[NVPTX] Fix lit test issue from used_bytes_mask

Author: dakersnar

llvm/test/CodeGen/NVPTX/LoadStoreVectorizer.ll

@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
-; RUN: llc < %s | FileCheck -check-prefix=ENABLED %s
-; RUN: llc -disable-nvptx-load-store-vectorizer < %s | FileCheck -check-prefix=DISABLED %s
-; RUN: %if ptxas %{ llc < %s | %ptxas-verify %}
-; RUN: %if ptxas %{ llc -disable-nvptx-load-store-vectorizer < %s | %ptxas-verify %}
+; RUN: llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 < %s | FileCheck -check-prefix=ENABLED %s
+; RUN: llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 -disable-nvptx-load-store-vectorizer < %s | FileCheck -check-prefix=DISABLED %s
+; RUN: %if ptxas-sm_90 && ptxas-isa-8.8 %{ llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 < %s | %ptxas-verify -arch=sm_90 %}
+; RUN: %if ptxas-sm_90 && ptxas-isa-8.8 %{ llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 -disable-nvptx-load-store-vectorizer < %s | %ptxas-verify -arch=sm_90 %}
 
 target triple = "nvptx64-nvidia-cuda"
 
@@ -52,7 +52,7 @@ define half @fh(ptr %p) {
 ; ENABLED-NEXT:    ld.param.b64 %rd1, [fh_param_0];
 ; ENABLED-NEXT:    .pragma "used_bytes_mask 0x3ff";
 ; ENABLED-NEXT:    ld.v4.b32 {%r1, %r2, %r3, %r4}, [%rd1];
-; ENABLED-NEXT:    { .reg .b16 tmp; mov.b32 {%rs1, tmp}, %r3; }
+; ENABLED-NEXT:    mov.b32 {%rs1, _}, %r3;
 ; ENABLED-NEXT:    mov.b32 {%rs2, %rs3}, %r2;
 ; ENABLED-NEXT:    mov.b32 {%rs4, %rs5}, %r1;
 ; ENABLED-NEXT:    cvt.f32.f16 %r5, %rs5;

llvm/test/CodeGen/NVPTX/param-vectorize-device.ll

@@ -1,5 +1,5 @@
-; RUN: llc < %s -mtriple=nvptx64-unknown-unknown | FileCheck %s
-; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64-unknown-unknown | %ptxas-verify %}
+; RUN: llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 < %s | FileCheck %s
+; RUN: %if ptxas-sm_90 && ptxas-isa-8.8 %{ llc -march=nvptx64 -mcpu=sm_90 -mattr=+ptx88 < %s | %ptxas-verify -arch=sm_90 %}
 ;
 ; Check that parameters of a __device__ function with private or internal
 ; linkage called from a __global__ (kernel) function get increased alignment,