| 0,0 → 1,98 |
|---|
| .section #gk110_builtin_code |
| // DIV U32 |
| // |
| // UNR recurrence (q = a / b): |
| // look for z such that 2^32 - b <= b * z < 2^32 |
| // then q - 1 <= (a * z) / 2^32 <= q |
| // |
| // INPUT: $r0: dividend, $r1: divisor |
| // OUTPUT: $r0: result, $r1: modulus |
| // CLOBBER: $r2 - $r3, $p0 - $p1 |
| // SIZE: 22 / 14 * 8 bytes |
| // |
| gk110_div_u32: |
| sched 0x28 0x04 0x28 0x04 0x28 0x28 0x28 |
| bfind u32 $r2 $r1 |
| xor b32 $r2 $r2 0x1f |
| mov b32 $r3 0x1 |
| shl b32 $r2 $r3 clamp $r2 |
| cvt u32 $r1 neg u32 $r1 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| sched 0x28 0x28 0x28 0x28 0x28 0x28 0x28 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| sched 0x04 0x28 0x04 0x28 0x28 0x2c 0x04 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mov b32 $r3 $r0 |
| mul high $r0 u32 $r0 u32 $r2 |
| cvt u32 $r2 neg u32 $r1 |
| add $r1 (mul u32 $r1 u32 $r0) $r3 |
| set $p0 0x1 ge u32 $r1 $r2 |
| $p0 sub b32 $r1 $r1 $r2 |
| sched 0x28 0x2c 0x04 0x20 0x2e 0x28 0x20 |
| $p0 add b32 $r0 $r0 0x1 |
| $p0 set $p0 0x1 ge u32 $r1 $r2 |
| $p0 sub b32 $r1 $r1 $r2 |
| $p0 add b32 $r0 $r0 0x1 |
| ret |
| |
| // DIV S32, like DIV U32 after taking ABS(inputs) |
| // |
| // INPUT: $r0: dividend, $r1: divisor |
| // OUTPUT: $r0: result, $r1: modulus |
| // CLOBBER: $r2 - $r3, $p0 - $p3 |
| // |
| gk110_div_s32: |
| set $p2 0x1 lt s32 $r0 0x0 |
| set $p3 0x1 lt s32 $r1 0x0 xor $p2 |
| sched 0x20 0x28 0x28 0x04 0x28 0x04 0x28 |
| cvt s32 $r0 abs s32 $r0 |
| cvt s32 $r1 abs s32 $r1 |
| bfind u32 $r2 $r1 |
| xor b32 $r2 $r2 0x1f |
| mov b32 $r3 0x1 |
| shl b32 $r2 $r3 clamp $r2 |
| cvt u32 $r1 neg u32 $r1 |
| sched 0x28 0x28 0x28 0x28 0x28 0x28 0x28 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| sched 0x28 0x28 0x04 0x28 0x04 0x28 0x28 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mul $r3 u32 $r1 u32 $r2 |
| add $r2 (mul high u32 $r2 u32 $r3) $r2 |
| mov b32 $r3 $r0 |
| mul high $r0 u32 $r0 u32 $r2 |
| cvt u32 $r2 neg u32 $r1 |
| add $r1 (mul u32 $r1 u32 $r0) $r3 |
| sched 0x2c 0x04 0x28 0x2c 0x04 0x28 0x20 |
| set $p0 0x1 ge u32 $r1 $r2 |
| $p0 sub b32 $r1 $r1 $r2 |
| $p0 add b32 $r0 $r0 0x1 |
| $p0 set $p0 0x1 ge u32 $r1 $r2 |
| $p0 sub b32 $r1 $r1 $r2 |
| $p0 add b32 $r0 $r0 0x1 |
| $p3 cvt s32 $r0 neg s32 $r0 |
| sched 0x04 0x2e 0x04 0x28 0x04 0x20 0x2c |
| $p2 cvt s32 $r1 neg s32 $r1 |
| ret |
| |
| gk110_rcp_f64: |
| gk110_rsq_f64: |
| ret |
| |
| .section #gk110_builtin_offsets |
| .b64 #gk110_div_u32 |
| .b64 #gk110_div_s32 |
| .b64 #gk110_rcp_f64 |
| .b64 #gk110_rsq_f64 |