From: Lancelot SIX <Lancelot.Six@amd.com>
To: Luis Machado <luis.machado@amd.com>, gdb-patches@sourceware.org
Subject: Re: [PATCH] gdb/testsuite: deduplicate --offload-arch targets in hcc_amdgpu_targets
Date: Wed, 1 Jul 2026 17:00:20 +0100 [thread overview]
Message-ID: <0cc2ad49-8ced-4f1b-a236-5f0efff00ddf@amd.com> (raw)
In-Reply-To: <20260629135613.2738058-1-luis.machado@amd.com>
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)
> ---
> 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 {} {
next prev parent reply other threads:[~2026-07-01 16:01 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 [this message]
2026-07-03 8:33 ` Luis
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=0cc2ad49-8ced-4f1b-a236-5f0efff00ddf@amd.com \
--to=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