Support -march=sapphirerapids for x86.
Compare with Icelake Server, it includes 14 more new features. They are
amxtile, amxint8, amxbf16, avx512bf16, avx512vp2intersect, cldemote,
enqcmd, movdir64b, movdiri, ptwrite, serialize, shstk, tsxldtrk, waitpkg.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D86503
gcc errors on this, but I'm nervous that since -mtune has been
ignored by clang for so long that there may be code bases out
there that pass 32-bit cpus to clang.
Remove _COMPAT. Drop the ARCHNAME. Remove the non-COMPAT versions
that are no longer needed.
We now only use these macros in places where we need compatibility
with libgcc/compiler-rt. So we don't need to call out _COMPAT
specifically.
setFeatureEnabled is a virtual function. setFeatureEnabledImpl
was its implementation. This split was to avoid virtual calls
when we need to call setFeatureEnabled in initFeatureMap.
With C++11 we can use 'final' on setFeatureEnabled to enable
the compiler to perform de-virtualization for the initFeatureMap
calls.
Previously we had to specify the forward and backwards feature dependencies separately which was error prone. And as dependencies have gotten more complex it was hard to be sure the transitive dependencies were handled correctly. The way it was written was also not super readable.
This patch replaces everything with a table that lists what features a feature is dependent on directly. Then we can recursively walk through the table to find the transitive dependencies. This is largely based on how we handle subtarget features in the MC layer from the tablegen descriptions.
Differential Revision: https://reviews.llvm.org/D83273
Instead of detecting the string in 2 places. Just swap the string
to 'sse4.1' or 'sse4.2' at the top of the function.
Prep work for a patch to switch the rest of this function to a
table based system. And I don't want to include 'sse4a' in the
table.
This replaces the switch statement implementation in the clang's
X86.cpp with a lookup table in X86TargetParser.cpp.
I've used constexpr and copy of the FeatureBitset from
SubtargetFeature.h to store the features in a lookup table.
After the lookup the bitset is translated into strings for use
by the rest of the frontend code.
I had to modify the implementation of the FeatureBitset to avoid
bugs in gcc 5.5 constexpr handling. It seems to not like the
same array entry to be used on the left side and right hand side
of an assignment or &= or |=. I've also used uint32_t instead of
uint64_t and sized based on the X86::CPU_FEATURE_MAX.
I've initialized the features for different CPUs outside of the
table so that we can express inheritance in an adhoc way. This
was one of the big limitations of the switch and we had resorted
to labels and gotos.
Differential Revision: https://reviews.llvm.org/D82731
Previously we inferred it if sse4.2 ended up being enabled after
all feature processing. But writing -march=nehalem -mno-sse4.2
should have popcnt enabled.
CPUs with avx always have xsave, but some CPUs without avx also
have xsave. So we shouldn't disable xsave just because avx is
disabled. This would prevent xsave from being enabled with
-march=native on CPUs with xsave and not avx.
But we also don't want -mavx -mno-avx to leave xsave eanabled.
So only enable xsave if avx is enabled after processing all features.
I thought about just not turning xsave on with avx at all, but
there might be someone out there depending on it.
These features implicitly enabled XSAVE in the frontend, but not
the backend. Disabling XSAVE in the frontend disabled XSAVEOPT, but
not the other 2. Nothing happened in the backend.
The PREFETCHW instruction was originally part of the 3DNow. But
it was given its own CPUID bit on later CPUs just before 3DNow
was deprecated.
We were setting the -mprfchw flag if -m3dnow was passed or the CPU
supported 3dnow unless -mno-prfchw was passed. But -march=native
on a CPU without the PRFCHW CPUID bit set will pass -mno-prfchw.
So -march=k8 will behave differently than -march=native on a K8
for example.
So remove this implicit setting from the frontend and instead
enable the backend to use PREFETCHW if 3dnow OR prfchw is enabled.
Also enable PRFCHW flag on amdfam10/barcelona which seems to be
where this CPUID bit was introduced. That CPU also supported
3dnow.
The PREFETCHW instruction was originally part of the 3DNow. But
it was given its own CPUID bit on later CPUs just before 3DNow
was deprecated.
We were setting the -mprfchw flag if -m3dnow was passed or the CPU
supported 3dnow unless -mno-prfchw was passed. But -march=native
on a CPU without the PRFCHW CPUID bit set will pass -mno-prfchw.
So -march=k8 will behave differently than -march=native on a K8
for example.
So remove this implicit setting from the frontend and instead
enable the backend to use PREFETCHW if 3dnow OR prfchw is enabled.
Also enable PRFCHW flag on amdfam10/barcelona which seems to be
where this CPUID bit was introduced. That CPU also supported
3dnow.
This patch removes the PROC macro in favor of CPUKind enum and a
table that contains information about CPUs.
The current information in the table is the CPU name, CPUKind enum
value, key feature for target multiversioning, and Is64Bit capable.
For the strings that are aliases, I've duplicated the information
in the table. This means there are more rows in the table than
CPUKind enums.
This replaces multiple StringSwitch's with loops through the table.
They are linear searches due to the table being more logically
ordered than alphabetical. The StringSwitch's would have also been
linear. I've used StringLiteral on the strings in the table so we
can quickly check the length while searching.
I contemplated having a CPUKind for each string so there was a 1:1
mapping, but didn't want to spread more names to the places that
use the enum.
My ultimate goal here is to store the features for each CPU as a
bitset within the table. Hoping to use constexpr to make this
composable so we can group features and inherit them. After the
table lookup we can turn the bitset into a list of strings for the
frontend. The current switch we have for selecting features for
CPUs has become difficult to maintain while trying to express
inheritance relationships.
Differential Revision: https://reviews.llvm.org/D82414
This was orignally done so we could separate the compatibility
values and the llvm internal only features into a separate entries
in the feature array. This was needed when we explicitly had to
convert the feature into the proper 32-bit chunk at every reference
and we didn't want things moving around.
Now everything is in an array and we have helper funtions or macros
to convert encoding to index. So we renumbering is no longer an
issue.
Similar to what some other targets have done. This information
could be reused by other frontends so doesn't make sense to live
in clang.
-Rename CK_Generic to CK_None to better reflect its illegalness.
-Move function for translating from string to enum into llvm.
-Call checkCPUKind directly from the string to enum translation
and update CPU kind to CK_None accordinly. Caller will use CK_None
as sentinel for bad CPU.
I'm planning to move all the CPU to feature mapping out next. As
part of that I want to devise a better way to express CPUs inheriting
features from an earlier CPU. Allowing this to be expressed in a
less rigid way than just falling through a switch. Or using gotos
as we've had to do lately.
Differential Revision: https://reviews.llvm.org/D81439
Neither gcc or icc support this. Split out from D79472. I want
to remove more, but it looks like icc does support some things
gcc doesn't and I need to double check our internal test suites.
Y is the start of several 2 letter constraints, but we also had
partial support to recognize it by itself. But it doesn't look
like it can get through clang as a single letter so the backend
support for this was effectively dead.
gcc supports selecting ymm0/zmm0 for the Yz constraint when used with 256 or 512 bit vector types.
Fixes PR45806
Differential Revision: https://reviews.llvm.org/D79448
Summary:
This patch adds a virtual method `getCPUCacheLineSize()` to `TargetInfo`. Currently, I've only implemented the method in `X86TargetInfo`. It's extremely important that each CPU's cache line size correct (e.g., we can't just define it as `64` across the board) so, it has been a little slow getting to this point.
I'll work on the ARM CPUs next, but that will probably come later in a different patch.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D74918
The validateOutputSize and validateInputSize need to check whether
AVX or AVX512 are enabled. But this can be affected by the
target attribute so we need to factor that in.
This patch moves some of the code from CodeGen to create an
appropriate feature map that we can pass to the function.
Differential Revision: https://reviews.llvm.org/D68627
The validateOutputSize and validateInputSize need to check whether
AVX or AVX512 are enabled. But this can be affected by the
target attribute so we need to factor that in.
This patch copies some of the code from CodeGen to create an
appropriate feature map that we can pass to the function. Probably
need some refactoring here to share more code with Codegen. Is
there a good place to do that? Also need to support the cpu_specific
attribute as well.
Differential Revision: https://reviews.llvm.org/D68627
All SSE capable CPUs have MMX. 3dnow implicitly enables MMX.
We have code that detects if sse is enabled and implicitly enables
MMX unless -mno-mmx is passed. So in most cases we were already
enabling MMX if march passed a CPU that supported SSE.
The exception to this is if you pass -march for a cpu supports SSE
and also pass -mno-sse. We should still enable MMX since its part
of the CPU capability.
-Deprecate -mmpx and -mno-mpx command line options
-Remove CPUID detection of mpx for -march=native
-Remove MPX from all CPUs
-Remove MPX preprocessor define
I've left the "mpx" string in the backend so we don't fail on old IR, but its not connected to anything.
gcc has also deprecated these command line options. https://www.phoronix.com/scan.php?page=news_item&px=GCC-Patch-To-Drop-MPX
Differential Revision: https://reviews.llvm.org/D66669
llvm-svn: 370393
Support -march=tigerlake for x86.
Compare with Icelake Client, It include 4 more new features ,they are
avx512vp2intersect, movdiri, movdir64b, shstk.
Patch by Xiang Zhang (xiangzhangllvm)
Differential Revision: https://reviews.llvm.org/D65840
llvm-svn: 368543
The recently added cooperlake CPU has made our already ugly switch statement even worse. There's a CPU exclusion list around the bf16 feature in the cooper lake block. I worry that we'll have to keep adding new CPUs to that until bf16 intercepts a client space CPU. We have several other exclusion lists in other parts of the switch due to skylakeserver, cascadelake, and cooperlake not having sgx. Another for cannonlake not having clwb but having all other features from skx.
This removes all these special ifs at the cost of some duplication of features and a goto. I've copied all of the skx features into either cannonlake or icelakeclient(for clwb). And pulled sklyakeserver, cascadelake, and cooperlake out of the main inheritance chain into their own chain. At the end of skylakeserver we merge back into the main chain at skylakeclient but below sgx. I think this is at least easier to follow.
Differential Revision: https://reviews.llvm.org/D63018
llvm-svn: 362965
We had an if statement that checked over every avx512* feature to see if it should enabled avx512f. Since they are all prefixed with avx512 just check for that instead.
llvm-svn: 361557
Previously we were doing this so that the 256 bit selectw builtin could be used in the implementation of the 512->256 bit conversion intrinsic.
After this commit we now use a masked convert builtin that will emit the intrinsic call and the 256-bit select from custom code in CGBuiltin. Then the header only needs to call that one intrinsic.
llvm-svn: 360924
Summary:
1. Enable infrastructure of AVX512_BF16, which is supported for BFLOAT16 in Cooper Lake;
2. Enable intrinsics for VCVTNE2PS2BF16, VCVTNEPS2BF16 and DPBF16PS instructions, which are Vector Neural Network Instructions supporting BFLOAT16 inputs and conversion instructions from IEEE single precision.
For more details about BF16 intrinsic, please refer to the latest ISE document: https://software.intel.com/en-us/download/intel-architecture-instruction-set-extensions-programming-reference
Patch by LiuTianle
Reviewers: craig.topper, smaslov, LuoYuanke, wxiao3, annita.zhang, spatel, RKSimon
Reviewed By: craig.topper
Subscribers: mgorny, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D60552
llvm-svn: 360018
Use the new cx8 feature flag that was added to the backend to represent support for cmpxchg8b. Use this flag to set the MaxAtomicInlineWidth.
This also assumes all the cmpxchg instructions are enabled for CK_Generic which is what cc1 defaults to when nothing is specified.
Differential Revision: https://reviews.llvm.org/D59566
llvm-svn: 356709
Freddy Ye