Leverage the method OpenCL uses that adds C intrinsics when the lookup
failed. There is no need to define C intrinsics in the header file any
more. It could help to avoid the large header file to speed up the
compilation of RVV source code. Besides that, only the C intrinsics used
by the users will be added into the declaration table.
This patch is based on https://reviews.llvm.org/D103228 and inspired by
OpenCL implementation.
### Experimental Results
#### TL;DR:
- Binary size of clang increase ~200k, which is +0.07% for debug build and +0.13% for release build.
- Single file compilation speed up ~33x for debug build and ~8.5x for release build
- Regression time reduce ~10% (`ninja check-all`, enable all targets)
#### Header size change
```
| size | LoC |
------------------------------
Before | 4,434,725 | 69,749 |
After | 6,140 | 162 |
```
#### Single File Compilation Time
Testcase:
```
#include <riscv_vector.h>
vint32m1_t test_vadd_vv_vfloat32m1_t(vint32m1_t op1, vint32m1_t op2, size_t vl) {
return vadd(op1, op2, vl);
}
```
##### Debug build:
Before:
```
real 0m19.352s
user 0m19.252s
sys 0m0.092s
```
After:
```
real 0m0.576s
user 0m0.552s
sys 0m0.024s
```
~33x speed up for debug build
##### Release build:
Before:
```
real 0m0.773s
user 0m0.741s
sys 0m0.032s
```
After:
```
real 0m0.092s
user 0m0.080s
sys 0m0.012s
```
~8.5x speed up for release build
#### Regression time
Note: the failed case is `tools/llvm-debuginfod-find/debuginfod.test` which is unrelated to this patch.
##### Debug build
Before:
```
Testing Time: 1358.38s
Skipped : 11
Unsupported : 446
Passed : 75767
Expectedly Failed: 190
Failed : 1
```
After
```
Testing Time: 1220.29s
Skipped : 11
Unsupported : 446
Passed : 75767
Expectedly Failed: 190
Failed : 1
```
##### Release build
Before:
```
Testing Time: 381.98s
Skipped : 12
Unsupported : 1407
Passed : 74765
Expectedly Failed: 176
Failed : 1
```
After:
```
Testing Time: 346.25s
Skipped : 12
Unsupported : 1407
Passed : 74765
Expectedly Failed: 176
Failed : 1
```
#### Binary size of clang
##### Debug build
Before
```
text data bss dec hex filename
335261851 12726004 552812 348540667 14c64efb bin/clang
```
After
```
text data bss dec hex filename
335442803 12798708 552940 348794451 14ca2e53 bin/clang
```
+253K, +0.07% code size
##### Release build
Before
```
text data bss dec hex filename
144123975 8374648 483140 152981763 91e5103 bin/clang
```
After
```
text data bss dec hex filename
144255762 8447296 483268 153186326 9217016 bin/clang
```
+204K, +0.13%
Authored-by: Kito Cheng <kito.cheng@sifive.com>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Reviewed By: khchen, aaron.ballman
Differential Revision: https://reviews.llvm.org/D111617
RVV C intrinsics use pointers to scalar for base address and their corresponding
IR intrinsics but use pointers to vector. It makes some vector load intrinsics
need specific ManualCodegen and MaskedManualCodegen to just add bitcast for
transforming to IR.
For simplifying riscv_vector.td, the patch make RISCVEmitter detect pointer
operands and bitcast them.
Reviewed By: kito-cheng
Differential Revision: https://reviews.llvm.org/D129043
This patch does the same thing as D125886 did.
- Use `Overloaded` rather than `Mangled`.
- Use `Prototype` or `Desc` rather than `Seq`, it's not just a string
sequence.
Reviewed By: fakepaper56
Differential Revision: https://reviews.llvm.org/D126634
- Use `Overloaded` rather than `Mangled`.
- Use `Prototype` or `Desc` rather than `Seq`, it's not just a string sequence.
- `Prototype` for those PrototypeDescriptor will used to evaluate as argument type.
- `Desc` for those PrototypeDescriptor will used to evaluate as part of function name.
Reviewed By: khchen
Differential Revision: https://reviews.llvm.org/D125886
This patch is preparation for D111617, use class/struct/enum rather than char/StringRef to present internal information as possible, that provide more compact way to store those info and also easier to serialize/deserialize.
And also that improve readability of the code, e.g. "v" vs TypeProfile::Vector.
Reviewed By: khchen
Differential Revision: https://reviews.llvm.org/D124730
We add a new clang library called `clangSupport` for putting those utils which can be used in clang table-gen and other clang component.
We tried to put that into `llvm/Support`, but actually those stuffs only used in clang* and clang-tblgen, so I think that might be better to create `clang/Support`
* clang will used that in https://reviews.llvm.org/D111617.
Reviewed By: khchen, MaskRay, aaron.ballman
Differential Revision: https://reviews.llvm.org/D121984
Reland Note: We've resolve the circular dependency issue on llvm/lib/Support and
llvm/TableGen.
Differential Revision: https://reviews.llvm.org/D121984
1. Rename nomask as unmasked to keep with the terminology in the spec.
2. Merge UnMaskpolicy and Maskedpolicy arguments into one in RVVBuiltin class.
3. Rename HasAutoDef as HasBuiltinAlias.
4. Move header definition code into one class.
Reviewed By: rogfer01
Differential Revision: https://reviews.llvm.org/D120870
Currently we allow half types in vectors if the scalar Zfh extension
is enabled. This behavior is not inline with the vector spec. For f32
and f64 types, the Zve32f, Zve64f, Zve64d, and V explicitly control
the availablity of floating point types in vectors.
In order to make our compiler compliant, we either need to remove all support
for half in vectors or we need an extension to control it.
Draft spec here https://github.com/riscv/riscv-v-spec/pull/780
Reviewed By: kito-cheng
Differential Revision: https://reviews.llvm.org/D121345
The goal is support tail and mask policy in RVV builtins.
We focus on IR part first.
The nomask vector Multiply-Add need a policy operand
because merge value could not be undef.
Reviewed By: monkchiang
Differential Revision: https://reviews.llvm.org/D119727
The goal is support tail and mask policy in RVV builtins.
We focus on IR part first.
If the passthru operand is undef, we use tail agnostic, otherwise
use tail undisturbed.
Add passthru operand for VSLIDE1UP_VL and VSLIDE1DOWN_VL to support
i64 scalar in rv32.
The masked VSLIDE1 would only emit mask undisturbed policy regardless
of giving mask agnostic policy until InsertVSETVLI supports mask agnostic.
Reviewed by: craig.topper, rogfer01
Differential Revision: https://reviews.llvm.org/D117989
According to v-spec 1.0, `vmulh`, `vmulhu`, `vmulhsu` and `vsmul` are
NOT supported for EEW=64 in Zve64*.
This patch tries to guard it correctly.
Authored by: Craig Topper <craig.topper@sifive.com> @craig.topper
Co-Authored by: Eop Chen <eop.chen@sifive.com> @eopXD
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D117913
The field 'RequiredExtensions' is used to specify the constraint for rvv builtin, and it contains something which is not a sub-extension or extension such as 'RV64'. So the word 'extension' is not accurate now, 'feature' seems better.
Differential Revision: https://reviews.llvm.org/D118015
According to the spec, there are some difference between V and Zve64d. For example, the vmulh integer multiply variants that return the high word of the product (vmulh.vv, vmulh.vx, vmulhu.vv, vmulhu.vx, vmulhsu.vv, vmulhsu.vx) are not included for EEW=64 in Zve64*, but V extension does support these instructions. So we should decouple Zve* extensions and the V extension.
Differential Revision: https://reviews.llvm.org/D117854
The `zve` extension specifies the maximum ELEN for both integer and floating
point mode - defined by macro `__riscv_v_elen` and `__riscv_v_elen_fp`.
This commit restricts the functions in riscv_vector.h by the zve defined
macro-s.
Change enum `RISCVExtension` to `RISCVPredefinedMacro` since now it
contains not only extensions. Also added type alignment to it.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D112986
This string no longer appears in the Vector Extension specification.
The segment load/store instructions are just part of the vector
instruction set.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D117724
According to v-spec v1.0, `zve-32x` is the new minimum extension to include
to have vector instructions.
Reviewed By: kito-cheng
Differential Revision: https://reviews.llvm.org/D112613
All but 2 of the vector builtins are only used by clang_builtin_alias.
When using clang_builtin_alias, the type string of the builtin is never
checked. Only the types in the function definition used for the alias
are checked.
This patch takes advantage of this to share a single builtin for
many different types. We already used type overloads on the IR intrinsic
so the codegen for the builtins that are being merge were already
the same. This extends the type overloading to the builtins.
I had to make a few tweaks to make this work.
-Floating point vector-vector vmerge now uses the vmerge intrinsic
instead of the vfmerge intrinsic. New isel patterns and tests are
added to support this.
-The SemaChecking for the immediate of vset_v/vget_v has been removed.
Determining the valid range is harder now. I've added masking to
ManualCodegen to ensure valid IR for invalid input.
This reduces the number of builtins from ~25000 to ~1100.
Reviewed By: HsiangKai
Differential Revision: https://reviews.llvm.org/D112102
Zvamo is not part of the 1.0 V spec. Remove the intrinsics
for now. This helps reduce clang binary size and lit test time.
Reviewed By: HsiangKai
Differential Revision: https://reviews.llvm.org/D111692
Previously we used builtin_alias for overloaded intrinsics, but
macros for the non-overloaded version. This patch changes the
non-overloaded versions to also use builtin_alias, but without
the overloadable attribute.
Reviewed By: khchen, HsiangKai
Differential Revision: https://reviews.llvm.org/D112020
It was being used to control the nothrow attribute on the builtins. The
nothrow attribute is for C++ exceptions. Even if the vector builtins
have side effects in IR, that's different than the nothrow attribute.
Reviewed By: HsiangKai
Differential Revision: https://reviews.llvm.org/D112028
Similar to SVE, this separates the RVV builtlins into their own
region of builtin IDs. Only those IDs are allowed to be used by
the builtin_alias attribute now.
Reviewed By: HsiangKai
Differential Revision: https://reviews.llvm.org/D111923
This reverts commit 97f0c63783.
As discussed in https://reviews.llvm.org/D110684, it increased the
compile time and the binary size of clang more than 1%. I reverted
this patch first to think about a better way to do it.
In the original design, we levarage _mt intrinsics to define macros for
_m intrinsics. Such as,
```
__builtin_rvv_vadd_vv_i8m1_mt((vbool8_t)(op0), (vint8m1_t)(op1), (vint8m1_t)(op2), (vint8m1_t)(op3), (size_t)(op4), (size_t)VE_TAIL_AGNOSTIC)
```
However, we could not define generic interface for mask intrinsics any
more due to clang_builtin_alias only accepts clang builtins as its
argument.
In the example,
```
__rvv_overloaded
__attribute__((clang_builtin_alias(__builtin_rvv_vadd_vv_i8m1_mt)))
vint8m1_t vadd(vbool8_t op0, vint8m1_t op1, vint8m1_t op2, vint8m1_t
op3, size_t op4, size_t op5);
```
op5 is the tail policy argument. When users want to use vadd generic
interface for masked vector add, they need to specify tail policy in the
previous design. In this patch, we define _m intrinsics as clang
builtins to solve the problem.
Differential Revision: https://reviews.llvm.org/D110684
The IR intrinsics use ImmArg for the policy operand so this needs to be enforced as a constant in the frontend.
Differential Revision: https://reviews.llvm.org/D110779
Add the tail policy argument to Clang builtins. There
are two policies for tail elements. Tail agnostic means users do not
care about the values in the tail elements and tail undisturbed means
the values in the tail elements need to be kept after the operation. In
order to let users control the tail policy, we add an additional
argument at the end of the argument list.
For unmasked operations, we have no maskedoff and the tail policy is
always tail agnostic. If users want to keep tail elements under unmasked
operations, they could use all one mask in the masked operations to do
it. So, we only add the additional argument for masked operations for
most cases. There are exceptions listed below.
In this patch, we do not handle the following cases to reduce the
complexity of the patch. There could be two separate patches for them.
Use dest argument to control tail policy
vmerge.vvm/vmerge.vxm/vmerge.vim (add _t builtins with additional dest
argument)
vfmerge.vfm (add _t builtins with additional dest argument)
vmv.v.v (add _t builtins with additional dest argument)
vmv.v.x (add _t builtins with additional dest argument)
vmv.v.i (add _t builtins with additional dest argument)
vfmv.v.f (add _t builtins with additional dest argument)
vadc.vvm/vadc.vxm/vadc.vim (add _t builtins with additional dest
argument)
vsbc.vvm/vsbc.vxm (add _t builtins with additional dest argument)
Always has tail argument for masked/unmasked intrinsics
Vector Single-Width Integer Multiply-Add Instructions (add _t and _mt
builtins)
Vector Widening Integer Multiply-Add Instructions (add _t and _mt
builtins)
Vector Single-Width Floating-Point Fused Multiply-Add Instructions (add
_t and _mt builtins)
Vector Widening Floating-Point Fused Multiply-Add Instructions (add _t
and _mt builtins)
Vector Reduction Operations (add _t and _mt builtins)
Vector Slideup Instructions (add _t and _mt builtins)
Vector Slidedown Instructions (add _t and _mt builtins)
Discussion: https://github.com/riscv/rvv-intrinsic-doc/pull/101
Differential Revision: https://reviews.llvm.org/D109322
Use _Float16 as the half-precision floating point type. Define a new
type specifier 'x' for the _Float16 type.
Differential Revision: https://reviews.llvm.org/D105001
Kito Cheng