2014-04-22 03:34:27 +08:00
|
|
|
; RUN: opt -codegenprepare -mtriple=arm64-apple=ios -S -o - %s | FileCheck --check-prefix=OPT %s
|
2016-07-19 21:35:11 +08:00
|
|
|
; RUN: llc < %s -mtriple=arm64-eabi | FileCheck %s
|
2014-03-29 18:18:08 +08:00
|
|
|
%struct.X = type { i8, i8, [2 x i8] }
|
|
|
|
|
%struct.Y = type { i32, i8 }
|
|
|
|
|
%struct.Z = type { i8, i8, [2 x i8], i16 }
|
|
|
|
|
%struct.A = type { i64, i8 }
|
|
|
|
|
|
|
|
|
|
define void @foo(%struct.X* nocapture %x, %struct.Y* nocapture %y) nounwind optsize ssp {
|
|
|
|
|
; CHECK-LABEL: foo:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: ubfx
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NOT: and
|
|
|
|
|
; CHECK: ret
|
|
|
|
|
|
|
|
|
|
%tmp = bitcast %struct.X* %x to i32*
|
2015-02-28 05:17:42 +08:00
|
|
|
%tmp1 = load i32, i32* %tmp, 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
|
|
|
%b = getelementptr inbounds %struct.Y, %struct.Y* %y, i64 0, i32 1
|
2014-03-29 18:18:08 +08:00
|
|
|
%bf.clear = lshr i32 %tmp1, 3
|
|
|
|
|
%bf.clear.lobit = and i32 %bf.clear, 1
|
|
|
|
|
%frombool = trunc i32 %bf.clear.lobit to i8
|
|
|
|
|
store i8 %frombool, i8* %b, align 1
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i32 @baz(i64 %cav1.coerce) nounwind {
|
|
|
|
|
; CHECK-LABEL: baz:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: sbfx w0, w0, #0, #4
|
2014-03-29 18:18:08 +08:00
|
|
|
%tmp = trunc i64 %cav1.coerce to i32
|
|
|
|
|
%tmp1 = shl i32 %tmp, 28
|
|
|
|
|
%bf.val.sext = ashr exact i32 %tmp1, 28
|
|
|
|
|
ret i32 %bf.val.sext
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i32 @bar(i64 %cav1.coerce) nounwind {
|
|
|
|
|
; CHECK-LABEL: bar:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: sbfx w0, w0, #4, #6
|
2014-03-29 18:18:08 +08:00
|
|
|
%tmp = trunc i64 %cav1.coerce to i32
|
|
|
|
|
%cav1.sroa.0.1.insert = shl i32 %tmp, 22
|
|
|
|
|
%tmp1 = ashr i32 %cav1.sroa.0.1.insert, 26
|
|
|
|
|
ret i32 %tmp1
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define void @fct1(%struct.Z* nocapture %x, %struct.A* nocapture %y) nounwind optsize ssp {
|
|
|
|
|
; CHECK-LABEL: fct1:
|
2016-04-22 02:03:06 +08:00
|
|
|
; CHECK: ubfx x{{[0-9]+}}, x{{[0-9]+}}
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NOT: and
|
|
|
|
|
; CHECK: ret
|
|
|
|
|
|
|
|
|
|
%tmp = bitcast %struct.Z* %x to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%tmp1 = load i64, i64* %tmp, 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
|
|
|
%b = getelementptr inbounds %struct.A, %struct.A* %y, i64 0, i32 0
|
2014-03-29 18:18:08 +08:00
|
|
|
%bf.clear = lshr i64 %tmp1, 3
|
|
|
|
|
%bf.clear.lobit = and i64 %bf.clear, 1
|
|
|
|
|
store i64 %bf.clear.lobit, i64* %b, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i64 @fct2(i64 %cav1.coerce) nounwind {
|
|
|
|
|
; CHECK-LABEL: fct2:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: sbfx x0, x0, #0, #36
|
2014-03-29 18:18:08 +08:00
|
|
|
%tmp = shl i64 %cav1.coerce, 28
|
|
|
|
|
%bf.val.sext = ashr exact i64 %tmp, 28
|
|
|
|
|
ret i64 %bf.val.sext
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i64 @fct3(i64 %cav1.coerce) nounwind {
|
|
|
|
|
; CHECK-LABEL: fct3:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: sbfx x0, x0, #4, #38
|
2014-03-29 18:18:08 +08:00
|
|
|
%cav1.sroa.0.1.insert = shl i64 %cav1.coerce, 22
|
|
|
|
|
%tmp1 = ashr i64 %cav1.sroa.0.1.insert, 26
|
|
|
|
|
ret i64 %tmp1
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define void @fct4(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct4:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #16, #24
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG1]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -16777216
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 16777215
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
store i64 %or, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define void @fct5(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct5:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG1]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -8
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
store i32 %or, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some low bits
|
|
|
|
|
define void @fct6(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct6:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsr [[REG2:w[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -8
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shr1 = lshr i32 %or, 2
|
|
|
|
|
store i32 %shr1, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
define void @fct7(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct7:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsl is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG2:w[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -8
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shl = shl i32 %or, 2
|
|
|
|
|
store i32 %shl, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some low bits
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct8(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct8:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsr [[REG2:x[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -8
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 7
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shr1 = lshr i64 %or, 2
|
|
|
|
|
store i64 %shr1, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct9(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct9:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG2:x[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -8
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 7
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shl = shl i64 %or, 2
|
|
|
|
|
store i64 %shl, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can catch bfm instruction when lsb is 0 (i.e., no lshr)
|
|
|
|
|
; (i32 version)
|
|
|
|
|
define void @fct10(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct10:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #0, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsl is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG2:w[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -8
|
|
|
|
|
%and1 = and i32 %x, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shl = shl i32 %or, 2
|
|
|
|
|
store i32 %shl, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can catch bfm instruction when lsb is 0 (i.e., no lshr)
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct11(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct11:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #0, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsl is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG2:x[0-9]+]], [[REG1]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -8
|
|
|
|
|
%and1 = and i64 %x, 7
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shl = shl i64 %or, 2
|
|
|
|
|
store i64 %shl, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define zeroext i1 @fct12bis(i32 %tmp2) unnamed_addr nounwind ssp align 2 {
|
|
|
|
|
; CHECK-LABEL: fct12bis:
|
|
|
|
|
; CHECK-NOT: and
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: ubfx w0, w0, #11, #1
|
2014-03-29 18:18:08 +08:00
|
|
|
%and.i.i = and i32 %tmp2, 2048
|
|
|
|
|
%tobool.i.i = icmp ne i32 %and.i.i, 0
|
|
|
|
|
ret i1 %tobool.i.i
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits
|
|
|
|
|
define void @fct12(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct12:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK-NEXT: ubfx [[REG2:w[0-9]+]], [[REG1]], #2, #28
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -8
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shl = shl i32 %or, 2
|
|
|
|
|
%shr2 = lshr i32 %shl, 4
|
|
|
|
|
store i32 %shr2, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct13(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct13:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK-NEXT: ubfx [[REG2:x[0-9]+]], [[REG1]], #2, #60
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG2]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -8
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 7
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shl = shl i64 %or, 2
|
|
|
|
|
%shr2 = lshr i64 %shl, 4
|
|
|
|
|
store i64 %shr2, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits
|
|
|
|
|
define void @fct14(i32* nocapture %y, i32 %x, i32 %x1) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct14:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], w1, #16, #8
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsr [[REG2:w[0-9]+]], [[REG1]], #4
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG2]], w2, #5, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsl is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG3:w[0-9]+]], [[REG2]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG3]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, -256
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 255
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shl = lshr i32 %or, 4
|
|
|
|
|
%and2 = and i32 %shl, -8
|
|
|
|
|
%shr1 = lshr i32 %x1, 5
|
|
|
|
|
%and3 = and i32 %shr1, 7
|
|
|
|
|
%or1 = or i32 %and2, %and3
|
|
|
|
|
%shl1 = shl i32 %or1, 2
|
|
|
|
|
store i32 %shl1, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct15(i64* nocapture %y, i64 %x, i64 %x1) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct15:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG1]], x1, #16, #8
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsr [[REG2:x[0-9]+]], [[REG1]], #4
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG2]], x2, #5, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsl is an alias of ubfm
|
|
|
|
|
; CHECK-NEXT: lsl [[REG3:x[0-9]+]], [[REG2]], #2
|
|
|
|
|
; CHECK-NEXT: str [[REG3]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, -256
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 255
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shl = lshr i64 %or, 4
|
|
|
|
|
%and2 = and i64 %shl, -8
|
|
|
|
|
%shr1 = lshr i64 %x1, 5
|
|
|
|
|
%and3 = and i64 %shr1, 7
|
|
|
|
|
%or1 = or i64 %and2, %and3
|
|
|
|
|
%shl1 = shl i64 %or1, 2
|
|
|
|
|
store i64 %shl1, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits and a masking operation has to be kept
|
|
|
|
|
define void @fct16(i32* nocapture %y, i32 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct16:
|
|
|
|
|
; CHECK: ldr [[REG1:w[0-9]+]],
|
|
|
|
|
; Create the constant
|
2017-01-19 02:57:08 +08:00
|
|
|
; CHECK: mov [[REGCST:w[0-9]+]], #33120
|
|
|
|
|
; CHECK: movk [[REGCST]], #26, lsl #16
|
2014-03-29 18:18:08 +08:00
|
|
|
; Do the masking
|
|
|
|
|
; CHECK: and [[REG2:w[0-9]+]], [[REG1]], [[REGCST]]
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG2]], w1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK-NEXT: ubfx [[REG3:w[0-9]+]], [[REG2]], #2, #28
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG3]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i32, i32* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i32 %0, 1737056
|
|
|
|
|
%shr = lshr i32 %x, 16
|
|
|
|
|
%and1 = and i32 %shr, 7
|
|
|
|
|
%or = or i32 %and, %and1
|
|
|
|
|
%shl = shl i32 %or, 2
|
|
|
|
|
%shr2 = lshr i32 %shl, 4
|
|
|
|
|
store i32 %shr2, i32* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
; Check if we can still catch bfm instruction when we drop some high bits
|
|
|
|
|
; and some low bits and a masking operation has to be kept
|
|
|
|
|
; (i64 version)
|
|
|
|
|
define void @fct17(i64* nocapture %y, i64 %x) nounwind optsize inlinehint ssp {
|
|
|
|
|
entry:
|
|
|
|
|
; CHECK-LABEL: fct17:
|
|
|
|
|
; CHECK: ldr [[REG1:x[0-9]+]],
|
|
|
|
|
; Create the constant
|
2017-01-19 02:57:08 +08:00
|
|
|
; CHECK: mov w[[REGCST:[0-9]+]], #33120
|
|
|
|
|
; CHECK: movk w[[REGCST]], #26, lsl #16
|
2014-03-29 18:18:08 +08:00
|
|
|
; Do the masking
|
2014-04-16 19:52:51 +08:00
|
|
|
; CHECK: and [[REG2:x[0-9]+]], [[REG1]], x[[REGCST]]
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK-NEXT: bfxil [[REG2]], x1, #16, #3
|
2014-03-29 18:18:08 +08:00
|
|
|
; lsr is an alias of ubfm
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK-NEXT: ubfx [[REG3:x[0-9]+]], [[REG2]], #2, #60
|
2014-03-29 18:18:08 +08:00
|
|
|
; CHECK-NEXT: str [[REG3]],
|
|
|
|
|
; CHECK-NEXT: ret
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %y, align 8
|
2014-03-29 18:18:08 +08:00
|
|
|
%and = and i64 %0, 1737056
|
|
|
|
|
%shr = lshr i64 %x, 16
|
|
|
|
|
%and1 = and i64 %shr, 7
|
|
|
|
|
%or = or i64 %and, %and1
|
|
|
|
|
%shl = shl i64 %or, 2
|
|
|
|
|
%shr2 = lshr i64 %shl, 4
|
|
|
|
|
store i64 %shr2, i64* %y, align 8
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i64 @fct18(i32 %xor72) nounwind ssp {
|
|
|
|
|
; CHECK-LABEL: fct18:
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: ubfx x0, x0, #9, #8
|
2014-03-29 18:18:08 +08:00
|
|
|
%shr81 = lshr i32 %xor72, 9
|
|
|
|
|
%conv82 = zext i32 %shr81 to i64
|
|
|
|
|
%result = and i64 %conv82, 255
|
|
|
|
|
ret i64 %result
|
|
|
|
|
}
|
2014-04-22 03:34:27 +08:00
|
|
|
|
|
|
|
|
; Using the access to the global array to keep the instruction and control flow.
|
|
|
|
|
@first_ones = external global [65536 x i8]
|
|
|
|
|
|
|
|
|
|
; Function Attrs: nounwind readonly ssp
|
|
|
|
|
define i32 @fct19(i64 %arg1) nounwind readonly ssp {
|
|
|
|
|
; CHECK-LABEL: fct19:
|
|
|
|
|
entry:
|
|
|
|
|
%x.sroa.1.0.extract.shift = lshr i64 %arg1, 16
|
|
|
|
|
%x.sroa.1.0.extract.trunc = trunc i64 %x.sroa.1.0.extract.shift to i16
|
|
|
|
|
%x.sroa.3.0.extract.shift = lshr i64 %arg1, 32
|
|
|
|
|
%x.sroa.5.0.extract.shift = lshr i64 %arg1, 48
|
|
|
|
|
%tobool = icmp eq i64 %x.sroa.5.0.extract.shift, 0
|
|
|
|
|
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
|
|
|
%arrayidx3 = getelementptr inbounds [65536 x i8], [65536 x i8]* @first_ones, i64 0, i64 %x.sroa.5.0.extract.shift
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i8, i8* %arrayidx3, align 1
|
2014-04-22 03:34:27 +08:00
|
|
|
%conv = zext i8 %0 to i32
|
|
|
|
|
br label %return
|
|
|
|
|
|
|
|
|
|
; OPT-LABEL: if.end
|
|
|
|
|
if.end: ; preds = %entry
|
|
|
|
|
; OPT: lshr
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: ubfx [[REG1:x[0-9]+]], [[REG2:x[0-9]+]], #32, #16
|
2014-04-22 03:34:27 +08:00
|
|
|
%x.sroa.3.0.extract.trunc = trunc i64 %x.sroa.3.0.extract.shift to i16
|
|
|
|
|
%tobool6 = icmp eq i16 %x.sroa.3.0.extract.trunc, 0
|
|
|
|
|
; CHECK: cbz
|
|
|
|
|
br i1 %tobool6, label %if.end13, label %if.then7
|
|
|
|
|
|
|
|
|
|
; OPT-LABEL: if.then7
|
|
|
|
|
if.then7: ; preds = %if.end
|
|
|
|
|
; OPT: lshr
|
|
|
|
|
; "and" should be combined to "ubfm" while "ubfm" should be removed by cse.
|
|
|
|
|
; So neither of them should be in the assemble code.
|
|
|
|
|
; CHECK-NOT: and
|
|
|
|
|
; CHECK-NOT: ubfm
|
|
|
|
|
%idxprom10 = and i64 %x.sroa.3.0.extract.shift, 65535
|
[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
|
|
|
%arrayidx11 = getelementptr inbounds [65536 x i8], [65536 x i8]* @first_ones, i64 0, i64 %idxprom10
|
2015-02-28 05:17:42 +08:00
|
|
|
%1 = load i8, i8* %arrayidx11, align 1
|
2014-04-22 03:34:27 +08:00
|
|
|
%conv12 = zext i8 %1 to i32
|
|
|
|
|
%add = add nsw i32 %conv12, 16
|
|
|
|
|
br label %return
|
|
|
|
|
|
|
|
|
|
; OPT-LABEL: if.end13
|
|
|
|
|
if.end13: ; preds = %if.end
|
|
|
|
|
; OPT: lshr
|
|
|
|
|
; OPT: trunc
|
2014-04-25 18:25:29 +08:00
|
|
|
; CHECK: ubfx [[REG3:x[0-9]+]], [[REG4:x[0-9]+]], #16, #16
|
2014-04-22 03:34:27 +08:00
|
|
|
%tobool16 = icmp eq i16 %x.sroa.1.0.extract.trunc, 0
|
|
|
|
|
; CHECK: cbz
|
|
|
|
|
br i1 %tobool16, label %return, label %if.then17
|
|
|
|
|
|
|
|
|
|
; OPT-LABEL: if.then17
|
|
|
|
|
if.then17: ; preds = %if.end13
|
|
|
|
|
; OPT: lshr
|
|
|
|
|
; "and" should be combined to "ubfm" while "ubfm" should be removed by cse.
|
|
|
|
|
; So neither of them should be in the assemble code.
|
|
|
|
|
; CHECK-NOT: and
|
|
|
|
|
; CHECK-NOT: ubfm
|
|
|
|
|
%idxprom20 = and i64 %x.sroa.1.0.extract.shift, 65535
|
[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
|
|
|
%arrayidx21 = getelementptr inbounds [65536 x i8], [65536 x i8]* @first_ones, i64 0, i64 %idxprom20
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load i8, i8* %arrayidx21, align 1
|
2014-04-22 03:34:27 +08:00
|
|
|
%conv22 = zext i8 %2 to i32
|
|
|
|
|
%add23 = add nsw i32 %conv22, 32
|
|
|
|
|
br label %return
|
|
|
|
|
|
|
|
|
|
return: ; preds = %if.end13, %if.then17, %if.then7, %if.then
|
|
|
|
|
; CHECK: ret
|
|
|
|
|
%retval.0 = phi i32 [ %conv, %if.then ], [ %add, %if.then7 ], [ %add23, %if.then17 ], [ 64, %if.end13 ]
|
|
|
|
|
ret i32 %retval.0
|
|
|
|
|
}
|
2014-04-22 09:20:34 +08:00
|
|
|
|
|
|
|
|
; Make sure we do not assert if the immediate in and is bigger than i64.
|
|
|
|
|
; PR19503.
|
|
|
|
|
; OPT-LABEL: @fct20
|
|
|
|
|
; OPT: lshr
|
|
|
|
|
; OPT-NOT: lshr
|
|
|
|
|
; OPT: ret
|
|
|
|
|
; CHECK-LABEL: fct20:
|
|
|
|
|
; CHECK: ret
|
|
|
|
|
define i80 @fct20(i128 %a, i128 %b) {
|
|
|
|
|
entry:
|
|
|
|
|
%shr = lshr i128 %a, 18
|
|
|
|
|
%conv = trunc i128 %shr to i80
|
|
|
|
|
%tobool = icmp eq i128 %b, 0
|
|
|
|
|
br i1 %tobool, label %then, label %end
|
|
|
|
|
then:
|
|
|
|
|
%and = and i128 %shr, 483673642326615442599424
|
|
|
|
|
%conv2 = trunc i128 %and to i80
|
|
|
|
|
br label %end
|
|
|
|
|
end:
|
|
|
|
|
%conv3 = phi i80 [%conv, %entry], [%conv2, %then]
|
|
|
|
|
ret i80 %conv3
|
|
|
|
|
}
|
2014-04-25 18:48:47 +08:00
|
|
|
|
[ARM64] Prevent bit extraction to be adjusted by following shift
For pattern like ((x >> C1) & Mask) << C2, DAG combiner may convert it
into (x >> (C1-C2)) & (Mask << C2), which makes pattern matching of ubfx
more difficult.
For example:
Given
%shr = lshr i64 %x, 4
%and = and i64 %shr, 15
%arrayidx = getelementptr inbounds [8 x [64 x i64]]* @arr, i64 0, %i64 2, i64 %and
%0 = load i64* %arrayidx
With current shift folding, it takes 3 instrs to compute base address:
lsr x8, x0, #1
and x8, x8, #0x78
add x8, x9, x8
If using ubfx, it only needs 2 instrs:
ubfx x8, x0, #4, #4
add x8, x9, x8, lsl #3
This fixes bug 19589
llvm-svn: 207702
2014-05-01 05:07:24 +08:00
|
|
|
; Check if we can still catch UBFX when "AND" is used by SHL.
|
|
|
|
|
; CHECK-LABEL: fct21:
|
|
|
|
|
; CHECK: ubfx
|
|
|
|
|
@arr = external global [8 x [64 x i64]]
|
|
|
|
|
define i64 @fct21(i64 %x) {
|
|
|
|
|
entry:
|
|
|
|
|
%shr = lshr i64 %x, 4
|
|
|
|
|
%and = and i64 %shr, 15
|
[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 [8 x [64 x i64]], [8 x [64 x i64]]* @arr, i64 0, i64 0, i64 %and
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %arrayidx, align 8
|
[ARM64] Prevent bit extraction to be adjusted by following shift
For pattern like ((x >> C1) & Mask) << C2, DAG combiner may convert it
into (x >> (C1-C2)) & (Mask << C2), which makes pattern matching of ubfx
more difficult.
For example:
Given
%shr = lshr i64 %x, 4
%and = and i64 %shr, 15
%arrayidx = getelementptr inbounds [8 x [64 x i64]]* @arr, i64 0, %i64 2, i64 %and
%0 = load i64* %arrayidx
With current shift folding, it takes 3 instrs to compute base address:
lsr x8, x0, #1
and x8, x8, #0x78
add x8, x9, x8
If using ubfx, it only needs 2 instrs:
ubfx x8, x0, #4, #4
add x8, x9, x8, lsl #3
This fixes bug 19589
llvm-svn: 207702
2014-05-01 05:07:24 +08:00
|
|
|
ret i64 %0
|
|
|
|
|
}
|
|
|
|
|
|
2014-04-25 18:48:47 +08:00
|
|
|
define i16 @test_ignored_rightbits(i32 %dst, i32 %in) {
|
|
|
|
|
; CHECK-LABEL: test_ignored_rightbits:
|
|
|
|
|
|
|
|
|
|
%positioned_field = shl i32 %in, 3
|
|
|
|
|
%positioned_masked_field = and i32 %positioned_field, 120
|
|
|
|
|
%masked_dst = and i32 %dst, 7
|
|
|
|
|
%insertion = or i32 %masked_dst, %positioned_masked_field
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK: {{bfm|bfi|bfxil}}
|
2014-04-25 18:48:47 +08:00
|
|
|
|
|
|
|
|
%shl16 = shl i32 %insertion, 8
|
|
|
|
|
%or18 = or i32 %shl16, %insertion
|
|
|
|
|
%conv19 = trunc i32 %or18 to i16
|
2014-05-01 20:29:38 +08:00
|
|
|
; CHECK: bfi {{w[0-9]+}}, {{w[0-9]+}}, #8, #7
|
2014-04-25 18:48:47 +08:00
|
|
|
|
|
|
|
|
ret i16 %conv19
|
|
|
|
|
}
|
2016-12-01 01:04:22 +08:00
|
|
|
|
|
|
|
|
; The following test excercises the case where we have a BFI
|
|
|
|
|
; instruction with the same input in both operands. We need to
|
|
|
|
|
; track the useful bits through both operands.
|
|
|
|
|
; CHECK-LABEL: sameOperandBFI
|
|
|
|
|
; CHECK: lsr
|
|
|
|
|
; CHECK: and
|
|
|
|
|
; CHECK: bfi
|
|
|
|
|
; CHECK: bfi
|
|
|
|
|
define void @sameOperandBFI(i64 %src, i64 %src2, i16 *%ptr) {
|
|
|
|
|
entry:
|
|
|
|
|
%shr47 = lshr i64 %src, 47
|
|
|
|
|
%src2.trunc = trunc i64 %src2 to i32
|
|
|
|
|
br i1 undef, label %end, label %if.else
|
|
|
|
|
|
|
|
|
|
if.else:
|
|
|
|
|
%and3 = and i32 %src2.trunc, 3
|
|
|
|
|
%shl2 = shl nuw nsw i64 %shr47, 2
|
|
|
|
|
%shl2.trunc = trunc i64 %shl2 to i32
|
|
|
|
|
%and12 = and i32 %shl2.trunc, 12
|
|
|
|
|
%BFISource = or i32 %and3, %and12 ; ...00000ABCD
|
|
|
|
|
%BFIRHS = shl nuw nsw i32 %BFISource, 4 ; ...0ABCD0000
|
|
|
|
|
%BFI = or i32 %BFIRHS, %BFISource ; ...0ABCDABCD
|
|
|
|
|
%BFItrunc = trunc i32 %BFI to i16
|
|
|
|
|
store i16 %BFItrunc, i16* %ptr, align 4
|
|
|
|
|
br label %end
|
|
|
|
|
|
|
|
|
|
end:
|
|
|
|
|
ret void
|
|
|
|
|
}
|