0e9725bcb9
[skip ci] Reviewed-by: Matt Caswell <matt@openssl.org> (Merged from https://github.com/openssl/openssl/pull/7827)
297 lines
6.3 KiB
ArmAsm
297 lines
6.3 KiB
ArmAsm
// Copyright 2004-2017 The OpenSSL Project Authors. All Rights Reserved.
|
|
//
|
|
// Licensed under the Apache License 2.0 (the "License"). You may not use
|
|
// this file except in compliance with the License. You can obtain a copy
|
|
// in the file LICENSE in the source distribution or at
|
|
// https://www.openssl.org/source/license.html
|
|
// Works on all IA-64 platforms: Linux, HP-UX, Win64i...
|
|
// On Win64i compile with ias.exe.
|
|
.text
|
|
|
|
#if defined(_HPUX_SOURCE) && !defined(_LP64)
|
|
#define ADDP addp4
|
|
#else
|
|
#define ADDP add
|
|
#endif
|
|
|
|
.global OPENSSL_cpuid_setup#
|
|
.proc OPENSSL_cpuid_setup#
|
|
OPENSSL_cpuid_setup:
|
|
{ .mib; br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_cpuid_setup#
|
|
|
|
.global OPENSSL_rdtsc#
|
|
.proc OPENSSL_rdtsc#
|
|
OPENSSL_rdtsc:
|
|
{ .mib; mov r8=ar.itc
|
|
br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_rdtsc#
|
|
|
|
.global OPENSSL_atomic_add#
|
|
.proc OPENSSL_atomic_add#
|
|
.align 32
|
|
OPENSSL_atomic_add:
|
|
{ .mii; ld4 r2=[r32]
|
|
nop.i 0
|
|
nop.i 0 };;
|
|
.Lspin:
|
|
{ .mii; mov ar.ccv=r2
|
|
add r8=r2,r33
|
|
mov r3=r2 };;
|
|
{ .mmi; mf;;
|
|
cmpxchg4.acq r2=[r32],r8,ar.ccv
|
|
nop.i 0 };;
|
|
{ .mib; cmp.ne p6,p0=r2,r3
|
|
nop.i 0
|
|
(p6) br.dpnt .Lspin };;
|
|
{ .mib; nop.m 0
|
|
sxt4 r8=r8
|
|
br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_atomic_add#
|
|
|
|
// Returns a structure comprising pointer to the top of stack of
|
|
// the caller and pointer beyond backing storage for the current
|
|
// register frame. The latter is required, because it might be
|
|
// insufficient to wipe backing storage for the current frame
|
|
// (as this procedure does), one might have to go further, toward
|
|
// higher addresses to reach for whole "retroactively" saved
|
|
// context...
|
|
.global OPENSSL_wipe_cpu#
|
|
.proc OPENSSL_wipe_cpu#
|
|
.align 32
|
|
OPENSSL_wipe_cpu:
|
|
.prologue
|
|
.fframe 0
|
|
.save ar.pfs,r2
|
|
.save ar.lc,r3
|
|
{ .mib; alloc r2=ar.pfs,0,96,0,96
|
|
mov r3=ar.lc
|
|
brp.loop.imp .L_wipe_top,.L_wipe_end-16
|
|
};;
|
|
{ .mii; mov r9=ar.bsp
|
|
mov r8=pr
|
|
mov ar.lc=96 };;
|
|
.body
|
|
{ .mii; add r9=96*8-8,r9
|
|
mov ar.ec=1 };;
|
|
|
|
// One can sweep double as fast, but then we can't guarantee
|
|
// that backing storage is wiped...
|
|
.L_wipe_top:
|
|
{ .mfi; st8 [r9]=r0,-8
|
|
mov f127=f0
|
|
mov r127=r0 }
|
|
{ .mfb; nop.m 0
|
|
nop.f 0
|
|
br.ctop.sptk .L_wipe_top };;
|
|
.L_wipe_end:
|
|
|
|
{ .mfi; mov r11=r0
|
|
mov f6=f0
|
|
mov r14=r0 }
|
|
{ .mfi; mov r15=r0
|
|
mov f7=f0
|
|
mov r16=r0 }
|
|
{ .mfi; mov r17=r0
|
|
mov f8=f0
|
|
mov r18=r0 }
|
|
{ .mfi; mov r19=r0
|
|
mov f9=f0
|
|
mov r20=r0 }
|
|
{ .mfi; mov r21=r0
|
|
mov f10=f0
|
|
mov r22=r0 }
|
|
{ .mfi; mov r23=r0
|
|
mov f11=f0
|
|
mov r24=r0 }
|
|
{ .mfi; mov r25=r0
|
|
mov f12=f0
|
|
mov r26=r0 }
|
|
{ .mfi; mov r27=r0
|
|
mov f13=f0
|
|
mov r28=r0 }
|
|
{ .mfi; mov r29=r0
|
|
mov f14=f0
|
|
mov r30=r0 }
|
|
{ .mfi; mov r31=r0
|
|
mov f15=f0
|
|
nop.i 0 }
|
|
{ .mfi; mov f16=f0 }
|
|
{ .mfi; mov f17=f0 }
|
|
{ .mfi; mov f18=f0 }
|
|
{ .mfi; mov f19=f0 }
|
|
{ .mfi; mov f20=f0 }
|
|
{ .mfi; mov f21=f0 }
|
|
{ .mfi; mov f22=f0 }
|
|
{ .mfi; mov f23=f0 }
|
|
{ .mfi; mov f24=f0 }
|
|
{ .mfi; mov f25=f0 }
|
|
{ .mfi; mov f26=f0 }
|
|
{ .mfi; mov f27=f0 }
|
|
{ .mfi; mov f28=f0 }
|
|
{ .mfi; mov f29=f0 }
|
|
{ .mfi; mov f30=f0 }
|
|
{ .mfi; add r9=96*8+8,r9
|
|
mov f31=f0
|
|
mov pr=r8,0x1ffff }
|
|
{ .mib; mov r8=sp
|
|
mov ar.lc=r3
|
|
br.ret.sptk b0 };;
|
|
.endp OPENSSL_wipe_cpu#
|
|
|
|
.global OPENSSL_cleanse#
|
|
.proc OPENSSL_cleanse#
|
|
OPENSSL_cleanse:
|
|
{ .mib; cmp.eq p6,p0=0,r33 // len==0
|
|
ADDP r32=0,r32
|
|
(p6) br.ret.spnt b0 };;
|
|
{ .mib; and r2=7,r32
|
|
cmp.leu p6,p0=15,r33 // len>=15
|
|
(p6) br.cond.dptk .Lot };;
|
|
|
|
.Little:
|
|
{ .mib; st1 [r32]=r0,1
|
|
cmp.ltu p6,p7=1,r33 } // len>1
|
|
{ .mbb; add r33=-1,r33 // len--
|
|
(p6) br.cond.dptk .Little
|
|
(p7) br.ret.sptk.many b0 };;
|
|
|
|
.Lot:
|
|
{ .mib; cmp.eq p6,p0=0,r2
|
|
(p6) br.cond.dptk .Laligned };;
|
|
{ .mmi; st1 [r32]=r0,1;;
|
|
and r2=7,r32 }
|
|
{ .mib; add r33=-1,r33
|
|
br .Lot };;
|
|
|
|
.Laligned:
|
|
{ .mmi; st8 [r32]=r0,8
|
|
and r2=-8,r33 // len&~7
|
|
add r33=-8,r33 };; // len-=8
|
|
{ .mib; cmp.ltu p6,p0=8,r2 // ((len+8)&~7)>8
|
|
(p6) br.cond.dptk .Laligned };;
|
|
|
|
{ .mbb; cmp.eq p6,p7=r0,r33
|
|
(p7) br.cond.dpnt .Little
|
|
(p6) br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_cleanse#
|
|
|
|
.global CRYPTO_memcmp#
|
|
.proc CRYPTO_memcmp#
|
|
.align 32
|
|
.skip 16
|
|
CRYPTO_memcmp:
|
|
.prologue
|
|
{ .mib; mov r8=0
|
|
cmp.eq p6,p0=0,r34 // len==0?
|
|
(p6) br.ret.spnt b0 };;
|
|
.save ar.pfs,r2
|
|
{ .mib; alloc r2=ar.pfs,3,5,0,8
|
|
.save ar.lc,r3
|
|
mov r3=ar.lc
|
|
brp.loop.imp .Loop_cmp_ctop,.Loop_cmp_cend-16
|
|
}
|
|
{ .mib; sub r10=r34,r0,1
|
|
.save pr,r9
|
|
mov r9=pr };;
|
|
{ .mii; ADDP r16=0,r32
|
|
mov ar.lc=r10
|
|
mov ar.ec=4 }
|
|
{ .mib; ADDP r17=0,r33
|
|
mov pr.rot=1<<16 };;
|
|
|
|
.Loop_cmp_ctop:
|
|
{ .mib; (p16) ld1 r32=[r16],1
|
|
(p18) xor r34=r34,r38 }
|
|
{ .mib; (p16) ld1 r36=[r17],1
|
|
(p19) or r8=r8,r35
|
|
br.ctop.sptk .Loop_cmp_ctop };;
|
|
.Loop_cmp_cend:
|
|
|
|
{ .mib; cmp.ne p6,p0=0,r8
|
|
mov ar.lc=r3 };;
|
|
{ .mib;
|
|
(p6) mov r8=1
|
|
mov pr=r9,0x1ffff
|
|
br.ret.sptk.many b0 };;
|
|
.endp CRYPTO_memcmp#
|
|
|
|
.global OPENSSL_instrument_bus#
|
|
.proc OPENSSL_instrument_bus#
|
|
OPENSSL_instrument_bus:
|
|
{ .mmi; mov r2=r33
|
|
ADDP r32=0,r32 }
|
|
{ .mmi; mov r8=ar.itc;;
|
|
mov r10=r0
|
|
mov r9=r8 };;
|
|
|
|
{ .mmi; fc r32;;
|
|
ld4 r8=[r32] };;
|
|
{ .mmi; mf
|
|
mov ar.ccv=r8
|
|
add r8=r8,r10 };;
|
|
{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
|
|
};;
|
|
.Loop:
|
|
{ .mmi; mov r8=ar.itc;;
|
|
sub r10=r8,r9 // diff=tick-lasttick
|
|
mov r9=r8 };; // lasttick=tick
|
|
{ .mmi; fc r32;;
|
|
ld4 r8=[r32] };;
|
|
{ .mmi; mf
|
|
mov ar.ccv=r8
|
|
add r8=r8,r10 };;
|
|
{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
|
|
add r33=-1,r33
|
|
add r32=4,r32 };;
|
|
{ .mib; cmp4.ne p6,p0=0,r33
|
|
(p6) br.cond.dptk .Loop };;
|
|
|
|
{ .mib; sub r8=r2,r33
|
|
br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_instrument_bus#
|
|
|
|
.global OPENSSL_instrument_bus2#
|
|
.proc OPENSSL_instrument_bus2#
|
|
OPENSSL_instrument_bus2:
|
|
{ .mmi; mov r2=r33 // put aside cnt
|
|
ADDP r32=0,r32 }
|
|
{ .mmi; mov r8=ar.itc;;
|
|
mov r10=r0
|
|
mov r9=r8 };;
|
|
|
|
{ .mmi; fc r32;;
|
|
ld4 r8=[r32] };;
|
|
{ .mmi; mf
|
|
mov ar.ccv=r8
|
|
add r8=r8,r10 };;
|
|
{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
|
|
};;
|
|
|
|
{ .mmi; mov r8=ar.itc;;
|
|
sub r10=r8,r9
|
|
mov r9=r8 };;
|
|
.Loop2:
|
|
{ .mmi; mov r11=r10 // lastdiff=diff
|
|
add r34=-1,r34 };; // --max
|
|
{ .mmi; fc r32;;
|
|
ld4 r8=[r32]
|
|
cmp4.eq p6,p0=0,r34 };;
|
|
{ .mmi; mf
|
|
mov ar.ccv=r8
|
|
add r8=r8,r10 };;
|
|
{ .mmb; cmpxchg4.acq r3=[r32],r8,ar.ccv
|
|
(p6) br.cond.spnt .Ldone2 };;
|
|
|
|
{ .mmi; mov r8=ar.itc;;
|
|
sub r10=r8,r9 // diff=tick-lasttick
|
|
mov r9=r8 };; // lasttick=tick
|
|
{ .mmi; cmp.ne p6,p0=r10,r11;; // diff!=lastdiff
|
|
(p6) add r33=-1,r33 };; // conditional --cnt
|
|
{ .mib; cmp4.ne p7,p0=0,r33
|
|
(p6) add r32=4,r32 // conditional ++out
|
|
(p7) br.cond.dptk .Loop2 };;
|
|
.Ldone2:
|
|
{ .mib; sub r8=r2,r33
|
|
br.ret.sptk.many b0 };;
|
|
.endp OPENSSL_instrument_bus2#
|