From: Luis <luis.machado.foss@gmail.com>
To: Lancelot SIX <Lancelot.Six@amd.com>,
Luis Machado <luis.machado@amd.com>,
gdb-patches@sourceware.org
Subject: Re: [PATCH] gdb/testsuite: deduplicate --offload-arch targets in hcc_amdgpu_targets
Date: Fri, 3 Jul 2026 09:33:51 +0100 [thread overview]
Message-ID: <4dfd7ace-ade6-434f-8a63-f447f383dfb0@gmail.com> (raw)
In-Reply-To: <0cc2ad49-8ced-4f1b-a236-5f0efff00ddf@amd.com>
On 01/07/2026 17:00, Lancelot SIX wrote:
>
>
> On 29/06/2026 14:56, Luis Machado wrote:
>> On systems with multiple identical GPUs, find_amdgpu_devices returns one
>> entry per device, causing hcc_amdgpu_targets to emit redundant
>> --offload-arch flags when compiling HIP test programs (e.g. eight
>> --offload-arch=gfx942 flags on an 8-GPU MI300 system).
>>
>> Deduplicate the target list in hcc_amdgpu_targets, preserving order, so
>> each unique architecture appears exactly once regardless of how many
>> physical devices share it. The same deduplication is applied when the
>> list comes from the HCC_AMDGPU_TARGET environment variable. An array
>> is used for the seen-set to give O(1) membership tests.
>>
>> Add gdb.rocm/hcc-amdgpu-targets.exp to unit-test the deduplication
>> logic across both the env-var and device-enumeration code paths, without
>> requiring a GPU.
>
> Hi Luis,
>
> This LGTM. Thanks for doing this.
>
> Best,
> Lancelot.
>
> Approved-by: Lancelot Six <lancelot.six@amd.com> (amdgpu)
>
Thanks. Pushed now.
>> ---
>> gdb/testsuite/gdb.rocm/hcc-amdgpu-targets.exp | 80 +++++++++++++++++++
>> gdb/testsuite/lib/rocm.exp | 20 ++++-
>> 2 files changed, 97 insertions(+), 3 deletions(-)
>> create mode 100644 gdb/testsuite/gdb.rocm/hcc-amdgpu-targets.exp
>>
>> diff --git a/gdb/testsuite/gdb.rocm/hcc-amdgpu-targets.exp b/gdb/
>> testsuite/gdb.rocm/hcc-amdgpu-targets.exp
>> new file mode 100644
>> index 00000000000..8f04cca7981
>> --- /dev/null
>> +++ b/gdb/testsuite/gdb.rocm/hcc-amdgpu-targets.exp
>> @@ -0,0 +1,80 @@
>> +# Copyright (C) 2026 Free Software Foundation, Inc.
>> +
>> +# This file is part of GDB.
>> +
>> +# This program is free software; you can redistribute it and/or modify
>> +# it under the terms of the GNU General Public License as published by
>> +# the Free Software Foundation; either version 3 of the License, or
>> +# (at your option) any later version.
>> +
>> +# This program is distributed in the hope that it will be useful,
>> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
>> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
>> +# GNU General Public License for more details.
>> +
>> +# You should have received a copy of the GNU General Public License
>> +# along with this program. If not, see <http://www.gnu.org/licenses/>.
>> +
>> +# Unit tests for hcc_amdgpu_targets. These exercise the deduplication
>> +# logic across both the env-var and device-enumeration code paths, and
>> +# do not require a GPU or a running GDB instance.
>> +
>> +load_lib rocm.exp
>> +
>> +# Run BODY with find_amdgpu_devices stubbed to return DEVICES, then
>> +# restore the original proc.
>> +proc with_stub_devices {devices body} {
>> + rename find_amdgpu_devices __saved_find_amdgpu_devices
>> + # tclint-disable-next-line command-args
>> + proc find_amdgpu_devices {} [list return $devices]
>> + set code [catch {uplevel 1 $body} result]
>> + rename find_amdgpu_devices {}
>> + rename __saved_find_amdgpu_devices find_amdgpu_devices
>> + return -code $code $result
>> +}
>> +
>> +# Tests using find_amdgpu_devices (no HCC_AMDGPU_TARGET env var).
>> +
>> +save_vars { env(HCC_AMDGPU_TARGET) } {
>> + unset -nocomplain ::env(HCC_AMDGPU_TARGET)
>> +
>> + with_stub_devices {gfx942 gfx942 gfx942} {
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {gfx942}} "duplicates removed: 3x gfx942
>> produces single entry"
>> + }
>> +
>> + with_stub_devices {gfx942 gfx942 gfx1100 gfx1100 gfx942} {
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {gfx942 gfx1100}} "duplicates removed,
>> order preserved: mixed list"
>> + }
>> +
>> + with_stub_devices {gfx906 gfx90a} {
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {gfx906 gfx90a}} "no duplicates: distinct
>> devices unchanged"
>> + }
>> +
>> + with_stub_devices {} {
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {}} "empty device list returned as-is"
>> + }
>> +}
>> +
>> +# Tests using HCC_AMDGPU_TARGET env var.
>> +
>> +save_vars { env(HCC_AMDGPU_TARGET) } {
>> + set ::env(HCC_AMDGPU_TARGET) "gfx942,gfx942,gfx942"
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {gfx942}} "env var: duplicates removed"
>> +}
>> +
>> +save_vars { env(HCC_AMDGPU_TARGET) } {
>> + set ::env(HCC_AMDGPU_TARGET) "gfx906,gfx90a,gfx906"
>> + set result [hcc_amdgpu_targets]
>> + verbose -log "hcc_amdgpu_targets: got $result"
>> + gdb_assert {$result eq {gfx906 gfx90a}} "env var: order preserved
>> when deduplicating"
>> +}
>> diff --git a/gdb/testsuite/lib/rocm.exp b/gdb/testsuite/lib/rocm.exp
>> index 15ee22dad26..ca1ae3f1a8b 100644
>> --- a/gdb/testsuite/lib/rocm.exp
>> +++ b/gdb/testsuite/lib/rocm.exp
>> @@ -80,20 +80,34 @@ gdb_caching_proc find_amdgpu_devices {} {
>> return $hip_gpu_devices
>> }
>> -# Get the list of GPU targets to compile for.
>> +# Get the list of unique GPU targets to compile for.
>> #
>> # If HCC_AMDGPU_TARGET is set in the environment, use it.
>> # Otherwise, consider the devices available on the system.
>> +#
>> +# Duplicates are removed so that systems with multiple identical
>> +# GPUs do not produce redundant --offload-arch flags.
>> proc hcc_amdgpu_targets {} {
>> # First, look for HCC_AMDGPU_TARGET (same env var hipcc uses).
>> if {[info exists ::env(HCC_AMDGPU_TARGET)]} {
>> # We don't verify the contents of HCC_AMDGPU_TARGET.
>> # That's the toolchain's job.
>> - return [split $::env(HCC_AMDGPU_TARGET) ","]
>> + set targets [split $::env(HCC_AMDGPU_TARGET) ","]
>> + } else {
>> + set targets [find_amdgpu_devices]
>> }
>> - return [find_amdgpu_devices]
>> + # Remove duplicates while preserving order.
>> + array set seen {}
>> + set unique {}
>> + foreach t $targets {
>> + if {![info exists seen($t)]} {
>> + set seen($t) 1
>> + lappend unique $t
>> + }
>> + }
>> + return $unique
>> }
>> gdb_caching_proc allow_hipcc_tests {} {
>
prev parent reply other threads:[~2026-07-03 8:34 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-29 13:56 Luis Machado
2026-07-01 16:00 ` Lancelot SIX
2026-07-03 8:33 ` Luis [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4dfd7ace-ade6-434f-8a63-f447f383dfb0@gmail.com \
--to=luis.machado.foss@gmail.com \
--cc=Lancelot.Six@amd.com \
--cc=gdb-patches@sourceware.org \
--cc=luis.machado@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox