0f1b5f115a
ptxas is a proprietary compiler from Nvidia that can compile PTX to machine code (SASS). It has a lot of diagnostics to catch errors in PTX, which can be used to verify PTX output from llc. Set -DPXTAS_EXECUTABLE=/path/to/ptxas CMake option to enable it. If this option is not set, then ptxas is substituted to true which effectively disables all ptxas RUN lines. LLVM_PTXAS_EXECUTABLE environment variable takes precedence over the CMake option, and allows to override ptxas executable that is used for LIT without complete re-configuration. Differential Revision: https://reviews.llvm.org/D121727
75 lines
1.9 KiB
LLVM
75 lines
1.9 KiB
LLVM
; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast | FileCheck %s
|
|
; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 -fp-contract=fast | FileCheck %s
|
|
; RUN: %if ptxas %{ llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast | %ptxas-verify %}
|
|
; RUN: %if ptxas %{ llc < %s -march=nvptx64 -mcpu=sm_20 -fp-contract=fast | %ptxas-verify %}
|
|
|
|
;; These tests should run for all targets
|
|
|
|
;;===-- Basic instruction selection tests ---------------------------------===;;
|
|
|
|
|
|
;;; f64
|
|
|
|
define double @fadd_f64(double %a, double %b) {
|
|
; CHECK: add.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fadd double %a, %b
|
|
ret double %ret
|
|
}
|
|
|
|
define double @fsub_f64(double %a, double %b) {
|
|
; CHECK: sub.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fsub double %a, %b
|
|
ret double %ret
|
|
}
|
|
|
|
define double @fmul_f64(double %a, double %b) {
|
|
; CHECK: mul.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fmul double %a, %b
|
|
ret double %ret
|
|
}
|
|
|
|
define double @fdiv_f64(double %a, double %b) {
|
|
; CHECK: div.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fdiv double %a, %b
|
|
ret double %ret
|
|
}
|
|
|
|
;; PTX does not have a floating-point rem instruction
|
|
|
|
|
|
;;; f32
|
|
|
|
define float @fadd_f32(float %a, float %b) {
|
|
; CHECK: add.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fadd float %a, %b
|
|
ret float %ret
|
|
}
|
|
|
|
define float @fsub_f32(float %a, float %b) {
|
|
; CHECK: sub.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fsub float %a, %b
|
|
ret float %ret
|
|
}
|
|
|
|
define float @fmul_f32(float %a, float %b) {
|
|
; CHECK: mul.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fmul float %a, %b
|
|
ret float %ret
|
|
}
|
|
|
|
define float @fdiv_f32(float %a, float %b) {
|
|
; CHECK: div.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}
|
|
; CHECK: ret
|
|
%ret = fdiv float %a, %b
|
|
ret float %ret
|
|
}
|
|
|
|
;; PTX does not have a floating-point rem instruction
|