On 09/11/2017 02:07 AM, Alex Bennée wrote: > > Richard Henderson <richard.hender...@linaro.org> writes: > >> On 09/08/2017 06:10 AM, Alex Bennée wrote: >>> Also this commit breaks RISU: >>> >>> qemu-aarch64 build/aarch64-linux-gnu/risu >>> testcases.aarch64/insn_ANDSi_RES8_ANDS_RES_ANDv_ASRV__INC.risu.bin \ >>> -t >>> testcases.aarch64/insn_ANDSi_RES8_ANDS_RES_ANDv_ASRV__INC.risu.bin.trace >>> >>> Gives: >>> >>> mismatch detail (master : apprentice): >>> V29 : 000000000000000005388083c1444242 vs >>> 00000000000000002a000e0416a30018 >>> >>> The insn is: >>> >>> 37c: 6f56a29d umull2 v29.4s, v20.8h, v6.h[1] >>> >>> Which is odd because I didn't think we'd touched that. >> >> Indeed we didn't. Still, I'll check it out next week. > > OK it would help if I had objdumped the right file: > > 36c: 0e781fdd bic v29.8b, v30.8b, v24.8b > 370: 00005af0 .inst 0x00005af0 ; undefined
Thanks. The sse pandn operand order is ... surprising. Even though I know that I still managed to get it wrong. Fixed for v2. r~