Don't use vec_disp8 encoding with the .d32 suffix
authorH.J. Lu <hjl.tools@gmail.com>
Tue, 5 Apr 2016 04:19:27 +0000 (21:19 -0700)
committerH.J. Lu <hjl.tools@gmail.com>
Tue, 5 Apr 2016 04:19:27 +0000 (21:19 -0700)
commit5be33403c378ceb9487ad14a1309392dc03a1263
treefe6ab152a83e85bc529ee95ea4091a3e42110075
parent5fdf38727b4381d074dc929fd4b027a8779108fe
Don't use vec_disp8 encoding with the .d32 suffix

Since the .d32 suffix prefers 32-bit displacement in encoding, try
vec_disp8 encoding only if i.disp_encoding != disp_encoding_32bit.

PR gas/19909
* config/tc-i386.c (check_VecOperands): Try vec_disp8 encoding
only if i.disp_encoding != disp_encoding_32bit.
* gas/testsuite/gas/i386/disp32.s: Add tests for vmovdqu64.d32.
* gas/testsuite/gas/i386/x86-64-disp32.s: Likewise.
* gas/testsuite/gas/i386/disp32.d: Updated.
* gas/testsuite/gas/i386/x86-64-disp32.d: Likewise.
gas/config/tc-i386.c
gas/testsuite/gas/i386/disp32.d
gas/testsuite/gas/i386/disp32.s
gas/testsuite/gas/i386/x86-64-disp32.d
gas/testsuite/gas/i386/x86-64-disp32.s