2014-11-19 07:38:19 +08:00
|
|
|
; RUN: llc -O3 -disable-peephole -mcpu=corei7-avx -mattr=+avx < %s | FileCheck %s
|
|
|
|
|
|
|
|
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
target triple = "x86_64-unknown-unknown"
|
|
|
|
|
|
|
|
; pr18846 - needless avx spill/reload
|
|
|
|
; Test for unnecessary repeated spills due to eliminateRedundantSpills failing
|
|
|
|
; to recognise unaligned ymm load/stores to the stack.
|
|
|
|
; Bugpoint reduced testcase.
|
|
|
|
|
|
|
|
;CHECK-LABEL: _Z16opt_kernel_cachePfS_S_
|
|
|
|
;CHECK-NOT: vmovups {{.*#+}} 32-byte Folded Spill
|
|
|
|
;CHECK-NOT: vmovups {{.*#+}} 32-byte Folded Reload
|
|
|
|
|
|
|
|
; Function Attrs: uwtable
|
|
|
|
define void @_Z16opt_kernel_cachePfS_S_() #0 {
|
|
|
|
entry:
|
|
|
|
br label %for.body29
|
|
|
|
|
|
|
|
for.body29: ; preds = %for.body29, %entry
|
|
|
|
br i1 undef, label %for.body29, label %for.body65
|
|
|
|
|
|
|
|
for.body65: ; preds = %for.body29
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load float, float* undef, align 4, !tbaa !1
|
2014-11-19 07:38:19 +08:00
|
|
|
%vecinit7.i4448 = insertelement <8 x float> undef, float %0, i32 7
|
2015-02-28 05:17:42 +08:00
|
|
|
%1 = load float, float* null, align 4, !tbaa !1
|
2014-11-19 07:38:19 +08:00
|
|
|
%vecinit7.i4304 = insertelement <8 x float> undef, float %1, i32 7
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load float, float* undef, align 4, !tbaa !1
|
2014-11-19 07:38:19 +08:00
|
|
|
%vecinit7.i4196 = insertelement <8 x float> undef, float %2, i32 7
|
|
|
|
%3 = or i64 0, 16
|
|
|
|
%add.ptr111.sum4096 = add i64 %3, 0
|
2015-02-28 05:17:42 +08:00
|
|
|
%4 = load <8 x float>, <8 x float>* null, align 16, !tbaa !5
|
[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.ptr162 = getelementptr inbounds [65536 x float], [65536 x float]* null, i64 0, i64 %add.ptr111.sum4096
|
2014-11-19 07:38:19 +08:00
|
|
|
%__v.i4158 = bitcast float* %add.ptr162 to <8 x float>*
|
2015-02-28 05:17:42 +08:00
|
|
|
%5 = load <8 x float>, <8 x float>* %__v.i4158, align 16, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%add.ptr158.sum40975066 = or i64 %add.ptr111.sum4096, 8
|
[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.ptr183 = getelementptr inbounds [65536 x float], [65536 x float]* null, i64 0, i64 %add.ptr158.sum40975066
|
2014-11-19 07:38:19 +08:00
|
|
|
%__v.i4162 = bitcast float* %add.ptr183 to <8 x float>*
|
2015-02-28 05:17:42 +08:00
|
|
|
%6 = load <8 x float>, <8 x float>* %__v.i4162, align 16, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%add.ptr200.sum40995067 = or i64 undef, 8
|
[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.ptr225 = getelementptr inbounds [65536 x float], [65536 x float]* null, i64 0, i64 %add.ptr200.sum40995067
|
2014-11-19 07:38:19 +08:00
|
|
|
%__v.i4167 = bitcast float* %add.ptr225 to <8 x float>*
|
2015-02-28 05:17:42 +08:00
|
|
|
%7 = load <8 x float>, <8 x float>* %__v.i4167, align 4, !tbaa !5
|
|
|
|
%8 = load <8 x float>, <8 x float>* undef, align 16, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%add.ptr242.sum41015068 = or i64 0, 8
|
[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.ptr267 = getelementptr inbounds [65536 x float], [65536 x float]* null, i64 0, i64 %add.ptr242.sum41015068
|
2014-11-19 07:38:19 +08:00
|
|
|
%__v.i4171 = bitcast float* %add.ptr267 to <8 x float>*
|
2015-02-28 05:17:42 +08:00
|
|
|
%9 = load <8 x float>, <8 x float>* %__v.i4171, align 4, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%mul.i4690 = fmul <8 x float> %7, undef
|
|
|
|
%add.i4665 = fadd <8 x float> undef, undef
|
|
|
|
%mul.i4616 = fmul <8 x float> %8, undef
|
|
|
|
%mul.i4598 = fmul <8 x float> undef, undef
|
|
|
|
%add.i4597 = fadd <8 x float> undef, %mul.i4598
|
|
|
|
%mul.i4594 = fmul <8 x float> %6, undef
|
|
|
|
%add.i4593 = fadd <8 x float> undef, %mul.i4594
|
|
|
|
%mul.i4578 = fmul <8 x float> %9, undef
|
|
|
|
%add.i4577 = fadd <8 x float> %add.i4593, %mul.i4578
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4577) #1
|
2015-02-28 05:17:42 +08:00
|
|
|
%10 = load <8 x float>, <8 x float>* null, align 16, !tbaa !5
|
|
|
|
%11 = load <8 x float>, <8 x float>* undef, align 16, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%mul.i4564 = fmul <8 x float> %4, undef
|
|
|
|
%add.i4563 = fadd <8 x float> %10, %mul.i4564
|
|
|
|
%mul.i4560 = fmul <8 x float> %5, undef
|
|
|
|
%add.i4559 = fadd <8 x float> %11, %mul.i4560
|
|
|
|
%add.i4547 = fadd <8 x float> %add.i4563, undef
|
|
|
|
%mul.i4546 = fmul <8 x float> %7, undef
|
|
|
|
%add.i4545 = fadd <8 x float> undef, %mul.i4546
|
|
|
|
%mul.i4544 = fmul <8 x float> %8, undef
|
|
|
|
%add.i4543 = fadd <8 x float> %add.i4559, %mul.i4544
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4547) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4545) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4543) #1
|
|
|
|
%add.i4455 = fadd <8 x float> undef, undef
|
|
|
|
%mul.i4454 = fmul <8 x float> undef, undef
|
|
|
|
%add.i4453 = fadd <8 x float> undef, %mul.i4454
|
|
|
|
%mul.i4440 = fmul <8 x float> zeroinitializer, %vecinit7.i4448
|
|
|
|
%add.i4439 = fadd <8 x float> %add.i4455, %mul.i4440
|
|
|
|
%mul.i4438 = fmul <8 x float> %7, %vecinit7.i4448
|
|
|
|
%add.i4437 = fadd <8 x float> %add.i4453, %mul.i4438
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4439) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4437) #1
|
|
|
|
%add.i4413 = fadd <8 x float> zeroinitializer, undef
|
|
|
|
%mul.i4400 = fmul <8 x float> %8, undef
|
|
|
|
%add.i4399 = fadd <8 x float> undef, %mul.i4400
|
|
|
|
%add.i4397 = fadd <8 x float> %add.i4413, zeroinitializer
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> zeroinitializer) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4399) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4397) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> undef) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> undef) #1
|
|
|
|
%mul.i4330 = fmul <8 x float> %7, undef
|
|
|
|
%add.i4329 = fadd <8 x float> undef, %mul.i4330
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4329) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> undef) #1
|
|
|
|
%mul.i4312 = fmul <8 x float> %4, undef
|
|
|
|
%add.i4311 = fadd <8 x float> undef, %mul.i4312
|
|
|
|
%mul.i4306 = fmul <8 x float> %6, undef
|
|
|
|
%add.i4305 = fadd <8 x float> undef, %mul.i4306
|
|
|
|
%add.i4295 = fadd <8 x float> %add.i4311, undef
|
|
|
|
%mul.i4294 = fmul <8 x float> %7, %vecinit7.i4304
|
|
|
|
%add.i4293 = fadd <8 x float> undef, %mul.i4294
|
|
|
|
%mul.i4292 = fmul <8 x float> %8, %vecinit7.i4304
|
|
|
|
%add.i4291 = fadd <8 x float> undef, %mul.i4292
|
|
|
|
%mul.i4290 = fmul <8 x float> %9, %vecinit7.i4304
|
|
|
|
%add.i4289 = fadd <8 x float> %add.i4305, %mul.i4290
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4295) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4293) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4291) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4289) #1
|
2015-02-28 05:17:42 +08:00
|
|
|
%12 = load <8 x float>, <8 x float>* undef, align 16, !tbaa !5
|
2014-11-19 07:38:19 +08:00
|
|
|
%mul.i4274 = fmul <8 x float> undef, undef
|
|
|
|
%add.i4273 = fadd <8 x float> %12, %mul.i4274
|
|
|
|
%mul.i4258 = fmul <8 x float> %7, undef
|
|
|
|
%add.i4257 = fadd <8 x float> %add.i4273, %mul.i4258
|
|
|
|
%mul.i4254 = fmul <8 x float> %9, undef
|
|
|
|
%add.i4253 = fadd <8 x float> undef, %mul.i4254
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4257) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i4253) #1
|
|
|
|
%mul.i = fmul <8 x float> %9, %vecinit7.i4196
|
|
|
|
%add.i = fadd <8 x float> undef, %mul.i
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> zeroinitializer) #1
|
|
|
|
call void @llvm.x86.avx.storeu.ps.256(i8* undef, <8 x float> %add.i) #1
|
|
|
|
unreachable
|
|
|
|
}
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
|
|
|
declare void @llvm.x86.avx.storeu.ps.256(i8*, <8 x float>) #1
|
|
|
|
|
|
|
|
attributes #0 = { uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #1 = { nounwind }
|
|
|
|
|
|
|
|
!llvm.ident = !{!0}
|
|
|
|
|
IR: Make metadata typeless in assembly
Now that `Metadata` is typeless, reflect that in the assembly. These
are the matching assembly changes for the metadata/value split in
r223802.
- Only use the `metadata` type when referencing metadata from a call
intrinsic -- i.e., only when it's used as a `Value`.
- Stop pretending that `ValueAsMetadata` is wrapped in an `MDNode`
when referencing it from call intrinsics.
So, assembly like this:
define @foo(i32 %v) {
call void @llvm.foo(metadata !{i32 %v}, metadata !0)
call void @llvm.foo(metadata !{i32 7}, metadata !0)
call void @llvm.foo(metadata !1, metadata !0)
call void @llvm.foo(metadata !3, metadata !0)
call void @llvm.foo(metadata !{metadata !3}, metadata !0)
ret void, !bar !2
}
!0 = metadata !{metadata !2}
!1 = metadata !{i32* @global}
!2 = metadata !{metadata !3}
!3 = metadata !{}
turns into this:
define @foo(i32 %v) {
call void @llvm.foo(metadata i32 %v, metadata !0)
call void @llvm.foo(metadata i32 7, metadata !0)
call void @llvm.foo(metadata i32* @global, metadata !0)
call void @llvm.foo(metadata !3, metadata !0)
call void @llvm.foo(metadata !{!3}, metadata !0)
ret void, !bar !2
}
!0 = !{!2}
!1 = !{i32* @global}
!2 = !{!3}
!3 = !{}
I wrote an upgrade script that handled almost all of the tests in llvm
and many of the tests in cfe (even handling many `CHECK` lines). I've
attached it (or will attach it in a moment if you're speedy) to PR21532
to help everyone update their out-of-tree testcases.
This is part of PR21532.
llvm-svn: 224257
2014-12-16 03:07:53 +08:00
|
|
|
!0 = !{!"clang version 3.5 "}
|
|
|
|
!1 = !{!2, !2, i64 0}
|
|
|
|
!2 = !{!"float", !3, i64 0}
|
|
|
|
!3 = !{!"omnipotent char", !4, i64 0}
|
|
|
|
!4 = !{!"Simple C/C++ TBAA"}
|
|
|
|
!5 = !{!3, !3, i64 0}
|