On 3/21/24 5:20 AM, Thomas Schwinge wrote:
Hi!
On 2024-02-16T10:48:53-0800, Mike Stump <mikest...@comcast.net> wrote:
On Feb 16, 2024, at 2:16 AM, Jakub Jelinek <ja...@redhat.com> wrote:
There is one special case, NVPTX, which is a TARGET_NO_REGISTER_ALLOCATION
target. I think claiming for it that it is a lra target is strange (even
though it effectively returns true for targetm.lra_p ()), unsure if it
supports asm goto with outputs or not, if it does and we want to test it,
perhaps we should introduce asm_goto_outputs effective target and use
lra || nvptx-*-* for that?
Since the port people have to maintain that code in general, I usually leave it
to them to try and select a cheap, maintainable way to manage it.
If people want to pave the way, I'd tend to defer to them, having thought about
more than I.
Here I am. ;-)
After commit e16f90be2dc8af6c371fe79044c3e668fa3dda62
"testsuite: Fix up lra effective target", we get for nvptx target:
-PASS: gcc.c-torture/compile/asmgoto-2.c -O0 (test for excess errors)
+ERROR: gcc.c-torture/compile/asmgoto-2.c -O0 : no files matched glob pattern
"lra1020113.c.[0-9][0-9][0-9]r.reload" for " dg-do 2 compile { target lra } "
Etc.
That is, the current effective-target 'lra' is not suitable for nvptx --
which, I suppose, is OK, given that nvptx neither uses LRA nor doesn't
use LRA. ;-) (Therefore, effective-target 'lra' shouldn't get used in
test cases that are active for nvptx.)
However, nvptx appears to support 'asm goto' with outputs, including the
new execution test case:
PASS: gcc.dg/pr107385.c execution test
I'm attaching "[WIP] New effective-target 'asm_goto_with_outputs'", which
does address the effective-target check for nvptx, and otherwise does
's%lra%asm_goto_with_outputs'. (I have not yet actually merged
'check_effective_target_lra' into
'check_effective_target_asm_goto_with_outputs'.)
I have verified that all current effective-target 'lra' test cases
actually use 'asm goto' with outputs, there is just one exception:
'gcc.dg/pr110079.c' (see
<https://inbox.sourceware.org/Zel5TMMr/3BHgl0g@tucnak>
"bb-reorder: Fix -freorder-blocks-and-partition ICEs on aarch64 with asm goto
[PR110079]",
<https://gcc.gnu.org/PR110079>
"ICE with -freorder-blocks-and-partition and inline-asm goto"). That
test case, 'gcc.dg/pr110079.c', currently uses 'target lra', and uses
'asm goto' -- but not with outputs, so is 'asm_goto_with_outputs' not
really applicable? The test case does PASS for nvptx target (but I've
not verified what it's actually doing/testing). How to handle that one?
I'd just make target_lra return false for nvptx rather than creating a
new selector -- I'm not aware of any features other than asm goto that
LRA provides that aren't supported reload.
Or perhaps rename the selector entirely to target_asm_goto?
jeff
