[1/2] x86: correct ISA-used version recording
Checks
Context |
Check |
Description |
linaro-tcwg-bot/tcwg_binutils_build--master-arm |
success
|
Build passed
|
linaro-tcwg-bot/tcwg_binutils_build--master-aarch64 |
success
|
Build passed
|
linaro-tcwg-bot/tcwg_binutils_check--master-aarch64 |
success
|
Test passed
|
linaro-tcwg-bot/tcwg_binutils_check--master-arm |
success
|
Test passed
|
Commit Message
Updating should be based solely on the current instruction. For example,
recording of VEX-encoded insns as v3 should be independent of there
being earlier AMX insns.
Further for BASELINE only a very limited set of the
GNU_PROPERTY_X86_FEATURE_2_* bits should actually be taken into account:
Most of the bits represent advanced (later) features (XSAVE, XSAVEOPT,
and XSAVEC for example being part of v3).
---
The latter part still feels wrong: xstate_xmm is entirely insn operand
dependent, so e.g. an AVX512VL insn would wrongly set baseline when it
really is a v4 insn.
There's more severe bitrotting here when it comes to - first and
foremost - APX, but also other more or less recent additions like
USER-MSR. APX insns like LZCNT with eGPR operand(s) would wrongly set
v3. VEX-encoded URDMSR would also wrongly set v3. EVEX-encoded insns
would wrongly set v4. Question is whether this machinery (a) should be
kept at all, and if so (b) won't better be re-written from scratch.
Comments
> From: Jan Beulich <jbeulich@suse.com>
> Sent: Tuesday, January 21, 2025 9:39 PM
>
> Updating should be based solely on the current instruction. For example,
> recording of VEX-encoded insns as v3 should be independent of there being
> earlier AMX insns.
>
> Further for BASELINE only a very limited set of the
> GNU_PROPERTY_X86_FEATURE_2_* bits should actually be taken into
> account:
> Most of the bits represent advanced (later) features (XSAVE, XSAVEOPT, and
> XSAVEC for example being part of v3).
> ---
> The latter part still feels wrong: xstate_xmm is entirely insn operand
> dependent, so e.g. an AVX512VL insn would wrongly set baseline when it
> really is a v4 insn.
>
> There's more severe bitrotting here when it comes to - first and foremost -
> APX, but also other more or less recent additions like USER-MSR. APX insns like
> LZCNT with eGPR operand(s) would wrongly set v3. VEX-encoded URDMSR
> would also wrongly set v3. EVEX-encoded insns would wrongly set v4.
> Question is whether this machinery (a) should be kept at all, and if so (b)
> won't better be re-written from scratch.
>
> --- a/gas/config/tc-i386.c
> +++ b/gas/config/tc-i386.c
> @@ -12173,21 +12173,23 @@ output_insn (const struct last_insn *las
> #ifdef OBJ_ELF
> if (x86_used_note && now_seg != absolute_section)
> {
> + unsigned int feature_2_used = 0;
> +
> if ((i.xstate & xstate_tmm) == xstate_tmm
> || is_cpu (&i.tm, CpuAMX_TILE))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_TMM;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_TMM;
>
> if (is_cpu (&i.tm, Cpu8087)
> || is_cpu (&i.tm, Cpu287)
> || is_cpu (&i.tm, Cpu387)
> || is_cpu (&i.tm, Cpu687)
> || is_cpu (&i.tm, CpuFISTTP))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_X87;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_X87;
>
> if ((i.xstate & xstate_mmx)
> || i.tm.mnem_off == MN_emms
> || i.tm.mnem_off == MN_femms)
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MMX;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MMX;
>
> if (i.index_reg)
> {
> @@ -12210,25 +12212,29 @@ output_insn (const struct last_insn *las
> || is_cpu (&i.tm, CpuAVX)))
> || is_cpu (&i.tm, CpuWideKL)
> || is_cpu (&i.tm, CpuKL))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XMM;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XMM;
>
> if ((i.xstate & xstate_ymm) == xstate_ymm)
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_YMM;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_YMM;
> if ((i.xstate & xstate_zmm) == xstate_zmm)
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_ZMM;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_ZMM;
> if (i.mask.reg || (i.xstate & xstate_mask) == xstate_mask)
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MASK;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MASK;
> if (is_cpu (&i.tm, CpuFXSR))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_FXSR;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_FXSR;
> if (is_cpu (&i.tm, CpuXsave))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVE;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVE;
> if (is_cpu (&i.tm, CpuXsaveopt))
> - x86_feature_2_used |=
> GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT;
> if (is_cpu (&i.tm, CpuXSAVEC))
> - x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEC;
> + feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEC;
> +
> + x86_feature_2_used |= feature_2_used;
>
> if (object_64bit
> - || x86_feature_2_used
> + || (feature_2_used
> + & (GNU_PROPERTY_X86_FEATURE_2_XMM
> + | GNU_PROPERTY_X86_FEATURE_2_FXSR)) != 0
> || is_cpu (&i.tm, CpuCMOV)
> || is_cpu (&i.tm, CpuSYSCALL)
> || i.tm.mnem_off == MN_cmpxchg8b)
> @@ -12256,13 +12262,13 @@ output_insn (const struct last_insn *las
> && !is_cpu (&i.tm, CpuFMA4)
> && !is_cpu (&i.tm, CpuLWP)
> && !is_cpu (&i.tm, CpuTBM)
> - && !(x86_feature_2_used &
> GNU_PROPERTY_X86_FEATURE_2_TMM))
> + && !(feature_2_used & GNU_PROPERTY_X86_FEATURE_2_TMM))
> || is_cpu (&i.tm, CpuF16C)
> || is_cpu (&i.tm, CpuFMA)
> || is_cpu (&i.tm, CpuLZCNT)
> || is_cpu (&i.tm, CpuMovbe)
> || is_cpu (&i.tm, CpuXSAVES)
> - || (x86_feature_2_used
> + || (feature_2_used
> & (GNU_PROPERTY_X86_FEATURE_2_XSAVE
> | GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT
> | GNU_PROPERTY_X86_FEATURE_2_XSAVEC)) != 0)
> --- a/gas/testsuite/gas/i386/property-6.d
> +++ b/gas/testsuite/gas/i386/property-6.d
> @@ -5,5 +5,5 @@
> Displaying notes found in: .note.gnu.property
> [ ]+Owner[ ]+Data size[ ]+Description
> GNU 0x[0-9a-f]+ NT_GNU_PROPERTY_TYPE_0
> - Properties: x86 ISA used: x86-64-baseline, x86-64-v4
> + Properties: x86 ISA used: (x86-64-baseline, )?x86-64-v4
> x86 feature used: x86, MASK
+ Sunil who is also familiar with might related part in glibc
Sunil, could you also have a look here?
Thx,
Haochen
@@ -12173,21 +12173,23 @@ output_insn (const struct last_insn *las
#ifdef OBJ_ELF
if (x86_used_note && now_seg != absolute_section)
{
+ unsigned int feature_2_used = 0;
+
if ((i.xstate & xstate_tmm) == xstate_tmm
|| is_cpu (&i.tm, CpuAMX_TILE))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_TMM;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_TMM;
if (is_cpu (&i.tm, Cpu8087)
|| is_cpu (&i.tm, Cpu287)
|| is_cpu (&i.tm, Cpu387)
|| is_cpu (&i.tm, Cpu687)
|| is_cpu (&i.tm, CpuFISTTP))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_X87;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_X87;
if ((i.xstate & xstate_mmx)
|| i.tm.mnem_off == MN_emms
|| i.tm.mnem_off == MN_femms)
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MMX;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MMX;
if (i.index_reg)
{
@@ -12210,25 +12212,29 @@ output_insn (const struct last_insn *las
|| is_cpu (&i.tm, CpuAVX)))
|| is_cpu (&i.tm, CpuWideKL)
|| is_cpu (&i.tm, CpuKL))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XMM;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XMM;
if ((i.xstate & xstate_ymm) == xstate_ymm)
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_YMM;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_YMM;
if ((i.xstate & xstate_zmm) == xstate_zmm)
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_ZMM;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_ZMM;
if (i.mask.reg || (i.xstate & xstate_mask) == xstate_mask)
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MASK;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_MASK;
if (is_cpu (&i.tm, CpuFXSR))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_FXSR;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_FXSR;
if (is_cpu (&i.tm, CpuXsave))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVE;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVE;
if (is_cpu (&i.tm, CpuXsaveopt))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT;
if (is_cpu (&i.tm, CpuXSAVEC))
- x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEC;
+ feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_XSAVEC;
+
+ x86_feature_2_used |= feature_2_used;
if (object_64bit
- || x86_feature_2_used
+ || (feature_2_used
+ & (GNU_PROPERTY_X86_FEATURE_2_XMM
+ | GNU_PROPERTY_X86_FEATURE_2_FXSR)) != 0
|| is_cpu (&i.tm, CpuCMOV)
|| is_cpu (&i.tm, CpuSYSCALL)
|| i.tm.mnem_off == MN_cmpxchg8b)
@@ -12256,13 +12262,13 @@ output_insn (const struct last_insn *las
&& !is_cpu (&i.tm, CpuFMA4)
&& !is_cpu (&i.tm, CpuLWP)
&& !is_cpu (&i.tm, CpuTBM)
- && !(x86_feature_2_used & GNU_PROPERTY_X86_FEATURE_2_TMM))
+ && !(feature_2_used & GNU_PROPERTY_X86_FEATURE_2_TMM))
|| is_cpu (&i.tm, CpuF16C)
|| is_cpu (&i.tm, CpuFMA)
|| is_cpu (&i.tm, CpuLZCNT)
|| is_cpu (&i.tm, CpuMovbe)
|| is_cpu (&i.tm, CpuXSAVES)
- || (x86_feature_2_used
+ || (feature_2_used
& (GNU_PROPERTY_X86_FEATURE_2_XSAVE
| GNU_PROPERTY_X86_FEATURE_2_XSAVEOPT
| GNU_PROPERTY_X86_FEATURE_2_XSAVEC)) != 0)
@@ -5,5 +5,5 @@
Displaying notes found in: .note.gnu.property
[ ]+Owner[ ]+Data size[ ]+Description
GNU 0x[0-9a-f]+ NT_GNU_PROPERTY_TYPE_0
- Properties: x86 ISA used: x86-64-baseline, x86-64-v4
+ Properties: x86 ISA used: (x86-64-baseline, )?x86-64-v4
x86 feature used: x86, MASK