i#4550: Support EVEX encoded vsqrt instructions. #4739
Conversation
Adds support for the EVEX encodings of vsqrtps, vsqrtpd, vsqrtss, and vsqrtsd. Also adds API tests and uncomments the relevant sections of the binutils decoder tests, with updated expectation files. Fixes #4550
|
I don't think that test failure is real. |
The convention is to list the test name with a link to the filed issue (or to file a new one if convinced it's a flake). |
| {OP_vsqrtpd, 0x660f5140, "vsqrtpd", Ved, xx, KEb, Wed, xx, mrm|evex|ttfv, x, modx[119][0]}, | ||
| {MOD_EXT, 0x660f5150, "(mod ext 119)", xx, xx, xx, xx, xx, mrm|evex, x, 119}, | ||
| }, { /* evex_W_ext 266 */ | ||
| {OP_vsqrtss, 0xf30f5100, "vsqrtss", Vdq, xx, KE1b, Hdq, Wss, mrm|evex|ttt1s, x, tevexwb[266][1]}, |
There was a problem hiding this comment.
(OK, tuple types of Full Vector and Typle 1 Scalar seem to match.)
I had trouble finding a useful error message in the log, other than just that decode-stress failed. |
It seems to be there. The output is both above as verbose original and below as why it didn't match: That is #4723. |
Ah, great. |
The EVEX-prefixed vsqrtss/vsqrtsd entries should take three-quarter and half width merge operands respectively, like their VEX-prefixed counterparts. This was an oversight when I wrote DynamoRIO#4739.
The EVEX-prefixed vsqrtss/vsqrtsd entries should take three-quarter and half width merge operands respectively, like their VEX-prefixed counterparts. This was an oversight when I wrote #4739.
Adds support for the EVEX encodings of vsqrtps, vsqrtpd, vsqrtss, and vsqrtsd. Also adds API tests and uncomments the relevant sections of the binutils decoder tests, with updated expectation files.
Fixes #4550