File: amdgpu_function_alt.test

package info (click to toggle)
llvm-toolchain-21 1%3A21.1.6-3
  • links: PTS, VCS
  • area: main
  • in suites: sid
  • size: 2,245,028 kB
  • sloc: cpp: 7,619,726; ansic: 1,434,018; asm: 1,058,748; python: 252,740; f90: 94,671; objc: 70,685; lisp: 42,813; pascal: 18,401; sh: 8,601; ml: 5,111; perl: 4,720; makefile: 3,675; awk: 3,523; javascript: 2,409; xml: 892; fortran: 770
file content (22 lines) | stat: -rw-r--r-- 1,193 bytes parent folder | download | duplicates (10)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
## Checks two things:
# - matching functions in AMDGPU assembly output without certain comments in the assembly
# - using the --tool and --default-march arguments

# Using `cat` as a tool requires some unusual setup (compared to the other test
# of UpdateTestChecks tools), because update_llc_test_checks only reliably
# substitutes %S (source directory) when evaluating RUN lines in the .ll file.
# The .ll file used during the test is a temporary copy, so we also need to
# provide a temporary copy of the cat'd .s file to avoid relying on
# implementation details of how llvm-lit chooses %t.

# RUN: rm -rf %t.dir && mkdir %t.dir
# RUN: cp -f %S/Inputs/amdgpu_function_alt.s %t.dir/
# RUN: cp -f %S/Inputs/amdgpu_function_alt.ll %t.dir/

# RUN: %update_llc_test_checks --llc-binary "" --tool cat --default-march=amdgcn --version 4 %t.dir/amdgpu_function_alt.ll
# RUN: diff -u %S/Inputs/amdgpu_function_alt.ll.expected %t.dir/amdgpu_function_alt.ll

# Check that re-running remembers the arguments and doesn't change anything

# RUN: %update_llc_test_checks --llc-binary "" %t.dir/amdgpu_function_alt.ll
# RUN: diff -u %S/Inputs/amdgpu_function_alt.ll.expected %t.dir/amdgpu_function_alt.ll