Code gen bug
The code:
struct Ele { long target; size_t n; }
struct NumWays {
long[] using;
long[Ele] lookup;
auto find(long target, size_t n) {
if(n==0) { return target==0; }
if(target < using[n]) { return 0; }
auto ret = find(target-using[n],n-1) + find(target,n-1);
if(auto ptr = Ele(target,n) in lookup) { return *ptr; }
return ret;
}
}
void main(string[] args) {
}
when compiled with ldc2 produces:
Invalid bitcast
%54 = bitcast i64 %53 to i32
LLVM ERROR: Broken module found, compilation aborted!
#0 0x000058d081240cd7 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6dcbcd7)
#1 0x000058d08123eacc llvm::sys::RunSignalHandlers() (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6dc9acc)
#2 0x000058d08124137f SignalHandler(int) Signals.cpp:0:0
#3 0x0000713c7bc42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)
#4 0x0000713c7bc969fc __pthread_kill_implementation ./nptl/./nptl/pthread_kill.c:44:76
#5 0x0000713c7bc969fc __pthread_kill_internal ./nptl/./nptl/pthread_kill.c:78:10
#6 0x0000713c7bc969fc pthread_kill ./nptl/./nptl/pthread_kill.c:89:10
#7 0x0000713c7bc42476 gsignal ./signal/../sysdeps/posix/raise.c:27:6
#8 0x0000713c7bc287f3 abort ./stdlib/./stdlib/abort.c:81:7
#9 0x000058d0811a99b2 llvm::report_fatal_error(llvm::Twine const&, bool) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6d349b2)
#10 0x000058d0811a97e6 (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6d347e6)
#11 0x000058d081038bcc (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6bc3bcc)
#12 0x000058d08151cc4d llvm::detail::PassModel<llvm::Module, llvm::VerifierPass, llvm::PreservedAnalyses, llvm::AnalysisManager<llvm::Module> >::run(llvm::Module&, llvm::AnalysisManager<llvm::Module>&) ld-temp.o:0:0
#13 0x000058d08100f834 llvm::PassManager<llvm::Module, llvm::AnalysisManager<llvm::Module> >::run(llvm::Module&, llvm::AnalysisManager<llvm::Module>&) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x6b9a834)
#14 0x000058d081516ec6 runOptimizationPasses(llvm::Module*) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x70a1ec6)
#15 0x000058d0815a6ade writeModule(llvm::Module*, char const*) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x7131ade)
#16 0x000058d0815a5baa ldc::CodeGenerator::writeAndFreeLLModule(char const*) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x7130baa)
#17 0x000058d07dfe4e90 codegenModules(Array<Module*>&) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x3b6fe90)
#18 0x000058d07df83dc1 mars_tryMain(Param&, Array<char const*>&) (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x3b0edc1)
#19 0x000058d07dfe7ac0 cppmain() (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x3b72ac0)
#20 0x000058d0817610fd _D2rt6dmain212_d_run_main2UAAamPUQgZiZ6runAllMFZv (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x72ec0fd)
#21 0x000058d081760ed8 _d_run_main2 (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x72ebed8)
#22 0x000058d081760ced _d_run_main (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x72ebced)
#23 0x000058d0815a31c8 main (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x712e1c8)
#24 0x0000713c7bc29d90 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:58:16
#25 0x0000713c7bc29e40 call_init ./csu/../csu/libc-start.c:128:20
#26 0x0000713c7bc29e40 __libc_start_main ./csu/../csu/libc-start.c:379:5
#27 0x000058d07dfec37e _start (/home/james/dlang/ldc-1.39.0/bin/ldc2+0x3b7737e)
Aborted (core dumped)
ldc2 version produces:
LDC - the LLVM D compiler (1.39.0):
based on DMD v2.109.1 and LLVM 18.1.6
built with LDC - the LLVM D compiler (1.39.0)
Default target: x86_64-unknown-linux-gnu
Host CPU: znver2
http://dlang.org - http://wiki.dlang.org/LDC
Registered Targets:
aarch64 - AArch64 (little endian)
aarch64_32 - AArch64 (little endian ILP32)
aarch64_be - AArch64 (big endian)
amdgcn - AMD GCN GPUs
arm - ARM
arm64 - ARM64 (little endian)
arm64_32 - ARM64 (little endian ILP32)
armeb - ARM (big endian)
avr - Atmel AVR Microcontroller
bpf - BPF (host endian)
bpfeb - BPF (big endian)
bpfel - BPF (little endian)
hexagon - Hexagon
lanai - Lanai
loongarch32 - 32-bit LoongArch
loongarch64 - 64-bit LoongArch
mips - MIPS (32-bit big endian)
mips64 - MIPS (64-bit big endian)
mips64el - MIPS (64-bit little endian)
mipsel - MIPS (32-bit little endian)
msp430 - MSP430 [experimental]
nvptx - NVIDIA PTX 32-bit
nvptx64 - NVIDIA PTX 64-bit
ppc32 - PowerPC 32
ppc32le - PowerPC 32 LE
ppc64 - PowerPC 64
ppc64le - PowerPC 64 LE
r600 - AMD GPUs HD2XXX-HD6XXX
riscv32 - 32-bit RISC-V
riscv64 - 64-bit RISC-V
sparc - Sparc
sparcel - Sparc LE
sparcv9 - Sparc V9
spirv - SPIR-V Logical
spirv32 - SPIR-V 32-bit
spirv64 - SPIR-V 64-bit
systemz - SystemZ
thumb - Thumb
thumbeb - Thumb (big endian)
ve - VE
wasm32 - WebAssembly 32-bit
wasm64 - WebAssembly 64-bit
x86 - 32-bit X86: Pentium-Pro and above
x86-64 - 64-bit X86: EM64T and AMD64
xcore - XCore
Also replacing auto find... by long find... causes it to go away.
Looks like a frontend issue at first glance - the type of ret is a 32-bit integer, making no sense.
Edit: To be precise, even the recursive find() invocation results are 32-bit.
AST output does correctly infer long: https://d.godbolt.org/z/TdW41qn81
AST output does correctly infer
long: https://d.godbolt.org/z/TdW41qn81
I think the problem is this:
int ret = this.find(target - this.using[n], n - 1LU) + this.find(target, n - 1LU);
This code is lowered to
%29 = sub i64 %28, 1, !dbg !68 ; [#uses = 1] [debug line = bug.d:9:3]
%30 = call i64 @_D3bug7NumWays4findMFNaNbNiNflmZl(ptr nonnull %.this_arg, i64 %27, i64 %29) #0, !dbg !68 ; [#uses = 1] [debug line = bug.d:9:3] [display name = find]
%31 = add i64 %26, %30, !dbg !68 ; [#uses = 1] [debug line = bug.d:9:3]
%32 = bitcast i64 %31 to i32, !dbg !68 ; [#uses = 1] [debug line = bug.d:9:3]
The problem is definitely in the frontend.
I have made a simpler reproducer:
extern (C) long stub();
auto recursive(long a, size_t n)
{
if (a < n)
return 0;
auto ret = recursive(a, n);
if (n < 10)
return stub();
return ret;
}
The condition to trigger the bug is this: the recursive function call-site is sandwiched between two other usages that also need type inference.