:45:11 -08:00

36 lines
1.1 KiB
LLVM

; RUN: llc < %s -mtriple=nvptx64 | FileCheck %s
; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 | %ptxas-verify %}
; CHECK-LABEL: .func{{.*}}test1
define float @test1(float %in) local_unnamed_addr {
; CHECK: rsqrt.approx.f32
%call = call float @llvm.nvvm.rsqrt.approx.f(float %in)
ret float %call
}
; CHECK-LABEL: .func{{.*}}test2
define double @test2(double %in) local_unnamed_addr {
; CHECK: rsqrt.approx.f64
%call = call double @llvm.nvvm.rsqrt.approx.d(double %in)
ret double %call
}
; CHECK-LABEL: .func{{.*}}test3
define float @test3(float %in) local_unnamed_addr {
; CHECK: rsqrt.approx.ftz.f32
%call = tail call float @llvm.nvvm.rsqrt.approx.ftz.f(float %in)
ret float %call
}
; CHECK-LABEL: .func{{.*}}test4
define double @test4(double %in) local_unnamed_addr {
; CHECK: rsqrt.approx.ftz.f64
%call = tail call double @llvm.nvvm.rsqrt.approx.ftz.d(double %in)
ret double %call
}
declare float @llvm.nvvm.rsqrt.approx.ftz.f(float)
declare double @llvm.nvvm.rsqrt.approx.ftz.d(double)
declare float @llvm.nvvm.rsqrt.approx.f(float)
declare double @llvm.nvvm.rsqrt.approx.d(double)