| Message ID | 20210201093846.10975-1-kito.cheng@sifive.com |
|---|---|
| State | New |
| Headers | show |
| Series | PR target/98743: Fix ICE in convert_move for RISC-V | expand |
On Mon, Feb 01, 2021 at 05:38:46PM +0800, Kito Cheng wrote: > - Check `TO` mode is not BLMmode before call store_expr, calling store_expr > with BLKmode will cause ICE. How do you end up with a SUBREG_PROMOTED* of something that has bitsize of 0 (GET_MODE_BITSIZE of BLKmode is 0, right)? That makes no sense. Jakub
> > - Check `TO` mode is not BLMmode before call store_expr, calling store_expr > > with BLKmode will cause ICE. > > How do you end up with a SUBREG_PROMOTED* of something that has bitsize of 0 > (GET_MODE_BITSIZE of BLKmode is 0, right)? to_rtx is already having a mode other than BLKmode in this point, it's SImode for riscv64*-*-*, so bitsize is 32 rather than 0. I guess my comment isn't clear enough, the root cause why `store_expr (from, to_rtx, 0, nontemporal, false)` ICE is because `from` is still BLKmode here. > That makes no sense. > > Jakub >
On Mon, Feb 01, 2021 at 05:57:28PM +0800, Kito Cheng wrote: > > > - Check `TO` mode is not BLMmode before call store_expr, calling store_expr > > > with BLKmode will cause ICE. > > > > How do you end up with a SUBREG_PROMOTED* of something that has bitsize of 0 > > (GET_MODE_BITSIZE of BLKmode is 0, right)? > > to_rtx is already having a mode other than BLKmode in this point, > it's SImode for riscv64*-*-*, so bitsize is 32 rather than 0. > > I guess my comment isn't clear enough, the root cause why > `store_expr (from, to_rtx, 0, nontemporal, false)` ICE is > because `from` is still BLKmode here. But mode is TYPE_MODE (TREE_TYPE (to)), that is expanded already at this point and got some different mode. So, if anything, you care about TYPE_MODE (TREE_TYPE (from)) rather than mode, don't you? Jakub
On Mon, Feb 1, 2021 at 6:10 PM Jakub Jelinek <jakub@redhat.com> wrote: > > On Mon, Feb 01, 2021 at 05:57:28PM +0800, Kito Cheng wrote: > > > > - Check `TO` mode is not BLMmode before call store_expr, calling store_expr > > > > with BLKmode will cause ICE. > > > > > > How do you end up with a SUBREG_PROMOTED* of something that has bitsize of 0 > > > (GET_MODE_BITSIZE of BLKmode is 0, right)? > > > > to_rtx is already having a mode other than BLKmode in this point, > > it's SImode for riscv64*-*-*, so bitsize is 32 rather than 0. > > > > I guess my comment isn't clear enough, the root cause why > > `store_expr (from, to_rtx, 0, nontemporal, false)` ICE is > > because `from` is still BLKmode here. > > But mode is TYPE_MODE (TREE_TYPE (to)), that is expanded already at this > point and got some different mode. So, if anything, you care about > TYPE_MODE (TREE_TYPE (from)) rather than mode, don't you? Yeah, that's correct, let me send v2 for updating the checking and comment. > > Jakub >
diff --git a/gcc/expr.c b/gcc/expr.c index 04ef5ad114d..3cf492acea3 100644 --- a/gcc/expr.c +++ b/gcc/expr.c @@ -5459,6 +5459,7 @@ expand_assignment (tree to, tree from, bool nontemporal) /* If to_rtx is a promoted subreg, we need to zero or sign extend the value afterwards. */ if (TREE_CODE (to) == MEM_REF + && mode != BLKmode && !REF_REVERSE_STORAGE_ORDER (to) && known_eq (bitpos, 0) && known_eq (bitsize, GET_MODE_BITSIZE (GET_MODE (to_rtx)))) diff --git a/gcc/testsuite/g++.target/riscv/pr98743.C b/gcc/testsuite/g++.target/riscv/pr98743.C new file mode 100644 index 00000000000..1b94597ac68 --- /dev/null +++ b/gcc/testsuite/g++.target/riscv/pr98743.C @@ -0,0 +1,27 @@ +// Test for value-initialization via {} +// { dg-do run { target c++11 } } +/* { dg-options "-Og -version -fno-early-inlining -finline-small-functions -fpack-struct" */ +void * operator new (__SIZE_TYPE__, void *p) { return p; } +void * operator new[] (__SIZE_TYPE__, void *p) { return p; } + +// Empty base so A isn't an aggregate +struct B {}; +struct A: B { + int i; +}; + +struct C: A { + C(): A{} {} +}; + +int main() +{ + int space = 42; + A* ap = new (&space) A{}; + int space1[1] = { 42 }; + A* a1p = new (space1) A[1]{}; + if (ap->i != 0 || + a1p[0].i != 0) + return 1; + return 0; +}
- Check `TO` mode is not BLMmode before call store_expr, calling store_expr with BLKmode will cause ICE. - Verified with riscv64, x86_64 and aarch64, no introduce new regression. Note: Those logic was introduced by 3e60ddeb8220ed388819bb3f14e8caa9309fd3c2, so I cc Jakub for reivew. gcc/ChangeLog: PR target/98743 * expr.c: Check mode before calling store_expr. gcc/testsuite/ChangeLog: PR target/98743 * g++.target/riscv/pr98743.C: New. --- gcc/expr.c | 1 + gcc/testsuite/g++.target/riscv/pr98743.C | 27 ++++++++++++++++++++++++ 2 files changed, 28 insertions(+) create mode 100644 gcc/testsuite/g++.target/riscv/pr98743.C