2012-10-17 06:51:55 +08:00
|
|
|
; RUN: llc < %s -mcpu=cortex-a9 -verify-coalescing -verify-machineinstrs | FileCheck %s
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:32:64-v128:32:128-a0:0:32-n32-S32"
|
|
|
|
|
target triple = "thumbv7-apple-ios0.0.0"
|
|
|
|
|
|
|
|
|
|
; CHECK: f
|
|
|
|
|
; The vld2 and vst2 are not aligned wrt each other, the second Q loaded is the
|
|
|
|
|
; first one stored.
|
|
|
|
|
; The coalescer must find a super-register larger than QQ to eliminate the copy
|
|
|
|
|
; setting up the vst2 data.
|
|
|
|
|
; CHECK: vld2
|
|
|
|
|
; CHECK-NOT: vorr
|
|
|
|
|
; CHECK-NOT: vmov
|
|
|
|
|
; CHECK: vst2
|
|
|
|
|
define void @f(float* %p, i32 %c) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = bitcast float* %p to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld2 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8* %0, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%vld221 = extractvalue { <4 x float>, <4 x float> } %vld2, 1
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%add.ptr = getelementptr inbounds float, float* %p, i32 8
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%1 = bitcast float* %add.ptr to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst2.p0i8.v4f32(i8* %1, <4 x float> %vld221, <4 x float> undef, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; CHECK: f1
|
|
|
|
|
; FIXME: This function still has copies.
|
|
|
|
|
define void @f1(float* %p, i32 %c) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = bitcast float* %p to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld2 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8* %0, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%vld221 = extractvalue { <4 x float>, <4 x float> } %vld2, 1
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%add.ptr = getelementptr inbounds float, float* %p, i32 8
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%1 = bitcast float* %add.ptr to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld22 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8* %1, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%vld2215 = extractvalue { <4 x float>, <4 x float> } %vld22, 0
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst2.p0i8.v4f32(i8* %1, <4 x float> %vld221, <4 x float> %vld2215, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; CHECK: f2
|
|
|
|
|
; FIXME: This function still has copies.
|
|
|
|
|
define void @f2(float* %p, i32 %c) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = bitcast float* %p to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld2 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8* %0, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%vld224 = extractvalue { <4 x float>, <4 x float> } %vld2, 1
|
|
|
|
|
br label %do.body
|
|
|
|
|
|
|
|
|
|
do.body: ; preds = %do.body, %entry
|
|
|
|
|
%qq0.0.1.0 = phi <4 x float> [ %vld224, %entry ], [ %vld2216, %do.body ]
|
|
|
|
|
%c.addr.0 = phi i32 [ %c, %entry ], [ %dec, %do.body ]
|
|
|
|
|
%p.addr.0 = phi float* [ %p, %entry ], [ %add.ptr, %do.body ]
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%add.ptr = getelementptr inbounds float, float* %p.addr.0, i32 8
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%1 = bitcast float* %add.ptr to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld22 = tail call { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8* %1, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%vld2215 = extractvalue { <4 x float>, <4 x float> } %vld22, 0
|
|
|
|
|
%vld2216 = extractvalue { <4 x float>, <4 x float> } %vld22, 1
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst2.p0i8.v4f32(i8* %1, <4 x float> %qq0.0.1.0, <4 x float> %vld2215, i32 4)
|
Enable sub-sub-register copy coalescing.
It is now possible to coalesce weird skewed sub-register copies by
picking a super-register class larger than both original registers. The
included test case produces code like this:
vld2.32 {d16, d17, d18, d19}, [r0]!
vst2.32 {d18, d19, d20, d21}, [r0]
We still perform interference checking as if it were a normal full copy
join, so this is still quite conservative. In particular, the f1 and f2
functions in the included test case still have remaining copies because
of false interference.
llvm-svn: 156878
2012-05-16 07:31:35 +08:00
|
|
|
%dec = add nsw i32 %c.addr.0, -1
|
|
|
|
|
%tobool = icmp eq i32 %dec, 0
|
|
|
|
|
br i1 %tobool, label %do.end, label %do.body
|
|
|
|
|
|
|
|
|
|
do.end: ; preds = %do.body
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2015-09-30 18:56:37 +08:00
|
|
|
declare { <4 x float>, <4 x float> } @llvm.arm.neon.vld2.v4f32.p0i8(i8*, i32) nounwind readonly
|
|
|
|
|
declare void @llvm.arm.neon.vst2.p0i8.v4f32(i8*, <4 x float>, <4 x float>, i32) nounwind
|
2012-09-18 07:03:25 +08:00
|
|
|
|
|
|
|
|
; CHECK: f3
|
|
|
|
|
; This function has lane insertions that span basic blocks.
|
|
|
|
|
; The trivial REG_SEQUENCE lowering can't handle that, but the coalescer can.
|
|
|
|
|
;
|
|
|
|
|
; void f3(float *p, float *q) {
|
|
|
|
|
; float32x2_t x;
|
|
|
|
|
; x[1] = p[3];
|
|
|
|
|
; if (q)
|
|
|
|
|
; x[0] = q[0] + q[1];
|
|
|
|
|
; else
|
|
|
|
|
; x[0] = p[2];
|
|
|
|
|
; vst1_f32(p+4, x);
|
|
|
|
|
; }
|
|
|
|
|
;
|
|
|
|
|
; CHECK-NOT: vmov
|
|
|
|
|
; CHECK-NOT: vorr
|
|
|
|
|
define void @f3(float* %p, float* %q) nounwind ssp {
|
|
|
|
|
entry:
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx = getelementptr inbounds float, float* %p, i32 3
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load float, float* %arrayidx, align 4
|
2012-09-18 07:03:25 +08:00
|
|
|
%vecins = insertelement <2 x float> undef, float %0, i32 1
|
|
|
|
|
%tobool = icmp eq float* %q, null
|
|
|
|
|
br i1 %tobool, label %if.else, label %if.then
|
|
|
|
|
|
|
|
|
|
if.then: ; preds = %entry
|
2015-02-28 05:17:42 +08:00
|
|
|
%1 = load float, float* %q, align 4
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx2 = getelementptr inbounds float, float* %q, i32 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load float, float* %arrayidx2, align 4
|
2012-09-18 07:03:25 +08:00
|
|
|
%add = fadd float %1, %2
|
|
|
|
|
%vecins3 = insertelement <2 x float> %vecins, float %add, i32 0
|
|
|
|
|
br label %if.end
|
|
|
|
|
|
|
|
|
|
if.else: ; preds = %entry
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx4 = getelementptr inbounds float, float* %p, i32 2
|
2015-02-28 05:17:42 +08:00
|
|
|
%3 = load float, float* %arrayidx4, align 4
|
2012-09-18 07:03:25 +08:00
|
|
|
%vecins5 = insertelement <2 x float> %vecins, float %3, i32 0
|
|
|
|
|
br label %if.end
|
|
|
|
|
|
|
|
|
|
if.end: ; preds = %if.else, %if.then
|
|
|
|
|
%x.0 = phi <2 x float> [ %vecins3, %if.then ], [ %vecins5, %if.else ]
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%add.ptr = getelementptr inbounds float, float* %p, i32 4
|
2012-09-18 07:03:25 +08:00
|
|
|
%4 = bitcast float* %add.ptr to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst1.p0i8.v2f32(i8* %4, <2 x float> %x.0, i32 4)
|
2012-09-18 07:03:25 +08:00
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2015-09-30 18:56:37 +08:00
|
|
|
declare void @llvm.arm.neon.vst1.p0i8.v2f32(i8*, <2 x float>, i32) nounwind
|
|
|
|
|
declare <2 x float> @llvm.arm.neon.vld1.v2f32.p0i8(i8*, i32) nounwind readonly
|
2012-09-20 05:29:18 +08:00
|
|
|
|
|
|
|
|
; CHECK: f4
|
|
|
|
|
; This function inserts a lane into a fully defined vector.
|
|
|
|
|
; The destination lane isn't read, so the subregs can coalesce.
|
|
|
|
|
; CHECK-NOT: vmov
|
|
|
|
|
; CHECK-NOT: vorr
|
|
|
|
|
define void @f4(float* %p, float* %q) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = bitcast float* %p to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld1 = tail call <2 x float> @llvm.arm.neon.vld1.v2f32.p0i8(i8* %0, i32 4)
|
2012-09-20 05:29:18 +08:00
|
|
|
%tobool = icmp eq float* %q, null
|
|
|
|
|
br i1 %tobool, label %if.end, label %if.then
|
|
|
|
|
|
|
|
|
|
if.then: ; preds = %entry
|
2015-02-28 05:17:42 +08:00
|
|
|
%1 = load float, float* %q, align 4
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx1 = getelementptr inbounds float, float* %q, i32 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load float, float* %arrayidx1, align 4
|
2012-09-20 05:29:18 +08:00
|
|
|
%add = fadd float %1, %2
|
|
|
|
|
%vecins = insertelement <2 x float> %vld1, float %add, i32 1
|
|
|
|
|
br label %if.end
|
|
|
|
|
|
|
|
|
|
if.end: ; preds = %entry, %if.then
|
|
|
|
|
%x.0 = phi <2 x float> [ %vecins, %if.then ], [ %vld1, %entry ]
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst1.p0i8.v2f32(i8* %0, <2 x float> %x.0, i32 4)
|
2012-09-20 05:29:18 +08:00
|
|
|
ret void
|
|
|
|
|
}
|
2012-09-21 07:08:42 +08:00
|
|
|
|
|
|
|
|
; CHECK: f5
|
|
|
|
|
; Coalesce vector lanes through phis.
|
|
|
|
|
; CHECK: vmov.f32 {{.*}}, #1.0
|
|
|
|
|
; CHECK-NOT: vmov
|
|
|
|
|
; CHECK-NOT: vorr
|
2012-11-13 08:34:44 +08:00
|
|
|
; CHECK: bx
|
2012-09-21 07:08:42 +08:00
|
|
|
; We may leave the last insertelement in the if.end block.
|
|
|
|
|
; It is inserting the %add value into a dead lane, but %add causes interference
|
|
|
|
|
; in the entry block, and we don't do dead lane checks across basic blocks.
|
|
|
|
|
define void @f5(float* %p, float* %q) nounwind ssp {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = bitcast float* %p to i8*
|
2015-09-30 18:56:37 +08:00
|
|
|
%vld1 = tail call <4 x float> @llvm.arm.neon.vld1.v4f32.p0i8(i8* %0, i32 4)
|
2012-09-21 07:08:42 +08:00
|
|
|
%vecext = extractelement <4 x float> %vld1, i32 0
|
|
|
|
|
%vecext1 = extractelement <4 x float> %vld1, i32 1
|
|
|
|
|
%vecext2 = extractelement <4 x float> %vld1, i32 2
|
|
|
|
|
%vecext3 = extractelement <4 x float> %vld1, i32 3
|
|
|
|
|
%add = fadd float %vecext3, 1.000000e+00
|
|
|
|
|
%tobool = icmp eq float* %q, null
|
|
|
|
|
br i1 %tobool, label %if.end, label %if.then
|
|
|
|
|
|
|
|
|
|
if.then: ; preds = %entry
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx = getelementptr inbounds float, float* %q, i32 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%1 = load float, float* %arrayidx, align 4
|
2012-09-21 07:08:42 +08:00
|
|
|
%add4 = fadd float %vecext, %1
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load float, float* %q, align 4
|
2012-09-21 07:08:42 +08:00
|
|
|
%add6 = fadd float %vecext1, %2
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx7 = getelementptr inbounds float, float* %q, i32 2
|
2015-02-28 05:17:42 +08:00
|
|
|
%3 = load float, float* %arrayidx7, align 4
|
2012-09-21 07:08:42 +08:00
|
|
|
%add8 = fadd float %vecext2, %3
|
|
|
|
|
br label %if.end
|
|
|
|
|
|
|
|
|
|
if.end: ; preds = %entry, %if.then
|
|
|
|
|
%a.0 = phi float [ %add4, %if.then ], [ %vecext, %entry ]
|
|
|
|
|
%b.0 = phi float [ %add6, %if.then ], [ %vecext1, %entry ]
|
|
|
|
|
%c.0 = phi float [ %add8, %if.then ], [ %vecext2, %entry ]
|
|
|
|
|
%vecinit = insertelement <4 x float> undef, float %a.0, i32 0
|
|
|
|
|
%vecinit9 = insertelement <4 x float> %vecinit, float %b.0, i32 1
|
|
|
|
|
%vecinit10 = insertelement <4 x float> %vecinit9, float %c.0, i32 2
|
|
|
|
|
%vecinit11 = insertelement <4 x float> %vecinit10, float %add, i32 3
|
2015-09-30 18:56:37 +08:00
|
|
|
tail call void @llvm.arm.neon.vst1.p0i8.v4f32(i8* %0, <4 x float> %vecinit11, i32 4)
|
2012-09-21 07:08:42 +08:00
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2015-09-30 18:56:37 +08:00
|
|
|
declare <4 x float> @llvm.arm.neon.vld1.v4f32.p0i8(i8*, i32) nounwind readonly
|
2012-09-21 07:08:42 +08:00
|
|
|
|
2015-09-30 18:56:37 +08:00
|
|
|
declare void @llvm.arm.neon.vst1.p0i8.v4f32(i8*, <4 x float>, i32) nounwind
|
2012-10-03 05:46:39 +08:00
|
|
|
|
|
|
|
|
; CHECK: pr13999
|
|
|
|
|
define void @pr13999() nounwind readonly {
|
|
|
|
|
entry:
|
|
|
|
|
br i1 true, label %outer_loop, label %loop.end
|
|
|
|
|
|
|
|
|
|
outer_loop:
|
|
|
|
|
%d = phi double [ 0.0, %entry ], [ %add, %after_inner_loop ]
|
|
|
|
|
%0 = insertelement <2 x double> <double 0.0, double 0.0>, double %d, i32 0
|
|
|
|
|
br i1 undef, label %after_inner_loop, label %inner_loop
|
|
|
|
|
|
|
|
|
|
inner_loop:
|
|
|
|
|
br i1 true, label %after_inner_loop, label %inner_loop
|
|
|
|
|
|
|
|
|
|
after_inner_loop:
|
|
|
|
|
%1 = phi <2 x double> [ %0, %outer_loop ], [ <double 0.0, double 0.0>,
|
|
|
|
|
%inner_loop ]
|
|
|
|
|
%2 = extractelement <2 x double> %1, i32 1
|
|
|
|
|
%add = fadd double 1.0, %2
|
|
|
|
|
br i1 false, label %loop.end, label %outer_loop
|
|
|
|
|
|
|
|
|
|
loop.end:
|
|
|
|
|
%d.end = phi double [ 0.0, %entry ], [ %add, %after_inner_loop ]
|
|
|
|
|
ret void
|
|
|
|
|
}
|
2012-10-14 00:15:31 +08:00
|
|
|
|
|
|
|
|
; CHECK: pr14078
|
|
|
|
|
define arm_aapcs_vfpcc i32 @pr14078(i8* nocapture %arg, i8* nocapture %arg1, i32 %arg2) nounwind uwtable readonly {
|
|
|
|
|
bb:
|
|
|
|
|
br i1 undef, label %bb31, label %bb3
|
|
|
|
|
|
|
|
|
|
bb3: ; preds = %bb12, %bb
|
|
|
|
|
%tmp = shufflevector <2 x i64> undef, <2 x i64> undef, <1 x i32> zeroinitializer
|
|
|
|
|
%tmp4 = bitcast <1 x i64> %tmp to <2 x float>
|
|
|
|
|
%tmp5 = shufflevector <2 x float> %tmp4, <2 x float> undef, <4 x i32> zeroinitializer
|
|
|
|
|
%tmp6 = bitcast <4 x float> %tmp5 to <2 x i64>
|
|
|
|
|
%tmp7 = shufflevector <2 x i64> %tmp6, <2 x i64> undef, <1 x i32> zeroinitializer
|
|
|
|
|
%tmp8 = bitcast <1 x i64> %tmp7 to <2 x float>
|
|
|
|
|
%tmp9 = tail call <2 x float> @baz(<2 x float> <float 0xFFFFFFFFE0000000, float 0.000000e+00>, <2 x float> %tmp8, <2 x float> zeroinitializer) nounwind
|
|
|
|
|
br i1 undef, label %bb10, label %bb12
|
|
|
|
|
|
|
|
|
|
bb10: ; preds = %bb3
|
2015-02-28 05:17:42 +08:00
|
|
|
%tmp11 = load <4 x float>, <4 x float>* undef, align 8
|
2012-10-14 00:15:31 +08:00
|
|
|
br label %bb12
|
|
|
|
|
|
|
|
|
|
bb12: ; preds = %bb10, %bb3
|
|
|
|
|
%tmp13 = shufflevector <2 x float> %tmp9, <2 x float> zeroinitializer, <2 x i32> <i32 0, i32 2>
|
|
|
|
|
%tmp14 = bitcast <2 x float> %tmp13 to <1 x i64>
|
|
|
|
|
%tmp15 = shufflevector <1 x i64> %tmp14, <1 x i64> zeroinitializer, <2 x i32> <i32 0, i32 1>
|
|
|
|
|
%tmp16 = bitcast <2 x i64> %tmp15 to <4 x float>
|
|
|
|
|
%tmp17 = fmul <4 x float> zeroinitializer, %tmp16
|
|
|
|
|
%tmp18 = bitcast <4 x float> %tmp17 to <2 x i64>
|
|
|
|
|
%tmp19 = shufflevector <2 x i64> %tmp18, <2 x i64> undef, <1 x i32> zeroinitializer
|
|
|
|
|
%tmp20 = bitcast <1 x i64> %tmp19 to <2 x float>
|
|
|
|
|
%tmp21 = tail call <2 x float> @baz67(<2 x float> %tmp20, <2 x float> undef) nounwind
|
|
|
|
|
%tmp22 = tail call <2 x float> @baz67(<2 x float> %tmp21, <2 x float> %tmp21) nounwind
|
|
|
|
|
%tmp23 = shufflevector <2 x float> %tmp22, <2 x float> undef, <4 x i32> zeroinitializer
|
|
|
|
|
%tmp24 = bitcast <4 x float> %tmp23 to <2 x i64>
|
|
|
|
|
%tmp25 = shufflevector <2 x i64> %tmp24, <2 x i64> undef, <1 x i32> zeroinitializer
|
|
|
|
|
%tmp26 = bitcast <1 x i64> %tmp25 to <2 x float>
|
|
|
|
|
%tmp27 = extractelement <2 x float> %tmp26, i32 0
|
|
|
|
|
%tmp28 = fcmp olt float %tmp27, 0.000000e+00
|
|
|
|
|
%tmp29 = select i1 %tmp28, i32 0, i32 undef
|
|
|
|
|
%tmp30 = icmp ult i32 undef, %arg2
|
|
|
|
|
br i1 %tmp30, label %bb3, label %bb31
|
|
|
|
|
|
|
|
|
|
bb31: ; preds = %bb12, %bb
|
|
|
|
|
%tmp32 = phi i32 [ 1, %bb ], [ %tmp29, %bb12 ]
|
|
|
|
|
ret i32 %tmp32
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
declare <2 x float> @baz(<2 x float>, <2 x float>, <2 x float>) nounwind readnone
|
|
|
|
|
|
|
|
|
|
declare <2 x float> @baz67(<2 x float>, <2 x float>) nounwind readnone
|
2012-10-14 01:26:47 +08:00
|
|
|
|
|
|
|
|
%struct.wombat.5 = type { %struct.quux, %struct.quux, %struct.quux, %struct.quux }
|
|
|
|
|
%struct.quux = type { <4 x float> }
|
|
|
|
|
|
|
|
|
|
; CHECK: pr14079
|
|
|
|
|
define linkonce_odr arm_aapcs_vfpcc %struct.wombat.5 @pr14079(i8* nocapture %arg, i8* nocapture %arg1, i8* nocapture %arg2) nounwind uwtable inlinehint {
|
|
|
|
|
bb:
|
|
|
|
|
%tmp = shufflevector <2 x i64> zeroinitializer, <2 x i64> undef, <1 x i32> zeroinitializer
|
|
|
|
|
%tmp3 = bitcast <1 x i64> %tmp to <2 x float>
|
|
|
|
|
%tmp4 = shufflevector <2 x float> %tmp3, <2 x float> zeroinitializer, <2 x i32> <i32 1, i32 3>
|
|
|
|
|
%tmp5 = shufflevector <2 x float> %tmp4, <2 x float> undef, <2 x i32> <i32 1, i32 3>
|
|
|
|
|
%tmp6 = bitcast <2 x float> %tmp5 to <1 x i64>
|
|
|
|
|
%tmp7 = shufflevector <1 x i64> undef, <1 x i64> %tmp6, <2 x i32> <i32 0, i32 1>
|
|
|
|
|
%tmp8 = bitcast <2 x i64> %tmp7 to <4 x float>
|
|
|
|
|
%tmp9 = shufflevector <2 x i64> zeroinitializer, <2 x i64> undef, <1 x i32> <i32 1>
|
|
|
|
|
%tmp10 = bitcast <1 x i64> %tmp9 to <2 x float>
|
|
|
|
|
%tmp11 = shufflevector <2 x float> %tmp10, <2 x float> undef, <2 x i32> <i32 0, i32 2>
|
|
|
|
|
%tmp12 = shufflevector <2 x float> %tmp11, <2 x float> undef, <2 x i32> <i32 0, i32 2>
|
|
|
|
|
%tmp13 = bitcast <2 x float> %tmp12 to <1 x i64>
|
|
|
|
|
%tmp14 = shufflevector <1 x i64> %tmp13, <1 x i64> undef, <2 x i32> <i32 0, i32 1>
|
|
|
|
|
%tmp15 = bitcast <2 x i64> %tmp14 to <4 x float>
|
|
|
|
|
%tmp16 = insertvalue %struct.wombat.5 undef, <4 x float> %tmp8, 1, 0
|
|
|
|
|
%tmp17 = insertvalue %struct.wombat.5 %tmp16, <4 x float> %tmp15, 2, 0
|
|
|
|
|
%tmp18 = insertvalue %struct.wombat.5 %tmp17, <4 x float> undef, 3, 0
|
|
|
|
|
ret %struct.wombat.5 %tmp18
|
|
|
|
|
}
|
2012-10-30 01:51:52 +08:00
|
|
|
|
|
|
|
|
; CHECK: adjustCopiesBackFrom
|
|
|
|
|
; The shuffle in if.else3 must be preserved even though adjustCopiesBackFrom
|
|
|
|
|
; is tempted to remove it.
|
|
|
|
|
; CHECK: vorr d
|
|
|
|
|
define internal void @adjustCopiesBackFrom(<2 x i64>* noalias nocapture sret %agg.result, <2 x i64> %in) {
|
|
|
|
|
entry:
|
|
|
|
|
%0 = extractelement <2 x i64> %in, i32 0
|
|
|
|
|
%cmp = icmp slt i64 %0, 1
|
|
|
|
|
%.in = select i1 %cmp, <2 x i64> <i64 0, i64 undef>, <2 x i64> %in
|
|
|
|
|
%1 = extractelement <2 x i64> %in, i32 1
|
|
|
|
|
%cmp1 = icmp slt i64 %1, 1
|
|
|
|
|
br i1 %cmp1, label %if.then2, label %if.else3
|
|
|
|
|
|
|
|
|
|
if.then2: ; preds = %entry
|
|
|
|
|
%2 = insertelement <2 x i64> %.in, i64 0, i32 1
|
|
|
|
|
br label %if.end4
|
|
|
|
|
|
|
|
|
|
if.else3: ; preds = %entry
|
|
|
|
|
%3 = shufflevector <2 x i64> %.in, <2 x i64> %in, <2 x i32> <i32 0, i32 3>
|
|
|
|
|
br label %if.end4
|
|
|
|
|
|
|
|
|
|
if.end4: ; preds = %if.else3, %if.then2
|
|
|
|
|
%result.2 = phi <2 x i64> [ %2, %if.then2 ], [ %3, %if.else3 ]
|
|
|
|
|
store <2 x i64> %result.2, <2 x i64>* %agg.result, align 128
|
|
|
|
|
ret void
|
|
|
|
|
}
|
2012-11-29 08:26:11 +08:00
|
|
|
|
|
|
|
|
; <rdar://problem/12758887>
|
|
|
|
|
; RegisterCoalescer::updateRegDefsUses() could visit an instruction more than
|
|
|
|
|
; once under rare circumstances. When widening a register from QPR to DTriple
|
|
|
|
|
; with the original virtual register in dsub_1_dsub_2, the double rewrite would
|
|
|
|
|
; produce an invalid sub-register.
|
|
|
|
|
;
|
|
|
|
|
; This is because dsub_1_dsub_2 is not an idempotent sub-register index.
|
|
|
|
|
; It will translate %vr:dsub_0 -> %vr:dsub_1.
|
|
|
|
|
define hidden fastcc void @radar12758887() nounwind optsize ssp {
|
|
|
|
|
entry:
|
|
|
|
|
br i1 undef, label %for.body, label %for.end70
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %for.end, %entry
|
|
|
|
|
br i1 undef, label %for.body29, label %for.end
|
|
|
|
|
|
|
|
|
|
for.body29: ; preds = %for.body29, %for.body
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load <2 x double>, <2 x double>* null, align 1
|
2012-11-29 08:26:11 +08:00
|
|
|
%splat40 = shufflevector <2 x double> %0, <2 x double> undef, <2 x i32> zeroinitializer
|
|
|
|
|
%mul41 = fmul <2 x double> undef, %splat40
|
|
|
|
|
%add42 = fadd <2 x double> undef, %mul41
|
|
|
|
|
%splat44 = shufflevector <2 x double> %0, <2 x double> undef, <2 x i32> <i32 1, i32 1>
|
|
|
|
|
%mul45 = fmul <2 x double> undef, %splat44
|
|
|
|
|
%add46 = fadd <2 x double> undef, %mul45
|
|
|
|
|
br i1 undef, label %for.end, label %for.body29
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body29, %for.body
|
|
|
|
|
%accumR2.0.lcssa = phi <2 x double> [ zeroinitializer, %for.body ], [ %add42, %for.body29 ]
|
|
|
|
|
%accumI2.0.lcssa = phi <2 x double> [ zeroinitializer, %for.body ], [ %add46, %for.body29 ]
|
|
|
|
|
%1 = shufflevector <2 x double> %accumI2.0.lcssa, <2 x double> undef, <2 x i32> <i32 1, i32 0>
|
|
|
|
|
%add58 = fadd <2 x double> undef, %1
|
|
|
|
|
%mul61 = fmul <2 x double> %add58, undef
|
|
|
|
|
%add63 = fadd <2 x double> undef, %mul61
|
|
|
|
|
%add64 = fadd <2 x double> undef, %add63
|
|
|
|
|
%add67 = fadd <2 x double> undef, %add64
|
|
|
|
|
store <2 x double> %add67, <2 x double>* undef, align 1
|
|
|
|
|
br i1 undef, label %for.end70, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end70: ; preds = %for.end, %entry
|
|
|
|
|
ret void
|
|
|
|
|
}
|