Ithaca College 1 Machine-Level Programming XI: inline Assembly Comp 21000: Introduction to Computer Systems & Assembly Lang On-Line resources* * See see

Ithaca College 2 Today Inline assembly  Example

Ithaca College 3 Why use assembly? Assembly can express very low-level things:  you can access machine-dependent registers and I/O  you can control the exact code behavior in critical sections that might otherwise involve deadlock between multiple software threads or hardware devices  you can break the conventions of your usual compiler, which might allow some optimizations (like temporarily breaking rules about memory allocation, threading, calling conventions, etc)  you can build interfaces between code fragments using incompatible conventions (e.g. produced by different compilers, or separated by a low-level interface)  you can get access to unusual programming modes of your processor (e.g. 16 bit mode to interface startup, firmware, or legacy code on Intel PCs)  you can produce reasonably fast code for tight loops to cope with a bad non- optimizing compiler (but then, there are free optimizing compilers available!)  you can produce hand-optimized code perfectly tuned for your particular hardware setup, though not to someone else's  you can write some code for your new language's optimizing compiler (that is something what very few ones will ever do, and even they not often)  i.e. you can be in complete control of your code From the linux assembly howto

Ithaca College 4 Why use assembly? Speed  Be careful! Optimizing compilers are almost always better!  useful when you know an assembly language instruction that can replace a library call  Example: transcendental function computation  has macros for some inline assembly sequences  Example: if spend most of the time in a loop computing the sine and cosine of the same angles, could use the fsincos assembly function  We’ll see an example later

Ithaca College 5 inline assembly code* Example:  mov instruction /* put this line in your C program*/ asm("assembly code "); /* alternative syntax */ __asm__ ("assembly code ") ; asm("movl %rbx, %rax"); /* moves the contents of ebx register to eax */ __asm__("movb %ch, (%rbx)") ; /* moves the byte from ch to the memory pointed by ebx */ * see

Ithaca College 6 inline assembly More sophisticated assembly  For more than one assembly instruction, use semicolon at the end of each instruction of inserted code  see example on next slide

Ithaca College 7 Example 0 #include int main() { /* Add 10 and 20 and store result into register %eax */ __asm__ ( "movl $10, %rax;" "movl $20, %rbx;" "addl %rbx, %rax;" ) ; /* Subtract 20 from 10 and store result into register %eax */ __asm__ ( "movl $10, %rax;" "movl $20, %rbx;" "subl %rbx, %rax;" ) ; /* Multiply 10 and 20 and store result into register %eax */ __asm__ ( "movl $10, %rax;" "movl $20, %rbx;" "imull %rbx, %rax;" ) ; return 0 ; } compile with –O1 flag trace in gdb to see registers change. compile with –O1 flag trace in gdb to see registers change.

Ithaca College 8 Extended inline assembly Idea  In extended assembly, we can also specify the operands  can specify the input registers, output registers and a list of clobbered registers.  If there are no output operands but there are input operands, we must place two consecutive colons surrounding the place where the output operands would go.  Can omit list of clobbered registers to use,  GCC and GCC’s optimization scheme will take care of the reg.  In general, it’s a bad idea to omit these asm ( "assembly code " : output operands /* optional */ : input operands /* optional */ : list of clobbered registers /* optional */ );

Ithaca College 9 Example 1 The variable "val" is kept in a register  val is a C variable that must be declared earlier in the C program the value in register %rax is copied onto that register, and the value of "val" is updated into the memory from that register. note that rax is preceded by 2 percent signs  differentiate from a asm parameter (asm works like printf) asm ("movq %rax, %0;" : "=r" ( val )); Assm instr Output operands see ~barr/Student/Comp210/Resources/inline directory for all examples asm ( "assembly code " : output operands /* optional */ : input operands /* optional */ : list of clobbered registers /* optional */ );

Ithaca College 10 Example 1 (continued) the %0 indicates the first operand of asm, it is associated with the first parameter, i.e., val (similar to printf) “=r” indicates a register constraint  see chart on next page for all possible register specifications  “r” indicates that gcc may keep the variable in any available General Purpose Register (see next slide)  the “=“ indicates write only mode. If our instruction can alter the condition code register, we have to add "cc" to the list of clobbered registers. "g" : Any register, memory or immediate integer operand is allowed, except for registers that are not general registers. asm ("movq %rax, %0;" : "=r" ( val ));

Ithaca College 11 Example 1 (continued) Try the example on nori Why doesn’t this work? Fill out the worksheet with the corresponding C code and explain what the problem is.

Ithaca College 12 register specifiers rregister RGeneral register (RAX, RBX, RCX, RDX, RSI, RDI, RBP, RSP) qGeneral register for data (RAX, RBX, RCX, RDX) fFloating point reg a%rax,%eax, %ax, %al b%rbx, %ebx, %bx, %bl c%rcx, %ecx, %cx, %cl d%rdx, %edx, %dx, %dl S%rsi, %esi, %si D%rdi, %edi, %di

Ithaca College 13 Example 2 Two variables are declared in the C code, no and val %0 is the first operand to asm and thus refers to the C variable val (the output variable) %1 is the second operand and thus refers to the C variable no (the input variable) “=r” and “r” say that gcc can use any registers to store the corresponding variable (either val or no ) the clobbered variable is %rbx so gcc should not use that variable anywhere else. long no = 100, val ; asm ("movq %1, %rbx;" "movl %rbx, %0;" : "=r" ( val ) /* output */ : "r" ( no ) /* input */ : ”%rbx" /* clobbered register */ );

Ithaca College 14 Example 3 The C code declares three variables: arg1, arg2, add The input variables will use %rax (for arg1 ) and %rbx (for arg2 ) the output variable is add and will use register %rax Note that don’t need the position parameters (%0, %1) if specify registers for the input/output variables. no clobber register is set; gcc can determine long arg1, arg2, add ; arg1 = 10; arg2 = 25; __asm__ ( "addq %rbx, %rax;" : "=a" (add) /* output */ : "a" (arg1), "b" (arg2) ) ; /* input */

Ithaca College 15 Example 4 #include int main() { long arg1, arg2, add, sub, mul, quo, rem ; printf( "Enter two integer numbers : " ); scanf( "%ld%ld", &arg1, &arg2 ); /* Perform Addition, Subtraction, Multiplication & Division */ __asm__ ( "addq %rbx, %rax;" : "=a" (add) : "a" (arg1), "b" (arg2) ) ; __asm__ ( "subq %rbx, %rax;" : "=a" (sub) : "a" (arg1), "b" (arg2) ) ; __asm__ ( "imulq %rbx, %rax;" : "=a" (mul) : "a" (arg1), "b" (arg2) ) ; __asm__ ( "movq $0x0, %rdx;" "movq %2, %rax;" "movq %3, %rbx;" "idivq %rbx;" : "=a" (quo), "=d" (rem) : "g" (arg1), "g" (arg2) ) ; printf( "%ld + %ld = %ld\n", arg1, arg2, add ) ; printf( "%ld - %ld = %ld\n", arg1, arg2, sub ) ; printf( "%ld * %ld = %ld\n", arg1, arg2, mul ) ; printf( "%ld / %ld = %ld\n", arg1, arg2, quo ) ; printf( "%ld % %ld = %ld\n", arg1, arg2, rem ) ; return 0 ; } Example 4 idivq S # Signed divide R[%rdx]  R[%rdx]:R[%rax] mod S; R[%rax]  R[%rdx]:R[%rax] / S idivq S # Signed divide R[%rdx]  R[%rdx]:R[%rax] mod S; R[%rax]  R[%rdx]:R[%rax] / S

Ithaca College 16 Example 4 #include int main() { long arg1, arg2, arg3, rslt; printf( "Enter three integer numbers : " ); scanf( "%ld%ld%ld", &arg1, &arg2, &arg3 ); /* Change this line to assembly */ rslt = arg1 + 2 * arg2 - arg3 / 2; printf( "%ld + 2 * %ld – %ld / 2 = %ld\n", arg1, arg2, arg3, rslt ) ; return 0 ; } Exercise4 idivq S # Signed divide R[%rdx]  R[%rdx]:R[%rax] mod S; R[%rax]  R[%rdx]:R[%rax] / S idivq S # Signed divide R[%rdx]  R[%rdx]:R[%rax] mod S; R[%rax]  R[%rdx]:R[%rax] / S

Ithaca College 17 Example 4 #include int main (int argc, char* argv[]) { long max = atoi (argv[1]); long number; long i; unsigned position; volatile unsigned result; /* Repeat the operation for a large number of values. */ for (number = 1; number <= max; ++number) { /* Repeatedly shift the number to the right, until the result is zero. Keep count of the number of shifts this requires. */ for (i = (number >> 1), position = 0; i != 0; ++position) i >>= 1; /* The position of the most significant set bit is the number of shifts we needed after the first one. */ result = position; } // end outer for loop return 0; } // end main Example 5 why use inline assembly? no assembly in this code See on arda: /home/barr/Student/Comp210 /Resources/inline/bit-pos-loop.c Example 5 why use inline assembly? no assembly in this code See on arda: /home/barr/Student/Comp210 /Resources/inline/bit-pos-loop.c % gcc -O2 -o bit-pos-loop bit-pos-loop.c % time./bit-pos-loop user 0.00 system 0:20.40 elapsed 95%CPU (0avgtext+0avgdata 0maxresident)k0inputs+0outputs (73major+11minor)pagefaults 0swaps % gcc -O2 -o bit-pos-loop bit-pos-loop.c % time./bit-pos-loop user 0.00 system 0:20.40 elapsed 95%CPU (0avgtext+0avgdata 0maxresident)k0inputs+0outputs (73major+11minor)pagefaults 0swaps

Ithaca College 18 Example 4 #include int main (int argc, char* argv[]) { long max = atoi (argv[1]); long number; unsigned position; volatile unsigned result; /* Repeat the operation for a large number of values. */ for (number = 1; number <= max; ++number) { /* Compute the position of the most significant set bit using the bsrl assembly instruction. */ asm (“bsrl %1, %0” : “=r” (position) : “r” (number)); result = position; } // end for loop return 0; } Example 5 why use inline assembly? assembly used for inner loop See on arda: /home/barr/Student/ Comp210/Resources/inline/ bit-pos-asm.c Example 5 why use inline assembly? assembly used for inner loop See on arda: /home/barr/Student/ Comp210/Resources/inline/ bit-pos-asm.c %gcc -O2 -o bit-pos-asm bit-pos-asm.c % time./bit-pos-asm user 0.00system 0:03.32elapsed 95%CPU (0avgtext+0avgdata 0maxresident)k0inputs+0outputs (73major+11minor)pagefaults 0swaps %gcc -O2 -o bit-pos-asm bit-pos-asm.c % time./bit-pos-asm user 0.00system 0:03.32elapsed 95%CPU (0avgtext+0avgdata 0maxresident)k0inputs+0outputs (73major+11minor)pagefaults 0swaps Compare to the 19.51user of the previous C only example! bsr: Scans source operand for first bit set. Sets ZF if a bit is found set and loads the destination with an index to first set bit. Clears ZF is no bits are found set. BSF scans forward across bit pattern (0-n) while BSR scans in reverse (n-0).

Ithaca College 19 Volitile If our assembly statement must execute where we put it, (e.g. must not be moved out of a loop as an optimization), put the keyword "volatile" or "__volatile__" after "asm" or "__asm__" and before the ()s. asm volatile ( "...;” "...;" :... ); __asm__ __volatile__ ( "...;" "...;" :... ) ; asm volatile ( "...;” "...;" :... ); __asm__ __volatile__ ( "...;" "...;" :... ) ;

Ithaca College 20 Example 4 #include long gcd( long a, long b ) { long result ; /* Compute Greatest Common Divisor using Euclid's Algorithm */ __asm__ __volatile__ ( "movq %1, %rax;" "movq %2, %rbx;" "CONTD: cmpq $0, %rbx;" "je DONE;" "xorq %rdx, %rdx;" "idivq %rbx;" "movq %rbx, %rax;" "movq %rdx, %rbx;" "jmp CONTD;" "DONE: movq %rax, %0;" : "=g" (result) : "g" (a), "g" (b) ) ; return result ; } int main() { long first, second ; printf( "Enter two integers : " ) ; scanf( "%ld%ld", &first, &second ); printf( "GCD of %ld & %ld is %ld\n", first, second, gcd(first, second) ) ; return 0 ; } Example 5.5 Compute GCD with Euclid’s Algm gcd.c Example 5.5 Compute GCD with Euclid’s Algm gcd.c Note that we can put labels and jump to them! Note that we can put several asm instructions in one __asm__ function call Do NOT use the optimization parameter when you compile this!

Ithaca College 21 #include int main() { int x, y, rslt, rtnval; int i; printf("Enter two integers\n"); rslt = scanf("%d%d", &x, &y); /* change into assembly code */ rtnval = 0; for(i = 0; i < y; i++) { rtnval += x; } /* end change */ printf("%d * %d = %d\n", x, y, rtnval); return 0; } Example 6 convert the for loop to assembly code (should be no more than 11 lines of assembly code) Example 6 convert the for loop to assembly code (should be no more than 11 lines of assembly code)