Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
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 {} {


  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