Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

loongarch64 half (f16) failure to select with +f and +d #93894

Closed
tgross35 opened this issue May 30, 2024 · 2 comments · Fixed by #94456
Closed

loongarch64 half (f16) failure to select with +f and +d #93894

tgross35 opened this issue May 30, 2024 · 2 comments · Fixed by #94456
Labels
backend:loongarch crash Prefer [crash-on-valid] or [crash-on-invalid]

Comments

@tgross35
Copy link
Contributor

define half @half_add(half %a, half %b) unnamed_addr #0 {
start:
  %_0 = fadd half %a, %b
  ret half %_0
}

With -mtriple=loongarch64-unknown-linux-gnu -mattr=+f,+d:

LLVM ERROR: Cannot select: t23: f32 = fp16_to_fp t21
  t21: i64 = fp_to_fp16 t4
    t4: f32,ch = CopyFromReg t0, Register:f32 %1
      t3: f32 = Register %1
In function: half_add
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
Stack dump:
0.	Program arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/llc -o /app/output.s -x86-asm-syntax=intel -mtriple=loongarch64-unknown-linux-gnu -mattr=+f,+d <source>
1.	Running pass 'Function Pass Manager' on module '<source>'.
2.	Running pass 'LoongArch DAG->DAG Pattern Instruction Selection' on function '@half_add'
 #0 0x00000000039e9678 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x39e9678)
 #1 0x00000000039e6dbc SignalHandler(int) Signals.cpp:0:0
 #2 0x00007e2b90a42520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)
 #3 0x00007e2b90a969fc pthread_kill (/lib/x86_64-linux-gnu/libc.so.6+0x969fc)
 #4 0x00007e2b90a42476 gsignal (/lib/x86_64-linux-gnu/libc.so.6+0x42476)
 #5 0x00007e2b90a287f3 abort (/lib/x86_64-linux-gnu/libc.so.6+0x287f3)
 #6 0x0000000000720e07 (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x720e07)
 #7 0x000000000379ccad llvm::SelectionDAGISel::CannotYetSelect(llvm::SDNode*) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x379ccad)
 #8 0x00000000037a1c9a llvm::SelectionDAGISel::SelectCodeCommon(llvm::SDNode*, unsigned char const*, unsigned int) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x37a1c9a)
 #9 0x00000000016a2837 llvm::LoongArchDAGToDAGISel::Select(llvm::SDNode*) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x16a2837)
#10 0x0000000003799aee llvm::SelectionDAGISel::DoInstructionSelection() (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x3799aee)
#11 0x00000000037a7db6 llvm::SelectionDAGISel::CodeGenAndEmitDAG() (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x37a7db6)
#12 0x00000000037ab0bd llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x37ab0bd)
#13 0x00000000037acaff llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) (.part.0) SelectionDAGISel.cpp:0:0
#14 0x00000000029e6721 llvm::MachineFunctionPass::runOnFunction(llvm::Function&) (.part.0) MachineFunctionPass.cpp:0:0
#15 0x0000000002fad603 llvm::FPPassManager::runOnFunction(llvm::Function&) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x2fad603)
#16 0x0000000002fad841 llvm::FPPassManager::runOnModule(llvm::Module&) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x2fad841)
#17 0x0000000002fae0a5 llvm::legacy::PassManagerImpl::run(llvm::Module&) (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x2fae0a5)
#18 0x000000000083129c compileModule(char**, llvm::LLVMContext&) llc.cpp:0:0
#19 0x0000000000728536 main (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x728536)
#20 0x00007e2b90a29d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)
#21 0x00007e2b90a29e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)
#22 0x000000000082804e _start (/opt/compiler-explorer/clang-assertions-trunk/bin/llc+0x82804e)
Program terminated with signal: SIGSEGV
Compiler returned: 139

There is no error if the -mattr flag is omitted

Reproduction: https://llvm.godbolt.org/z/faenafnzh

@tgross35
Copy link
Contributor Author

tgross35 commented Jun 4, 2024

The other variants of this fail here as well, e.g. i64 = fp_to_fp16, rust-lang/rust#125016 (comment)

@andrewrk
Copy link
Member

andrewrk commented Jun 5, 2024

Zig blocker too ziglang/zig#20193

xen0n added a commit to xen0n/llvm-project that referenced this issue Jun 5, 2024
The test case is adapted from llvm/test/CodeGen/RISCV/fp16-promote.ll,
because it covers some more IR patterns that ought to be common.

Fixes llvm#93894
heiher pushed a commit that referenced this issue Jun 6, 2024
The test case is adapted from llvm/test/CodeGen/RISCV/fp16-promote.ll,
because it covers some more IR patterns that ought to be common.

Fixes #93894
@EugeneZelenko EugeneZelenko removed the llvm:SelectionDAG SelectionDAGISel as well label Jun 6, 2024
tgross35 added a commit to tgross35/rust that referenced this issue Aug 22, 2024
Loongarch previously had a selection failure for `f16` math [1]. This
was fixed in [2], which Rust got with the update to LLVM19.

Enable it in `std/build.rs` so we start running tests.

[1]: llvm/llvm-project#93894
[2]: llvm/llvm-project#94456
tgross35 added a commit to tgross35/rust that referenced this issue Aug 22, 2024
Loongarch previously had a selection failure for `f16` math [1]. This
was fixed in [2], which Rust got with the update to LLVM 19 [3].

Enable loongarch in `std/build.rs` so we start running tests.

[1]: llvm/llvm-project#93894
[2]: llvm/llvm-project#94456
[3]: rust-lang#127513
bors added a commit to rust-lang-ci/rust that referenced this issue Aug 22, 2024
Enable `f16` tests on loongarch

Loongarch previously had a selection failure for `f16` math llvm/llvm-project#93894. This was fixed in llvm/llvm-project#94456, which Rust got with the update to LLVM 19 rust-lang#127513.

Enable loongarch in `std/build.rs` so we start running tests.

try-job: dist-loongarch64-linux
try-job: dist-loongarch64-musl
SixWeining pushed a commit that referenced this issue Sep 13, 2024
…ongarch (#107791)

For zig with LLVM 19.1.0rc4, we are seeing the following error when
bootstrapping a `loongarch64-linux-musl` target.


ziglang/zig-bootstrap#164 (comment)

It seems that this issue is caused by `PromoteFloatResult` is not
handling FREEZE OP on loongarch.

Here is the reproduction of the error: https://godbolt.org/z/PPfvWjjG5

~~This patch adds the FREEZE OP handling with `PromoteFloatRes_UnaryOp`
and adds a test case.~~

This patch changes loongarch's way of floating point promotion to soft
promotion to avoid this problem.

See: loongarch's handling of `half`:
- #93894
- #94456

Also see: other float promotion FREEZE handling
-
0019c2f
llvmbot pushed a commit to llvmbot/llvm-project that referenced this issue Sep 18, 2024
…ongarch (llvm#107791)

For zig with LLVM 19.1.0rc4, we are seeing the following error when
bootstrapping a `loongarch64-linux-musl` target.

ziglang/zig-bootstrap#164 (comment)

It seems that this issue is caused by `PromoteFloatResult` is not
handling FREEZE OP on loongarch.

Here is the reproduction of the error: https://godbolt.org/z/PPfvWjjG5

~~This patch adds the FREEZE OP handling with `PromoteFloatRes_UnaryOp`
and adds a test case.~~

This patch changes loongarch's way of floating point promotion to soft
promotion to avoid this problem.

See: loongarch's handling of `half`:
- llvm#93894
- llvm#94456

Also see: other float promotion FREEZE handling
-
llvm@0019c2f

(cherry picked from commit 13280d9)
heiher pushed a commit to llvmbot/llvm-project that referenced this issue Oct 19, 2024
…ongarch (llvm#107791)

For zig with LLVM 19.1.0rc4, we are seeing the following error when
bootstrapping a `loongarch64-linux-musl` target.

ziglang/zig-bootstrap#164 (comment)

It seems that this issue is caused by `PromoteFloatResult` is not
handling FREEZE OP on loongarch.

Here is the reproduction of the error: https://godbolt.org/z/PPfvWjjG5

~~This patch adds the FREEZE OP handling with `PromoteFloatRes_UnaryOp`
and adds a test case.~~

This patch changes loongarch's way of floating point promotion to soft
promotion to avoid this problem.

See: loongarch's handling of `half`:
- llvm#93894
- llvm#94456

Also see: other float promotion FREEZE handling
-
llvm@0019c2f

(cherry picked from commit 13280d9)
# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
backend:loongarch crash Prefer [crash-on-valid] or [crash-on-invalid]
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants