Floating Point Arithmetic in AT&T syntax?

Implement it in C first, test it, then use gcc -S to generate assembler source, then either use that as is or use it as a template for your own code.

Up vote 0 down vote favorite share g+ share fb share tw.

Basically, all I have to do is multiply a matrix by a vector, using floating point operations in assembly. My first draft looks like this: . Data mat: .

Float 1.0, 2.0, 3.0, 4.0, 5.0, 6.0 vek: . Float 1.0, 2.0, 3.0 res: . Float 0.0, 0.0 .

Text . Globl main main: ; here goes the magic call exit Unfortunately, all I've tried doesn't work. I know how to do the basic arithmetic operations (like faddp, fmulp and so on), but I'm still not able to store it efficiently.

Fstpl doesn't seem to work at all. Can anybody give me a rough sketch (not on how to make this matrix - vector multiplication) how to use the FPU commands and how to store a computed result in a register? Thanks in advance, assembly x86 floating-point link|improve this question edited Nov 15 '11 at 12:26Paul R48.9k43288 asked Nov 15 '11 at 12:25Sebastian R.1.

SSE is much easier to use. – Stephen Canon Nov 15 '11 at 12:35.

Implement it in C first, test it, then use gcc -S to generate assembler source, then either use that as is or use it as a template for your own code. E.g. Here is a C code implementation, mat_vec.

C: #include int main(void) { // note: use "volatile" qualifier for input data otherwise gcc will // just optimise all the arithmetic away... volatile float mat23 = { { 1.0f, 2.0f, 3.0f }, { 4.0f, 5.0f, 6.0f } }; volatile float vek3 = { 1.0f, 2.0f, 3.0f }; float res2 = { 0.0f, 0.0f }; res0 = mat00 * vek0 + mat01 * vek1 + mat02 * vek2; res1 = mat10 * vek0 + mat11 * vek1 + mat12 * vek2; printf("res = { %g, %g }\n", res0, res1); return 0; } Let's make sure it works: $ gcc -Wall -Os -m32 -march=i686 mat_vec. C -o mat_vec $ . /mat_vec res = { 14, 32 } $ Looks good, so let's generate assembler source: $ gcc -Wall -Os -m32 -march=i686 -S mat_vec.

C -o mat_vec. S $ cat mat_vec. S .

Cstring LC6: . Ascii "res = { %g, %g }\12\0" . Text .

Globl _main _main: pushl %ebp movl $0x40000000, %ecx movl %esp, %ebp movl $0x40400000, %edx pushl %esi movl $0x40800000, %eax pushl %ebx movl $0x3f800000, %esi subl $96, %esp movl %esi, -44(%ebp) movl %ecx, -40(%ebp) movl %edx, -36(%ebp) movl %eax, -32(%ebp) movl $0x40a00000, %eax movl %eax, -28(%ebp) movl $0x40c00000, %eax movl %eax, -24(%ebp) movl %esi, -20(%ebp) movl %ecx, -16(%ebp) movl %edx, -12(%ebp) flds -44(%ebp) flds -20(%ebp) fstps -72(%ebp) flds -40(%ebp) flds -16(%ebp) fstps -68(%ebp) flds -36(%ebp) fstps -64(%ebp) flds -12(%ebp) fstps -60(%ebp) flds -32(%ebp) flds -20(%ebp) flds -28(%ebp) flds -16(%ebp) fxch %st(3) fmulp %st, %st(2) flds -24(%ebp) flds -12(%ebp) fxch %st(2) fmulp %st, %st(4) call L3 "L00000000001$pb": L3: popl %ebx fmulp %st, %st(1) fxch %st(3) fmuls -68(%ebp) fxch %st(1) faddp %st, %st(2) fxch %st(3) fmuls -72(%ebp) fxch %st(1) faddp %st, %st(2) fxch %st(1) leal LC6-"L00000000001$pb"(%ebx), %eax fstpl 12(%esp) flds -64(%ebp) fmuls -60(%ebp) fxch %st(1) faddp %st, %st(2) faddp %st, %st(1) fstpl 4(%esp) movl %eax, (%esp) call L_printf$stub addl $96, %esp xorl %eax, %eax popl %ebx popl %esi leave ret . Section __IMPORT,__jump_table,symbol_stubs,self_modifying_code+pure_instructions,5 L_printf$stub: . Indirect_symbol _printf hlt ; hlt ; hlt ; hlt ; hlt .

Subsections_via_symbols The part you're interested in starts just before label L3.

I cant really gove you an answer,but what I can give you is a way to a solution, that is you have to find the anglde that you relate to or peaks your interest. A good paper is one that people get drawn into because it reaches them ln some way.As for me WW11 to me, I think of the holocaust and the effect it had on the survivors, their families and those who stood by and did nothing until it was too late.