[do not merge] rhealstone benchmark #1240

lukileczo · 2024-11-26T14:17:17Z

Description

Motivation and Context

Types of changes

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)
Chore (refactoring, style fixes, git/CI config, submodule management, no code logic changes)

How Has This Been Tested?

Already covered by automatic testing.
New test added: (add PR link here).
Tested by hand on: (list targets here).

Checklist:

My change requires a change to the documentation.
I have updated the documentation accordingly.
I have added tests to cover my changes.
All new and existing linter checks and tests passed.
My changes generate no new compilation warnings for any of the targets.

Special treatment

This PR needs additional PRs to work (list the PRs, preferably in merge-order).
I will merge this PR by myself when appropriate.

JIRA: RTOS-883

Timer count can be read from a configuration register. JIRA: RTOS-883

github-actions · 2024-11-26T14:17:37Z

_projects/sparcv8leon-gr740-mini/board_config.h

+#define FTMCTRL_BASE 0xff903000
+
+
+#define RAM_ADDR 0x07000000


[clang-format-pr] _{reported by reviewdog 🐶}
suggested fix

Suggested change

#define RAM_ADDR 0x07000000

#define RAM_ADDR 0x07000000

github-actions · 2024-11-26T14:18:26Z

_targets/sparcv8leon/gr740/build.project

+
+	ROOTFS="$PREFIX_BOOT/rootfs.jffs2"
+
+	local erase_sz=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0._meta.block_size }}')


⚠️ [shellcheck] _{reported by reviewdog 🐶}
Declare and assign separately to avoid masking return values. SC2155

github-actions · 2024-11-26T14:18:26Z

_targets/sparcv8leon/gr740/build.project

+		mv "$ROOTFS.tmp" "$ROOTFS"
+	fi
+
+	local FS_OFFS=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0.rootfs.offs }}')


⚠️ [shellcheck] _{reported by reviewdog 🐶}
Declare and assign separately to avoid masking return values. SC2155

github-actions · 2024-11-26T14:18:26Z

_targets/sparcv8leon/gr740/build.project

+	fi
+
+	local FS_OFFS=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0.rootfs.offs }}')
+	local FS_SZ=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0.rootfs.size }}')


⚠️ [shellcheck] _{reported by reviewdog 🐶}
Declare and assign separately to avoid masking return values. SC2155

github-actions · 2024-11-26T14:18:26Z

_targets/sparcv8leon/gr712rc/build.project

@@ -63,10 +61,11 @@ b_image_target() {

 	ROOTFS="$PREFIX_BOOT/rootfs.jffs2"

+	local erase_sz=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0._meta.block_size }}')


⚠️ [shellcheck] _{reported by reviewdog 🐶}
Declare and assign separately to avoid masking return values. SC2155

github-actions · 2024-11-26T14:18:26Z

_projects/sparcv8leon-gr740-mini/build.project

+[ "${BASH_SOURCE[0]}" -ef "$0" ] && echo "You should source this script, not execute it!" && exit 1
+
+
+FLASH_SZ=$((0x8000000))


⚠️ [shellcheck] _{reported by reviewdog 🐶}
FLASH_SZ appears unused. Verify use (or export if used externally). SC2034

github-actions · 2024-11-26T14:18:26Z

_projects/sparcv8leon-gr740-mini/build.project

+
+
+FLASH_SZ=$((0x8000000))
+ROOTFS_SZ=$((0x800000))


⚠️ [shellcheck] _{reported by reviewdog 🐶}
ROOTFS_SZ appears unused. Verify use (or export if used externally). SC2034

agkaminski

General remark - watch out for atomics, these might have some heavy overheads (including unexpected mutex in some cases). Perhaps it's ok, but we can't be sure without inspecting resulting binary

agkaminski · 2024-11-26T16:35:34Z

_user/rhealstone/deadlock-break/main.c

+
+bool deadBrk;
+atomic_int count = 0;
+atomic_bool done = false;


I think this should be volatile, but I'm not sure what exact effect it has on atomic_s

agkaminski · 2024-11-26T16:38:32Z

_user/rhealstone/deadlock-break/main.c

+	benchStart = getCntr();
+
+	priority(4);
+
+	usleep(0);
+
+	threadJoin(tid1, 0);
+
+	benchEnd = getCntr();


We introduce 2 syscalls overhead here that skews the result. Perhaps we could measure the time in task1 instead?

agkaminski · 2024-11-26T16:41:24Z

_user/rhealstone/irq-latency/main.c

+
+#define BENCHMARKS 10000
+
+atomic_uint_least64_t benchEnd = 0;


Atomic might give unwanted overhead, especially 64 bit one - we trigger the interrupt via SW, so should be fine to use plain volatile

agkaminski · 2024-11-26T16:43:40Z

_user/rhealstone/irq-latency/main.c

+
+int irqHandler(unsigned int n, void *arg)
+{
+	if (n == IRQ_UNUSED) {


This if is not strictly needed - we provide n argument as a way for one handler to handle more than one interrupt, check on n is not enforced

agkaminski · 2024-11-26T16:45:28Z

_user/rhealstone/irq-latency/main.c

+{
+	BENCH_NAME("Interrupt latency");
+
+	uint32_t *irqCtrl = mmap(NULL, _PAGE_SIZE, PROT_READ | PROT_WRITE, MAP_DEVICE | MAP_PHYSMEM | MAP_ANONYMOUS, -1, (uintptr_t)INT_CTRL_BASE);


Should be volatile, actually it's kinda weird that the access to the irqCtrl hasn't been optimized out. Are we sure we have -O2?

agkaminski · 2024-11-26T16:48:24Z

_user/rhealstone/msg-latency/main.c

+	uint64_t loopOverhead = getCntr();
+
+	for (volatile int i = 0; i < BENCHMARKS; i++) {
+	}
+
+	for (volatile int i = 0; i < BENCHMARKS; i++) {
+	}
+
+	loopOverhead = getCntr() - loopOverhead;
+
+	uint64_t joinOverhead = threadJoinOverhead();


for performance calibration, I assume. nop version might be better, we use non-volatile iterator in the use-case

agkaminski · 2024-11-26T16:49:39Z

_user/rhealstone/msg-latency/main.c

+	uint64_t benchStart = getCntr();
+
+	priority(4);
+	usleep(0);
+
+	threadJoin(tid1, 0);
+	threadJoin(tid2, 0);
+
+	uint64_t benchEnd = getCntr();


Perhaps we could store start time in task1 and end time in task2? This would eliminate 3 syscalls and context switch overheads from the test

agkaminski · 2024-11-26T16:50:37Z

_user/rhealstone/preempt/main.c

+	for (volatile int cnt2 = 0; cnt2 < MAX_LOOPS; cnt2++) {
+	}


I guess this one is for task2. Then nop version is more adequate

agkaminski · 2024-11-26T16:51:12Z

_user/rhealstone/preempt/main.c

+	benchStart = getCntr();
+	priority(4);
+
+	usleep(0);
+
+	threadJoin(tid1, 0);
+	threadJoin(tid2, 0);
+
+	benchEnd = getCntr();


Same as other cases, perhaps time the execution in threads instead

agkaminski · 2024-11-26T16:52:52Z

_user/rhealstone/task-switch/main.c

+	for (volatile unsigned int i = 0; i < MAX_LOOPS; i++) {
+		/* usleep(0); */
+	}
+	for (volatile unsigned int i = 0; i < MAX_LOOPS; i++) {
+		/* usleep(0); */
+	}


imho volatile -> nop

lukileczo added 5 commits November 19, 2024 14:27

_targets/sparcv8leon/gr712rc: update configuration

ad72a2e

JIRA: RTOS-883

add sparcv8leon-gr740-mini target

44954a3

JIRA: RTOS-883

sparcv8leon-gr712rc/board_config: add flash and FTMCTRL_BASE addresses

7782789

JIRA: RTOS-883

_project/sparcv8leon: remove explicit timer count defines

4cb6dec

Timer count can be read from a configuration register. JIRA: RTOS-883

gr740 no root

8c96571

lukileczo requested a review from agkaminski November 26, 2024 14:17

github-actions bot reviewed Nov 26, 2024

View reviewed changes

lukileczo force-pushed the lukileczo/rhealstone branch from c77cf08 to 2227a5f Compare November 26, 2024 14:49

agkaminski reviewed Nov 26, 2024

View reviewed changes

add rhealstone benchmark

f61c6ee

lukileczo force-pushed the lukileczo/rhealstone branch from 2227a5f to f61c6ee Compare November 27, 2024 12:38

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[do not merge] rhealstone benchmark #1240

[do not merge] rhealstone benchmark #1240

lukileczo commented Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

github-actions bot Nov 26, 2024

agkaminski left a comment

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024

agkaminski Nov 26, 2024


		ROOTFS="$PREFIX_BOOT/rootfs.jffs2"

		local erase_sz=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0._meta.block_size }}')

		@@ -63,10 +61,11 @@ b_image_target() {

		ROOTFS="$PREFIX_BOOT/rootfs.jffs2"

		local erase_sz=$(image_builder.py query --nvm "$NVM_CONFIG" '{{ nvm.flash0._meta.block_size }}')

		[ "${BASH_SOURCE[0]}" -ef "$0" ] && echo "You should source this script, not execute it!" && exit 1


		FLASH_SZ=$((0x8000000))



		FLASH_SZ=$((0x8000000))
		ROOTFS_SZ=$((0x800000))

[do not merge] rhealstone benchmark #1240

Are you sure you want to change the base?

[do not merge] rhealstone benchmark #1240

Conversation

lukileczo commented Nov 26, 2024

Description

Motivation and Context

Types of changes

How Has This Been Tested?

Checklist:

Special treatment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

github-actions bot Nov 26, 2024

Choose a reason for hiding this comment

agkaminski left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment