[thirdparty/openssl.git] / crypto / ia64cpuid.S

// 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#
Commit	Line	Data
c9a41d7d	1	// Copyright 2004-2017 The OpenSSL Project Authors. All Rights Reserved.
44c8a5e2	2	//
0e9725bc	3	// Licensed under the Apache License 2.0 (the "License"). You may not use
44c8a5e2 RS	4	// this file except in compliance with the License. You can obtain a copy
	5	// in the file LICENSE in the source distribution or at
	6	// https://www.openssl.org/source/license.html
14e21f86 AP	7	// Works on all IA-64 platforms: Linux, HP-UX, Win64i...
	8	// On Win64i compile with ias.exe.
	9	.text
b2dba9bf	10
e33826f0 AP	11	#if defined(_HPUX_SOURCE) && !defined(_LP64)
	12	#define ADDP addp4
	13	#else
	14	#define ADDP add
	15	#endif
	16
591e85e9 AP	17	.global OPENSSL_cpuid_setup#
	18	.proc OPENSSL_cpuid_setup#
	19	OPENSSL_cpuid_setup:
	20	{ .mib; br.ret.sptk.many b0 };;
	21	.endp OPENSSL_cpuid_setup#
b2dba9bf	22
14e21f86 AP	23	.global OPENSSL_rdtsc#
	24	.proc OPENSSL_rdtsc#
	25	OPENSSL_rdtsc:
cee73df3 AP	26	{ .mib; mov r8=ar.itc
cee73df3 AP	27	br.ret.sptk.many b0 };;
14e21f86	28	.endp OPENSSL_rdtsc#
cee73df3 AP	29
	30	.global OPENSSL_atomic_add#
	31	.proc OPENSSL_atomic_add#
	32	.align 32
	33	OPENSSL_atomic_add:
	34	{ .mii; ld4 r2=[r32]
	35	nop.i 0
	36	nop.i 0 };;
	37	.Lspin:
	38	{ .mii; mov ar.ccv=r2
	39	add r8=r2,r33
	40	mov r3=r2 };;
5fabb88a	41	{ .mmi; mf;;
cee73df3 AP	42	cmpxchg4.acq r2=[r32],r8,ar.ccv
	43	nop.i 0 };;
	44	{ .mib; cmp.ne p6,p0=r2,r3
	45	nop.i 0
	46	(p6) br.dpnt .Lspin };;
	47	{ .mib; nop.m 0
	48	sxt4 r8=r8
	49	br.ret.sptk.many b0 };;
	50	.endp OPENSSL_atomic_add#
	51
	52	// Returns a structure comprising pointer to the top of stack of
	53	// the caller and pointer beyond backing storage for the current
	54	// register frame. The latter is required, because it might be
	55	// insufficient to wipe backing storage for the current frame
	56	// (as this procedure does), one might have to go further, toward
	57	// higher addresses to reach for whole "retroactively" saved
	58	// context...
	59	.global OPENSSL_wipe_cpu#
	60	.proc OPENSSL_wipe_cpu#
	61	.align 32
	62	OPENSSL_wipe_cpu:
	63	.prologue
	64	.fframe 0
	65	.save ar.pfs,r2
	66	.save ar.lc,r3
	67	{ .mib; alloc r2=ar.pfs,0,96,0,96
	68	mov r3=ar.lc
	69	brp.loop.imp .L_wipe_top,.L_wipe_end-16
	70	};;
	71	{ .mii; mov r9=ar.bsp
	72	mov r8=pr
	73	mov ar.lc=96 };;
	74	.body
	75	{ .mii; add r9=96*8-8,r9
	76	mov ar.ec=1 };;
	77
c9a41d7d	78	// One can sweep double as fast, but then we can't guarantee
cee73df3 AP	79	// that backing storage is wiped...
	80	.L_wipe_top:
	81	{ .mfi; st8 [r9]=r0,-8
	82	mov f127=f0
	83	mov r127=r0 }
	84	{ .mfb; nop.m 0
	85	nop.f 0
	86	br.ctop.sptk .L_wipe_top };;
	87	.L_wipe_end:
	88
	89	{ .mfi; mov r11=r0
	90	mov f6=f0
	91	mov r14=r0 }
	92	{ .mfi; mov r15=r0
	93	mov f7=f0
	94	mov r16=r0 }
	95	{ .mfi; mov r17=r0
	96	mov f8=f0
	97	mov r18=r0 }
	98	{ .mfi; mov r19=r0
	99	mov f9=f0
	100	mov r20=r0 }
	101	{ .mfi; mov r21=r0
	102	mov f10=f0
	103	mov r22=r0 }
	104	{ .mfi; mov r23=r0
	105	mov f11=f0
	106	mov r24=r0 }
	107	{ .mfi; mov r25=r0
	108	mov f12=f0
	109	mov r26=r0 }
	110	{ .mfi; mov r27=r0
	111	mov f13=f0
	112	mov r28=r0 }
	113	{ .mfi; mov r29=r0
	114	mov f14=f0
	115	mov r30=r0 }
	116	{ .mfi; mov r31=r0
	117	mov f15=f0
	118	nop.i 0 }
	119	{ .mfi; mov f16=f0 }
	120	{ .mfi; mov f17=f0 }
	121	{ .mfi; mov f18=f0 }
	122	{ .mfi; mov f19=f0 }
	123	{ .mfi; mov f20=f0 }
	124	{ .mfi; mov f21=f0 }
	125	{ .mfi; mov f22=f0 }
	126	{ .mfi; mov f23=f0 }
	127	{ .mfi; mov f24=f0 }
	128	{ .mfi; mov f25=f0 }
	129	{ .mfi; mov f26=f0 }
	130	{ .mfi; mov f27=f0 }
	131	{ .mfi; mov f28=f0 }
	132	{ .mfi; mov f29=f0 }
	133	{ .mfi; mov f30=f0 }
	134	{ .mfi; add r9=96*8+8,r9
	135	mov f31=f0
	136	mov pr=r8,0x1ffff }
	137	{ .mib; mov r8=sp
	138	mov ar.lc=r3
	139	br.ret.sptk b0 };;
	140	.endp OPENSSL_wipe_cpu#
b2dba9bf AP	141
	142	.global OPENSSL_cleanse#
	143	.proc OPENSSL_cleanse#
	144	OPENSSL_cleanse:
3f2a98ac	145	{ .mib; cmp.eq p6,p0=0,r33 // len==0
e33826f0	146	ADDP r32=0,r32
3f2a98ac	147	(p6) br.ret.spnt b0 };;
b2dba9bf AP	148	{ .mib; and r2=7,r32
	149	cmp.leu p6,p0=15,r33 // len>=15
	150	(p6) br.cond.dptk .Lot };;
	151
	152	.Little:
	153	{ .mib; st1 [r32]=r0,1
	154	cmp.ltu p6,p7=1,r33 } // len>1
	155	{ .mbb; add r33=-1,r33 // len--
	156	(p6) br.cond.dptk .Little
	157	(p7) br.ret.sptk.many b0 };;
	158
	159	.Lot:
	160	{ .mib; cmp.eq p6,p0=0,r2
	161	(p6) br.cond.dptk .Laligned };;
	162	{ .mmi; st1 [r32]=r0,1;;
	163	and r2=7,r32 }
	164	{ .mib; add r33=-1,r33
	165	br .Lot };;
	166
	167	.Laligned:
	168	{ .mmi; st8 [r32]=r0,8
	169	and r2=-8,r33 // len&~7
	170	add r33=-8,r33 };; // len-=8
	171	{ .mib; cmp.ltu p6,p0=8,r2 // ((len+8)&~7)>8
	172	(p6) br.cond.dptk .Laligned };;
	173
	174	{ .mbb; cmp.eq p6,p7=r0,r33
	175	(p7) br.cond.dpnt .Little
	176	(p6) br.ret.sptk.many b0 };;
	177	.endp OPENSSL_cleanse#
5fabb88a	178
e33826f0 AP	179	.global CRYPTO_memcmp#
	180	.proc CRYPTO_memcmp#
	181	.align 32
	182	.skip 16
	183	CRYPTO_memcmp:
	184	.prologue
	185	{ .mib; mov r8=0
	186	cmp.eq p6,p0=0,r34 // len==0?
	187	(p6) br.ret.spnt b0 };;
	188	.save ar.pfs,r2
	189	{ .mib; alloc r2=ar.pfs,3,5,0,8
	190	.save ar.lc,r3
	191	mov r3=ar.lc
	192	brp.loop.imp .Loop_cmp_ctop,.Loop_cmp_cend-16
	193	}
	194	{ .mib; sub r10=r34,r0,1
	195	.save pr,r9
	196	mov r9=pr };;
	197	{ .mii; ADDP r16=0,r32
	198	mov ar.lc=r10
	199	mov ar.ec=4 }
	200	{ .mib; ADDP r17=0,r33
	201	mov pr.rot=1<<16 };;
	202
	203	.Loop_cmp_ctop:
	204	{ .mib; (p16) ld1 r32=[r16],1
	205	(p18) xor r34=r34,r38 }
	206	{ .mib; (p16) ld1 r36=[r17],1
	207	(p19) or r8=r8,r35
	208	br.ctop.sptk .Loop_cmp_ctop };;
	209	.Loop_cmp_cend:
	210
	211	{ .mib; cmp.ne p6,p0=0,r8
	212	mov ar.lc=r3 };;
	213	{ .mib;
	214	(p6) mov r8=1
	215	mov pr=r9,0x1ffff
	216	br.ret.sptk.many b0 };;
	217	.endp CRYPTO_memcmp#
	218
5fabb88a AP	219	.global OPENSSL_instrument_bus#
5fabb88a AP	220	.proc OPENSSL_instrument_bus#
56c5f703	221	OPENSSL_instrument_bus:
5fabb88a	222	{ .mmi; mov r2=r33
e33826f0	223	ADDP r32=0,r32 }
5fabb88a AP	224	{ .mmi; mov r8=ar.itc;;
	225	mov r10=r0
	226	mov r9=r8 };;
	227
	228	{ .mmi; fc r32;;
	229	ld4 r8=[r32] };;
	230	{ .mmi; mf
	231	mov ar.ccv=r8
	232	add r8=r8,r10 };;
	233	{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
	234	};;
	235	.Loop:
	236	{ .mmi; mov r8=ar.itc;;
	237	sub r10=r8,r9 // diff=tick-lasttick
	238	mov r9=r8 };; // lasttick=tick
	239	{ .mmi; fc r32;;
	240	ld4 r8=[r32] };;
	241	{ .mmi; mf
	242	mov ar.ccv=r8
	243	add r8=r8,r10 };;
	244	{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
	245	add r33=-1,r33
	246	add r32=4,r32 };;
	247	{ .mib; cmp4.ne p6,p0=0,r33
	248	(p6) br.cond.dptk .Loop };;
	249
	250	{ .mib; sub r8=r2,r33
	251	br.ret.sptk.many b0 };;
	252	.endp OPENSSL_instrument_bus#
	253
	254	.global OPENSSL_instrument_bus2#
	255	.proc OPENSSL_instrument_bus2#
56c5f703	256	OPENSSL_instrument_bus2:
5fabb88a	257	{ .mmi; mov r2=r33 // put aside cnt
e33826f0	258	ADDP r32=0,r32 }
5fabb88a AP	259	{ .mmi; mov r8=ar.itc;;
	260	mov r10=r0
	261	mov r9=r8 };;
	262
	263	{ .mmi; fc r32;;
	264	ld4 r8=[r32] };;
	265	{ .mmi; mf
	266	mov ar.ccv=r8
	267	add r8=r8,r10 };;
	268	{ .mmi; cmpxchg4.acq r3=[r32],r8,ar.ccv
	269	};;
	270
	271	{ .mmi; mov r8=ar.itc;;
	272	sub r10=r8,r9
	273	mov r9=r8 };;
	274	.Loop2:
	275	{ .mmi; mov r11=r10 // lastdiff=diff
	276	add r34=-1,r34 };; // --max
	277	{ .mmi; fc r32;;
	278	ld4 r8=[r32]
	279	cmp4.eq p6,p0=0,r34 };;
	280	{ .mmi; mf
	281	mov ar.ccv=r8
	282	add r8=r8,r10 };;
	283	{ .mmb; cmpxchg4.acq r3=[r32],r8,ar.ccv
	284	(p6) br.cond.spnt .Ldone2 };;
	285
	286	{ .mmi; mov r8=ar.itc;;
	287	sub r10=r8,r9 // diff=tick-lasttick
	288	mov r9=r8 };; // lasttick=tick
	289	{ .mmi; cmp.ne p6,p0=r10,r11;; // diff!=lastdiff
	290	(p6) add r33=-1,r33 };; // conditional --cnt
	291	{ .mib; cmp4.ne p7,p0=0,r33
	292	(p6) add r32=4,r32 // conditional ++out
	293	(p7) br.cond.dptk .Loop2 };;
	294	.Ldone2:
	295	{ .mib; sub r8=r2,r33
	296	br.ret.sptk.many b0 };;
	297	.endp OPENSSL_instrument_bus2#