llvm-project/llvm/test/CodeGen/X86/2008-02-18-TailMergingBug.ll

221 lines
9.1 KiB
LLVM
Raw Normal View History

; REQUIRES: asserts
; RUN: llc < %s -march=x86 -mcpu=yonah -stats 2>&1 | grep "Number of block tails merged" | grep 16
2008-02-19 10:09:58 +08:00
; PR1909
@.str = internal constant [48 x i8] c"transformed bounds: (%.2f, %.2f), (%.2f, %.2f)\0A\00" ; <[48 x i8]*> [#uses=1]
define void @minmax(float* %result) nounwind optsize {
2008-02-19 10:09:58 +08:00
entry:
%tmp2 = load float, float* %result, align 4 ; <float> [#uses=6]
[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
%tmp4 = getelementptr float, float* %result, i32 2 ; <float*> [#uses=5]
%tmp5 = load float, float* %tmp4, align 4 ; <float> [#uses=10]
[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
%tmp7 = getelementptr float, float* %result, i32 4 ; <float*> [#uses=5]
%tmp8 = load float, float* %tmp7, align 4 ; <float> [#uses=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
%tmp10 = getelementptr float, float* %result, i32 6 ; <float*> [#uses=3]
%tmp11 = load float, float* %tmp10, align 4 ; <float> [#uses=8]
2008-02-19 10:09:58 +08:00
%tmp12 = fcmp olt float %tmp8, %tmp11 ; <i1> [#uses=5]
br i1 %tmp12, label %bb, label %bb21
bb: ; preds = %entry
%tmp23469 = fcmp olt float %tmp5, %tmp8 ; <i1> [#uses=1]
br i1 %tmp23469, label %bb26, label %bb30
bb21: ; preds = %entry
%tmp23 = fcmp olt float %tmp5, %tmp11 ; <i1> [#uses=1]
br i1 %tmp23, label %bb26, label %bb30
bb26: ; preds = %bb21, %bb
%tmp52471 = fcmp olt float %tmp2, %tmp5 ; <i1> [#uses=1]
br i1 %tmp52471, label %bb111, label %bb59
bb30: ; preds = %bb21, %bb
br i1 %tmp12, label %bb40, label %bb50
bb40: ; preds = %bb30
%tmp52473 = fcmp olt float %tmp2, %tmp8 ; <i1> [#uses=1]
br i1 %tmp52473, label %bb111, label %bb59
bb50: ; preds = %bb30
%tmp52 = fcmp olt float %tmp2, %tmp11 ; <i1> [#uses=1]
br i1 %tmp52, label %bb111, label %bb59
bb59: ; preds = %bb50, %bb40, %bb26
br i1 %tmp12, label %bb72, label %bb80
bb72: ; preds = %bb59
%tmp82475 = fcmp olt float %tmp5, %tmp8 ; <i1> [#uses=2]
%brmerge786 = or i1 %tmp82475, %tmp12 ; <i1> [#uses=1]
%tmp4.mux787 = select i1 %tmp82475, float* %tmp4, float* %tmp7 ; <float*> [#uses=1]
br i1 %brmerge786, label %bb111, label %bb103
bb80: ; preds = %bb59
%tmp82 = fcmp olt float %tmp5, %tmp11 ; <i1> [#uses=2]
%brmerge = or i1 %tmp82, %tmp12 ; <i1> [#uses=1]
%tmp4.mux = select i1 %tmp82, float* %tmp4, float* %tmp7 ; <float*> [#uses=1]
br i1 %brmerge, label %bb111, label %bb103
bb103: ; preds = %bb80, %bb72
br label %bb111
bb111: ; preds = %bb103, %bb80, %bb72, %bb50, %bb40, %bb26
%iftmp.0.0.in = phi float* [ %tmp10, %bb103 ], [ %result, %bb26 ], [ %result, %bb40 ], [ %result, %bb50 ], [ %tmp4.mux, %bb80 ], [ %tmp4.mux787, %bb72 ] ; <float*> [#uses=1]
%iftmp.0.0 = load float, float* %iftmp.0.0.in ; <float> [#uses=1]
2008-02-19 10:09:58 +08:00
%tmp125 = fcmp ogt float %tmp8, %tmp11 ; <i1> [#uses=5]
br i1 %tmp125, label %bb128, label %bb136
bb128: ; preds = %bb111
%tmp138477 = fcmp ogt float %tmp5, %tmp8 ; <i1> [#uses=1]
br i1 %tmp138477, label %bb141, label %bb145
bb136: ; preds = %bb111
%tmp138 = fcmp ogt float %tmp5, %tmp11 ; <i1> [#uses=1]
br i1 %tmp138, label %bb141, label %bb145
bb141: ; preds = %bb136, %bb128
%tmp167479 = fcmp ogt float %tmp2, %tmp5 ; <i1> [#uses=1]
br i1 %tmp167479, label %bb226, label %bb174
bb145: ; preds = %bb136, %bb128
br i1 %tmp125, label %bb155, label %bb165
bb155: ; preds = %bb145
%tmp167481 = fcmp ogt float %tmp2, %tmp8 ; <i1> [#uses=1]
br i1 %tmp167481, label %bb226, label %bb174
bb165: ; preds = %bb145
%tmp167 = fcmp ogt float %tmp2, %tmp11 ; <i1> [#uses=1]
br i1 %tmp167, label %bb226, label %bb174
bb174: ; preds = %bb165, %bb155, %bb141
br i1 %tmp125, label %bb187, label %bb195
bb187: ; preds = %bb174
%tmp197483 = fcmp ogt float %tmp5, %tmp8 ; <i1> [#uses=2]
%brmerge790 = or i1 %tmp197483, %tmp125 ; <i1> [#uses=1]
%tmp4.mux791 = select i1 %tmp197483, float* %tmp4, float* %tmp7 ; <float*> [#uses=1]
br i1 %brmerge790, label %bb226, label %bb218
bb195: ; preds = %bb174
%tmp197 = fcmp ogt float %tmp5, %tmp11 ; <i1> [#uses=2]
%brmerge788 = or i1 %tmp197, %tmp125 ; <i1> [#uses=1]
%tmp4.mux789 = select i1 %tmp197, float* %tmp4, float* %tmp7 ; <float*> [#uses=1]
br i1 %brmerge788, label %bb226, label %bb218
bb218: ; preds = %bb195, %bb187
br label %bb226
bb226: ; preds = %bb218, %bb195, %bb187, %bb165, %bb155, %bb141
%iftmp.7.0.in = phi float* [ %tmp10, %bb218 ], [ %result, %bb141 ], [ %result, %bb155 ], [ %result, %bb165 ], [ %tmp4.mux789, %bb195 ], [ %tmp4.mux791, %bb187 ] ; <float*> [#uses=1]
%iftmp.7.0 = load float, float* %iftmp.7.0.in ; <float> [#uses=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
%tmp229 = getelementptr float, float* %result, i32 1 ; <float*> [#uses=7]
%tmp230 = load float, float* %tmp229, align 4 ; <float> [#uses=6]
[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
%tmp232 = getelementptr float, float* %result, i32 3 ; <float*> [#uses=5]
%tmp233 = load float, float* %tmp232, align 4 ; <float> [#uses=10]
[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
%tmp235 = getelementptr float, float* %result, i32 5 ; <float*> [#uses=5]
%tmp236 = load float, float* %tmp235, align 4 ; <float> [#uses=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
%tmp238 = getelementptr float, float* %result, i32 7 ; <float*> [#uses=3]
%tmp239 = load float, float* %tmp238, align 4 ; <float> [#uses=8]
2008-02-19 10:09:58 +08:00
%tmp240 = fcmp olt float %tmp236, %tmp239 ; <i1> [#uses=5]
br i1 %tmp240, label %bb243, label %bb251
bb243: ; preds = %bb226
%tmp253485 = fcmp olt float %tmp233, %tmp236 ; <i1> [#uses=1]
br i1 %tmp253485, label %bb256, label %bb260
bb251: ; preds = %bb226
%tmp253 = fcmp olt float %tmp233, %tmp239 ; <i1> [#uses=1]
br i1 %tmp253, label %bb256, label %bb260
bb256: ; preds = %bb251, %bb243
%tmp282487 = fcmp olt float %tmp230, %tmp233 ; <i1> [#uses=1]
br i1 %tmp282487, label %bb341, label %bb289
bb260: ; preds = %bb251, %bb243
br i1 %tmp240, label %bb270, label %bb280
bb270: ; preds = %bb260
%tmp282489 = fcmp olt float %tmp230, %tmp236 ; <i1> [#uses=1]
br i1 %tmp282489, label %bb341, label %bb289
bb280: ; preds = %bb260
%tmp282 = fcmp olt float %tmp230, %tmp239 ; <i1> [#uses=1]
br i1 %tmp282, label %bb341, label %bb289
bb289: ; preds = %bb280, %bb270, %bb256
br i1 %tmp240, label %bb302, label %bb310
bb302: ; preds = %bb289
%tmp312491 = fcmp olt float %tmp233, %tmp236 ; <i1> [#uses=2]
%brmerge793 = or i1 %tmp312491, %tmp240 ; <i1> [#uses=1]
%tmp232.mux794 = select i1 %tmp312491, float* %tmp232, float* %tmp235 ; <float*> [#uses=1]
br i1 %brmerge793, label %bb341, label %bb333
bb310: ; preds = %bb289
%tmp312 = fcmp olt float %tmp233, %tmp239 ; <i1> [#uses=2]
%brmerge792 = or i1 %tmp312, %tmp240 ; <i1> [#uses=1]
%tmp232.mux = select i1 %tmp312, float* %tmp232, float* %tmp235 ; <float*> [#uses=1]
br i1 %brmerge792, label %bb341, label %bb333
bb333: ; preds = %bb310, %bb302
br label %bb341
bb341: ; preds = %bb333, %bb310, %bb302, %bb280, %bb270, %bb256
%iftmp.14.0.in = phi float* [ %tmp238, %bb333 ], [ %tmp229, %bb280 ], [ %tmp229, %bb270 ], [ %tmp229, %bb256 ], [ %tmp232.mux, %bb310 ], [ %tmp232.mux794, %bb302 ] ; <float*> [#uses=1]
%iftmp.14.0 = load float, float* %iftmp.14.0.in ; <float> [#uses=1]
2008-02-19 10:09:58 +08:00
%tmp355 = fcmp ogt float %tmp236, %tmp239 ; <i1> [#uses=5]
br i1 %tmp355, label %bb358, label %bb366
bb358: ; preds = %bb341
%tmp368493 = fcmp ogt float %tmp233, %tmp236 ; <i1> [#uses=1]
br i1 %tmp368493, label %bb371, label %bb375
bb366: ; preds = %bb341
%tmp368 = fcmp ogt float %tmp233, %tmp239 ; <i1> [#uses=1]
br i1 %tmp368, label %bb371, label %bb375
bb371: ; preds = %bb366, %bb358
%tmp397495 = fcmp ogt float %tmp230, %tmp233 ; <i1> [#uses=1]
br i1 %tmp397495, label %bb456, label %bb404
bb375: ; preds = %bb366, %bb358
br i1 %tmp355, label %bb385, label %bb395
bb385: ; preds = %bb375
%tmp397497 = fcmp ogt float %tmp230, %tmp236 ; <i1> [#uses=1]
br i1 %tmp397497, label %bb456, label %bb404
bb395: ; preds = %bb375
%tmp397 = fcmp ogt float %tmp230, %tmp239 ; <i1> [#uses=1]
br i1 %tmp397, label %bb456, label %bb404
bb404: ; preds = %bb395, %bb385, %bb371
br i1 %tmp355, label %bb417, label %bb425
bb417: ; preds = %bb404
%tmp427499 = fcmp ogt float %tmp233, %tmp236 ; <i1> [#uses=2]
%brmerge797 = or i1 %tmp427499, %tmp355 ; <i1> [#uses=1]
%tmp232.mux798 = select i1 %tmp427499, float* %tmp232, float* %tmp235 ; <float*> [#uses=1]
br i1 %brmerge797, label %bb456, label %bb448
bb425: ; preds = %bb404
%tmp427 = fcmp ogt float %tmp233, %tmp239 ; <i1> [#uses=2]
%brmerge795 = or i1 %tmp427, %tmp355 ; <i1> [#uses=1]
%tmp232.mux796 = select i1 %tmp427, float* %tmp232, float* %tmp235 ; <float*> [#uses=1]
br i1 %brmerge795, label %bb456, label %bb448
bb448: ; preds = %bb425, %bb417
br label %bb456
bb456: ; preds = %bb448, %bb425, %bb417, %bb395, %bb385, %bb371
%iftmp.21.0.in = phi float* [ %tmp238, %bb448 ], [ %tmp229, %bb395 ], [ %tmp229, %bb385 ], [ %tmp229, %bb371 ], [ %tmp232.mux796, %bb425 ], [ %tmp232.mux798, %bb417 ] ; <float*> [#uses=1]
%iftmp.21.0 = load float, float* %iftmp.21.0.in ; <float> [#uses=1]
2008-02-19 10:09:58 +08:00
%tmp458459 = fpext float %iftmp.21.0 to double ; <double> [#uses=1]
%tmp460461 = fpext float %iftmp.7.0 to double ; <double> [#uses=1]
%tmp462463 = fpext float %iftmp.14.0 to double ; <double> [#uses=1]
%tmp464465 = fpext float %iftmp.0.0 to double ; <double> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to the call instruction See r230786 and r230794 for similar changes to gep and load respectively. Call is a bit different because it often doesn't have a single explicit type - usually the type is deduced from the arguments, and just the return type is explicit. In those cases there's no need to change the IR. When that's not the case, the IR usually contains the pointer type of the first operand - but since typed pointers are going away, that representation is insufficient so I'm just stripping the "pointerness" of the explicit type away. This does make the IR a bit weird - it /sort of/ reads like the type of the first operand: "call void () %x(" but %x is actually of type "void ()*" and will eventually be just of type "ptr". But this seems not too bad and I don't think it would benefit from repeating the type ("void (), void () * %x(" and then eventually "void (), ptr %x(") as has been done with gep and load. This also has a side benefit: since the explicit type is no longer a pointer, there's no ambiguity between an explicit type and a function that returns a function pointer. Previously this case needed an explicit type (eg: a function returning a void() function was written as "call void () () * @x(" rather than "call void () * @x(" because of the ambiguity between a function returning a pointer to a void() function and a function returning void). No ambiguity means even function pointer return types can just be written alone, without writing the whole function's type. This leaves /only/ the varargs case where the explicit type is required. Given the special type syntax in call instructions, the regex-fu used for migration was a bit more involved in its own unique way (as every one of these is) so here it is. Use it in conjunction with the apply.sh script and associated find/xargs commands I've provided in rr230786 to migrate your out of tree tests. Do let me know if any of this doesn't cover your cases & we can iterate on a more general script/regexes to help others with out of tree tests. About 9 test cases couldn't be automatically migrated - half of those were functions returning function pointers, where I just had to manually delete the function argument types now that we didn't need an explicit function type there. The other half were typedefs of function types used in calls - just had to manually drop the * from those. import fileinput import sys import re pat = re.compile(r'((?:=|:|^|\s)call\s(?:[^@]*?))(\s*$|\s*(?:(?:\[\[[a-zA-Z0-9_]+\]\]|[@%](?:(")?[\\\?@a-zA-Z0-9_.]*?(?(3)"|)|{{.*}}))(?:\(|$)|undef|inttoptr|bitcast|null|asm).*$)') addrspace_end = re.compile(r"addrspace\(\d+\)\s*\*$") func_end = re.compile("(?:void.*|\)\s*)\*$") def conv(match, line): if not match or re.search(addrspace_end, match.group(1)) or not re.search(func_end, match.group(1)): return line return line[:match.start()] + match.group(1)[:match.group(1).rfind('*')].rstrip() + match.group(2) + line[match.end():] for line in sys.stdin: sys.stdout.write(conv(re.search(pat, line), line)) llvm-svn: 235145
2015-04-17 07:24:18 +08:00
%tmp467 = tail call i32 (i8*, ...) @printf( i8* getelementptr ([48 x i8], [48 x i8]* @.str, i32 0, i32 0), double %tmp464465, double %tmp462463, double %tmp460461, double %tmp458459 ) nounwind ; <i32> [#uses=0]
2008-02-19 10:09:58 +08:00
ret void
}
declare i32 @printf(i8*, ...) nounwind