|
From: <sv...@va...> - 2012-06-03 23:14:30
|
sewardj 2012-06-04 00:14:23 +0100 (Mon, 04 Jun 2012)
New Revision: 12608
Log:
Update.
Modified files:
trunk/docs/internals/avx-notes.txt
Modified: trunk/docs/internals/avx-notes.txt (+4 -23)
===================================================================
--- trunk/docs/internals/avx-notes.txt 2012-06-04 00:13:08 -23:00 (rev 12607)
+++ trunk/docs/internals/avx-notes.txt 2012-06-04 00:14:23 -23:00 (rev 12608)
@@ -2,27 +2,8 @@
Cleanups
~~~~~~~~
-FXSAVE/FXRSTOR: can no longer say (w.r.t the guest state
-effects declaration) that the SSE regs are written/read
-in one single block. Instead need to make a declaration
-for each bottom-half independently :-(
+* convert none/tests/amd64/avx-1.c into a proper test
-in fact, re-check everything that assumes the XMM regs form
-an array, because they no longer do. Done: PCMPISTRI et al,
-Also AESENC et al.
-
-* guest state alignment, all targets -- will probably fail now
-
-* FXSAVE/FXRSTOR on amd64, as noted above
-
-* tools other than memcheck -- now fail w/ AVX insns
-
-* remove regclass HRc256
-
-* disable Avx insns in backend (or rm this code, will we
- ever need it?)
-
-* change amd64 getAllocableRegs back to what it was originally
- [DONE]
-
-* fix up none/tests/amd64/avx-1.c
+* can dis_AVX128_E_V_to_G be replaced by some variant of
+ dis_VEX_NDS_128_AnySimdPfx_0F_WIG_simple ? Seems like the former is
+ merely a subset of the latter.
|