Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 474fd6e8 authored by Martin Schwidefsky's avatar Martin Schwidefsky
Browse files

RAID/s390: add SIMD implementation for raid6 gen/xor 



Using vector registers is slightly faster:

raid6: vx128x8  gen() 19705 MB/s
raid6: vx128x8  xor() 11886 MB/s
raid6: using algorithm vx128x8 gen() 19705 MB/s
raid6: .... xor() 11886 MB/s, rmw enabled

vs the software algorithms:

raid6: int64x1  gen()  3018 MB/s
raid6: int64x1  xor()  1429 MB/s
raid6: int64x2  gen()  4661 MB/s
raid6: int64x2  xor()  3143 MB/s
raid6: int64x4  gen()  5392 MB/s
raid6: int64x4  xor()  3509 MB/s
raid6: int64x8  gen()  4441 MB/s
raid6: int64x8  xor()  3207 MB/s
raid6: using algorithm int64x4 gen() 5392 MB/s
raid6: .... xor() 3509 MB/s, rmw enabled

Signed-off-by: default avatarMartin Schwidefsky <schwidefsky@de.ibm.com>
parent 8f149ea6

arch/s390/include/asm/vx-insn.h

+86 −0
Original line number Diff line number Diff line
@@ -278,6 +278,15 @@ 
	VLVG	\v, \gr, \index, 3

.endm



/* VECTOR LOAD REGISTER */

.macro	VLR	v1, v2

	VX_NUM	v1, \v1

	VX_NUM	v2, \v2

	.word	0xE700 | ((v1&15) << 4) | (v2&15)

	.word	0

	MRXBOPC	0, 0x56, v1, v2

.endm



/* VECTOR LOAD */

.macro	VL	v, disp, index="%r0", base

	VX_NUM	v1, \v
@@ -404,6 +413,16 @@ 


/* Vector integer instructions */



/* VECTOR AND */

.macro	VN	vr1, vr2, vr3

	VX_NUM	v1, \vr1

	VX_NUM	v2, \vr2

	VX_NUM	v3, \vr3

	.word	0xE700 | ((v1&15) << 4) | (v2&15)

	.word	((v3&15) << 12)

	MRXBOPC	0, 0x68, v1, v2, v3

.endm



/* VECTOR EXCLUSIVE OR */

.macro	VX	vr1, vr2, vr3

	VX_NUM	v1, \vr1
@@ -469,6 +488,73 @@ 
	MRXBOPC	0, 0x7D, v1, v2, v3

.endm



/* VECTOR REPLICATE IMMEDIATE */

.macro	VREPI	vr1, imm2, m3

	VX_NUM	v1, \vr1

	.word	0xE700 | ((v1&15) << 4)

	.word	\imm2

	MRXBOPC	\m3, 0x45, v1

.endm

.macro	VREPIB	vr1, imm2

	VREPI	\vr1, \imm2, 0

.endm

.macro	VREPIH	vr1, imm2

	VREPI	\vr1, \imm2, 1

.endm

.macro	VREPIF	vr1, imm2

	VREPI	\vr1, \imm2, 2

.endm

.macro	VREPIG	vr1, imm2

	VREP	\vr1, \imm2, 3

.endm



/* VECTOR ADD */

.macro	VA	vr1, vr2, vr3, m4

	VX_NUM	v1, \vr1

	VX_NUM	v2, \vr2

	VX_NUM	v3, \vr3

	.word	0xE700 | ((v1&15) << 4) | (v2&15)

	.word	((v3&15) << 12)

	MRXBOPC	\m4, 0xF3, v1, v2, v3

.endm

.macro	VAB	vr1, vr2, vr3

	VA	\vr1, \vr2, \vr3, 0

.endm

.macro	VAH	vr1, vr2, vr3

	VA	\vr1, \vr2, \vr3, 1

.endm

.macro	VAF	vr1, vr2, vr3

	VA	\vr1, \vr2, \vr3, 2

.endm

.macro	VAG	vr1, vr2, vr3

	VA	\vr1, \vr2, \vr3, 3

.endm

.macro	VAQ	vr1, vr2, vr3

	VA	\vr1, \vr2, \vr3, 4

.endm



/* VECTOR ELEMENT SHIFT RIGHT ARITHMETIC */

.macro	VESRAV	vr1, vr2, vr3, m4

	VX_NUM	v1, \vr1

	VX_NUM	v2, \vr2

	VX_NUM	v3, \vr3

	.word	0xE700 | ((v1&15) << 4) | (v2&15)

	.word	((v3&15) << 12)

	MRXBOPC \m4, 0x7A, v1, v2, v3

.endm



.macro	VESRAVB	vr1, vr2, vr3

	VESRAV	\vr1, \vr2, \vr3, 0

.endm

.macro	VESRAVH	vr1, vr2, vr3

	VESRAV	\vr1, \vr2, \vr3, 1

.endm

.macro	VESRAVF	vr1, vr2, vr3

	VESRAV	\vr1, \vr2, \vr3, 2

.endm

.macro	VESRAVG	vr1, vr2, vr3

	VESRAV	\vr1, \vr2, \vr3, 3

.endm



#endif	/* __ASSEMBLY__ */

#endif	/* __ASM_S390_VX_INSN_H */

include/linux/raid/pq.h

+1 −0
Original line number Diff line number Diff line
@@ -103,6 +103,7 @@ extern const struct raid6_calls raid6_avx2x1; 
extern const struct raid6_calls raid6_avx2x2;

extern const struct raid6_calls raid6_avx2x4;

extern const struct raid6_calls raid6_tilegx8;

extern const struct raid6_calls raid6_s390vx8;



struct raid6_recov_calls {

	void (*data2)(int, size_t, int, int, void **);

lib/raid6/.gitignore

+1 −0
Original line number Diff line number Diff line
@@ -3,3 +3,4 @@ altivec*.c 
int*.c

tables.c

neon?.c

s390vx?.c

lib/raid6/Makefile

+6 −0
Original line number Diff line number Diff line
@@ -7,6 +7,7 @@ raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o 
raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o

raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o

raid6_pq-$(CONFIG_TILEGX) += tilegx8.o

raid6_pq-$(CONFIG_S390) += s390vx8.o



hostprogs-y	+= mktables
@@ -116,6 +117,11 @@ $(obj)/tilegx8.c: UNROLL := 8 
$(obj)/tilegx8.c:   $(src)/tilegx.uc $(src)/unroll.awk FORCE

	$(call if_changed,unroll)



targets += s390vx8.c

$(obj)/s390vx8.c:   UNROLL := 8

$(obj)/s390vx8.c:   $(src)/s390vx.uc $(src)/unroll.awk FORCE

	$(call if_changed,unroll)



quiet_cmd_mktable = TABLE   $@

      cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 )

lib/raid6/algos.c

+3 −0
Original line number Diff line number Diff line
@@ -68,6 +68,9 @@ const struct raid6_calls * const raid6_algos[] = { 
#endif

#if defined(CONFIG_TILEGX)

	&raid6_tilegx8,

#endif

#if defined(CONFIG_S390)

	&raid6_s390vx8,

#endif

	&raid6_intx1,

	&raid6_intx2,