# Copyright 2026 Free Software Foundation, Inc.
# Copyright (C) 2026 Advanced Micro Devices, Inc. All rights reserved.
#
# 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/>.

# 'target teams distribute parallel for' spreads the loop across GPU
# lanes, with consecutive iterations on consecutive lanes.  Verify this
# by reading the loop index on two different lanes (the difference must
# match the lane distance) and checking a per-lane computed value.

load_lib rocm.exp

require allow_omp_offload_tests

# Reading OpenMP mapped/firstprivate data inside a target region
# (and single-stepping it) requires unoptimized device code; those
# values are optimized out at -O1 and above.  Skip under optimization.
require no_optimized_code

standard_testfile

set break_line [gdb_get_line_number "teams-line"]

if {[prepare_for_testing "failed to prepare" $testfile $srcfile \
	 {debug omp-rocm}] == -1} {
    return -1
}

with_rocm_gpu_lock {
    if {![runto_main]} {
	return
    }

    gdb_breakpoint "$srcfile:$break_line" allow-pending

    gdb_continue_to_breakpoint "teams breakpoint"

    # 'distribute parallel for' runs consecutive iterations on
    # consecutive lanes, so lane L sees loop index i = i0 + L.  Switch
    # to lane 0 explicitly (rather than relying on which lane the stop
    # selected), read i, then read it on lane 7, and check the two
    # lanes are 7 iterations apart.
    gdb_test "lane 0" "Switching to .*lane 0 .*" "switch to lane 0"
    set i0 [get_integer_valueof "i" -1 "loop index on lane 0"]
    gdb_test "lane 7" "Switching to .*lane 7 .*" "switch to lane 7"
    set i7 [get_integer_valueof "i" -1 "loop index on lane 7"]
    gdb_assert {$i7 - $i0 == 7} "lane 7 runs 7 iterations past lane 0"

    # a[i] == i and b[i] == N - i, so a[i] + b[i] == N (256) on any
    # lane (evaluated here on lane 7).
    gdb_test "print a\[i\] + b\[i\]" "= 256" "a\[i\] + b\[i\] == N"

    delete_breakpoints
    # The OpenMP runtime's host threads emit "[Thread ... exited]"
    # lines before the program exits, so allow extra output.
    gdb_continue_to_end "" "continue" 1
}