[4970] Generate tuning inputs on GPU via splitmix64 device RNG

[itikhono]

This PR covers 1st part of the issue #4970.
Eliminates "input-gen + H2D (CPU waste)" part , GPU part (caused by bundle increase 1->10) remains

• During op/program tuning, candidate inputs were generated on the host (xorshf96 PRNG) and copied to the device for every candidate. This replaces that with a device kernel that fills tuning inputs directly on the GPU, removing the per-candidate host PRNG + H2D copy.
• New device::generate_random uses a counter-based splitmix64 RNG (seed + i * golden_ratio_step → splitmix64), so output is deterministic per seed and reproducible across candidates for fair comparison.
• time_program now allocates inputs with allocate_gpu and fills them via gpu_generate_random (recurses tuple sub-objects), while fill_map inputs keep the host-fill path.

Behavior parity with the old host path

• bool: handled by visit_all → normalize<bool> → 0/1, identical to the old special-case.
• fp4x2 (only non-computable type): visit_all would throw, so generation falls back to a raw byte fill — matching the old uint8 host behavior.
• tuples: same seed across sub-objects, same as the previous generate_argument.

Performance

• No FPS regression across the YOLO model family (within noise).
• Compile/tuning time improved up to ~6.6x at batch 64 on MI350, and ~10x at batch 32 on R9700 (measured together with reverting the bundle increase 1->10)

Test plan

• test_gpu_generate_random: seed determinism + range, half type, empty shape no-op, tuple fills every sub-buffer, non-computable (fp4x2) raw-byte fill — 5/5 pass.
• YOLO compile + inference sweep (fork vs develop).

Perf testing for YOLO-family models (MI350):

Used migraphx-driver perf, no actual diff detected, the results are quite noisy

different models, batch 4

Model	Fixed, img/s	Develop (before), img/s	Δ
yolov8m	1255.7	1242.5	+1.1%
yolov9m	1177.1	1111.2	+5.9%
yolov10m	1337.1	1303.4	+2.6%
yolo11m	1583.1	1564.2	+1.2%
yolo12m	1341.4	1350.1	−0.6%
yolo26m	1415.0	1407.2	+0.6%

[github-actions]

Thank you for your contribution! Since this is an external pull request, a maintainer must review PR and add the "ok-to-test" label if it is approved for testing.

[Copilot]

Pull request overview

This PR improves GPU compile/tuning throughput by generating candidate input buffers directly on the GPU (splitmix64 counter-based RNG) instead of generating on the host and copying H2D per candidate.

Changes:

• Added a GPU-side random-fill kernel (device::generate_random) and a host wrapper (gpu_generate_random) that recurses into tuple sub-objects.
• Updated time_program tuning path to allocate parameter buffers on GPU and fill them via gpu_generate_random (keeping fill_map on the host-fill path).
• Added a GPU unit test covering determinism, supported types, empty shapes, tuples, and non-computable types.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/gpu/generate_random.cpp	New GPU test coverage for deterministic RNG fill, tuples, and non-computable types.
src/targets/gpu/time_op.cpp	Switches tuning input creation to GPU allocation + GPU RNG fill (except fill_map).
src/targets/gpu/include/migraphx/gpu/hip.hpp	Exposes gpu_generate_random API.
src/targets/gpu/include/migraphx/gpu/device/generate_random.hpp	Declares new device-side RNG entrypoint.
src/targets/gpu/hip.cpp	Implements gpu_generate_random wrapper with tuple recursion.
src/targets/gpu/device/generate_random.cpp	Implements splitmix64-based device kernel to fill buffers.

[itikhono]

@pfultz2 @causten could you help to run [MIGraphX Performance Tests / security_gate (pull_request_target)] target? I think I don't have access/rights to trigger this job

[pfultz2]

could you help to run [MIGraphX Performance Tests / security_gate (pull_request_target)] target?

Dont worry about that one, I think it missing a label. Just fix the licensing.

[gh-app-migraphx-bot-pr-write]

Test	Batch	New Rate (5a3a3e)	Old Rate (f54ca3)	Diff	Status
torchvision-resnet50	64	669.58	3,151.54	-78.75%	🔴
torchvision-resnet50_fp16	64	1,766.59	6,672.57	-73.52%	🔴
torchvision-densenet121	32	557.68	2,669.98	-79.11%	🔴
torchvision-densenet121_fp16	32	644.55	4,553.52	-85.85%	🔴
torchvision-inceptionv3	32	416.47	1,794.69	-76.79%	🔴
torchvision-inceptionv3_fp16	32	501.54	2,710.69	-81.50%	🔴
cadene-inceptionv4	16	222.10	224.10	-0.89%	✅
cadene-resnext64x4	16	176.18	409.63	-56.99%	🔴
slim-mobilenet	64	4,428.97	8,288.90	-46.57%	🔴
slim-nasnetalarge	64	nan	229.24	nan	❌
slim-resnet50v2	64	776.52	3,329.55	-76.68%	🔴
bert-mrpc-onnx	8	140.05	1,170.13	-88.03%	🔴
bert-mrpc-tf	1	37.33	482.99	-92.27%	🔴
pytorch-examples-wlang-gru	1	210.63	335.86	-37.28%	🔴
pytorch-examples-wlang-lstm	1	220.15	464.03	-52.56%	🔴
torchvision-resnet50_1	1	25.53	771.72	-96.69%	🔴
cadene-dpn92_1	1	48.67	451.76	-89.23%	🔴
cadene-resnext101_1	1	44.39	363.79	-87.80%	🔴
onnx-taau-downsample	1	35.45	400.76	-91.15%	🔴
dlrm-criteoterabyte	1	6.16	32.67	-81.14%	🔴
dlrm-criteoterabyte_fp16	1	nan	52.58	nan	❌
agentmodel	1	688.21	9,679.70	-92.89%	🔴
unet_fp16	2	11.24	57.23	-80.35%	🔴
resnet50v1_fp16	1	40.31	972.36	-95.85%	🔴
resnet50v1_int8	1	300.05	960.27	-68.75%	🔴
bert_base_cased_fp16	64	323.80	1,102.92	-70.64%	🔴
bert_large_uncased_fp16	32	82.58	347.52	-76.24%	🔴
bert_large_fp16	1	4.49	205.28	-97.82%	🔴
distilgpt2_fp16	16	303.52	2,094.45	-85.51%	🔴
yolov5s	1	95.46	568.20	-83.20%	🔴
tinyllama	1	3.76	46.01	-91.83%	🔴
vicuna-fastchat	1	8.54	44.04	-80.60%	🔴
whisper-tiny-encoder	1	51.11	419.49	-87.82%	🔴
whisper-tiny-decoder	1	13.59	420.06	-96.76%	🔴
llama2_7b	1	2.06	20.46	-89.95%	🔴
qwen1.5-7b	1	11.39	23.67	-51.88%	🔴
phi3-3.8b	1	2.67	26.98	-90.09%	🔴
llama3-8b	1	2.11	21.83	-90.31%	🔴
whisper-large-encoder	1	2.50	10.32	-75.76%	🔴
whisper-large-decoder	1	3.09	107.05	-97.11%	🔴
mistral-7b	1	5.08	23.86	-78.72%	🔴
FLUX.1-schnell	1	43.74	761.61	-94.26%	🔴

Regressions detected 🔴

[gh-app-migraphx-bot-pr-write]

Test	Status	Result
bert-mrpc-onnx	✅	PASSED: MIGraphX meets tolerance
bert-mrpc-tf	❌	ERROR - check error output traceback Traceback (most recent call last): File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 377, in main() File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 313, in main import tensorflow as tf File "/usr/local/lib/python3.10/dist-packages/tensorflow/init.py", line 38, in from tensorflow.python.tools import module_util as _module_util File "/usr/local/lib/python3.10/dist-packages/tensorflow/python/init.py", line 36, in from tensorflow.python import pywrap_tensorflow as _pywrap_tensorflow File "/usr/local/lib/python3.10/dist-packages/tensorflow/python/pywrap_tensorflow.py", line 26, in self_check.preload_check() File "/usr/local/lib/python3.10/dist-packages/tensorflow/python/platform/self_check.py", line 63, in preload_check from tensorflow.python.platform import _pywrap_cpu_feature_guard ImportError: libamdhip64.so.6: cannot open shared object file: No such file or directory
pytorch-examples-wlang-gru	✅	PASSED: MIGraphX meets tolerance
pytorch-examples-wlang-lstm	✅	PASSED: MIGraphX meets tolerance
dlrm-criteoterabyte	✅	PASSED: MIGraphX meets tolerance
agentmodel	✅	PASSED: MIGraphX meets tolerance
unet	✅	PASSED: MIGraphX meets tolerance
resnet50v1	✅	PASSED: MIGraphX meets tolerance
bert_base_cased_fp16	✅	PASSED: MIGraphX meets tolerance
bert_large_uncased_fp16	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
bert_large	✅	PASSED: MIGraphX meets tolerance
yolov5s	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
tinyllama	✅	PASSED: MIGraphX meets tolerance
vicuna-fastchat	✅	PASSED: MIGraphX meets tolerance
whisper-tiny-encoder	✅	PASSED: MIGraphX meets tolerance
whisper-tiny-decoder	✅	PASSED: MIGraphX meets tolerance
distilgpt2_fp16	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
llama2_7b	✅	PASSED: MIGraphX meets tolerance
qwen1.5-7b	✅	PASSED: MIGraphX meets tolerance
phi3-3.8b	✅	PASSED: MIGraphX meets tolerance
llama3-8b	✅	PASSED: MIGraphX meets tolerance
whisper-large-encoder	❌	ERROR - check error output traceback 2026-06-18 18:02:07.178313 [WARN] [/data/src/onnx/onnx_parser.cpp:282] Model has unbound symbolic dimension(s): batch_size, encoder_sequence_length, feature_size. These default to 1 and may cause unexpected behavior. Try setting --dim-param @<name> <value> or --input-dim @<input> <dims> if program compilation fails. Traceback (most recent call last): File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 377, in main() File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 224, in main model = migraphx.parse_onnx(model_name, default_dim_value=batch) RuntimeError: /data/src/include/migraphx/op/convolution.hpp:113: normalize_compute_shape: CONVOLUTION: mismatched channel numbers: input channels (1) != weights channels (80) * group (1)
whisper-large-decoder	❌	ERROR - check error output traceback 2026-06-18 18:02:09.983074 [WARN] [/data/src/onnx/onnx_parser.cpp:282] Model has unbound symbolic dimension(s): batch_size, decoder_sequence_length, encoder_sequence_length / 2. These default to 1 and may cause unexpected behavior. Try setting --dim-param @<name> <value> or --input-dim @<input> <dims> if program compilation fails.
mistral-7b	✅	PASSED: MIGraphX meets tolerance
FLUX.1-schnell	✅	PASSED: MIGraphX meets tolerance

[itikhono]

@kahmed10 @shivadbhavsar could you help with reviewing this PR please?