ustb-owl / mimic Goto Github PK

When I try to compile some simple program, like (https://github.com/pku-minic/compiler-dev-test-cases/blob/master/testcases/lv9/16_sort2.c), the compiler always gives the following error:

demo.c:4:14: error: id: putint, conflicted function type

the putint definition is defined in src/main.cpp, but I don't know why it is a conflicted function type.

Hello! I apologize for bothering you during the holiday, but I would like to report a bug.

I have discovered that the compiler is an excellent resource for learning. Before delving into the source code, I plan to test it with the provided test cases from this link: https://gitlab.eduxiji.net/windcome/sysyruntimelibrary. To do so, I am using the Sysy RISCV library found here: https://github.com/pku-minic/sysy-runtime-lib/.

For testing purposes, I have chosen to test 3 backends (C, AARCH32, RISCV32) and enable or disable optimizations. As I don't currently possess the ARM or RISCV board, I'm utilizing QEMU-static to test the cross-compiled binary executable files. Here is my current testing environment:

$ uname -osr
Linux 5.10.102.1-microsoft-standard-WSL2 GNU/Linux
$ clang --version
clang version 15.0.7
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/sbin
$ qemu-riscv32-static --version
qemu-riscv32 version 8.0.0
Copyright (c) 2003-2022 Fabrice Bellard and the QEMU Project developers

Almost all the testcases passed, except for the riscv backend with -O2. There are three tests that failed, which are fft0, fft1, fft2. Since all backends generate the same IR, I don't think the bug is in the middle-end optimization passes.

Here is the build script:

mmcc = '../build/mmcc -S -ta riscv32'
exe = 'temp-rv'
asm = 'temp-rv.s'
obj = 'temp-rv.o'
asm_flags = '-target riscv32-unknown-linux-elf -march=rv32im -mabi=ilp32'
cc = f'clang {asm} -c -o {obj} {asm_flags}'
lib_flags = f'-L../sysy_lib/riscv32 -lsysy'
link = f'ld.lld {obj} {lib_flags} -o {exe}'
simu = f'qemu-riscv32-static ./{exe}'

After conducting some debugging, I found that the invalid instruction in fft0 is as follows:

00011c44 <.Lpcrel_hi30>:
   11c44: 97 22 00 00  	auipc	t0, 2
   11c48: 83 a2 02 0b  	lw	t0, 176(t0)
   11c4c: 23 22 54 e0  	sw	t0, -508(s0)
   11c50: 13 83 02 00  	mv	t1, t0
   11c54: b3 82 62 00  	add	t0, t0, t1
   11c58: 23 22 54 e0  	sw	t0, -508(s0)
   11c5c: b3 02 b5 01  	add	t0, a0, s11
   11c60: 23 20 54 f4  	sw	t0, -192(s0)
   11c64: 93 92 22 00  	slli	t0, t0, 2
   11c68: 23 22 54 f4  	sw	t0, -188(s0)
   11c6c: 83 22 c4 de  	lw	t0, -532(s0)
   11c70: 03 23 44 f4  	lw	t1, -188(s0)
   11c74: b3 82 62 00  	add	t0, t0, t1
   11c78: 23 2a 54 f4  	sw	t0, -172(s0)
   11c7c: 83 a2 02 00  	lw	t0, 0(t0)
   11c80: 23 24 54 f4  	sw	t0, -184(s0)
   11c84: 03 23 44 e0  	lw	t1, -508(s0)
   11c88: 23 20 53 00  	sw	t0, 0(t1)      <--------- invalid

This can be verified using gdb:

(gdb) target remote :1234
Remote debugging using :1234
0x00012c64 in ?? ()
(gdb) b *0x11c84
Breakpoint 1 at 0x11c84
(gdb) c
Continuing.
Breakpoint 1, 0x00011c84 in ?? ()
(gdb) si
0x00011c88 in ?? ()
(gdb) si
Program received signal SIGSEGV, Segmentation fault.
0x00011c88 in ?? ()

However, I am currently unsure of how to proceed and locate the bug. If you have the time, could you please take a look? Your assistance would be greatly appreciated!