1 2=pod 3 4=head1 NAME 5 6FileCheck - Flexible pattern matching file verifier 7 8=head1 SYNOPSIS 9 10B<FileCheck> I<match-filename> [I<--check-prefix=XXX>] [I<--strict-whitespace>] 11 12=head1 DESCRIPTION 13 14B<FileCheck> reads two files (one from standard input, and one specified on the 15command line) and uses one to verify the other. This behavior is particularly 16useful for the testsuite, which wants to verify that the output of some tool 17(e.g. llc) contains the expected information (for example, a movsd from esp or 18whatever is interesting). This is similar to using grep, but it is optimized 19for matching multiple different inputs in one file in a specific order. 20 21The I<match-filename> file specifies the file that contains the patterns to 22match. The file to verify is always read from standard input. 23 24=head1 OPTIONS 25 26=over 27 28=item B<-help> 29 30Print a summary of command line options. 31 32=item B<--check-prefix> I<prefix> 33 34FileCheck searches the contents of I<match-filename> for patterns to match. By 35default, these patterns are prefixed with "CHECK:". If you'd like to use a 36different prefix (e.g. because the same input file is checking multiple 37different tool or options), the B<--check-prefix> argument allows you to specify 38a specific prefix to match. 39 40=item B<--strict-whitespace> 41 42By default, FileCheck canonicalizes input horizontal whitespace (spaces and 43tabs) which causes it to ignore these differences (a space will match a tab). 44The --strict-whitespace argument disables this behavior. 45 46=item B<-version> 47 48Show the version number of this program. 49 50=back 51 52=head1 EXIT STATUS 53 54If B<FileCheck> verifies that the file matches the expected contents, it exits 55with 0. Otherwise, if not, or if an error occurs, it will exit with a non-zero 56value. 57 58=head1 TUTORIAL 59 60FileCheck is typically used from LLVM regression tests, being invoked on the RUN 61line of the test. A simple example of using FileCheck from a RUN line looks 62like this: 63 64 ; RUN: llvm-as < %s | llc -march=x86-64 | FileCheck %s 65 66This syntax says to pipe the current file ("%s") into llvm-as, pipe that into 67llc, then pipe the output of llc into FileCheck. This means that FileCheck will 68be verifying its standard input (the llc output) against the filename argument 69specified (the original .ll file specified by "%s"). To see how this works, 70lets look at the rest of the .ll file (after the RUN line): 71 72 define void @sub1(i32* %p, i32 %v) { 73 entry: 74 ; <b>CHECK: sub1:</b> 75 ; <b>CHECK: subl</b> 76 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v) 77 ret void 78 } 79 80 define void @inc4(i64* %p) { 81 entry: 82 ; <b>CHECK: inc4:</b> 83 ; <b>CHECK: incq</b> 84 %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1) 85 ret void 86 } 87 88Here you can see some "CHECK:" lines specified in comments. Now you can see 89how the file is piped into llvm-as, then llc, and the machine code output is 90what we are verifying. FileCheck checks the machine code output to verify that 91it matches what the "CHECK:" lines specify. 92 93The syntax of the CHECK: lines is very simple: they are fixed strings that 94must occur in order. FileCheck defaults to ignoring horizontal whitespace 95differences (e.g. a space is allowed to match a tab) but otherwise, the contents 96of the CHECK: line is required to match some thing in the test file exactly. 97 98One nice thing about FileCheck (compared to grep) is that it allows merging 99test cases together into logical groups. For example, because the test above 100is checking for the "sub1:" and "inc4:" labels, it will not match unless there 101is a "subl" in between those labels. If it existed somewhere else in the file, 102that would not count: "grep subl" matches if subl exists anywhere in the 103file. 104 105 106 107=head2 The FileCheck -check-prefix option 108 109The FileCheck -check-prefix option allows multiple test configurations to be 110driven from one .ll file. This is useful in many circumstances, for example, 111testing different architectural variants with llc. Here's a simple example: 112 113 ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \ 114 ; RUN: | <b>FileCheck %s -check-prefix=X32</b> 115 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \ 116 ; RUN: | <b>FileCheck %s -check-prefix=X64</b> 117 118 define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind { 119 %tmp1 = insertelement <4 x i32>; %tmp, i32 %s, i32 1 120 ret <4 x i32> %tmp1 121 ; <b>X32:</b> pinsrd_1: 122 ; <b>X32:</b> pinsrd $1, 4(%esp), %xmm0 123 124 ; <b>X64:</b> pinsrd_1: 125 ; <b>X64:</b> pinsrd $1, %edi, %xmm0 126 } 127 128In this case, we're testing that we get the expected code generation with 129both 32-bit and 64-bit code generation. 130 131 132 133=head2 The "CHECK-NEXT:" directive 134 135Sometimes you want to match lines and would like to verify that matches 136happen on exactly consecutive lines with no other lines in between them. In 137this case, you can use CHECK: and CHECK-NEXT: directives to specify this. If 138you specified a custom check prefix, just use "<PREFIX>-NEXT:". For 139example, something like this works as you'd expect: 140 141 define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) { 142 %tmp3 = load <2 x double>* %A, align 16 143 %tmp7 = insertelement <2 x double> undef, double %B, i32 0 144 %tmp9 = shufflevector <2 x double> %tmp3, 145 <2 x double> %tmp7, 146 <2 x i32> < i32 0, i32 2 > 147 store <2 x double> %tmp9, <2 x double>* %r, align 16 148 ret void 149 150 ; <b>CHECK:</b> t2: 151 ; <b>CHECK:</b> movl 8(%esp), %eax 152 ; <b>CHECK-NEXT:</b> movapd (%eax), %xmm0 153 ; <b>CHECK-NEXT:</b> movhpd 12(%esp), %xmm0 154 ; <b>CHECK-NEXT:</b> movl 4(%esp), %eax 155 ; <b>CHECK-NEXT:</b> movapd %xmm0, (%eax) 156 ; <b>CHECK-NEXT:</b> ret 157 } 158 159CHECK-NEXT: directives reject the input unless there is exactly one newline 160between it an the previous directive. A CHECK-NEXT cannot be the first 161directive in a file. 162 163 164 165=head2 The "CHECK-NOT:" directive 166 167The CHECK-NOT: directive is used to verify that a string doesn't occur 168between two matches (or before the first match, or after the last match). For 169example, to verify that a load is removed by a transformation, a test like this 170can be used: 171 172 define i8 @coerce_offset0(i32 %V, i32* %P) { 173 store i32 %V, i32* %P 174 175 %P2 = bitcast i32* %P to i8* 176 %P3 = getelementptr i8* %P2, i32 2 177 178 %A = load i8* %P3 179 ret i8 %A 180 ; <b>CHECK:</b> @coerce_offset0 181 ; <b>CHECK-NOT:</b> load 182 ; <b>CHECK:</b> ret i8 183 } 184 185 186 187=head2 FileCheck Pattern Matching Syntax 188 189The CHECK: and CHECK-NOT: directives both take a pattern to match. For most 190uses of FileCheck, fixed string matching is perfectly sufficient. For some 191things, a more flexible form of matching is desired. To support this, FileCheck 192allows you to specify regular expressions in matching strings, surrounded by 193double braces: B<{{yourregex}}>. Because we want to use fixed string 194matching for a majority of what we do, FileCheck has been designed to support 195mixing and matching fixed string matching with regular expressions. This allows 196you to write things like this: 197 198 ; CHECK: movhpd <b>{{[0-9]+}}</b>(%esp), <b>{{%xmm[0-7]}}</b> 199 200In this case, any offset from the ESP register will be allowed, and any xmm 201register will be allowed. 202 203Because regular expressions are enclosed with double braces, they are 204visually distinct, and you don't need to use escape characters within the double 205braces like you would in C. In the rare case that you want to match double 206braces explicitly from the input, you can use something ugly like 207B<{{[{][{]}}> as your pattern. 208 209 210 211=head2 FileCheck Variables 212 213It is often useful to match a pattern and then verify that it occurs again 214later in the file. For codegen tests, this can be useful to allow any register, 215but verify that that register is used consistently later. To do this, FileCheck 216allows named variables to be defined and substituted into patterns. Here is a 217simple example: 218 219 ; CHECK: test5: 220 ; CHECK: notw <b>[[REGISTER:%[a-z]+]]</b> 221 ; CHECK: andw {{.*}}<b>[[REGISTER]]</b> 222 223The first check line matches a regex (<tt>%[a-z]+</tt>) and captures it into 224the variables "REGISTER". The second line verifies that whatever is in REGISTER 225occurs later in the file after an "andw". FileCheck variable references are 226always contained in <tt>[[ ]]</tt> pairs, are named, and their names can be 227formed with the regex "<tt>[a-zA-Z_][a-zA-Z0-9_]*</tt>". If a colon follows the 228name, then it is a definition of the variable, if not, it is a use. 229 230FileCheck variables can be defined multiple times, and uses always get the 231latest value. Note that variables are all read at the start of a "CHECK" line 232and are all defined at the end. This means that if you have something like 233"<tt>CHECK: [[XYZ:.*]]x[[XYZ]]<tt>" that the check line will read the previous 234value of the XYZ variable and define a new one after the match is performed. If 235you need to do something like this you can probably take advantage of the fact 236that FileCheck is not actually line-oriented when it matches, this allows you to 237define two separate CHECK lines that match on the same line. 238 239 240 241=head1 AUTHORS 242 243Maintained by The LLVM Team (L<http://llvm.org/>). 244 245=cut 246