microblaze: Add libgcc function directly to kernel
Replaced libgcc functions with asm optimized implementation.
Signed-off-by: Michal Simek <monstr@monstr.eu>
diff --git a/arch/microblaze/lib/divsi3.S b/arch/microblaze/lib/divsi3.S
new file mode 100644
index 0000000..595b02d
--- /dev/null
+++ b/arch/microblaze/lib/divsi3.S
@@ -0,0 +1,73 @@
+#include <linux/linkage.h>
+
+/*
+* Divide operation for 32 bit integers.
+* Input : Dividend in Reg r5
+* Divisor in Reg r6
+* Output: Result in Reg r3
+*/
+ .text
+ .globl __divsi3
+ .type __divsi3, @function
+ .ent __divsi3
+__divsi3:
+ .frame r1, 0, r15
+
+ addik r1, r1, -16
+ swi r28, r1, 0
+ swi r29, r1, 4
+ swi r30, r1, 8
+ swi r31, r1, 12
+
+ beqi r6, div_by_zero /* div_by_zero - division error */
+ beqi r5, result_is_zero /* result is zero */
+ bgeid r5, r5_pos
+ xor r28, r5, r6 /* get the sign of the result */
+ rsubi r5, r5, 0 /* make r5 positive */
+r5_pos:
+ bgei r6, r6_pos
+ rsubi r6, r6, 0 /* make r6 positive */
+r6_pos:
+ addik r30, r0, 0 /* clear mod */
+ addik r3, r0, 0 /* clear div */
+ addik r29, r0, 32 /* initialize the loop count */
+
+ /* first part try to find the first '1' in the r5 */
+div0:
+ blti r5, div2 /* this traps r5 == 0x80000000 */
+div1:
+ add r5, r5, r5 /* left shift logical r5 */
+ bgtid r5, div1
+ addik r29, r29, -1
+div2:
+ /* left shift logical r5 get the '1' into the carry */
+ add r5, r5, r5
+ addc r30, r30, r30 /* move that bit into the mod register */
+ rsub r31, r6, r30 /* try to subtract (r30 a r6) */
+ blti r31, mod_too_small
+ /* move the r31 to mod since the result was positive */
+ or r30, r0, r31
+ addik r3, r3, 1
+mod_too_small:
+ addik r29, r29, -1
+ beqi r29, loop_end
+ add r3, r3, r3 /* shift in the '1' into div */
+ bri div2 /* div2 */
+loop_end:
+ bgei r28, return_here
+ brid return_here
+ rsubi r3, r3, 0 /* negate the result */
+div_by_zero:
+result_is_zero:
+ or r3, r0, r0 /* set result to 0 */
+return_here:
+/* restore values of csrs and that of r3 and the divisor and the dividend */
+ lwi r28, r1, 0
+ lwi r29, r1, 4
+ lwi r30, r1, 8
+ lwi r31, r1, 12
+ rtsd r15, 8
+ addik r1, r1, 16
+
+.size __divsi3, . - __divsi3
+.end __divsi3