Opteron Bottleneck?? - Page 3

dsouza123 · 2003-09-13, 02:05

How was the MASM code used before for linux ?

The object files it could make were COFF format or some other format (? CV ?).

What compiler, linker was used ?

Dresdenboy · 2003-09-13, 11:50

I read in the makefiles that masm was used with options /c and /coff.

Unfortunately the latter one doesn't work for MASM 8.0.

I tried it with gcc 3.3

objcopy (x86-64 version) doesn't recognize the MASM 8.0 obj file.

Prime95 · 2003-09-13, 15:23

Have you looked in the binutils config file to see if there is an entry for the new MASM object file format? (It sure would be nice if Microsoft told us what that format is!)

Dresdenboy · 2003-09-13, 17:55

Quote:

Originally posted by Prime95
Have you looked in the binutils config file to see if there is an entry for the new MASM object file format? (It sure would be nice if Microsoft told us what that format is!)

I searched the internet a bit but I found only a few occurences of that topic at all.. On some russian site I found a detailed description of the COFF and PE files plus some info on AMD64 specific things. The executable format for the latter is called PE32+.

I compared an obj file of MASM 7 with the one of MASM 8. They look very similar (path names, code and other things at the same locations) - only several section names like "_TEXT32" have changed.

In some binutils source for IA64 they mention COFF_WITH_PEP64. Maybe the file format for AMD64 is very similiar.

Dresdenboy · 2003-09-15, 08:32

I'm currently downloading the whole Windows Platform SDK in the hope to get some detailed information about the new COFF/PE+ (or PE32+) formats.

Because I don't know if there will be some success in being able to convert ML64 generated obj files to some elf64 format, I'll also look into the possibility to make some small script which translates the assembly sources. It would even be possible to handle some cases where instructions or special instruction formats aren't supported in 64bit mode. And the final output could be edited and assembled more quickly than always having to convert the obj. files.

Dresdenboy · 2003-09-15, 09:49

That's what I found in some Intel documentation:

New Portable Executable File Format (PE32+):

Win64* executables are almost identical to Win32* PE files, a few fields widened to 64 bit:
–Image Base, Stack and Heap sizes, Import Address Table entries

So it looks like only these widened fields (and some format ID bytes) changed. That's all what is preventing us from doing the conversion. It should be easily possible to change bfd to understand the new format. Some IA64 support is included but as I read somewhere else, it is just rudimentary support.

lunna · 2003-09-16, 15:35

If you are looking for an assembler, there is YASM (http://www.tortall.net/projects/yasm/) too.

I have never ever written any Intel machine code, so I don't know if the feature list of that fits the bill. At least it has AMD64 support and a few different object files output options.

Dresdenboy · 2003-09-16, 16:14

I've already planned to have a look at it when I find the time. Actually the most important feature besides AMD64 capabilities is to have similar possibilities of macro usage like in MASM.

As I read, NASM will also support AMD64 code some day and AMD internally already uses a modified NASM which knows the AMD64 ISA extensions.

Prime95 · 2003-09-16, 20:40

I asked on the binutils mailing list if there would be support for MASM 8.0 object files. The answer was: "Yes - when someone submits acceptable code for it".

Dresdenboy · 2003-09-17, 06:54

It seems, that we have to submit that code.

With the info I found about PE+/PE32+ it should be possible to convert that format into others.

But then I have to find out how to include it in bfd.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
The bandwidth bottleneck is apparently much older than I thought	Dubslow	Hardware	5	2017-11-16 19:50
Opteron is Hyperthreaded ?	bgbeuning	Information & Answers	3	2016-01-10 08:26
Modular Inversion Bottleneck	Sam Kennedy	Programming	4	2013-01-25 16:50
AMD Athlon 64 vs AMD Opteron for ecm	thomasn	Factoring	6	2004-11-08 13:25
AMD Opteron	naclosagc	Software	27	2003-08-10 19:14

2003-09-13, 02:05	#23
dsouza123 Sep 2002 2·331 Posts	How was the MASM code used before for linux ? The object files it could make were COFF format or some other format (? CV ?). What compiler, linker was used ?

2003-09-13, 11:50	#24
Dresdenboy Apr 2003 Berlin, Germany 551₈ Posts	I read in the makefiles that masm was used with options /c and /coff. Unfortunately the latter one doesn't work for MASM 8.0. I tried it with gcc 3.3 objcopy (x86-64 version) doesn't recognize the MASM 8.0 obj file.

2003-09-13, 15:23	#25
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 17×487 Posts	Have you looked in the binutils config file to see if there is an entry for the new MASM object file format? (It sure would be nice if Microsoft told us what that format is!)

2003-09-15, 08:32	#27
Dresdenboy Apr 2003 Berlin, Germany 19² Posts	I'm currently downloading the whole Windows Platform SDK in the hope to get some detailed information about the new COFF/PE+ (or PE32+) formats. Because I don't know if there will be some success in being able to convert ML64 generated obj files to some elf64 format, I'll also look into the possibility to make some small script which translates the assembly sources. It would even be possible to handle some cases where instructions or special instruction formats aren't supported in 64bit mode. And the final output could be edited and assembled more quickly than always having to convert the obj. files.

2003-09-15, 09:49	#28
Dresdenboy Apr 2003 Berlin, Germany 19² Posts	That's what I found in some Intel documentation: New Portable Executable File Format (PE32+): Win64* executables are almost identical to Win32* PE files, a few fields widened to 64 bit: –Image Base, Stack and Heap sizes, Import Address Table entries So it looks like only these widened fields (and some format ID bytes) changed. That's all what is preventing us from doing the conversion. It should be easily possible to change bfd to understand the new format. Some IA64 support is included but as I read somewhere else, it is just rudimentary support.

2003-09-16, 15:35	#29
lunna Mar 2003 2⁴ Posts	If you are looking for an assembler, there is YASM (http://www.tortall.net/projects/yasm/) too. I have never ever written any Intel machine code, so I don't know if the feature list of that fits the bill. At least it has AMD64 support and a few different object files output options.

2003-09-16, 16:14	#30
Dresdenboy Apr 2003 Berlin, Germany 361₁₀ Posts	I've already planned to have a look at it when I find the time. Actually the most important feature besides AMD64 capabilities is to have similar possibilities of macro usage like in MASM. As I read, NASM will also support AMD64 code some day and AMD internally already uses a modified NASM which knows the AMD64 ISA extensions.

2003-09-16, 20:40	#31
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 17×487 Posts	I asked on the binutils mailing list if there would be support for MASM 8.0 object files. The answer was: "Yes - when someone submits acceptable code for it".

2003-09-17, 06:54	#32
Dresdenboy Apr 2003 Berlin, Germany 361₁₀ Posts	It seems, that we have to submit that code. With the info I found about PE+/PE32+ it should be possible to convert that format into others. But then I have to find out how to include it in bfd.