jlb6740 commented on Issue #1797:
Can close this PR as it is now obsolete. I've pushed a new PR #1884 that I originally intended as an update to this PR.
jlb6740 deleted a comment on Issue #1797:
Can close this PR as it is now obsolete. I've pushed a new PR #1884 that I originally intended as an update to this PR.
jlb6740 commented on Issue #1797:
@bnjbvr For the change request, I simply removed the min/max lower and will update in a future patch.
julian-seward1 commented on Issue #1797:
As a totally drive-by comment on test cases, I saw this frag in email ..
"F3450F5C94C241010000", "subss 321(%r10,%rax,8), %xmm10",
If I could suggest .. for future tests ..
(1) Don't use the same register number twice in an example. This test case won't be able to distinguish an error where the encodings for
%r10
and%xmm10
are accidentally swapped (so to speak) both registers have the same encoding (10 !)(2) Not such a problem here, but .. try to have some registers with encoding 8 and above and some with encoding 7 or below. If regs are all above or all below the 7/8 boundary, it means that bit 3 of their encodings will be the same, so the test will be unable to detect confusion in the construction of bits 2:0 of the REX prefix.
I think these and other suggestions are documented in comments at the start of the x64 encoding tests.
julian-seward1 edited a comment on Issue #1797:
As a totally drive-by comment on test cases, I saw this frag in email ..
"F3450F5C94C241010000", "subss 321(%r10,%rax,8), %xmm10",
If I could suggest .. for future tests ..
(1) Don't use the same register number twice in an example. This test case won't be able to distinguish an error where the encodings for
%r10
and%xmm10
are accidentally swapped (so to speak) since both registers have the same encoding (10 !)(2) Not such a problem here, but .. try to have some registers with encoding 8 and above and some with encoding 7 or below. If regs are all above or all below the 7/8 boundary, it means that bit 3 of their encodings will be the same, so the test will be unable to detect confusion in the construction of bits 2:0 of the REX prefix.
I think these and other suggestions are documented in comments at the start of the x64 encoding tests.
Last updated: Dec 23 2024 at 13:07 UTC