# RUN: llvm-mc --disassemble %s -triple=x86_64-apple-darwin9 -mcpu=knl | FileCheck %s
# RUN: llvm-mc --disassemble %s -triple=x86_64-apple-darwin9 -mcpu=skx | FileCheck --check-prefix=CHECK-SKX %s

# CHECK: vpbroadcastd    %xmm18, %zmm28 {%k7} {z}
0x62 0x22 0x7d 0xcf 0x58 0xe2

# CHECK: vbroadcastss    (%rsp), %zmm28
0x62 0x62 0x7d 0x48 0x18 0x24 0x24

# CHECK: vblendmpd       (%rsi), %zmm2, %zmm8 {%k7}
0x62 0x72 0xed 0x4f 0x65 0x06

# CHECK: vpermpd (%rsi,%r10,4), %zmm2, %zmm8
0x62 0x32 0xed 0x48 0x16 0x04 0x96

# CHECK: vpbroadcastmw2d %k2, %zmm8
0x62 0x72 0x7e 0x48 0x3a 0xc2

# CHECK-SKX: vpbroadcastmw2d %k2, %xmm8
0x62 0x72 0x7e 0x08 0x3a 0xc2

# CHECK-SKX: vpbroadcastmw2d %k2, %ymm8
0x62 0x72 0x7e 0x28 0x3a 0xc2

# CHECK: vpbroadcastq    (%r9,%rax), %zmm28
0x62 0x42 0xfd 0x48 0x59 0x24 0x01

# CHECK: vbroadcastss    %xmm0, %zmm1
0x62 0xf2 0x7d 0x48 0x18 0xc8

# CHECK: vextracti32x4   $4, %zmm0, (%r10)
0x62 0xd3 0x7d 0x48 0x39 0x02 0x04

# CHECK: vextracti32x4   $4, %zmm0, %xmm1
0x62 0xf3 0x7d 0x48 0x39 0xc1 0x04

# CHECK: vinserti32x4    $1, %xmm21, %zmm5, %zmm17
0x62 0xa3 0x55 0x48 0x38 0xcd 0x01

# CHECK: vmovaps %zmm21, %zmm5 {%k3}
0x62 0xb1 0x7c 0x4b 0x28 0xed

# CHECK: vgatherdps      (%rsi,%zmm0,4), %zmm1 {%k2}
0x62 0xf2 0x7d 0x4a 0x92 0x0c 0x86

# CHECK: vgatherdpd      (%rsi,%ymm0,4), %zmm1 {%k2}
0x62 0xf2 0xfd 0x4a 0x92 0x0c 0x86

# CHECK: vgatherqps (%r13,%zmm4), %ymm5 {%k2}
0x62 0xd2 0x7d 0x4a 0x93 0x6c 0x25 0x00

# CHECK: vpslld $16, %zmm21, %zmm22
0x62 0xb1 0x4d 0x40 0x72 0xf5 0x10

# CHECK: vpord %zmm22, %zmm21, %zmm23
0x62 0xa1 0x55 0x40 0xeb 0xfe

# CHECK: vmovq   %xmm19, 1016(%rdx)
0x62 0xe1 0xfd 0x08 0x7e 0x5a 0x7f

#####################################################
#                MASK INSTRUCTIONS                  #
#####################################################

# CHECK: kshiftlw        $3, %k1, %k2
0xc4 0xe3 0xf9 0x32 0xd1 0x03

# CHECK: kmovw   (%rdi), %k1
0xc5 0xf8 0x90 0x0f

# CHECK: kmovw   %k1, %eax
0xc5 0xf8 0x93 0xc1

# CHECK: kandw   %k1, %k2, %k3
0xc5 0xec 0x41 0xd9

# CHECK: kmovw   %k5, %k1
0xc5 0xf8 0x90 0xcd

#####################################################
#             COMPRESSED DISPLACEMENT               #
#####################################################

# TupleType = FVM
# CHECK: vmovdqu32 %zmm0, -448(%rcx)
0x62 0xf1 0x7e 0x48 0x7f 0x41 0xf9

# TupleType = T1S, 64-bit eltsize
# CHECK: vaddsd 256(%rdx), %xmm0, %xmm16
0x62 0xe1 0xff 0x08 0x58 0x42 0x20

# TupleType = T1S, 32-bit eltsize
# CHECK: vaddss 256(%rdx), %xmm0, %xmm16
0x62 0xe1 0x7e 0x08 0x58 0x42 0x40

# TupleType = FV
# CHECK: vaddpd 256(%rdx), %zmm0, %zmm16
0x62 0xe1 0xfd 0x48 0x58 0x42 0x04

# TupleType = FV, broadcast, 64-bit eltsize
# CHECK: vaddpd 256(%rdx){1to8}, %zmm0, %zmm16
0x62 0xe1 0xfd 0x58 0x58 0x42 0x20

# TupleType = FV, broadcast, 32-bit eltsize
# CHECK: vaddps 256(%rdx){1to16}, %zmm0, %zmm16
0x62 0xe1 0x7c 0x58 0x58 0x42 0x40

# TupleType = T4
# CHECK: vbroadcasti32x4 256(%rdx), %zmm16
0x62 0xe2 0x7d 0x48 0x5a 0x42 0x10

# Cases where we can't use cdisp8
# CHECK: vaddss 255(%rdx), %xmm0, %xmm16
0x62 0xe1 0x7e 0x08 0x58 0x82 0xff 0x00 0x00 0x00

# CHECK: vaddss 1024(%rdx), %xmm0, %xmm16
0x62 0xe1 0x7e 0x08 0x58 0x82 0x00 0x04 0x00 0x00

# CHECK: vpcmpeqd %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x0

# CHECK: vpcmpltd %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x1

# CHECK: vpcmpled %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x2

# CHECK: vpcmpd $3, %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x3

# CHECK: vpcmpneqd %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x4

# CHECK: vpcmpnltd %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x5

# CHECK: vpcmpnled %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x6

# CHECK: vpcmpd $7, %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x7

# CHECK: vpcmpd $8, %zmm10, %zmm25, %k5
0x62 0xd3 0x35 0x40 0x1f 0xea 0x8

# CHECK: vcmppd	$127, {sae}, %zmm27, %zmm11, %k4
0x62 0x91 0xa5 0x58 0xc2 0xe3 0x7f

# CHECK: vcmpsd	$204, (%rax), %xmm4, %k5
0x62 0xf1 0xdf 0x8 0xc2 0x28 0xcc

# CHECK: vcmpss	$204, (%rax), %xmm4, %k5
0x62 0xf1 0x5e 0x08 0xc2 0x28 0xcc

# CHECK: vcmpsd	$204, %xmm3, %xmm4, %k5
0x62 0xf1 0xdf 0x08 0xc2 0xeb 0xcc

# CHECK: vcmpss	$204, %xmm3, %xmm4, %k5
0x62 0xf1 0x5e 0x08 0xc2 0xeb 0xcc

# CHECK: vcmpsd	$204, {sae}, %xmm3, %xmm4, %k5
0x62 0xf1 0xdf 0x18 0xc2 0xeb 0xcc

# CHECK: vcmpss	$204, {sae}, %xmm3, %xmm4, %k5
0x62 0xf1 0x5e 0x18 0xc2 0xeb 0xcc

# CHECK: vcmppd	$127, %ymm27, %ymm11, %k4
0x62 0x91 0xa5 0x28 0xc2 0xe3 0x7f

# CHECK: vcmpps	$127, %ymm27, %ymm11, %k4
0x62 0x91 0x24 0x28 0xc2 0xe3 0x7f

# CHECK: vcmppd	$127, %xmm27, %xmm11, %k4
0x62 0x91 0xa5 0x08 0xc2 0xe3 0x7f

# CHECK: vcmpps	$127, %xmm27, %xmm11, %k4
0x62 0x91 0x24 0x08 0xc2 0xe3 0x7f

# CHECK: vcmpps	$37, %xmm1, %xmm0, %k1 {%k7}
0x62 0xf1 0x7c 0x0f 0xc2 0xc9 0x25

# CHECK: vcmpps	$37, %ymm1, %ymm0, %k1 {%k7}
0x62 0xf1 0x7c 0x2f 0xc2 0xc9 0x25

# CHECK: vpgatherdd 256(%r9,%xmm31), %xmm17 {%k1}
0x62 0x82 0x7d 0x01 0x90 0x4c 0x39 0x40

# CHECK: vpgatherdd 256(%r9,%ymm31), %ymm19 {%k1}
0x62 0x82 0x7d 0x21 0x90 0x5c 0x39 0x40

# CHECK: vpgatherdq 256(%r9,%xmm31), %xmm17 {%k1}
0x62 0x82 0xfd 0x01 0x90 0x4c 0x39 0x20

# CHECK: vpgatherdq 256(%r9,%xmm31), %ymm26 {%k1}
0x62 0x02 0xfd 0x21 0x90 0x54 0x39 0x20

# CHECK: vpgatherqd 256(%r9,%xmm31), %xmm21 {%k1}
0x62 0x82 0x7d 0x01 0x91 0x6c 0x39 0x40

# CHECK: vpgatherqd 256(%r9,%ymm31), %xmm25 {%k1}
0x62 0x02 0x7d 0x21 0x91 0x4c 0x39 0x40

# CHECK: vpgatherqq 256(%r9,%xmm31), %xmm18 {%k1}
0x62 0x82 0xfd 0x01 0x91 0x54 0x39 0x20

# CHECK: vpgatherqq 256(%r9,%ymm31), %ymm19 {%k1}
0x62 0x82 0xfd 0x21 0x91 0x5c 0x39 0x20

# CHECK: vgatherdpd 256(%r9,%xmm31), %xmm17 {%k1}
0x62 0x82 0xfd 0x01 0x92 0x4c 0x39 0x20

# CHECK: vgatherdpd 256(%r9,%xmm31), %ymm23 {%k1}
0x62 0x82 0xfd 0x21 0x92 0x7c 0x39 0x20

# CHECK: vgatherdps 256(%r9,%xmm31), %xmm18 {%k1}
0x62 0x82 0x7d 0x01 0x92 0x54 0x39 0x40

# CHECK: vgatherdps 256(%r9,%ymm31), %ymm27 {%k1}
0x62 0x02 0x7d 0x21 0x92 0x5c 0x39 0x40

# CHECK: vgatherqpd 256(%r9,%xmm31), %xmm17 {%k1}
0x62 0x82 0xfd 0x01 0x93 0x4c 0x39 0x20

# CHECK: vgatherqpd 256(%r9,%ymm31), %ymm29 {%k1}
0x62 0x02 0xfd 0x21 0x93 0x6c 0x39 0x20

# CHECK: vgatherqps 256(%r9,%xmm31), %xmm21 {%k1}
0x62 0x82 0x7d 0x01 0x93 0x6c 0x39 0x40

# CHECK: vgatherqps 256(%r9,%ymm31), %xmm19 {%k1}
0x62 0x82 0x7d 0x21 0x93 0x5c 0x39 0x40

# CHECK: vpscatterdd %xmm20, 256(%r9,%xmm31) {%k1}
0x62 0x82 0x7d 0x01 0xa0 0x64 0x39 0x40

# CHECK: vpscatterdd %ymm28, 256(%r9,%ymm31) {%k1}
0x62 0x02 0x7d 0x21 0xa0 0x64 0x39 0x40

# CHECK: vpscatterdq %xmm21, 256(%r9,%xmm31) {%k1}
0x62 0x82 0xfd 0x01 0xa0 0x6c 0x39 0x20

# CHECK: vpscatterdq %ymm28, 256(%r9,%xmm31) {%k1}
0x62 0x02 0xfd 0x21 0xa0 0x64 0x39 0x20

# CHECK: vpscatterqd %xmm22, 256(%r9,%xmm31) {%k1}
0x62 0x82 0x7d 0x01 0xa1 0x74 0x39 0x40

# CHECK: vpscatterqd %xmm24, 256(%r9,%ymm31) {%k1}
0x62 0x02 0x7d 0x21 0xa1 0x44 0x39 0x40

# CHECK: vpscatterqq %xmm28, 256(%r9,%xmm31) {%k1}
0x62 0x02 0xfd 0x01 0xa1 0x64 0x39 0x20

# CHECK: vpscatterqq %ymm19, 256(%r9,%ymm31) {%k1}
0x62 0x82 0xfd 0x21 0xa1 0x5c 0x39 0x20

# CHECK: vscatterdps %xmm20, 256(%r9,%xmm31) {%k1}
0x62 0x82 0x7d 0x01 0xa2 0x64 0x39 0x40

# CHECK: vscatterdps %ymm28, 256(%r9,%ymm31) {%k1}
0x62 0x02 0x7d 0x21 0xa2 0x64 0x39 0x40

# CHECK: vscatterdpd %xmm21, 256(%r9,%xmm31) {%k1}
0x62 0x82 0xfd 0x01 0xa2 0x6c 0x39 0x20

# CHECK: vscatterdpd %ymm28, 256(%r9,%xmm31) {%k1}
0x62 0x02 0xfd 0x21 0xa2 0x64 0x39 0x20

# CHECK: vscatterqps %xmm22, 256(%r9,%xmm31) {%k1}
0x62 0x82 0x7d 0x01 0xa3 0x74 0x39 0x40

# CHECK: vscatterqps %xmm24, 256(%r9,%ymm31) {%k1}
0x62 0x02 0x7d 0x21 0xa3 0x44 0x39 0x40

# CHECK: vscatterqpd %xmm28, 256(%r9,%xmm31) {%k1}
0x62 0x02 0xfd 0x01 0xa3 0x64 0x39 0x20

# CHECK: vscatterqpd %ymm19, 256(%r9,%ymm31) {%k1}
0x62 0x82 0xfd 0x21 0xa3 0x5c 0x39 0x20

#####################################################
#                  SAE ATTRIBUTE                    #
#####################################################

# CHECK: vcomisd {sae}, %xmm2, %xmm1
0x62 0xf1 0xfd 0x18 0x2f 0xca

# Same as above but ignore EVEX L'L bits.
# CHECK: vcomisd {sae}, %xmm2, %xmm1
0x62 0xf1 0xfd 0x78 0x2f 0xca

# CHECK: vminpd  {sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0xf5 0x10 0x5d 0xda

# Ignore EVEX L'L bits.
# CHECK: vminpd  {sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0xf5 0x30 0x5d 0xda

# Ignore EVEX L'L bits.
# CHECK: vminpd  {sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0xf5 0x50 0x5d 0xda

# Ignore EVEX L'L bits.
# CHECK: vminpd  {sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0xf5 0x70 0x5d 0xda

# CHECK: vcmppd  $127, {sae}, %zmm27, %zmm11, %k4
0x62 0x91 0xa5 0x18 0xc2 0xe3 0x7f

# CHECK: vrsqrt28pd {sae}, %zmm2, %zmm17
0x62 0xe2 0xfd 0x18 0xcc 0xca

#####################################################
#                 ROUNDING CONTROL                  #
#####################################################

# Verify all rounding modes work.

# CHECK: vaddps {rn-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0x74 0x10 0x58 0xda

# CHECK: vaddps {rd-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0x74 0x30 0x58 0xda

# CHECK: vaddps {ru-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0x74 0x50 0x58 0xda

# CHECK: vaddps {rz-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe1 0x74 0x70 0x58 0xda

# CHECK: vmulss {rn-sae}, %xmm2, %xmm17, %xmm19
0x62 0xe1 0x76 0x10 0x59 0xda

# CHECK: vmulss {rd-sae}, %xmm2, %xmm17, %xmm19
0x62 0xe1 0x76 0x30 0x59 0xda

# CHECK: vmulss {ru-sae}, %xmm2, %xmm17, %xmm19
0x62 0xe1 0x76 0x50 0x59 0xda

# CHECK: vmulss {rz-sae}, %xmm2, %xmm17, %xmm19
0x62 0xe1 0x76 0x70 0x59 0xda

# CHECK: vscalefpd {rn-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe2 0xf5 0x10 0x2c 0xda

# CHECK: vscalefpd {rd-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe2 0xf5 0x30 0x2c 0xda

# CHECK: vscalefpd {ru-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe2 0xf5 0x50 0x2c 0xda

# CHECK: vscalefpd {rz-sae}, %zmm2, %zmm17, %zmm19
0x62 0xe2 0xf5 0x70 0x2c 0xda

# CHECK: vcvtqq2ps {rd-sae}, %zmm2, %ymm17
0x62 0xe1 0xfc 0x38 0x5b 0xca

# CHECK: vsqrtpd {rd-sae}, %zmm2, %zmm17
0x62 0xe1 0xfd 0x38 0x51 0xca