[NVPTX] disable combine-mad test for newer ptxas

[AlexMaclean]

No description provided.

[llvmbot]

@llvm/pr-subscribers-backend-nvptx

Author: Alex MacLean (AlexMaclean)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/93919.diff

1 Files Affected:

• (modified) llvm/test/CodeGen/NVPTX/combine-mad.ll (+2-2)

diff --git a/llvm/test/CodeGen/NVPTX/combine-mad.ll b/llvm/test/CodeGen/NVPTX/combine-mad.ll
index fba389afdca39..df4f842c822ce 100644
--- a/llvm/test/CodeGen/NVPTX/combine-mad.ll
+++ b/llvm/test/CodeGen/NVPTX/combine-mad.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc < %s -mtriple=nvptx -mcpu=sm_20 -O1 | FileCheck %s
 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_20 -O1 | FileCheck %s
-; RUN: %if ptxas %{ llc < %s -mtriple=nvptx -mcpu=sm_20 -O1 | %ptxas-verify %}
-; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 -mcpu=sm_20 -O1 | %ptxas-verify %}
+; RUN: %if ptxas && !ptxas-12.0 %{ llc < %s -mtriple=nvptx -mcpu=sm_20 -O1 | %ptxas-verify %}
+; RUN: %if ptxas && !ptxas-12.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_20 -O1 | %ptxas-verify %}
 
 define i32 @test1(i32 %n, i32 %m) {
 ;

[AlexMaclean]

@Artem-B, this addresses your request from #93628 (comment). Can you clarify what version of ptxas you're using and what error you're seeing?

[Artem-B]

@Artem-B, this addresses your request from #93628 (comment). Can you clarify what version of ptxas you're using and what error you're seeing?

The error:

[  4] ; RUN: %if ptxas %{ llc < %s -mtriple=nvptx -mcpu=sm_20 -O1 | %ptxas-verify %} [FAIL]
 llc < /build/work/ea8b3a3717bb294ac025e537a8aabcc8890c/google3/runfiles/google3/third_party/llvm/llvm-project/llvm/test/CodeGen/NVPTX/combine-mad.ll -mtriple=nvptx -mcpu=sm_20 -O1 | /build/work/ea8b3a3717bb294ac025e537a8aabcc8890c/google3/runfiles/google3/third_party/gpus/cuda_12_0/bin/ptxas -arch=sm_60 -c -o /dev/null - 
ptxas warning :  64 Bit host architecture (--machine) being used mismatches with .address_size of 32 bits
ptxas fatal   :  32-Bit compilation is no longer supported
Command failed: exit status 255

[Artem-B]

This line does not need the change. 64-bit variant works everywhere.

[AlexMaclean]

Fixed

[Artem-B]

The change should apply to line 4 only. The change itself is fine.

You can (and should run LLVM tests with LLVM_PTXAS_EXECUTABLE= set to point to ptxas from a recent CUDA version.