Skip to contentSkip to conversation

BYTE Unix Benchmarks Version 3.6

DISCUSSION Paolo Ciarrocchi 8 participants 14 messages Sep 4, 2002 – Sep 9, 2002

lore
Share & tools
lore
#0 lore raw

Hi all,
I've just ran the BYTE Unix Benchmarks Version 3.6 on the 2.4.19 and on the 2.5.33 kernel. Here it goes the results:

BYTE UNIX Benchmarks (Version 3.11)
System -- Linux localhost.localdomain 2.4.19 #10 Fri Aug 23 20:53:06 BST 2002 i686 unknown Start Benchmark Run: Wed Sep 4 22:11:32 BST 2002
1 interactive users.

Dhrystone 2 without register variables   1499020.6 lps   (10 secs, 6 samples)
Dhrystone 2 using register variables     1501168.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = arithoh)         3598100.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = register)        201521.0 lps   (10 secs, 6 samples)
Arithmetic Test (type = short)           190245.9 lps   (10 secs, 6 samples)
Arithmetic Test (type = int)             201904.5 lps   (10 secs, 6 samples)
Arithmetic Test (type = long)            201906.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = float)           210562.7 lps   (10 secs, 6 samples)
Arithmetic Test (type = double)          210385.9 lps   (10 secs, 6 samples)
System Call Overhead Test                407402.6 lps   (10 secs, 6 samples)
Pipe Throughput Test                     476268.6 lps   (10 secs, 6 samples)
Pipe-based Context Switching Test        218969.9 lps   (10 secs, 6 samples)

Process Creation Test 9078.6 lps (10 secs, 6 samples)
Execl Throughput Test 998.0 lps (9 secs, 6 samples)
File Read (10 seconds) 1571652.0 KBps (10 secs, 6 samples)
File Write (10 seconds) 109237.0 KBps (10 secs, 6 samples)
File Copy (10 seconds) 24329.0 KBps (10 secs, 6 samples)
File Read (30 seconds) 1562505.0 KBps (30 secs, 6 samples)

File Write (30 seconds)                  113152.0 KBps  (30 secs, 6 samples)
File Copy  (30 seconds)                   14334.0 KBps  (30 secs, 6 samples)
C Compiler Test                             470.9 lpm   (60 secs, 3 samples)
Shell scripts (1 concurrent)                980.4 lpm   (60 secs, 3 samples)
Shell scripts (2 concurrent)                544.1 lpm   (60 secs, 3 samples)
Shell scripts (4 concurrent)                287.0 lpm   (60 secs, 3 samples)
Shell scripts (8 concurrent)                147.0 lpm   (60 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places          42311.6 lpm   (60 secs, 6 samples)
Recursion Test--Tower of Hanoi            18915.4 lps   (10 secs, 6 samples)

INDEX VALUES
TEST BASELINE RESULT INDEX

Arithmetic Test (type = double)               2541.7   210385.9       82.8
Dhrystone 2 without register variables       22366.3  1499020.6       67.0
Execl Throughput Test                           16.5      998.0       60.5
File Copy  (30 seconds)                        179.0    14334.0       80.1

Pipe-based Context Switching Test 1318.5 218969.9 166.1
Shell scripts (8 concurrent) 4.0 147.0 36.8
=========
SUM of 6 items 493.2
AVERAGE 82.2

sh pgms/report.sh results/log > results/report
sh pgms/index.sh pgms/index.base results/log >> results/report cat results/report

BYTE UNIX Benchmarks (Version 3.11)
System -- Linux localhost.localdomain 2.5.33 #32 Tue Sep 3 22:18:19 BST 2002 i686 unknown Start Benchmark Run: Wed Sep 4 20:46:31 BST 2002
1 interactive users.

Dhrystone 2 without register variables   1488327.9 lps   (10 secs, 6 samples)
Dhrystone 2 using register variables     1488265.3 lps   (10 secs, 6 samples)
Arithmetic Test (type = arithoh)         3435944.6 lps   (10 secs, 6 samples)
Arithmetic Test (type = register)        197870.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = short)           145140.8 lps   (10 secs, 6 samples)
Arithmetic Test (type = int)             104440.5 lps   (10 secs, 6 samples)
Arithmetic Test (type = long)            177757.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = float)           208476.4 lps   (10 secs, 6 samples)
Arithmetic Test (type = double)          208443.3 lps   (10 secs, 6 samples)
System Call Overhead Test                397276.7 lps   (10 secs, 6 samples)
Pipe Throughput Test                     434561.9 lps   (10 secs, 6 samples)
Pipe-based Context Switching Test        148653.5 lps   (10 secs, 6 samples)

Process Creation Test 5422.1 lps (10 secs, 6 samples)
Execl Throughput Test 771.6 lps (10 secs, 6 samples)
File Read (10 seconds) 1553289.0 KBps (10 secs, 6 samples)
File Write (10 seconds) 132002.0 KBps (10 secs, 6 samples)
File Copy (10 seconds) 17994.0 KBps (10 secs, 6 samples)
File Read (30 seconds) 1540682.0 KBps (30 secs, 6 samples)

File Write (30 seconds)                  137781.0 KBps  (30 secs, 6 samples)
File Copy  (30 seconds)                   11460.0 KBps  (30 secs, 6 samples)
C Compiler Test                             450.9 lpm   (60 secs, 3 samples)
Shell scripts (1 concurrent)                876.7 lpm   (60 secs, 3 samples)
Shell scripts (2 concurrent)                480.3 lpm   (60 secs, 3 samples)
Shell scripts (4 concurrent)                251.0 lpm   (60 secs, 3 samples)
Shell scripts (8 concurrent)                126.0 lpm   (60 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places          33530.4 lpm   (60 secs, 6 samples)
Recursion Test--Tower of Hanoi            18514.3 lps   (10 secs, 6 samples)

INDEX VALUES
TEST BASELINE RESULT INDEX

Arithmetic Test (type = double)               2541.7   208443.3       82.0
Dhrystone 2 without register variables       22366.3  1488327.9       66.5
Execl Throughput Test                           16.5      771.6       46.8
File Copy  (30 seconds)                        179.0    11460.0       64.0

Pipe-based Context Switching Test 1318.5 148653.5 112.7
Shell scripts (8 concurrent) 4.0 126.0 31.5
=========
SUM of 6 items 403.6
AVERAGE 67.3

Comments?

Signature · 4 lines
Get your free email from www.linuxmail.org 


Powered by Outblaze
#1 lore raw
Quoted 4 lines

Hi all,
I've just ran the BYTE Unix Benchmarks Version 3.6 on the 2.4.19 and on the 2.5.33 kernel. Here it goes the results:

snipped:
Always useful to compare. I think the data from STP show something simular. I've pulled some reports, rather than send a huge email, here are the URL's (btw STP runs v4.0.1 UNIXbench)

runs made on 2-CPU platform
2.4.19: http://khack.osdl.org/stp/3877/results/report
2.5.33: http://khack.osdl.org/stp/4928/results/report

Another compare...why does pre5 appear a little worse than pre4 ? 2.4.20-pre5: http://khack.osdl.org/stp/4836/results/report
2.4.20-pre4: http://khack.osdl.org/stp/4581/results/report

cliffw

#2 lore raw
In reply to Paolo Ciarrocchi
Quoted 3 lines · Paolo Ciarrocchi

Hi all,
I've just ran the BYTE Unix Benchmarks Version 3.6 on the 2.4.19 and on the 2.5.33 kernel. Here it goes the results:

can you make a run on 2.4.20pre5aa1 too? thanks,

Quoted 1 line

Comments?

I'm just guessing but I think part of the global regression in most numbers is most due the increased HZ and rmap.

Andrea

#3 lore raw

side-by-side Re: BYTE Unix Benchmarks Version 3.6

In reply to Paolo Ciarrocchi

Side-by-side with some marked changes highlighted:

                                        2.4.19           2.5.33
-----------------------------------------------------------------------
Dhrystone 2 without register variable   1499020.6 lps    1488327.9 lps
Dhrystone 2 using register variables    1501168.4 lps    1488265.3 lps
Arithmetic Test (type = arithoh)        3598100.4 lps    3435944.6 lps
Arithmetic Test (type = register)        201521.0 lps     197870.4 lps
Arithmetic Test (type = short)           190245.9 lps     145140.8 lps
Arithmetic Test (type = int)             201904.5 lps     104440.5 lps
Arithmetic Test (type = long)            201906.4 lps     177757.4 lps
Arithmetic Test (type = float)           210562.7 lps     208476.4 lps
Arithmetic Test (type = double)          210385.9 lps     208443.3 lps
System Call Overhead Test                407402.6 lps     397276.7 lps
>Pipe Throughput Test                    476268.6 lps     434561.9 lps
>Pipe-based Context Switching Test       218969.9 lps     148653.5 lps
Quoted 1 line

Process Creation Test 9078.6 lps 5422.1 lps

Execl Throughput Test 998.0 lps 771.6 lps
File Read (10 seconds) 1571652.0 KBps 1553289.0 KBps
File Write (10 seconds) 109237.0 KBps 132002.0 KBps

Quoted 1 line

File Copy (10 seconds) 24329.0 KBps 17994.0 KBps

File Read (30 seconds) 1562505.0 KBps 1540682.0 KBps
File Write (30 seconds) 113152.0 KBps 137781.0 KBps
File Copy (30 seconds) 14334.0 KBps 11460.0 KBps

C Compiler Test                             470.9 lpm        450.9 lpm
Shell scripts (1 concurrent)                980.4 lpm        876.7 lpm
Shell scripts (2 concurrent)                544.1 lpm        480.3 lpm
Shell scripts (4 concurrent)                287.0 lpm        251.0 lpm
Shell scripts (8 concurrent)                147.0 lpm        126.0 lpm
Quoted 1 line

Dc: sqrt(2) to 99 decimal places 42311.6 lpm 33530.4 lpm

Recursion Test--Tower of Hanoi 18915.4 lps 18514.3 lps

                      INDEX VALUES           2.4.19     2.5      
 TEST                                        INDEX      INDEX
 Arithmetic Test (type = double)              82.8       82.0
 Dhrystone 2 without register variables       67.0       66.5
 Execl Throughput Test                        60.5       46.8
 File Copy  (30 seconds)                      80.1       64.0

Pipe-based Context Switching Test 166.1 112.7
Shell scripts (8 concurrent) 36.8 31.5
========= =========
SUM of 6 items 493.2 403.6
AVERAGE 82.2 67.3

Signature · 3 lines
http://www.PowerDNS.com          Versatile DNS Software & Services
http://www.tk                              the dot in .tk
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO
#4 lore raw

Re: side-by-side Re: BYTE Unix Benchmarks Version 3.6

I usually run byte bench regularly with every new kernel, so I see some strange results here.

Quoted 1 line

From your numbers, I would say you are using a PIII 600/900 Mhz (more or

less). It is not an AMD AThlon or a PIV, since float and double are too slow, not it is a K6 because they are too fast.

Quoted 6 lines · bert hubert

Date: Thu, 5 Sep 2002 15:48:30 +0200
From: bert hubert <[email protected]>
To: Paolo Ciarrocchi <[email protected]>
Cc: [email protected]
Subject: side-by-side Re: BYTE Unix Benchmarks Version 3.6

Quoted 7 lines

Side-by-side with some marked changes highlighted:

                                        2.4.19           2.5.33
-----------------------------------------------------------------------
Dhrystone 2 without register variable   1499020.6 lps    1488327.9 lps
Dhrystone 2 using register variables    1501168.4 lps    1488265.3 lps
Arithmetic Test (type = arithoh)        3598100.4 lps    3435944.6 lps

this could vary a little

Quoted 2 lines
Arithmetic Test (type = register)        201521.0 lps     197870.4 lps
Arithmetic Test (type = short)           190245.9 lps     145140.8 lps

the difference should never be so big

Quoted 1 line

Arithmetic Test (type = int) 201904.5 lps 104440.5 lps

the difference should never be so big

Quoted 1 line

Arithmetic Test (type = long) 201906.4 lps 177757.4 lps

the difference should never be so big

seeing this I think you had something running in background using your CPU while you where running int tests. if you loock at bm/results/log (log.accum if you did some other run recently)
should find lines like:

Arithmetic Test (type = int)|10.0|lps|227163.1|227158.7|6

that is a little more interesting if you are under load.

Quoted 7 lines
Arithmetic Test (type = float)           210562.7 lps     208476.4 lps
Arithmetic Test (type = double)          210385.9 lps     208443.3 lps
System Call Overhead Test                407402.6 lps     397276.7 lps
>Pipe Throughput Test                    476268.6 lps     434561.9 lps
Quoted 1 line

Pipe-based Context Switching Test 218969.9 lps 148653.5 lps

this could vary because of a lot of factors, starting from a bad page colouring going to sendmail activity.

Quoted 2 lines
Quoted 1 line

Process Creation Test 9078.6 lps 5422.1 lps

Execl Throughput Test 998.0 lps 771.6 lps

this is interesting, but seeing previous results about int and short, I am curious about your real load. I am quite curious if with 2.5 you are using kernel preemption.

Quoted 6 lines

File Read (10 seconds) 1571652.0 KBps 1553289.0 KBps
File Write (10 seconds) 109237.0 KBps 132002.0 KBps

Quoted 1 line

File Copy (10 seconds) 24329.0 KBps 17994.0 KBps

File Read (30 seconds) 1562505.0 KBps 1540682.0 KBps
File Write (30 seconds) 113152.0 KBps 137781.0 KBps
File Copy (30 seconds) 14334.0 KBps 11460.0 KBps

I saw the save with IDE disks... again, are you using kernel preemption?

Quoted 5 lines
C Compiler Test                             470.9 lpm        450.9 lpm
Shell scripts (1 concurrent)                980.4 lpm        876.7 lpm
Shell scripts (2 concurrent)                544.1 lpm        480.3 lpm
Shell scripts (4 concurrent)                287.0 lpm        251.0 lpm
Shell scripts (8 concurrent)                147.0 lpm        126.0 lpm

In my tests generally shell scripts are faster with 2.5 kernel.

Quoted 12 lines
Quoted 1 line

Dc: sqrt(2) to 99 decimal places 42311.6 lpm 33530.4 lpm

Recursion Test--Tower of Hanoi 18915.4 lps 18514.3 lps

                      INDEX VALUES           2.4.19     2.5
 TEST                                        INDEX      INDEX
 Arithmetic Test (type = double)              82.8       82.0
 Dhrystone 2 without register variables       67.0       66.5
 Execl Throughput Test                        60.5       46.8
 File Copy  (30 seconds)                      80.1       64.0
Quoted 6 lines

Pipe-based Context Switching Test 166.1 112.7
Shell scripts (8 concurrent) 36.8 31.5
========= =========
SUM of 6 items 493.2 403.6
AVERAGE 82.2 67.3

Luigi

Quoted 10 lines

--
http://www.PowerDNS.com Versatile DNS Software & Services
http://www.tk the dot in .tk
http://lartc.org Linux Advanced Routing & Traffic Control HOWTO
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#5 lore raw

Re: side-by-side Re: BYTE Unix Benchmarks Version 3.6

In reply to bert hubert
Quoted 7 lines · bert hubert

Arithmetic Test (type = arithoh) 3598100.4 lps 3435944.6 lps

Arithmetic Test (type = register)        201521.0 lps     197870.4 lps
Arithmetic Test (type = short)           190245.9 lps     145140.8 lps
Arithmetic Test (type = int)             201904.5 lps     104440.5 lps
Arithmetic Test (type = long)            201906.4 lps     177757.4 lps
Arithmetic Test (type = float)           210562.7 lps     208476.4 lps
Arithmetic Test (type = double)          210385.9 lps     208443.3 lps

What kind of arithmetic is this? Why on earth would arithmetic vary from one kernel to another?

Signature · 1 line
Daniel
#6 lore raw

Calculating kernel logical address ..

Hi,

I need help to correctly calculate kernel logical address for a user pointer which I mmaped from device driver. In mmap() file operation, I allocate some memory using kmalloc() and call :

remap_page_range(vma->vm_start,
virt_to_phys((void *)(Uint32)kmalloc_buffer),
size,
PAGE_SHARED);

after reserving all pages and doing some other stuff. Now when I get a user pointer and I need to calculate correspoding kernel logical address, I use following code:

adr = user_address;
pgd_offset(current->mm, adr);

if (!pgd_none(*pgd)) {
	pmd = pmd_offset(pgd, adr);
	if (!pmd_none(*pmd)) {
		ptep = pte_offset(pmd, adr);
		pte = *ptep;
		if(pte_present(pte)) {
			kaddr  = (unsigned long) page_address(pte_page(pte));
			kaddr |= (adr & (PAGE_SIZE - 1));
		}
	}
}

Will this code always give me correct kernel logical address?

I will really appreciate any guidance.

Thanks,
Imran.

#13 lore raw

Re: side-by-side Re: BYTE Unix Benchmarks Version 3.6

In reply to Daniel Phillips
Quoted 11 lines · Daniel Phillips
Quoted 7 lines · bert hubert

Arithmetic Test (type = arithoh) 3598100.4 lps 3435944.6 lps

Arithmetic Test (type = register)        201521.0 lps     197870.4 lps
Arithmetic Test (type = short)           190245.9 lps     145140.8 lps
Arithmetic Test (type = int)             201904.5 lps     104440.5 lps
Arithmetic Test (type = long)            201906.4 lps     177757.4 lps
Arithmetic Test (type = float)           210562.7 lps     208476.4 lps
Arithmetic Test (type = double)          210385.9 lps     208443.3 lps

What kind of arithmetic is this? Why on earth would arithmetic vary from one kernel to another?

I wasn't involved in this benchmark, I just reformatted the results. However, it might be that this benchmark is a tad braindead and 'suffers' from far better timing resolution because of HZ=1000. I'm unsure. I saw that this benchmark used something like a Sparcstation 5 as a reference platform, so maybe it is not geared for today's processors.

Regards,

bert hubert

Signature · 3 lines
http://www.PowerDNS.com          Versatile DNS Software & Services
http://www.tk                              the dot in .tk
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO
#7 lore raw

Re: Calculating kernel logical address ..

In reply to Imran Badr
Quoted 12 lines · Imran Badr
if (!pgd_none(*pgd)) {
	pmd = pmd_offset(pgd, adr);
	if (!pmd_none(*pmd)) {
		ptep = pte_offset(pmd, adr);
		pte = *ptep;
		if(pte_present(pte)) {
			kaddr  = (unsigned long) page_address(pte_page(pte));
			kaddr |= (adr & (PAGE_SIZE - 1));
		}
	}
}
Quoted 3 lines

Will this code always give me correct kernel logical address?

I will really appreciate any guidance.

It looks good to me. Note that somebody has added some new voodoo in 2.5 so that page table pages can be in highmem, with the result that the above code won't work in 2.5, whether or not highmem is configured.

Signature · 1 line
Daniel
#8 lore raw

Re: Calculating kernel logical address ..

From: Daniel Phillips <[email protected]>
Date: Sat, 7 Sep 2002 03:44:53 +0200

It looks good to me. Note that somebody has added some new voodoo in 2.5 so that page table pages can be in highmem, with the result that the above code won't work in 2.5, whether or not highmem is configured.

The example given won't work for kernel text/data addresses on a few platforms (sparc64 is one). And in fact on MIPS the KSEG0 pages lack any page tables.

There are only three things one can portably obtain a physical address of:

1) A user address, for a known MM

2) a kmalloc/get_free_page kernel page

3) A vmalloc page

For anything else you're in non-portablt land, including and in partiular:

1) kernel stack addresses
2) addresses within the main kernel image text/data/bss

#9 lore raw

Re: Calculating kernel logical address ..

Quoted 11 lines · David S. Miller

From: Daniel Phillips <[email protected]>
Date: Sat, 7 Sep 2002 03:44:53 +0200

It looks good to me. Note that somebody has added some new voodoo in 2.5 so that page table pages can be in highmem, with the result that the above code won't work in 2.5, whether or not highmem is configured.

The example given won't work for kernel text/data addresses on a few platforms (sparc64 is one). And in fact on MIPS the KSEG0 pages lack any page tables.

Quoted 8 lines

There are only three things one can portably obtain a physical address of:

1) A user address, for a known MM

2) a kmalloc/get_free_page kernel page

3) A vmalloc page

Actually, he was trying to obtain a kernel virtual address, which presents its own difficulties, particularly with respect to highmem.

Quoted 4 lines

For anything else you're in non-portablt land, including and in partiular:

1) kernel stack addresses

Could you elaborate on what bad things happen here?

Quoted 1 line

2) addresses within the main kernel image text/data/bss

Yep. MIPS's KSEG0 (a stupid design if there ever was one) and i386 large kernel image pages are just two examples of wrinkles that would need to be handled. The general principle is: mappings in the kernel's virtual address space are not maintained by the faulting mechanism, they are maintained 'by hand'; and the cross-platform N-level page tree is defined only for addresses that can fault.

Where the page tree does define a mapping to physical memory, obviously only a physical address may be obtained that way, and due to highmem, this address may not be mapped into the kernel's virtual address range, in which case a further kmapping step has to be performed to obtain a kernel-usable address, subject to bizarre rules that tend to change on a weekly basis.

Somebody needs to write a book about this ;-)

Signature · 1 line
Daniel
#10 lore raw

Re: Calculating kernel logical address ..

From: Daniel Phillips <[email protected]>
Date: Sun, 8 Sep 2002 20:44:17 +0200

Quoted 4 lines

For anything else you're in non-portablt land, including and in partiular:

1) kernel stack addresses

Could you elaborate on what bad things happen here?

Kernel stack allocation is defined per-architecture. On sun4c sparc systems, we carve virtual pages out from the kernel address space and hard map them into the TLB by hand.

Quoted 1 line

2) addresses within the main kernel image text/data/bss

Yep. MIPS's KSEG0 (a stupid design if there ever was one)

Actually, KSEG0 the most Linux friendly design in the world particularly in 64-bit mode. There is no need to have page tables at all for the main kernel physical memory map. It would shave a lot of code from the sparc64 TLB miss handlers if I didn't have to handle PAGE_OFFSET pages, for example.

Alpha does something akin to KSEG0 as well.

I pine constantly for it appearing some day on a future UltraSPARC revision :-)

#11 lore raw

Re: Calculating kernel logical address ..

Quoted 6 lines · David S. Miller
Quoted 1 line

2) addresses within the main kernel image text/data/bss

Yep. MIPS's KSEG0 (a stupid design if there ever was one)

Actually, KSEG0 the most Linux friendly design in the world particularly in 64-bit mode.

That's easy to say until you try and work with it (I assume you have, and forgot). Just try to do a 3G/1G split on it, for example.

Quoted 2 lines

I pine constantly for it appearing some day on a future UltraSPARC revision :-)

Don't wish too hard, you might get it ;-)

Signature · 1 line
Daniel
#12 lore raw

Re: Calculating kernel logical address ..

From: Daniel Phillips <[email protected]>
Date: Mon, 9 Sep 2002 07:17:30 +0200

Quoted 2 lines · David S. Miller

Actually, KSEG0 the most Linux friendly design in the world particularly in 64-bit mode.

That's easy to say until you try and work with it (I assume you have, and forgot). Just try to do a 3G/1G split on it, for example.

Maybe you missed the "64-bit mode" part of what I said. :-)

In 64-bit mode there is no need to do any kind of split. You just use the KSEG mapping with full cache coherency for all of physical memory as the PAGE_OFFSET area.

I forget if it was KSEG0 or some other number, but I know it works.