Hi!
Richard, maybe you have an opinion here, in particular about my
"SLP vectorizer" comment below? Please see
<87r1f2puss.fsf@euler.schwinge.homeip.net">http://mid.mail-archive.com/87r1f2puss.fsf@euler.schwinge.homeip.net>
for the full context.
On 2021-08-16T10:21:04+0200, Jakub Jelinek <ja...@redhat.com> wrote:
> On Mon, Aug 16, 2021 at 10:08:42AM +0200, Thomas Schwinge wrote:
>> /* Build COMPONENT_REF and set TREE_THIS_VOLATILE and TREE_READONLY on it
>> as appropriate. */
>>
>> tree
>> omp_build_component_ref (tree obj, tree field)
>> {
>> + tree field_type = TREE_TYPE (field);
>> + tree obj_type = TREE_TYPE (obj);
>> + if (!ADDR_SPACE_GENERIC_P (TYPE_ADDR_SPACE (obj_type)))
>> + field_type
>> + = build_qualified_type (field_type,
>> + KEEP_QUAL_ADDR_SPACE (TYPE_QUALS (obj_type)));
(For later reference: "Kwok's new code" here is to propagate to
'field_type' any non-generic address space of 'obj_type'.)
|> Concerning the current 'gcc/omp-low.c:omp_build_component_ref', for the
|> current set of offloading testcases, we never see a
|> '!ADDR_SPACE_GENERIC_P' there, so the address space handling doesn't seem
|> to be necessary there (but also won't do any harm: no-op).
>
> Are you sure this can't trigger?
> Say
> extern int __seg_fs a;
>
> void
> foo (void)
> {
> #pragma omp parallel private (a)
> a = 2;
> }
That test case doesn't run into 'omp_build_component_ref' at all,
but I'm attaching an altered and extended variant that does,
"Add 'libgomp.c/address-space-1.c'". OK to push to master branch?
In this case, 'omp_build_component_ref' called via host compilation
'pass_lower_omp', it's the 'field_type' that has 'address-space-1', not
'obj_type', so indeed Kwok's new code is a no-op:
(gdb) call debug_tree(field_type)
<pointer_type 0x7ffff7686b28
type <integer_type 0x7ffff7686498 int address-space-1 SI
size <integer_cst 0x7ffff7540f30 constant 32>
unit-size <integer_cst 0x7ffff7540f48 constant 4>
align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7686498 precision:32 min <integer_cst 0x7ffff7540ee8 -2147483648> max
<integer_cst 0x7ffff7540f00 2147483647>
pointer_to_this <pointer_type 0x7ffff7686b28>>
unsigned DI
size <integer_cst 0x7ffff7540cf0 type <integer_type 0x7ffff75590a8
bitsizetype> constant 64>
unit-size <integer_cst 0x7ffff7540d08 type <integer_type 0x7ffff7559000
sizetype> constant 8>
align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7686b28>
(gdb) call debug_tree(obj_type)
<record_type 0x7ffff7686bd0 .omp_data_t.0 readonly DI
size <integer_cst 0x7ffff7540cf0 type <integer_type 0x7ffff75590a8
bitsizetype> constant 64>
unit-size <integer_cst 0x7ffff7540d08 type <integer_type 0x7ffff7559000
sizetype> constant 8>
align:64 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7686bd0
fields <field_decl 0x7ffff7568428 a
type <pointer_type 0x7ffff7686b28 type <integer_type 0x7ffff7686498
int address-space-1>
unsigned DI size <integer_cst 0x7ffff7540cf0 64> unit-size
<integer_cst 0x7ffff7540d08 8>
align:64 warn_if_not_align:0 symtab:0 alias-set -1
canonical-type 0x7ffff7686b28>
unsigned DI /home/thomas/shared/gcc/omp/as.c:4:14 size <integer_cst
0x7ffff7540cf0 64> unit-size <integer_cst 0x7ffff7540d08 8>
align:64 warn_if_not_align:0 offset_align 128
offset <integer_cst 0x7ffff7540d20 constant 0>
bit-offset <integer_cst 0x7ffff7540d68 constant 0> context
<record_type 0x7ffff7686540 .omp_data_t.0>> reference_to_this <reference_type
0x7ffff7686c78>>
The case that Kwok's new code handles, however, is when 'obj_type' has a
non-generic address space, and then propagates that one to 'field_type'.
For a similar OpenACC example, 'omp_build_component_ref' called via GCN
offloading compilation 'pass_omp_oacc_neuter_broadcast', we've got
without Kwok's new code:
(gdb) call debug_tree(field_type)
<boolean_type 0x7ffff7550b28 bool public unsigned QI
size <integer_cst 0x7ffff754fa80 type <integer_type 0x7ffff75500a8
bitsizetype> constant 8>
unit-size <integer_cst 0x7ffff754fa98 type <integer_type 0x7ffff7550000
sizetype> constant 1>
align:8 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7550b28 precision:1 min <integer_cst 0x7ffff754fcd8 0> max <integer_cst
0x7ffff754fd08 1>>
(gdb) call debug_tree(obj_type)
<record_type 0x7ffff7631000 .oacc_ws_data_s.0 address-space-4 QI
size <integer_cst 0x7ffff754fa80 type <integer_type 0x7ffff75500a8
bitsizetype> constant 8>
unit-size <integer_cst 0x7ffff754fa98 type <integer_type 0x7ffff7550000
sizetype> constant 1>
align:8 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7631000
fields <field_decl 0x7ffff762e260 _52
type <boolean_type 0x7ffff7550b28 bool public unsigned QI size
<integer_cst 0x7ffff754fa80 8> unit-size <integer_cst 0x7ffff754fa98 1>
align:8 warn_if_not_align:0 symtab:0 alias-set -1
canonical-type 0x7ffff7550b28 precision:1 min <integer_cst 0x7ffff754fcd8 0>
max <integer_cst 0x7ffff754fd08 1>>
unsigned QI <built-in>:0:0 size <integer_cst 0x7ffff754fa80 8>
unit-size <integer_cst 0x7ffff754fa98 1>
align:8 warn_if_not_align:0 offset_align 64
offset <integer_cst 0x7ffff754f9c0 constant 0>
bit-offset <integer_cst 0x7ffff754fa08 constant 0> context
<record_type 0x7ffff7631000 .oacc_ws_data_s.0 address-space-4>>
pointer_to_this <pointer_type 0x7ffff7631498>>
..., and with Kwok's new code the 'address-space-4' of 'obj_type' is
propagated to 'field_type':
(gdb) call debug_tree(field_type)
<boolean_type 0x7ffff7631540 bool address-space-4 unsigned QI
size <integer_cst 0x7ffff754fa80 type <integer_type 0x7ffff75500a8
bitsizetype> constant 8>
unit-size <integer_cst 0x7ffff754fa98 type <integer_type 0x7ffff7550000
sizetype> constant 1>
align:8 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff7631540 precision:1 min <integer_cst 0x7ffff754fcd8 0> max <integer_cst
0x7ffff754fd08 1>>
I'm not familiar enough with these bits to tell whether Kwok's new code
is the right solution to this problem -- or if, for example, the problem
is rather in the SLP vectorizer, where the ICE seems to ultimately
emerge?
Without (ICEs later) vs. with (works) Kwok's new code, we see the
'a.xamdgcn-amdhsa.mkoffload.175t.slp1' dump change as follows (word-diff,
only additional '<address-space-4>', occasionally):
[...]
{+<address-space-4>+} vector(2) long int * vectp.58;
{+<address-space-4>+} vector(2) long int * vectp_.oacc_worker_o.57;
{+<address-space-4>+} vector(2) int * vectp.56;
{+<address-space-4>+} vector(2) int * vectp_.oacc_worker_o.55;
[...]
{+<address-space-4>+} long int * _104;
[...]
{+<address-space-4>+} long int * _108;
[...]
<address-space-4> void * _350;
[...]
_350 = __builtin_gcn_single_copy_start (&.oacc_worker_o.6);
[...]
MEM <{+<address-space-4>+} vector(2) long int> [(long int
*)&.oacc_worker_o.6] = _101;
_108 = &.oacc_worker_o.6._22 + 16;
MEM <{+<address-space-4>+} vector(2) long int> [(long int *)_108] = _100;
_104 = &.oacc_worker_o.6._22 + 32;
[...]
For example, with Kwok's new code, '_108' ('<address-space-4> long int *')
is cast into '(long int *)' -- presumably synthesized in the SLP
vectorizer? Is that correct or shouldn't that cast also include
'<address-space-4>'?
I see a similar issue has been fixed a while ago: r245772 (Git commit
c7d97b2846c5647a81548caa3264d77c0a595010) for PR79723
"Another case of dropped gs: prefix", changing
'gcc/tree-vect-stmts.c:get_vectype_for_scalar_type_and_size' as follows:
+ /* Re-attach the address-space qualifier if we canonicalized the scalar
+ type. */
+ if (TYPE_ADDR_SPACE (orig_scalar_type) != TYPE_ADDR_SPACE (vectype))
+ return build_qualified_type
+ (vectype, KEEP_QUAL_ADDR_SPACE (TYPE_QUALS
(orig_scalar_type)));
+
return vectype;
(It looks a bit like the address space handling is quite fragile in GCC's
'tree' types/interfaces? Do we have ideas about how to make that more
robust, less "bolt-on"?)
I did add a few 'assert's for non-generic address space to
'gcc/tree-vect*', but have not yet located where things may be going
wrong.
> I think keeping the qual addr space here is the wrong thing to do,
> it should keep the other quals and clear the address space instead,
> the whole struct is going to be in generic addres space, isn't it?
Correct for 'omp_build_component_ref' called via host compilation
'pass_lower_omp', but in the case of 'omp_build_component_ref' called via
GCN offloading compilation 'pass_omp_oacc_neuter_broadcast', 'obj_type'
has a non-generic address space.
However, regarding the former comment -- shouldn't we force generic
address space for all 'tree' types read in via LTO streaming for
offloading compilation? I assume that (in the general case) address
spaces are never compatible between host and offloading compilation?
For the attached "Add 'libgomp.c/address-space-1.c'", propagating the
'__seg_fs' address space across the offloading boundary (assuming I did
interpret the dumps correctly) doesn't seem to cause any problems, but
maybe it's problematic for other cases? (This is, however, a separate
issue from what I'm discussing here.)
>> + tree ret = build3 (COMPONENT_REF, field_type, obj, field, NULL);
>> + if (TREE_THIS_VOLATILE (field))
>> + TREE_THIS_VOLATILE (ret) |= 1;
>> + if (TREE_READONLY (field))
>> + TREE_READONLY (ret) |= 1;
>
> When touching these two, shouldn't it be better written as
> = 1; instead of |= 1; ? For a bitfield...
Yes, that was just copied from the original
'gcc/omp-general.c:omp_build_component_ref' -- but happy to simplify
that, of course.
Grüße
Thomas
-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634
München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas
Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht
München, HRB 106955
>From 6799543a4ff56360f3002931de1766a4448062a0 Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tho...@codesourcery.com>
Date: Thu, 19 Aug 2021 15:14:51 +0200
Subject: [PATCH] Add 'libgomp.c/address-space-1.c'
libgomp/
* testsuite/libgomp.c/address-space-1.c: New file.
Co-authored-by: Jakub Jelinek <ja...@redhat.com>
---
libgomp/testsuite/libgomp.c/address-space-1.c | 24 +++++++++++++++++++
1 file changed, 24 insertions(+)
create mode 100644 libgomp/testsuite/libgomp.c/address-space-1.c
diff --git a/libgomp/testsuite/libgomp.c/address-space-1.c b/libgomp/testsuite/libgomp.c/address-space-1.c
new file mode 100644
index 00000000000..90244db03b1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.c/address-space-1.c
@@ -0,0 +1,24 @@
+/* Verify OMP instances of variables with address space. */
+
+/* { dg-do run { target i?86-*-* x86_64-*-* } } */
+/* { dg-require-effective-target offload_device_nonshared_as } */
+
+#include <assert.h>
+
+int __seg_fs a;
+
+int
+main (void)
+{
+ // a = 123; // SIGSEGV
+ int b;
+#pragma omp target map(alloc: a) map(from: b)
+ {
+ a = 321; // no SIGSEGV (given 'offload_device_nonshared_as')
+ asm volatile ("" : : : "memory");
+ b = a;
+ }
+ assert (b == 321);
+
+ return 0;
+}
--
2.30.2
