maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/InstCombine/extractvalue.ll

108 lines
3.8 KiB
LLVM
Raw Normal View History

Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; RUN: opt < %s -instcombine -S | FileCheck %s
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
declare void @bar({i32, i32} %a)
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
declare i32 @baz(i32 %a)
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
Catch more CHECK that can be converted to CHECK-LABEL in Transforms for easier debugging. No functionality change. This conversion was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)define\([^@]*\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3define\4@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186269
2013-07-14 09:50:49 +08:00
; CHECK-LABEL: define i32 @foo(
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; CHECK-NOT: extractvalue
Update this test to keep testing the -instcombine transform it's supposed to be testing instead of triggering the improved constant folding for insertvalue and extractvalue. llvm-svn: 120319
2010-11-30 04:55:40 +08:00
define i32 @foo(i32 %a, i32 %b) {
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; Instcombine should fold various combinations of insertvalue and extractvalue
; together
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
; Build a simple struct and pull values out again
Update this test to keep testing the -instcombine transform it's supposed to be testing instead of triggering the improved constant folding for insertvalue and extractvalue. llvm-svn: 120319
2010-11-30 04:55:40 +08:00
%s1.1 = insertvalue {i32, i32} undef, i32 %a, 0
%s1 = insertvalue {i32, i32} %s1.1, i32 %b, 1
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
%v1 = extractvalue {i32, i32} %s1, 0
%v2 = extractvalue {i32, i32} %s1, 1
; Build a nested struct and pull a sub struct out of it
; This requires instcombine to insert a few insertvalue instructions
%ns1.1 = insertvalue {i32, {i32, i32}} undef, i32 %v1, 0
%ns1.2 = insertvalue {i32, {i32, i32}} %ns1.1, i32 %v1, 1, 0
%ns1 = insertvalue {i32, {i32, i32}} %ns1.2, i32 %v2, 1, 1
%s2 = extractvalue {i32, {i32, i32}} %ns1, 1
Add a few cases to instcombine's extractvalue testcase. llvm-svn: 53675
2008-07-16 20:57:25 +08:00
%v3 = extractvalue {i32, {i32, i32}} %ns1, 1, 1
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
call void @bar({i32, i32} %s2)
Add a few cases to instcombine's extractvalue testcase. llvm-svn: 53675
2008-07-16 20:57:25 +08:00
; Use nested extractvalues to get to a value
%s3 = extractvalue {i32, {i32, i32}} %ns1, 1
%v4 = extractvalue {i32, i32} %s3, 1
call void @bar({i32, i32} %s3)
; Use nested insertvalues to build a nested struct
%s4.1 = insertvalue {i32, i32} undef, i32 %v3, 0
%s4 = insertvalue {i32, i32} %s4.1, i32 %v4, 1
%ns2 = insertvalue {i32, {i32, i32}} undef, {i32, i32} %s4, 1
; And now extract a single value from there
%v5 = extractvalue {i32, {i32, i32}} %ns2, 1, 1
ret i32 %v5
Teach instruction combining about the extractvalue. It can succesfully fold useless insert-extract chains, similar to how it folds them for vectors. Add a testcase for this. llvm-svn: 52217
2008-06-11 22:05:05 +08:00
}
Catch more CHECK that can be converted to CHECK-LABEL in Transforms for easier debugging. No functionality change. This conversion was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)define\([^@]*\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3define\4@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186269
2013-07-14 09:50:49 +08:00
; CHECK-LABEL: define i32 @extract2gep(
[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
; CHECK-NEXT: [[GEP:%[a-z0-9]+]] = getelementptr inbounds {{.*}}, {{.*}}* %pair, i32 0, i32 1
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; CHECK-NEXT: [[LOAD:%[A-Za-z0-9]+]] = load i32* [[GEP]]
; CHECK-NEXT: store
; CHECK-NEXT: br label %loop
; CHECK-NOT: extractvalue
; CHECK: call {{.*}}(i32 [[LOAD]])
; CHECK-NOT: extractvalue
; CHECK: ret i32 [[LOAD]]
define i32 @extract2gep({i32, i32}* %pair, i32* %P) {
; The load + extractvalue should be converted
; to an inbounds gep + smaller load.
; The new load should be in the same spot as the old load.
%L = load {i32, i32}* %pair
store i32 0, i32* %P
br label %loop
loop:
%E = extractvalue {i32, i32} %L, 1
%C = call i32 @baz(i32 %E)
store i32 %C, i32* %P
%cond = icmp eq i32 %C, 0
br i1 %cond, label %end, label %loop
end:
ret i32 %E
}
Catch more CHECK that can be converted to CHECK-LABEL in Transforms for easier debugging. No functionality change. This conversion was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)define\([^@]*\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3define\4@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186269
2013-07-14 09:50:49 +08:00
; CHECK-LABEL: define i32 @doubleextract2gep(
[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
; CHECK-NEXT: [[GEP:%[a-z0-9]+]] = getelementptr inbounds {{.*}}, {{.*}}* %arg, i32 0, i32 1, i32 1
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; CHECK-NEXT: [[LOAD:%[A-Za-z0-9]+]] = load i32* [[GEP]]
; CHECK-NEXT: ret i32 [[LOAD]]
define i32 @doubleextract2gep({i32, {i32, i32}}* %arg) {
; The load + extractvalues should be converted
; to a 3-index inbounds gep + smaller load.
%L = load {i32, {i32, i32}}* %arg
%E1 = extractvalue {i32, {i32, i32}} %L, 1
%E2 = extractvalue {i32, i32} %E1, 1
ret i32 %E2
}
; CHECK: define i32 @nogep-multiuse
; CHECK-NEXT: load {{.*}} %pair
; CHECK-NEXT: extractvalue
; CHECK-NEXT: extractvalue
; CHECK-NEXT: add
; CHECK-NEXT: ret
define i32 @nogep-multiuse({i32, i32}* %pair) {
; The load should be left unchanged since both parts are needed.
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 14:54:59 +08:00
%L = load volatile {i32, i32}* %pair
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
%LHS = extractvalue {i32, i32} %L, 0
%RHS = extractvalue {i32, i32} %L, 1
%R = add i32 %LHS, %RHS
ret i32 %R
}
; CHECK: define i32 @nogep-volatile
Move "atomic" and "volatile" designations on instructions after the opcode of the instruction. Note that this change affects the existing non-atomic load and store instructions; the parser now accepts both forms, and the change is noted in the release notes. llvm-svn: 137527
2011-08-13 06:50:01 +08:00
; CHECK-NEXT: load volatile {{.*}} %pair
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
; CHECK-NEXT: extractvalue
; CHECK-NEXT: ret
define i32 @nogep-volatile({i32, i32}* %pair) {
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 14:54:59 +08:00
; The load volatile should be left unchanged.
%L = load volatile {i32, i32}* %pair
Transform (extractvalue (load P), ...) to (load (gep P, 0, ...)) if the load has no other uses, shrinking the load. llvm-svn: 120323
2010-11-30 05:56:20 +08:00
%E = extractvalue {i32, i32} %L, 1
ret i32 %E
}