AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improve chip info logic. (#581)

Browse Source 
* Clean up unused functions.

* Fix number of XCDs for MI300X CPX (core partition).

* Add support for memory partition mode.

* Modify total_xcd to adapt to all gpu models.

* Run black and isort.

* Make gpu_arch regex more generic.

* Add error checking for compute partition mode num xcds.

* Set gpu_chip_id as optional.

* Fix get_gpu_model.

---------

Signed-off-by: xuchen-amd <xuchen@amd.com>
This commit is contained in:
xuchen-amd
2025-03-21 02:02:58 -04:00
committed by GitHub
parent b596098d14
commit 2e7f82aa13
6 changed files with 594 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/utils/utils.py
+32 -19
View File
@@ -42,6 +42,7 @@ from pathlib import Path as path
import pandas as pd
import config
from utils.mi_gpu_spec import get_mi300_num_xcds
rocprof_cmd = ""
rocprof_args = ""
@@ -686,7 +687,7 @@ def run_prof(
if new_env and not using_v3():
# flatten tcc for applicable mi300 input
f = path(workload_dir + "/out/pmc_1/results_" + fbase + ".csv")
xcds = total_xcds(mspec.gpu_model, mspec.compute_partition)
xcds = get_mi300_num_xcds(mspec.gpu_model, mspec.compute_partition)
df = flatten_tcc_info_across_xcds(f, xcds, int(mspec._l2_banks))
df.to_csv(f, index=False)
@@ -835,6 +836,7 @@ def replace_timestamps(workload_dir):
def gen_sysinfo(
workload_name, workload_dir, ip_blocks, app_cmd, skip_roof, roof_only, mspec, soc
):
console_debug("[gen_sysinfo]")
df = mspec.get_class_members()
# Append workload information to machine specs
@@ -1051,47 +1053,58 @@ def flatten_tcc_info_across_xcds(file, xcds, tcc_channel_per_xcd):
return df
def total_xcds(archname, compute_partition):
def total_xcds(gpu_model, compute_partition):
"""
Returns the number of xcds for a gpu model and compute_partition pair.
"""
# For mi300 chips, return result from mi_gpu_spec
result = get_mi300_num_xcds(gpu_model, compute_partition)
if result:
return result
# For other systems, use manual check
# check MI300 has a valid compute partition
mi300a_archs = ["mi300a_a0", "mi300a_a1"]
mi300x_archs = ["mi300x_a0", "mi300x_a1"]
mi308x_archs = ["mi308x"]
mi300a_model = ["mi300a_a0", "mi300a_a1"]
mi300x_model = ["mi300x_a0", "mi300x_a1"]
mi308x_model = ["mi308x"]
if (
archname.lower() in mi300a_archs + mi300x_archs + mi308x_archs
gpu_model.lower() in mi300a_model + mi300x_model + mi308x_model
and compute_partition == "NA"
):
console_error("Invalid compute partition found for {}".format(archname))
if archname.lower() not in mi300a_archs + mi300x_archs + mi308x_archs:
console_error("Invalid compute partition found for {}".format(gpu_model))
if gpu_model.lower() not in mi300a_model + mi300x_model + mi308x_model:
return 1
# from the whitepaper
# https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-3-white-paper.pdf
if compute_partition.lower() == "spx":
if archname.lower() in mi300a_archs:
if gpu_model.lower() in mi300a_model:
return 6
if archname.lower() in mi300x_archs:
if gpu_model.lower() in mi300x_model:
return 8
if archname.lower() in mi308x_archs:
if gpu_model.lower() in mi308x_model:
return 4
if compute_partition.lower() == "tpx":
if archname.lower() in mi300a_archs:
if gpu_model.lower() in mi300a_model:
return 2
if compute_partition.lower() == "dpx":
if archname.lower() in mi300x_archs:
if gpu_model.lower() in mi300x_model:
return 4
if archname.lower() in mi308x_archs:
if gpu_model.lower() in mi308x_model:
return 2
if compute_partition.lower() == "qpx":
if archname.lower() in mi300x_archs:
if gpu_model.lower() in mi300x_model:
return 2
if compute_partition.lower() == "cpx":
if archname.lower() in mi300x_archs:
return 2
if archname.lower() in mi308x_archs:
if gpu_model.lower() in mi300x_model:
return 1
if gpu_model.lower() in mi308x_model:
return 1
# TODO implement other archs here as needed
console_error(
"Unknown compute partition / arch found for {} / {}".format(
compute_partition, archname
compute_partition, gpu_model
)
)