summaryrefslogtreecommitdiff
path: root/libc/string/ia64/memcpy.S
diff options
context:
space:
mode:
Diffstat (limited to 'libc/string/ia64/memcpy.S')
-rw-r--r--libc/string/ia64/memcpy.S160
1 files changed, 80 insertions, 80 deletions
diff --git a/libc/string/ia64/memcpy.S b/libc/string/ia64/memcpy.S
index 810eb0c0e..6c48a72d9 100644
--- a/libc/string/ia64/memcpy.S
+++ b/libc/string/ia64/memcpy.S
@@ -42,8 +42,8 @@
#define LFETCH_DIST 500
-#define ALIGN_UNROLL_no 4 // no. of elements
-#define ALIGN_UNROLL_sh 2 // (shift amount)
+#define ALIGN_UNROLL_no 4 /* no. of elements */
+#define ALIGN_UNROLL_sh 2 /* (shift amount) */
#define MEMLAT 8
#define Nrot ((4*(MEMLAT+2) + 7) & ~7)
@@ -168,76 +168,76 @@ ENTRY(memcpy)
.rotr r[MEMLAT+1], s[MEMLAT+2], q[MEMLAT+1], t[MEMLAT+1]
.rotp p[MEMLAT+2]
.rotf fr[MEMLAT+1], fq[MEMLAT+1], fs[MEMLAT+1], ft[MEMLAT+1]
- mov ret0 = in0 // return tmp2 = dest
+ mov ret0 = in0 /* return tmp2 = dest */
.save pr, saved_pr
- movi0 saved_pr = pr // save the predicate registers
+ movi0 saved_pr = pr /* save the predicate registers */
} { .mmi
- and tmp4 = 7, in0 // check if destination is aligned
- mov dest = in0 // dest
- mov src = in1 // src
+ and tmp4 = 7, in0 /* check if destination is aligned */
+ mov dest = in0 /* dest */
+ mov src = in1 /* src */
;; }
{ .mii
- cmp.eq p_scr, p0 = in2, r0 // if (len == 0)
+ cmp.eq p_scr, p0 = in2, r0 /* if (len == 0) */
.save ar.lc, saved_lc
- movi0 saved_lc = ar.lc // save the loop counter
+ movi0 saved_lc = ar.lc /* save the loop counter */
.body
- cmp.ge p_few, p0 = OP_T_THRES, in2 // is len <= OP_T_THRESH
+ cmp.ge p_few, p0 = OP_T_THRES, in2 /* is len <= OP_T_THRESH */
} { .mbb
- mov len = in2 // len
-(p_scr) br.cond.dpnt.few .restore_and_exit // Branch no. 1: return dest
-(p_few) br.cond.dpnt.many .copy_bytes // Branch no. 2: copy byte by byte
+ mov len = in2 /* len */
+(p_scr) br.cond.dpnt.few .restore_and_exit /* Branch no. 1: return dest */
+(p_few) br.cond.dpnt.many .copy_bytes /* Branch no. 2: copy byte by byte */
;; }
{ .mmi
#if defined(USE_LFETCH)
- lfetch.nt1 [dest] //
- lfetch.nt1 [src] //
+ lfetch.nt1 [dest] /* */
+ lfetch.nt1 [src] /* */
#endif
- shr.u elemcnt = len, 3 // elemcnt = len / 8
+ shr.u elemcnt = len, 3 /* elemcnt = len / 8 */
} { .mib
- cmp.eq p_scr, p0 = tmp4, r0 // is destination aligned?
- sub loopcnt = 7, tmp4 //
+ cmp.eq p_scr, p0 = tmp4, r0 /* is destination aligned? */
+ sub loopcnt = 7, tmp4 /* */
(p_scr) br.cond.dptk.many .dest_aligned
;; }
{ .mmi
- ld1 tmp2 = [src], 1 //
- sub len = len, loopcnt, 1 // reduce len
- movi0 ar.lc = loopcnt //
+ ld1 tmp2 = [src], 1 /* */
+ sub len = len, loopcnt, 1 /* reduce len */
+ movi0 ar.lc = loopcnt /* */
} { .mib
- cmp.ne p_scr, p0 = 0, loopcnt // avoid loading beyond end-point
+ cmp.ne p_scr, p0 = 0, loopcnt /* avoid loading beyond end-point */
;; }
-.l0: // ---------------------------- // L0: Align src on 8-byte boundary
+.l0: /* ---------------------------- L0: Align src on 8-byte boundary */
{ .mmi
- st1 [dest] = tmp2, 1 //
-(p_scr) ld1 tmp2 = [src], 1 //
+ st1 [dest] = tmp2, 1 /* */
+(p_scr) ld1 tmp2 = [src], 1 /* */
} { .mib
- cmp.lt p_scr, p0 = 1, loopcnt // avoid load beyond end-point
+ cmp.lt p_scr, p0 = 1, loopcnt /* avoid load beyond end-point */
add loopcnt = -1, loopcnt
- br.cloop.dptk.few .l0 //
+ br.cloop.dptk.few .l0 /* */
;; }
.dest_aligned:
{ .mmi
- and tmp4 = 7, src // ready for alignment check
- shr.u elemcnt = len, 3 // elemcnt = len / 8
+ and tmp4 = 7, src /* ready for alignment check */
+ shr.u elemcnt = len, 3 /* elemcnt = len / 8 */
;; }
{ .mib
- cmp.ne p_scr, p0 = tmp4, r0 // is source also aligned
- tbit.nz p_xtr, p_nxtr = src, 3 // prepare a separate move if src
-} { .mib // is not 16B aligned
- add ptr2 = LFETCH_DIST, dest // prefetch address
+ cmp.ne p_scr, p0 = tmp4, r0 /* is source also aligned */
+ tbit.nz p_xtr, p_nxtr = src, 3 /* prepare a separate move if src */
+} { .mib /* is not 16B aligned */
+ add ptr2 = LFETCH_DIST, dest /* prefetch address */
add ptr1 = LFETCH_DIST, src
(p_scr) br.cond.dptk.many .src_not_aligned
;; }
-// The optimal case, when dest, and src are aligned
+/* The optimal case, when dest, and src are aligned */
.both_aligned:
{ .mmi
.pred.rel "mutex",p_xtr,p_nxtr
-(p_xtr) cmp.gt p_scr, p0 = ALIGN_UNROLL_no+1, elemcnt // Need N + 1 to qualify
-(p_nxtr) cmp.gt p_scr, p0 = ALIGN_UNROLL_no, elemcnt // Need only N to qualify
- movi0 pr.rot = 1 << 16 // set rotating predicates
+(p_xtr) cmp.gt p_scr, p0 = ALIGN_UNROLL_no+1, elemcnt /* Need N + 1 to qualify */
+(p_nxtr) cmp.gt p_scr, p0 = ALIGN_UNROLL_no, elemcnt /* Need only N to qualify */
+ movi0 pr.rot = 1 << 16 /* set rotating predicates */
} { .mib
(p_scr) br.cond.dpnt.many .copy_full_words
;; }
@@ -245,21 +245,21 @@ ENTRY(memcpy)
{ .mmi
(p_xtr) load tempreg = [src], 8
(p_xtr) add elemcnt = -1, elemcnt
- movi0 ar.ec = MEMLAT + 1 // set the epilog counter
+ movi0 ar.ec = MEMLAT + 1 /* set the epilog counter */
;; }
{ .mmi
-(p_xtr) add len = -8, len //
- add asrc = 16, src // one bank apart (for USE_INT)
- shr.u loopcnt = elemcnt, ALIGN_UNROLL_sh // cater for unrolling
+(p_xtr) add len = -8, len /* */
+ add asrc = 16, src /* one bank apart (for USE_INT) */
+ shr.u loopcnt = elemcnt, ALIGN_UNROLL_sh /* cater for unrolling */
;;}
{ .mmi
add loopcnt = -1, loopcnt
-(p_xtr) store [dest] = tempreg, 8 // copy the "extra" word
+(p_xtr) store [dest] = tempreg, 8 /* copy the "extra" word */
nop.i 0
;; }
{ .mib
add adest = 16, dest
- movi0 ar.lc = loopcnt // set the loop counter
+ movi0 ar.lc = loopcnt /* set the loop counter */
;; }
#ifdef GAS_ALIGN_BREAKS_UNWIND_INFO
@@ -268,7 +268,7 @@ ENTRY(memcpy)
.align 32
#endif
#if defined(USE_FLP)
-.l1: // ------------------------------- // L1: Everything a multiple of 8
+.l1: /* ------------------------------- L1: Everything a multiple of 8 */
{ .mmi
#if defined(USE_LFETCH)
(p[0]) lfetch.nt1 [ptr2],32
@@ -290,7 +290,7 @@ ENTRY(memcpy)
br.ctop.dptk.many .l1
;; }
#elif defined(USE_INT)
-.l1: // ------------------------------- // L1: Everything a multiple of 8
+.l1: /* ------------------------------- L1: Everything a multiple of 8 */
{ .mmi
(p[0]) load the_r[0] = [src], 8
(p[0]) load the_q[0] = [asrc], 8
@@ -317,58 +317,58 @@ ENTRY(memcpy)
.copy_full_words:
{ .mib
- cmp.gt p_scr, p0 = 8, len //
- shr.u elemcnt = len, 3 //
+ cmp.gt p_scr, p0 = 8, len /* */
+ shr.u elemcnt = len, 3 /* */
(p_scr) br.cond.dpnt.many .copy_bytes
;; }
{ .mii
load tempreg = [src], 8
- add loopcnt = -1, elemcnt //
+ add loopcnt = -1, elemcnt /* */
;; }
{ .mii
- cmp.ne p_scr, p0 = 0, loopcnt //
- mov ar.lc = loopcnt //
+ cmp.ne p_scr, p0 = 0, loopcnt /* */
+ mov ar.lc = loopcnt /* */
;; }
-.l2: // ------------------------------- // L2: Max 4 words copied separately
+.l2: /* ------------------------------- L2: Max 4 words copied separately */
{ .mmi
store [dest] = tempreg, 8
-(p_scr) load tempreg = [src], 8 //
+(p_scr) load tempreg = [src], 8 /* */
add len = -8, len
} { .mib
- cmp.lt p_scr, p0 = 1, loopcnt // avoid load beyond end-point
+ cmp.lt p_scr, p0 = 1, loopcnt /* avoid load beyond end-point */
add loopcnt = -1, loopcnt
br.cloop.dptk.few .l2
;; }
.copy_bytes:
{ .mib
- cmp.eq p_scr, p0 = len, r0 // is len == 0 ?
- add loopcnt = -1, len // len--;
+ cmp.eq p_scr, p0 = len, r0 /* is len == 0 ? */
+ add loopcnt = -1, len /* len--; */
(p_scr) br.cond.spnt .restore_and_exit
;; }
{ .mii
ld1 tmp2 = [src], 1
movi0 ar.lc = loopcnt
- cmp.ne p_scr, p0 = 0, loopcnt // avoid load beyond end-point
+ cmp.ne p_scr, p0 = 0, loopcnt /* avoid load beyond end-point */
;; }
-.l3: // ------------------------------- // L3: Final byte move
+.l3: /* ------------------------------- L3: Final byte move */
{ .mmi
st1 [dest] = tmp2, 1
(p_scr) ld1 tmp2 = [src], 1
} { .mib
- cmp.lt p_scr, p0 = 1, loopcnt // avoid load beyond end-point
+ cmp.lt p_scr, p0 = 1, loopcnt /* avoid load beyond end-point */
add loopcnt = -1, loopcnt
br.cloop.dptk.few .l3
;; }
.restore_and_exit:
{ .mmi
- movi0 pr = saved_pr, -1 // restore the predicate registers
+ movi0 pr = saved_pr, -1 /* restore the predicate registers */
;; }
{ .mib
- movi0 ar.lc = saved_lc // restore the loop counter
+ movi0 ar.lc = saved_lc /* restore the loop counter */
br.ret.sptk.many b0
;; }
@@ -376,41 +376,41 @@ ENTRY(memcpy)
.src_not_aligned:
{ .mmi
cmp.gt p_scr, p0 = 16, len
- and sh1 = 7, src // sh1 = src % 8
- shr.u loopcnt = len, 4 // element-cnt = len / 16
+ and sh1 = 7, src /* sh1 = src % 8 */
+ shr.u loopcnt = len, 4 /* element-cnt = len / 16 */
} { .mib
add tmp4 = @ltoff(.table), gp
add tmp3 = @ltoff(.loop56), gp
-(p_scr) br.cond.dpnt.many .copy_bytes // do byte by byte if too few
+(p_scr) br.cond.dpnt.many .copy_bytes /* do byte by byte if too few */
;; }
{ .mmi
- and asrc = -8, src // asrc = (-8) -- align src for loop
- add loopcnt = -1, loopcnt // loopcnt--
- shl sh1 = sh1, 3 // sh1 = 8 * (src % 8)
+ and asrc = -8, src /* asrc = (-8) -- align src for loop */
+ add loopcnt = -1, loopcnt /* loopcnt-- */
+ shl sh1 = sh1, 3 /* sh1 = 8 * (src % 8) */
} { .mmi
- ld8 ptable = [tmp4] // ptable = &table
- ld8 ploop56 = [tmp3] // ploop56 = &loop56
- and tmp2 = -16, len // tmp2 = len & -OPSIZ
+ ld8 ptable = [tmp4] /* ptable = &table */
+ ld8 ploop56 = [tmp3] /* ploop56 = &loop56 */
+ and tmp2 = -16, len /* tmp2 = len & -OPSIZ */
;; }
{ .mmi
- add tmp3 = ptable, sh1 // tmp3 = &table + sh1
- add src = src, tmp2 // src += len & (-16)
- movi0 ar.lc = loopcnt // set LC
+ add tmp3 = ptable, sh1 /* tmp3 = &table + sh1 */
+ add src = src, tmp2 /* src += len & (-16) */
+ movi0 ar.lc = loopcnt /* set LC */
;; }
{ .mmi
- ld8 tmp4 = [tmp3] // tmp4 = loop offset
- sub len = len, tmp2 // len -= len & (-16)
- movi0 ar.ec = MEMLAT + 2 // one more pass needed
+ ld8 tmp4 = [tmp3] /* tmp4 = loop offset */
+ sub len = len, tmp2 /* len -= len & (-16) */
+ movi0 ar.ec = MEMLAT + 2 /* one more pass needed */
;; }
{ .mmi
- ld8 s[1] = [asrc], 8 // preload
- sub loopaddr = ploop56,tmp4 // loopadd = &loop56 - loop offset
- movi0 pr.rot = 1 << 16 // set rotating predicates
+ ld8 s[1] = [asrc], 8 /* preload */
+ sub loopaddr = ploop56,tmp4 /* loopadd = &loop56 - loop offset */
+ movi0 pr.rot = 1 << 16 /* set rotating predicates */
;; }
{ .mib
nop.m 0
movi0 b6 = loopaddr
- br b6 // jump to the appropriate loop
+ br b6 /* jump to the appropriate loop */
;; }
LOOP(8)
@@ -426,7 +426,7 @@ libc_hidden_def (memcpy)
.rodata
.align 8
.table:
- data8 0 // dummy entry
+ data8 0 /* dummy entry */
data8 .loop56 - .loop8
data8 .loop56 - .loop16
data8 .loop56 - .loop24
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-10 20:50:39 +0000