[x86 disassembler] vpblendvb operands incorrect

[llvmbot]

Bugzilla Link	10917
Resolution	FIXED
Resolved on	Nov 07, 2018 00:17
Version	trunk
OS	All
Reporter	LLVM Bugzilla Contributor
CC	@topperc

Extended Description

According to the "Intel® 64 and IA-32 Architectures Software Developer’s Manual
Volume 2 (2A & 2B): Instruction Set Reference, A-Z", p. 4-69:
VEX.NDS.128.66.0F3A.W0 4C /r /is4
VPBLENDVB xmm1, xmm2, xmm3/m128, xmm4

Select byte values from xmm2 and xmm3/m128 using mask bits in the
specified mask register, xmm4, and store the values into xmm1.

Clang on OSX assembles this:
vpblendvb %xmm0, %xmm1, %xmm2, %xmm3

as:
C4 E3 69 4C D9 00

But llvm-mc built from r139522 shows:
$ echo '0xc4 0xe3 0x69 0x4c 0xd9 0x00'| ./llvm-mc -disassemble -triple="x86_64" -x86-asm-syntax=intel
vpblendvb XMM3, XMM2, XMM1, 0

$ echo '0xc4 0xe3 0x69 0x4c 0xd9 0x00'| ./llvm-mc -disassemble -triple="x86_64"
vpblendvb $0, %xmm1, %xmm2, %xmm3

The mask register appears to be printed as an immediate operand.

[topperc]

Fixed in r139690.

[llvmbot]

Looks good with r140430:
$ echo '0xc4 0xe3 0x69 0x4c 0xd9 0x00'| ./llvm-mc -disassemble -triple="x86_64"
vpblendvb %xmm0, %xmm1, %xmm2, %xmm3

$ echo '0xc4 0xe3 0x69 0x4c 0xd9 0x00'| ./llvm-mc -disassemble -triple="x86_64" -x86-asm-syntax=intel
vpblendvb XMM3, XMM2, XMM1, XMM0

I couldn't find any documentation for the order of source operands in AT&T syntax, but it looks like it's always the exact reverse of Intel syntax?