Update to use rocprofiler-sdk (#55)

- Renames the CMake option "ROCPROFSYS_USE_HIP" to "ROCPROFSYS_USE_ROCM"
- Remove the "ROCPROFSYS_USE_ROCM_SMI option. Controlled with the "ROCPROFSYS_USE_ROCM" option, instead.
   - Runtime configuration can still toggle ROCPROFSYS_USE_ROCM_SMI to disable the sampling.
- Rename ROCPROFSYS_HIP_VERSION macro to ROCPROFSYS_ROCM_VERSION and remove blocks for `ROCPROFSYS_ROCM_VERSION < 60000`
- Remove ROCPROFSYS_USE_ROCTRACER and ROCPROFSYS_USE_ROCPROFILER
- Update test cases
- Update docker files and workflows to install cmake 3.21, which is required for the rocprofiler-sdk findPackage script.
- Removed rocm-6.2 from workflows due to a rocprofiler-sdk API change.
This commit is contained in:
David Galiffi
2024-12-13 18:48:39 -05:00
committed by GitHub
parent d3725df816
commit 88aa2d3cbe
87 changed files with 3842 additions and 6261 deletions
+11 -25
View File
@@ -4,9 +4,7 @@
#
# -------------------------------------------------------------------------------------- #
set(ROCPROFSYS_ROCM_EVENTS_TEST
"GRBM_COUNT,GPUBusy,SQ_WAVES,SQ_INSTS_VALU,VALUInsts,TCC_HIT_sum,TA_TA_BUSY[0]:device=0,TA_TA_BUSY[11]:device=0"
)
set(ROCPROFSYS_ROCM_EVENTS_TEST "GRBM_COUNT,SQ_WAVES,SQ_INSTS_VALU,TA_TA_BUSY:device=0")
rocprofiler_systems_add_test(
NAME transpose
@@ -26,7 +24,8 @@ rocprofiler_systems_add_test(
args
-E
uniform_int_distribution
ENVIRONMENT "${_base_environment}")
ENVIRONMENT "${_base_environment}"
RUNTIME_TIMEOUT 480)
rocprofiler_systems_add_test(
SKIP_REWRITE SKIP_RUNTIME
@@ -36,9 +35,7 @@ rocprofiler_systems_add_test(
GPU ON
NUM_PROCS 1
RUN_ARGS 1 2 2
ENVIRONMENT
"${_base_environment};ROCPROFSYS_ROCTRACER_HSA_ACTIVITY=OFF;ROCPROFSYS_ROCTRACER_HSA_API=OFF"
)
ENVIRONMENT "${_base_environment}")
rocprofiler_systems_add_test(
SKIP_BASELINE SKIP_RUNTIME
@@ -64,7 +61,11 @@ rocprofiler_systems_add_test(
ENVIRONMENT "${_base_environment}"
REWRITE_FAIL_REGEX "0 instrumented loops in procedure transpose")
if(ROCPROFSYS_USE_ROCPROFILER)
if(ROCPROFSYS_USE_ROCM)
set(_ROCP_PASS_REGEX
"rocprof-device-0-GRBM_COUNT.txt(.*)rocprof-device-0-SQ_INSTS_VALU.txt(.*)rocprof-device-0-SQ_WAVES.txt(.*)rocprof-device-0-TA_TA_BUSY.txt(.*)"
)
rocprofiler_systems_add_test(
SKIP_BASELINE SKIP_RUNTIME
NAME transpose-rocprofiler
@@ -76,22 +77,7 @@ if(ROCPROFSYS_USE_ROCPROFILER)
REWRITE_ARGS -e -v 2 -E uniform_int_distribution
ENVIRONMENT
"${_base_environment};ROCPROFSYS_ROCM_EVENTS=${ROCPROFSYS_ROCM_EVENTS_TEST}"
REWRITE_RUN_PASS_REGEX
"rocprof-device-0-GRBM_COUNT.txt(.*)rocprof-device-0-GPUBusy.txt(.*)rocprof-device-0-SQ_WAVES.txt(.*)rocprof-device-0-SQ_INSTS_VALU.txt(.*)rocprof-device-0-VALUInsts.txt(.*)rocprof-device-0-TCC_HIT_sum.txt(.*)rocprof-device-0-TA_TA_BUSY_0.txt(.*)rocprof-device-0-TA_TA_BUSY_11.txt"
)
REWRITE_RUN_PASS_REGEX "${_ROCP_PASS_REGEX}"
SAMPLING_PASS_REGEX "${_ROCP_PASS_REGEX}")
rocprofiler_systems_add_test(
SKIP_BASELINE SKIP_RUNTIME
NAME transpose-rocprofiler-no-roctracer
TARGET transpose
LABELS "rocprofiler"
MPI ${TRANSPOSE_USE_MPI}
GPU ON
NUM_PROCS ${NUM_PROCS}
REWRITE_ARGS -e -v 2 -E uniform_int_distribution
ENVIRONMENT
"${_base_environment};ROCPROFSYS_USE_ROCTRACER=OFF;ROCPROFSYS_ROCM_EVENTS=${ROCPROFSYS_ROCM_EVENTS_TEST}"
REWRITE_RUN_PASS_REGEX
"rocprof-device-0-GRBM_COUNT.txt(.*)rocprof-device-0-GPUBusy.txt(.*)rocprof-device-0-SQ_WAVES.txt(.*)rocprof-device-0-SQ_INSTS_VALU.txt(.*)rocprof-device-0-VALUInsts.txt(.*)rocprof-device-0-TCC_HIT_sum.txt(.*)rocprof-device-0-TA_TA_BUSY_0.txt(.*)rocprof-device-0-TA_TA_BUSY_11.txt"
REWRITE_RUN_FAIL_REGEX "roctracer.txt|ROCPROFSYS_ABORT_FAIL_REGEX")
endif()
+8 -15
View File
@@ -226,7 +226,7 @@ endif()
# -------------------------------------------------------------------------------------- #
set(_VALID_GPU OFF)
if(ROCPROFSYS_USE_HIP AND (NOT DEFINED ROCPROFSYS_CI_GPU OR ROCPROFSYS_CI_GPU))
if(ROCPROFSYS_USE_ROCM AND (NOT DEFINED ROCPROFSYS_CI_GPU OR ROCPROFSYS_CI_GPU))
set(_VALID_GPU ON)
find_program(
ROCPROFSYS_ROCM_SMI_EXE
@@ -254,7 +254,7 @@ if(ROCPROFSYS_USE_HIP AND (NOT DEFINED ROCPROFSYS_CI_GPU OR ROCPROFSYS_CI_GPU))
endif()
endif()
set(LULESH_USE_GPU ${LULESH_USE_HIP})
set(LULESH_USE_GPU ${LULESH_USE_ROCM})
if(LULESH_USE_CUDA)
set(LULESH_USE_GPU ON)
endif()
@@ -314,8 +314,6 @@ ROCPROFSYS_SAMPLING_FREQ = 300
ROCPROFSYS_SAMPLING_DELAY = 0.05
ROCPROFSYS_SAMPLING_CPUS = 0-${NUM_SAMPLING_PROCS}
ROCPROFSYS_SAMPLING_GPUS = $env:HIP_VISIBLE_DEVICES
ROCPROFSYS_ROCTRACER_HSA_API = ON
ROCPROFSYS_ROCTRACER_HSA_ACTIVITY = ON
# test-specific values
${_FILE_CONTENTS}
@@ -430,18 +428,18 @@ function(ROCPROFILER_SYSTEMS_ADD_TEST)
if(TEST_GPU)
list(APPEND TEST_LABELS "gpu")
if(NOT "ROCPROFSYS_USE_ROCTRACER=OFF" IN_LIST TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "roctracer")
if(NOT "ROCPROFSYS_USE_ROCM=OFF" IN_LIST TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "rocm")
endif()
if(NOT "ROCPROFSYS_USE_ROCM_SMI=OFF" IN_LIST TEST_ENVIRONMENT)
if(NOT "ROCPROFSYS_USE_ROCM=OFF" IN_LIST TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "rocm-smi")
endif()
endif()
if("ROCPROFSYS_USE_ROCTRACER=ON" IN_LIST TEST_ENVIRONMENT AND NOT "roctracer" IN_LIST
TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "roctracer")
if("ROCPROFSYS_USE_ROCM=ON" IN_LIST TEST_ENVIRONMENT AND NOT "rocm" IN_LIST
TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "rocm")
endif()
if("ROCPROFSYS_USE_ROCM_SMI=ON" IN_LIST TEST_ENVIRONMENT AND NOT "rocm-smi" IN_LIST
@@ -449,11 +447,6 @@ function(ROCPROFILER_SYSTEMS_ADD_TEST)
list(APPEND TEST_LABELS "rocm-smi")
endif()
if("ROCPROFSYS_USE_ROCPROFILER=ON" IN_LIST TEST_ENVIRONMENT
AND NOT "rocprofiler" IN_LIST TEST_ENVIRONMENT)
list(APPEND TEST_LABELS "rocprofiler")
endif()
if(TARGET ${TEST_TARGET})
if(DEFINED TEST_MPI
AND ${TEST_MPI}