This is an archive of the discontinued LLVM Phabricator instance.

Differential D59843

[PowerPC] Add support for __dcbf builtin
ClosedPublic

Authored by saghir on Mar 26 2019, 1:47 PM.

Details

Reviewers
kbarton
nemanjai
stefanp
lei
amyk
hfinkel
Commits
rL359517: Add __builtin_dcbf support for PPC
rG3962d6da1779: Add __builtin_dcbf support for PPC
rC359517: Add __builtin_dcbf support for PPC
Summary

This patch is intended to add support for __dcbf builtin.

Purpose:
Data Cache Block Flush: Copies the contents of a modified block from the data cache to main memory and flushes the copy from the data cache.

Prototype:
void __dcbf(const void* addr);

Diff Detail

Event Timeline

saghir created this revision.Mar 26 2019, 1:47 PM
Herald added a project: Restricted Project. · View Herald TranscriptMar 26 2019, 1:47 PM
Herald added a subscriber: jsji. · View Herald Transcript
saghir edited reviewers, added: hfinkel; removed: llvm-commits.Mar 26 2019, 1:51 PM
saghir added a subscriber: llvm-commits.
saghir updated this revision to Diff 192355.Mar 26 2019, 2:17 PM

Updated test in dcbf.ll

stefanp accepted this revision.Mar 28 2019, 11:19 AM

Minor nit that can be fixed when this patch is committed.
LGTM.

llvm/test/CodeGen/PowerPC/dcbf.ll
10

nit:
I don't think that you need to check the basic block label here. It does not have anything to do with what you are trying to test.

This revision is now accepted and ready to land.Mar 28 2019, 11:19 AM
amyk accepted this revision.Mar 28 2019, 11:26 AM

This looks good to me.

saghir updated this revision to Diff 193957.Apr 5 2019, 1:25 PM

Removed check for basic block in dcbf.ll

steven.zhang added a subscriber: steven.zhang.Apr 7 2019, 6:08 PM

It would be great if we could add some documentation for these new added builtins, to make more people to know what we have done. i.e. clang/docs/LanguageExtensions.rst

saghir updated this revision to Diff 194442.Apr 9 2019, 9:50 PM

Added documentation in clang/docs/LanguageExtensions.rst

amyk accepted this revision.Apr 10 2019, 1:52 PM

I think this still looks good to me, with the addition of the documentation.

steven.zhang added a comment.Apr 10 2019, 8:17 PM

Looks good, just one minor comments.
It seems that, we have already had the PowerPC section in the rst. It would be better to put them together.

PowerPC Language Extensions
------------------------------

Set the Floating Point Rounding Mode
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
saghir updated this revision to Diff 195094.Apr 14 2019, 10:22 PM

Moved cache builtins section to PowerPC Language extensions in LanguageExtensions.rst

steven.zhang added a comment.Apr 15 2019, 3:29 AM

I am ok now for the doc part change. Thank you.

stefanp accepted this revision.Apr 15 2019, 10:59 AM

Looks good to me.
Then only thing I noticed was that in the description in the LanguageExtensions.rst file you use plural to refer to the chache builtins. Of course, you only added one builtin but I know that in the future you plan to add more builtins of the same type so I wouldn't worry about that wording now.

saghir added a comment.Apr 15 2019, 1:49 PM
In D59843#1467249, @stefanp wrote:

Looks good to me.
Then only thing I noticed was that in the description in the LanguageExtensions.rst file you use plural to refer to the chache builtins. Of course, you only added one builtin but I know that in the future you plan to add more builtins of the same type so I wouldn't worry about that wording now.

That is precisely the reason for using plural since we plan to add support for more cache builtins.

Closed by commit rC359517: Add __builtin_dcbf support for PPC (authored by saghir). · Explain WhyApr 29 2019, 4:23 PM
This revision was automatically updated to reflect the committed changes.
Herald added a project: Restricted Project. · View Herald TranscriptApr 29 2019, 4:23 PM
Herald added subscribers: cfe-commits, kristina. · View Herald Transcript

Revision Contents

PathSize
clang/
include/
clang/
Basic/
3 lines
test/
CodeGen/
47 lines
llvm/
include/
llvm/
IR/
3 lines
test/
CodeGen/
PowerPC/
16 lines

Diff 192355

clang/include/clang/Basic/BuiltinsPPC.def

Show First 20 Lines • Show All 469 Lines • ▼ Show 20 Lines
BUILTIN(__builtin_divde, "SLLiSLLiSLLi", "") BUILTIN(__builtin_divde, "SLLiSLLiSLLi", "")
BUILTIN(__builtin_divdeu, "ULLiULLiULLi", "") BUILTIN(__builtin_divdeu, "ULLiULLiULLi", "")
BUILTIN(__builtin_bpermd, "SLLiSLLiSLLi", "") BUILTIN(__builtin_bpermd, "SLLiSLLiSLLi", "")
// Vector int128 (un)pack // Vector int128 (un)pack
BUILTIN(__builtin_unpack_vector_int128, "ULLiV1LLLii", "") BUILTIN(__builtin_unpack_vector_int128, "ULLiV1LLLii", "")
BUILTIN(__builtin_pack_vector_int128, "V1LLLiULLiULLi", "") BUILTIN(__builtin_pack_vector_int128, "V1LLLiULLiULLi", "")
// Cache built-ins
BUILTIN(__builtin_dcbf, "vvC*", "")
// FIXME: Obviously incomplete. // FIXME: Obviously incomplete.
#undef BUILTIN #undef BUILTIN

clang/test/CodeGen/builtins-ppc-cache.c

  • This file was added.
// RUN: %clang_cc1 -triple powerpc64le-unknown-linux-gnu -emit-llvm \
// RUN: -o - %s | FileCheck %s
int A;
int B[5];
float C;
float D[5];
double E;
double F[5];
void func(int a, int b[], float c, float d[], double e, double f[]) {
__builtin_dcbf (&a);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&A);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&b[2]);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&B[2]);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&c);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&C);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&d[2]);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&D[2]);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&e);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&E);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&f[0]);
// CHECK: @llvm.ppc.dcbf(i8*
__builtin_dcbf (&F[0]);
// CHECK: @llvm.ppc.dcbf(i8*
}

llvm/include/llvm/IR/IntrinsicsPowerPC.td

Show All 12 Lines
//===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===//
// Definitions for all PowerPC intrinsics. // Definitions for all PowerPC intrinsics.
// //
// Non-altivec intrinsics. // Non-altivec intrinsics.
let TargetPrefix = "ppc" in { // All intrinsics start with "llvm.ppc.". let TargetPrefix = "ppc" in { // All intrinsics start with "llvm.ppc.".
// dcba/dcbf/dcbi/dcbst/dcbt/dcbz/dcbzl(PPC970) instructions. // dcba/dcbf/dcbi/dcbst/dcbt/dcbz/dcbzl(PPC970) instructions.
def int_ppc_dcba : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcba : Intrinsic<[], [llvm_ptr_ty], []>;
def int_ppc_dcbf : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcbf : GCCBuiltin<"__builtin_dcbf">,
Intrinsic<[], [llvm_ptr_ty], []>;
def int_ppc_dcbi : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcbi : Intrinsic<[], [llvm_ptr_ty], []>;
def int_ppc_dcbst : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcbst : Intrinsic<[], [llvm_ptr_ty], []>;
def int_ppc_dcbt : Intrinsic<[], [llvm_ptr_ty], def int_ppc_dcbt : Intrinsic<[], [llvm_ptr_ty],
[IntrArgMemOnly, NoCapture<0>]>; [IntrArgMemOnly, NoCapture<0>]>;
def int_ppc_dcbtst: Intrinsic<[], [llvm_ptr_ty], def int_ppc_dcbtst: Intrinsic<[], [llvm_ptr_ty],
[IntrArgMemOnly, NoCapture<0>]>; [IntrArgMemOnly, NoCapture<0>]>;
def int_ppc_dcbz : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcbz : Intrinsic<[], [llvm_ptr_ty], []>;
def int_ppc_dcbzl : Intrinsic<[], [llvm_ptr_ty], []>; def int_ppc_dcbzl : Intrinsic<[], [llvm_ptr_ty], []>;
▲ Show 20 Lines • Show All 1,140 Lines • Show Last 20 Lines

llvm/test/CodeGen/PowerPC/dcbf.ll

  • This file was added.
; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu < %s \
; RUN: -verify-machineinstrs -ppc-asm-full-reg-names \
; RUN: -ppc-vsr-nums-as-vr | FileCheck %s
; Function Attrs: nounwind
define void @dcbf_test(i8* %a) {
entry:
tail call void @llvm.ppc.dcbf(i8* %a)
; CHECK-LABEL: @dcbf_test
; CHECK: # %bb.0:
stefanpUnsubmitted
Not Done
ReplyInline Actions

nit:
I don't think that you need to check the basic block label here. It does not have anything to do with what you are trying to test.

stefanp: nit: I don't think that you need to check the basic block label here. It does not have…
; CHECK: dcbf 0, r3
; CHECK-NEXT: blr
ret void
}
declare void @llvm.ppc.dcbf(i8*)