Sync with upstream 'oracle/release/graal-vm/24.0' for July 2024 CPU

Merged cleanly except for: 1. oracle@7c93032 which conflicts with our earlier backport in graalvm#722. To deviate as little as possible from upstream I dropped our backport in favour of the upstream one. 2. The suite files were updated to match the Mandrel version scheme, i.e. the version is set to 24.0.2.0 instead of 24.0.2, and the Release value is set to False. # modified: substratevm/src/com.oracle.svm.core.genscavenge/src/com/oracle/svm/core/genscavenge/SerialGCOptions.java
zakkak · Jul 17, 2024 · 68b478d · 68b478d
2 parents c35c2d5 + 1f72d53
commit 68b478d
Show file tree

Hide file tree

Showing 135 changed files with 3,050 additions and 1,236 deletions.
diff --git a/ci/ci_common/bench-common.libsonnet b/ci/ci_common/bench-common.libsonnet
@@ -31,6 +31,17 @@
       else true
   },
 
+  # max number of threads to use for benchmarking in general
+  # the goal being to limit parallelism on very large servers which may not be respresentative of real-world scenarios
+  bench_max_threads:: {
+   restrict_threads:: 36
+  },
+
+  bench_no_thread_cap:: {
+    restrict_threads:: null,
+    should_use_hwloc:: false
+  },
+
   bench_hw:: {
     _bench_machine:: {
       targets+: ["bench"],
@@ -40,15 +51,16 @@
       numa_nodes:: [],
       is_numa:: std.length(self.numa_nodes) > 0,
       num_threads:: error "num_threads must bet set!",
+      hyperthreading:: true,
       threads_per_node:: if self.is_numa then self.num_threads / std.length(self.numa_nodes) else self.num_threads,
     },
 
-    x52:: common.linux_amd64 + self._bench_machine + {
-      machine_name:: "x52",
-      capabilities+: ["tmpfs25g"],
+    e3:: common.linux_amd64 + self._bench_machine + {
+      machine_name:: "e3",
+      capabilities: ["e3", "tmpfs25g", "linux", "amd64"],
       numa_nodes:: [0, 1],
-      default_numa_node:: 0,
-      num_threads:: 72
+      default_numa_node:: 1,
+      num_threads:: 256
     },
     e4_8_64:: common.linux_amd64 + self._bench_machine + {
       machine_name:: "e4_8_64",
@@ -67,31 +79,29 @@
     xgene3:: common.linux_aarch64 + self._bench_machine + {
       machine_name:: "xgene3",
       capabilities+: [],
-      num_threads:: 32
+      num_threads:: 32,
+      hyperthreading:: false
     },
     a12c:: common.linux_aarch64 + self._bench_machine + {
       machine_name:: "a12c",
       capabilities+: ["tmpfs25g"],
       numa_nodes:: [0, 1],
       default_numa_node:: 0,
-      num_threads:: 160
+      num_threads:: 160,
+      hyperthreading:: false
     }
   },
 
-  hwlocIfNuma(numa, cmd, node=0)::
-    if numa then
+  hwloc_cmd(cmd, num_threads, node, hyperthreading, max_threads_per_node)::
+    if num_threads == null then
       ["hwloc-bind", "--cpubind", "node:"+node, "--membind", "node:"+node, "--"] + cmd
     else
-      cmd,
-
-  parallelHwloc(cmd_node0, cmd_node1)::
-    // Returns a list of commands that will run cmd_nod0 on NUMA node 0
-    // concurrently with cmd_node1 on NUMA node 1 and then wait for both to complete.
-    [
-      $.hwlocIfNuma(true, cmd_node0, node=0) + ["&"],
-      $.hwlocIfNuma(true, cmd_node1, node=1) + ["&"],
-      ["wait"]
-    ],
+      local threads = if num_threads != null then num_threads else max_threads_per_node;
+      assert if hyperthreading then threads % 2 == 0 else true: "It is required to bind to an even number of threads on hyperthreaded machines. Got requested "+threads+" threads";
+      assert threads <= max_threads_per_node: "Benchmarking must run on a single NUMA node for stability reasons. Got requested "+threads+" threads but the machine has only "+max_threads_per_node+" threads per node";      local cores = if hyperthreading then "0-"+((threads/2)-1)+".pu:0-1" else "0-"+(threads-1)+".pu:0";
+      local cpu_bind = if hyperthreading then "node:"+node+".core:"+cores else "node:"+node+".core:"+cores+".pu:0";
+      ["hwloc-bind", "--cpubind", cpu_bind, "--membind", "node:"+node, "--"] + cmd
+  ,
 
   // building block used to generate fork builds
   many_forks_benchmarking:: common.build_base + {

diff --git a/common.json b/common.json
@@ -4,7 +4,7 @@
     "Jsonnet files should not include this file directly but use ci/common.jsonnet instead."
   ],
 
-  "mx_version": "7.4.1",
+  "mx_version": "7.4.1.1",
 
   "COMMENT.jdks": "When adding or removing JDKs keep in sync with JDKs in ci/common.jsonnet",
   "jdks": {
@@ -44,13 +44,13 @@
     "labsjdk-ee-21Debug": {"name": "labsjdk",   "version": "ee-21.0.1+11-jvmci-23.1-b26-debug", "platformspecific": true },
     "labsjdk-ee-21-llvm": {"name": "labsjdk",   "version": "ee-21.0.1+11-jvmci-23.1-b26-sulong", "platformspecific": true },
 
-    "oraclejdk-latest":       {"name": "jpg-jdk",   "version": "22",      "build_id": "jdk-22.0.1+8", "platformspecific": true, "extrabundles": ["static-libs"]},
-    "labsjdk-ce-latest":      {"name": "labsjdk",   "version": "ce-22.0.1+8-jvmci-b01", "platformspecific": true },
-    "labsjdk-ce-latestDebug": {"name": "labsjdk",   "version": "ce-22.0.1+8-jvmci-b01-debug", "platformspecific": true },
-    "labsjdk-ce-latest-llvm": {"name": "labsjdk",   "version": "ce-22.0.1+8-jvmci-b01-sulong", "platformspecific": true },
-    "labsjdk-ee-latest":      {"name": "labsjdk",   "version": "ee-22.0.1+8-jvmci-b01", "platformspecific": true },
-    "labsjdk-ee-latestDebug": {"name": "labsjdk",   "version": "ee-22.0.1+8-jvmci-b01-debug", "platformspecific": true },
-    "labsjdk-ee-latest-llvm": {"name": "labsjdk",   "version": "ee-22.0.1+8-jvmci-b01-sulong", "platformspecific": true }
+    "oraclejdk-latest":       {"name": "jpg-jdk",   "version": "22",      "build_id": "jdk-22.0.2+9", "platformspecific": true, "extrabundles": ["static-libs"]},
+    "labsjdk-ce-latest":      {"name": "labsjdk",   "version": "ce-22.0.2+9-jvmci-b01", "platformspecific": true },
+    "labsjdk-ce-latestDebug": {"name": "labsjdk",   "version": "ce-22.0.2+9-jvmci-b01-debug", "platformspecific": true },
+    "labsjdk-ce-latest-llvm": {"name": "labsjdk",   "version": "ce-22.0.2+9-jvmci-b01-sulong", "platformspecific": true },
+    "labsjdk-ee-latest":      {"name": "labsjdk",   "version": "ee-22.0.2+9-jvmci-b01", "platformspecific": true },
+    "labsjdk-ee-latestDebug": {"name": "labsjdk",   "version": "ee-22.0.2+9-jvmci-b01-debug", "platformspecific": true },
+    "labsjdk-ee-latest-llvm": {"name": "labsjdk",   "version": "ee-22.0.2+9-jvmci-b01-sulong", "platformspecific": true }
   },
 
   "eclipse": {

diff --git a/compiler/ci/ci_common/benchmark-builders.jsonnet b/compiler/ci/ci_common/benchmark-builders.jsonnet
@@ -11,41 +11,35 @@
 
   local main_builds = std.flattenArrays([
     [
-    c.daily + c.opt_post_merge + hw.x52 + jdk + cc.libgraal + bench.dacapo + PR_bench_libgraal,
-    c.weekly                   + hw.x52 + jdk + cc.libgraal + bench.dacapo_size_variants,
-    c.daily + c.opt_post_merge + hw.x52 + jdk + cc.libgraal + bench.scala_dacapo + PR_bench_libgraal,
-    c.weekly                   + hw.x52 + jdk + cc.libgraal + bench.scala_dacapo_size_variants,
-    c.daily + c.opt_post_merge + hw.x52 + jdk + cc.libgraal + bench.renaissance + PR_bench_libgraal,
-    c.daily + c.opt_post_merge + hw.x52 + jdk + cc.libgraal + bench.specjvm2008 + PR_bench_libgraal,
-    c.weekly                   + hw.x52 + jdk + cc.libgraal + bench.specjbb2015,
-    c.daily + c.opt_post_merge + hw.x52 + jdk + cc.libgraal + bench.awfy + PR_bench_libgraal,
-    c.daily                    + hw.x52 + jdk + cc.libgraal + bench.microservice_benchmarks,
-    c.daily                    + hw.x52 + jdk + cc.libgraal + bench.micros_graal_whitebox,
-    c.daily                    + hw.x52 + jdk + cc.libgraal + bench.micros_graal_dist,
-    c.daily                    + hw.x52 + jdk + cc.libgraal + bench.micros_misc_graal_dist,
-    c.daily                    + hw.x52 + jdk + cc.libgraal + bench.micros_shootout_graal_dist,
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.awfy + {job_prefix:: "bench-e4vm-compiler"},
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.dacapo + {job_prefix:: "bench-e4vm-compiler"},
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.scala_dacapo + {job_prefix:: "bench-e4vm-compiler"},
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.renaissance + {job_prefix:: "bench-e4vm-compiler"},
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.specjvm2008 + {job_prefix:: "bench-e4vm-compiler"},
-    c.daily                    + hw.e4_8_64 + jdk + cc.libgraal + bench.microservice_benchmarks + {job_prefix:: "bench-e4vm-compiler"},
+    c.daily + c.opt_post_merge + hw.e3 + jdk + cc.libgraal + bench.dacapo + PR_bench_libgraal,
+    c.daily + c.opt_post_merge + hw.e3 + jdk + cc.libgraal + bench.scala_dacapo + PR_bench_libgraal,
+    c.daily + c.opt_post_merge + hw.e3 + jdk + cc.libgraal + bench.renaissance + PR_bench_libgraal,
+    c.daily + c.opt_post_merge + hw.e3 + jdk + cc.libgraal + bench.specjvm2008 + PR_bench_libgraal,
+    c.on_demand                + hw.e3 + jdk + cc.libgraal + bench.dacapo_size_variants,
+    c.on_demand                + hw.e3 + jdk + cc.libgraal + bench.scala_dacapo_size_variants,
+    c.monthly                  + hw.e3 + jdk + cc.libgraal + bench.specjbb2015,
+    c.daily + c.opt_post_merge + hw.e3 + jdk + cc.libgraal + bench.awfy + PR_bench_libgraal,
+    c.daily                    + hw.e3 + jdk + cc.libgraal + bench.microservice_benchmarks,
+    c.weekly                   + hw.e3 + jdk + cc.libgraal + bench.micros_graal_whitebox,
+    c.weekly                   + hw.e3 + jdk + cc.libgraal + bench.micros_graal_dist,
+    c.weekly                   + hw.e3 + jdk + cc.libgraal + bench.micros_misc_graal_dist,
+    c.weekly                   + hw.e3 + jdk + cc.libgraal + bench.micros_shootout_graal_dist,
     ]
   for jdk in cc.product_jdks
   ]),
 
   local profiling_builds = std.flattenArrays([
     [
-    c.weekly + hw.x52 + jdk + cc.libgraal + suite + cc.enable_profiling   + { job_prefix:: "bench-compiler-profiling" },
-    c.weekly + hw.x52 + jdk + cc.libgraal + suite + cc.footprint_tracking + { job_prefix:: "bench-compiler-footprint" }
+    c.monthly + hw.e3 + jdk + cc.libgraal + suite + cc.enable_profiling   + { job_prefix:: "bench-compiler-profiling" },
+    c.monthly + hw.e3 + jdk + cc.libgraal + suite + cc.footprint_tracking + { job_prefix:: "bench-compiler-footprint" }
     ]
   for jdk in cc.product_jdks
   for suite in bench.groups.main_suites
   ]),
 
   local weekly_amd64_forks_builds = std.flattenArrays([
-    bc.generate_fork_builds(c.weekly  + hw.x52  + jdk + cc.libgraal + suite, subdir='compiler') +
-    bc.generate_fork_builds(c.monthly + hw.x52  + jdk + cc.jargraal + suite, subdir='compiler')
+    bc.generate_fork_builds(c.weekly  + hw.e3  + jdk + cc.libgraal + suite, subdir='compiler') +
+    bc.generate_fork_builds(c.monthly + hw.e3  + jdk + cc.jargraal + suite, subdir='compiler')
   for jdk in cc.product_jdks
   for suite in bench.groups.weekly_forks_suites
   ]),
@@ -70,7 +64,7 @@
   ],
 
   local zgc_builds = [
-    c.weekly + hw.x52 + jdk + cc.libgraal + cc.zgc_mode + suite,
+    c.weekly + hw.e3 + jdk + cc.libgraal + cc.zgc_mode + suite,
   for jdk in cc.product_jdks
   for suite in bench.groups.main_suites + [bench.specjbb2015]
   ],
@@ -83,13 +77,13 @@
   ],
 
   local no_tiered_builds = [
-    c.weekly + hw.x52 + jdk + cc.libgraal + cc.no_tiered_comp + suite,
+    c.monthly + hw.e3 + jdk + cc.libgraal + cc.no_tiered_comp + suite,
   for jdk in cc.product_jdks
   for suite in bench.groups.main_suites
   ],
 
   local no_profile_info_builds = [
-    c.weekly + hw.x52 + jdk + cc.libgraal + cc.no_profile_info + suite,
+    c.monthly + hw.e3 + jdk + cc.libgraal + cc.no_profile_info + suite,
   for jdk in cc.product_jdks
   for suite in bench.groups.main_suites
   ],

diff --git a/compiler/ci/ci_common/benchmark-suites.libsonnet b/compiler/ci/ci_common/benchmark-suites.libsonnet
@@ -24,7 +24,7 @@
 
   // suite definitions
   // *****************
-  awfy: cc.compiler_benchmark + c.heap.small + {
+  awfy: cc.compiler_benchmark + c.heap.small + bc.bench_max_threads + {
     suite:: "awfy",
     run+: [
       self.benchmark_cmd + ["awfy:*", "--"] + self.extra_vm_args
@@ -36,7 +36,7 @@
     max_jdk_version:: null
   },
 
-  dacapo: cc.compiler_benchmark + c.heap.default + {
+  dacapo: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "dacapo",
     run+: [
       self.benchmark_cmd + ["dacapo:*", "--"] + self.extra_vm_args
@@ -48,7 +48,7 @@
     max_jdk_version:: null
   },
 
-  dacapo_size_variants: cc.compiler_benchmark + c.heap.default + {
+  dacapo_size_variants: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "dacapo-size-variants",
     run+: [
       self.benchmark_cmd + ["dacapo-small:*", "--"] + self.extra_vm_args,
@@ -64,7 +64,7 @@
     max_jdk_version:: null
   },
 
-  scala_dacapo: cc.compiler_benchmark + c.heap.default + {
+  scala_dacapo: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "scala-dacapo",
     run+: [
       self.benchmark_cmd + ["scala-dacapo:*", "--"] + self.extra_vm_args
@@ -76,7 +76,7 @@
     max_jdk_version:: null
   },
 
-  scala_dacapo_size_variants: cc.compiler_benchmark + c.heap.default + {
+  scala_dacapo_size_variants: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "scala-dacapo-size-variants",
     run+: [
       self.benchmark_cmd + ["scala-dacapo-tiny:*", "--"] + self.extra_vm_args,
@@ -97,7 +97,7 @@
     max_jdk_version:: null
   },
 
-  renaissance_template(suite_version=null, suite_name="renaissance", max_jdk_version=null):: cc.compiler_benchmark + c.heap.default + {
+  renaissance_template(suite_version=null, suite_name="renaissance", max_jdk_version=null):: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: suite_name,
     local suite_version_args = if suite_version != null then ["--bench-suite-version=" + suite_version] else [],
     run+: [
@@ -110,12 +110,9 @@
     max_jdk_version:: max_jdk_version
   },
 
-  renaissance: self.renaissance_template() + {
-    # [JDK-8303076] [GR-44499] requires extra stack size for C1
-    extra_vm_args+:: if self.platform == "c1" then ["-Xss1090K"] else []
-  },
+  renaissance: self.renaissance_template(),
 
-  specjbb2015: cc.compiler_benchmark + c.heap.large_with_large_young_gen + {
+  specjbb2015: cc.compiler_benchmark + c.heap.large_with_large_young_gen + bc.bench_max_threads + {
     suite:: "specjbb2015",
     downloads+: {
       "SPECJBB2015": { name: "specjbb2015", version: "1.03" }
@@ -130,7 +127,7 @@
     max_jdk_version:: null
   },
 
-  specjvm2008: cc.compiler_benchmark + c.heap.default + {
+  specjvm2008: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "specjvm2008",
     downloads+: {
       "SPECJVM2008": { name: "specjvm2008", version: "1.01" }
@@ -146,7 +143,7 @@
   },
 
   // Microservice benchmarks
-  microservice_benchmarks: cc.compiler_benchmark + {
+  microservice_benchmarks: cc.compiler_benchmark + bc.bench_no_thread_cap + {  # no thread cap here since hwloc is handled at the mx level for microservices
     suite:: "microservices",
     packages+: {
       "pip:psutil": "==5.8.0"
@@ -159,42 +156,26 @@
     local hwlocBind_16C_32T = ["--hwloc-bind=--cpubind node:0.core:0-15.pu:0-1 --membind node:0"],
     run+: [
       # shopcart-wrk
-      self.benchmark_cmd + ["shopcart-wrk:mixed-tiny"]                   + hwlocBind_1C_1T   + ["--"] + self.extra_vm_args + ["-Xms32m",   "-Xmx112m",  "-XX:ActiveProcessorCount=1",  "-XX:MaxDirectMemorySize=256m"],
-      bench_upload,
-      self.benchmark_cmd + ["shopcart-wrk:mixed-small"]                  + hwlocBind_2C_2T   + ["--"] + self.extra_vm_args + ["-Xms64m",   "-Xmx224m",  "-XX:ActiveProcessorCount=2",  "-XX:MaxDirectMemorySize=512m"],
-      bench_upload,
-      self.benchmark_cmd + ["shopcart-wrk:mixed-medium"]                 + hwlocBind_4C_4T   + ["--"] + self.extra_vm_args + ["-Xms128m",  "-Xmx512m",  "-XX:ActiveProcessorCount=4",  "-XX:MaxDirectMemorySize=1024m"],
-      bench_upload,
       self.benchmark_cmd + ["shopcart-wrk:mixed-large"]                  + hwlocBind_16C_16T + ["--"] + self.extra_vm_args + ["-Xms512m",  "-Xmx3072m", "-XX:ActiveProcessorCount=16", "-XX:MaxDirectMemorySize=4096m"],
       bench_upload,
-      self.benchmark_cmd + ["shopcart-wrk:mixed-huge"]                   + hwlocBind_16C_32T + ["--"] + self.extra_vm_args + ["-Xms1024m", "-Xmx8192m", "-XX:ActiveProcessorCount=32", "-XX:MaxDirectMemorySize=8192m"],
-      bench_upload,
 
       # petclinic-wrk
-      self.benchmark_cmd + ["petclinic-wrk:mixed-tiny"]                  + hwlocBind_1C_1T   + ["--"] + self.extra_vm_args + ["-Xms32m",   "-Xmx100m",  "-XX:ActiveProcessorCount=1"],
-      bench_upload,
-      self.benchmark_cmd + ["petclinic-wrk:mixed-small"]                 + hwlocBind_2C_2T   + ["--"] + self.extra_vm_args + ["-Xms40m",   "-Xmx144m",  "-XX:ActiveProcessorCount=2"],
-      bench_upload,
-      self.benchmark_cmd + ["petclinic-wrk:mixed-medium"]                + hwlocBind_4C_4T   + ["--"] + self.extra_vm_args + ["-Xms80m",   "-Xmx256m",  "-XX:ActiveProcessorCount=4"],
-      bench_upload,
       self.benchmark_cmd + ["petclinic-wrk:mixed-large"]                 + hwlocBind_16C_16T + ["--"] + self.extra_vm_args + ["-Xms128m",  "-Xmx512m", "-XX:ActiveProcessorCount=16"],
       bench_upload,
-      self.benchmark_cmd + ["petclinic-wrk:mixed-huge"]                  + hwlocBind_16C_32T + ["--"] + self.extra_vm_args + ["-Xms640m",  "-Xmx3072m", "-XX:ActiveProcessorCount=32"],
-      bench_upload,
 
       # helloworld-wrk
       self.benchmark_cmd + ["micronaut-helloworld-wrk:helloworld"]       + hwlocBind_1C_1T   + ["--"] + self.extra_vm_args + ["-Xms8m",    "-Xmx64m",   "-XX:ActiveProcessorCount=1", "-XX:MaxDirectMemorySize=256m"],
       bench_upload,
       self.benchmark_cmd + ["spring-helloworld-wrk:helloworld"]          + hwlocBind_1C_1T   + ["--"] + self.extra_vm_args + ["-Xms8m",    "-Xmx64m",   "-XX:ActiveProcessorCount=1", "-XX:MaxDirectMemorySize=256m"],
       bench_upload
     ],
-    timelimit: "7:00:00",
+    timelimit: "4:00:00",
     min_jdk_version:: 11,
     max_jdk_version:: null
   },
 
   // JMH microbenchmarks
-  micros_graal_whitebox: cc.compiler_benchmark + c.heap.default + {
+  micros_graal_whitebox: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "micros-graal-whitebox",
     run+: [
       self.benchmark_cmd + ["jmh-whitebox:*", "--"] + self.extra_vm_args + ["--", "jdk.graal.compiler"]
@@ -204,7 +185,7 @@
     max_jdk_version:: null
   },
 
-  micros_graal_dist: cc.compiler_benchmark + c.heap.default + {
+  micros_graal_dist: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "micros-graal-dist",
     run+: [
       self.benchmark_cmd + ["jmh-dist:GRAAL_COMPILER_MICRO_BENCHMARKS", "--"] + self.extra_vm_args
@@ -214,7 +195,7 @@
     max_jdk_version:: null
   },
 
-  micros_misc_graal_dist: cc.compiler_benchmark + c.heap.default + {
+  micros_misc_graal_dist: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "micros-misc-graal-dist",
     run+: [
       self.benchmark_cmd + ["jmh-dist:GRAAL_BENCH_MISC", "--"] + self.extra_vm_args
@@ -224,7 +205,7 @@
     max_jdk_version:: null
   },
 
-  micros_shootout_graal_dist: cc.compiler_benchmark + c.heap.default {
+  micros_shootout_graal_dist: cc.compiler_benchmark + c.heap.default + bc.bench_max_threads + {
     suite:: "micros-shootout-graal-dist",
     run+: [
       self.benchmark_cmd + ["jmh-dist:GRAAL_BENCH_SHOOTOUT", "--"] + self.extra_vm_args

diff --git a/compiler/ci/ci_common/compiler-common.libsonnet b/compiler/ci/ci_common/compiler-common.libsonnet
@@ -71,7 +71,8 @@
       "${BENCH_RESULTS_FILE_PATH}",
       "--machine-name=${MACHINE_NAME}"] +
       (if std.objectHasAll(self.environment, 'MX_TRACKER') then ["--tracker=" + self.environment['MX_TRACKER']] else ["--tracker=rss"]),
-    benchmark_cmd:: bench_common.hwlocIfNuma(self.should_use_hwloc, self.plain_benchmark_cmd, node=self.default_numa_node),
+    restrict_threads:: null,  # can be overridden to restrict the benchmark to the given number of threads. If null, will use one full NUMA node
+    benchmark_cmd:: if self.should_use_hwloc then bench_common.hwloc_cmd(self.plain_benchmark_cmd, self.restrict_threads, self.default_numa_node, self.hyperthreading, self.threads_per_node) else self.plain_benchmark_cmd,
     min_heap_size:: if std.objectHasAll(self.environment, 'XMS') then ["-Xms${XMS}"] else [],
     max_heap_size:: if std.objectHasAll(self.environment, 'XMX') then ["-Xmx${XMX}"] else [],
     extra_vm_args::