From 8980c2051531d4825f3a1392989cbe9c3c32b1e7 Mon Sep 17 00:00:00 2001
From: Simon Marchi <simon.marchi@efficios.com>
Date: Thu, 11 Jun 2026 16:01:27 -0400
Subject: [PATCH] gdb.rocm/names.exp: build test program with optimizations
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

When running on gfx90a, the test program for gdb.rocm/names.exp hangs
when running all kernels:

    $ GPU_MAX_HW_QUEUES=32 testsuite/outputs/gdb.rocm/names/names all
    *hangs*

When debugging with GDB, I see 4 running waves waiting in
function wait_all_kernels, and kernels_seen is 4, meaning that the other
kernels didn't start.  I noticed that if I compiled the test program at
-O2, it ran fine.  Claude identified this as a resource issue:

    ● I extracted the kernel metadata from both binaries to compare. Here's the cause:

      names-0 (-O0), per kernel:
      - vgpr_count: 43, sgpr_count: 42
      - private_segment_fixed_size: 72 (or 88/104)
      - uses_dynamic_stack: true

      names-2 (-O2), per kernel:
      - vgpr_count: 3, sgpr_count: 9
      - private_segment_fixed_size: 0
      - uses_dynamic_stack: false

Presumably, at -O0, each wave (or queue?) consumes more resources and we
can't get all 9 on the hardware at the same time.  But I am not
knowledgeable enough to say if that's correct.

Fix it by building with -O2 so all kernels can be in flight
concurrently.

At -O2, the argument values appear to be optimized out, so update the
expected frame function string.  I don't think that it matters for this
test that we are losing the actual argument values.  I just hope that
they will indeed always be optimized out so that we won't get FAILs
because of that (worst case we can expect .* for the argument values).

Change-Id: Ida07faf8b07bf9e0b21ce16bb6af7b019f01588b
---
 gdb/testsuite/gdb.rocm/names.exp | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/gdb/testsuite/gdb.rocm/names.exp b/gdb/testsuite/gdb.rocm/names.exp
index b18c1ac2f53..44b05438817 100644
--- a/gdb/testsuite/gdb.rocm/names.exp
+++ b/gdb/testsuite/gdb.rocm/names.exp
@@ -21,7 +21,10 @@ require allow_hipcc_tests
 
 standard_testfile .cpp
 
-if {[build_executable "failed to prepare" $testfile $srcfile {debug hip}]} {
+# Optimize: at -O0 the kernels use scratch memory, which limits GPU
+# concurrency below what wait_all_kernels needs and causes a deadlock.
+if {[build_executable "failed to prepare" $testfile $srcfile \
+	 {debug hip optimize=-O2}]} {
     return
 }
 
@@ -69,7 +72,7 @@ set all_kernels_info {
     {long_kernel_name_0123456789                  long_kernel-0245 {long_kernel_name_0123456789 ()}}
     {long_kernel_name_abcdefghijklmnopqrstuvwxyz  long_kernel-479c {long_kernel_name_abcdefghijklmnopqrstuvwxyz ()}}
     {templ_non_type                               templ_non_type   {templ_non_type<(E)0> ()}}
-    {templ_type                                   templ_type       {templ_type<int, long> (args=1, args=2)}}
+    {templ_type                                   templ_type       {templ_type<int, long> (args=<optimized out>, args=<optimized out>)}}
 }
 
 # By default, the runtime allows 4 hardware queues max, which limits