RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-26 14:33:02 +02:00
Code Issues Projects Releases Wiki Activity
277d7cfcc2
llvm-mirror/test/CodeGen/X86/lsr-delayed-fold.ll

179 lines
7.9 KiB
LLVM
Raw Normal View History

Fix LSR to tolerate cases where ScalarEvolution initially misses an opportunity to fold add operands, but folds them after LSR has separated them out. This fixes rdar://7886751. llvm-svn: 102157
2010-04-23 03:55:05 +02:00
; RUN: llc -march=x86-64 < %s > /dev/null
; ScalarEvolution misses an opportunity to fold ((trunc x) + (trunc -x) + y),
; but LSR should tolerate this.
When checking whether the special handling for an addrec increment which doesn't dominate the header is needed, don't check whether the increment expression has computable loop evolution. While the operands of an addrec are required to be loop-invariant, they're not required to dominate any part of the loop. This fixes PR6914. llvm-svn: 102389
2010-04-26 23:46:36 +02:00
; rdar://7886751
Fix LSR to tolerate cases where ScalarEvolution initially misses an opportunity to fold add operands, but folds them after LSR has separated them out. This fixes rdar://7886751. llvm-svn: 102157
2010-04-23 03:55:05 +02:00
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
target triple = "x86_64-apple-darwin11.0"
define fastcc void @formatValue(i64 %arg5) nounwind {
bb12: ; preds = %bb11
%t = trunc i64 %arg5 to i32 ; <i32> [#uses=1]
%t13 = sub i64 0, %arg5 ; <i64> [#uses=1]
%t14 = and i64 %t13, 4294967295 ; <i64> [#uses=1]
br label %bb15
bb15: ; preds = %bb21, %bb12
%t16 = phi i64 [ 0, %bb12 ], [ %t23, %bb15 ] ; <i64> [#uses=2]
%t17 = mul i64 %t14, %t16 ; <i64> [#uses=1]
%t18 = add i64 undef, %t17 ; <i64> [#uses=1]
%t19 = trunc i64 %t18 to i32 ; <i32> [#uses=1]
%t22 = icmp eq i32 %t19, %t ; <i1> [#uses=1]
%t23 = add i64 %t16, 1 ; <i64> [#uses=1]
br i1 %t22, label %bb24, label %bb15
bb24: ; preds = %bb21, %bb11
unreachable
}
When checking whether the special handling for an addrec increment which doesn't dominate the header is needed, don't check whether the increment expression has computable loop evolution. While the operands of an addrec are required to be loop-invariant, they're not required to dominate any part of the loop. This fixes PR6914. llvm-svn: 102389
2010-04-26 23:46:36 +02:00
; ScalarEvolution should be able to correctly expand the crazy addrec here.
; PR6914
define void @int323() nounwind {
entry:
br label %for.cond
for.cond: ; preds = %lbl_264, %for.inc, %entry
%g_263.tmp.1 = phi i8 [ undef, %entry ], [ %g_263.tmp.1, %for.cond ]
%p_95.addr.0 = phi i8 [ 0, %entry ], [ %add, %for.cond ]
%add = add i8 %p_95.addr.0, 1 ; <i8> [#uses=1]
br i1 undef, label %for.cond, label %lbl_264
lbl_264: ; preds = %if.end, %lbl_264.preheader
%g_263.tmp.0 = phi i8 [ %g_263.tmp.1, %for.cond ] ; <i8> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-27 22:17:42 +01:00
%tmp7 = load i16, i16* undef ; <i16> [#uses=1]
When checking whether the special handling for an addrec increment which doesn't dominate the header is needed, don't check whether the increment expression has computable loop evolution. While the operands of an addrec are required to be loop-invariant, they're not required to dominate any part of the loop. This fixes PR6914. llvm-svn: 102389
2010-04-26 23:46:36 +02:00
%conv8 = trunc i16 %tmp7 to i8 ; <i8> [#uses=1]
%mul.i = mul i8 %p_95.addr.0, %p_95.addr.0 ; <i8> [#uses=1]
%mul.i18 = mul i8 %mul.i, %conv8 ; <i8> [#uses=1]
%tobool12 = icmp eq i8 %mul.i18, 0 ; <i1> [#uses=1]
unreachable
}
When pruning candidate formulae out of an LSRUse, update the LSRUse's Regs set after all pruning is done, rather than trying to do it on the fly, which can produce an incomplete result. This fixes a case where heuristic pruning was stripping all formulae from a use, which led the solver to enter an infinite loop. Also, add a few asserts to diagnose this kind of situation. llvm-svn: 103328
2010-05-08 01:36:59 +02:00
; LSR ends up going into conservative pruning mode; don't prune the solution
; so far that it becomes unsolvable though.
; PR7077
%struct.Bu = type { i32, i32, i32 }
define void @_Z3fooP2Bui(%struct.Bu* nocapture %bu) {
entry:
br label %for.body
for.body: ; preds = %for.inc131, %entry
%indvar = phi i64 [ %indvar.next, %for.inc131 ], [ 0, %entry ] ; <i64> [#uses=3]
br i1 undef, label %for.inc131, label %lor.lhs.false
lor.lhs.false: ; preds = %for.body
%tmp15 = add i64 %indvar, 1 ; <i64> [#uses=1]
%tmp17 = add i64 %indvar, 2 ; <i64> [#uses=1]
%tmp19 = add i64 %indvar, 3 ; <i64> [#uses=1]
%tmp21 = add i64 %indvar, 4 ; <i64> [#uses=1]
%tmp23 = add i64 %indvar, 5 ; <i64> [#uses=1]
%tmp25 = add i64 %indvar, 6 ; <i64> [#uses=1]
%tmp27 = add i64 %indvar, 7 ; <i64> [#uses=1]
%tmp29 = add i64 %indvar, 8 ; <i64> [#uses=1]
%tmp31 = add i64 %indvar, 9 ; <i64> [#uses=1]
%tmp35 = add i64 %indvar, 11 ; <i64> [#uses=1]
%tmp37 = add i64 %indvar, 12 ; <i64> [#uses=1]
%tmp39 = add i64 %indvar, 13 ; <i64> [#uses=1]
%tmp41 = add i64 %indvar, 14 ; <i64> [#uses=1]
%tmp43 = add i64 %indvar, 15 ; <i64> [#uses=1]
%tmp45 = add i64 %indvar, 16 ; <i64> [#uses=1]
%tmp47 = add i64 %indvar, 17 ; <i64> [#uses=1]
%mul = trunc i64 %indvar to i32 ; <i32> [#uses=1]
%add22 = trunc i64 %tmp15 to i32 ; <i32> [#uses=1]
%add28 = trunc i64 %tmp17 to i32 ; <i32> [#uses=1]
%add34 = trunc i64 %tmp19 to i32 ; <i32> [#uses=1]
%add40 = trunc i64 %tmp21 to i32 ; <i32> [#uses=1]
%add46 = trunc i64 %tmp23 to i32 ; <i32> [#uses=1]
%add52 = trunc i64 %tmp25 to i32 ; <i32> [#uses=1]
%add58 = trunc i64 %tmp27 to i32 ; <i32> [#uses=1]
%add64 = trunc i64 %tmp29 to i32 ; <i32> [#uses=1]
%add70 = trunc i64 %tmp31 to i32 ; <i32> [#uses=1]
%add82 = trunc i64 %tmp35 to i32 ; <i32> [#uses=1]
%add88 = trunc i64 %tmp37 to i32 ; <i32> [#uses=1]
%add94 = trunc i64 %tmp39 to i32 ; <i32> [#uses=1]
%add100 = trunc i64 %tmp41 to i32 ; <i32> [#uses=1]
%add106 = trunc i64 %tmp43 to i32 ; <i32> [#uses=1]
%add112 = trunc i64 %tmp45 to i32 ; <i32> [#uses=1]
%add118 = trunc i64 %tmp47 to i32 ; <i32> [#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
2015-02-27 20:29:02 +01:00
%tmp10 = getelementptr %struct.Bu, %struct.Bu* %bu, i64 %indvar, i32 2 ; <i32*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-27 22:17:42 +01:00
%tmp11 = load i32, i32* %tmp10 ; <i32> [#uses=0]
When pruning candidate formulae out of an LSRUse, update the LSRUse's Regs set after all pruning is done, rather than trying to do it on the fly, which can produce an incomplete result. This fixes a case where heuristic pruning was stripping all formulae from a use, which led the solver to enter an infinite loop. Also, add a few asserts to diagnose this kind of situation. llvm-svn: 103328
2010-05-08 01:36:59 +02:00
tail call void undef(i32 %add22)
tail call void undef(i32 %add28)
tail call void undef(i32 %add34)
tail call void undef(i32 %add40)
tail call void undef(i32 %add46)
tail call void undef(i32 %add52)
tail call void undef(i32 %add58)
tail call void undef(i32 %add64)
tail call void undef(i32 %add70)
tail call void undef(i32 %add82)
tail call void undef(i32 %add88)
tail call void undef(i32 %add94)
tail call void undef(i32 %add100)
tail call void undef(i32 %add106)
tail call void undef(i32 %add112)
tail call void undef(i32 %add118)
br label %for.body123
for.body123: ; preds = %for.body123, %lor.lhs.false
%j.03 = phi i32 [ 0, %lor.lhs.false ], [ %inc, %for.body123 ] ; <i32> [#uses=2]
%add129 = add i32 %mul, %j.03 ; <i32> [#uses=1]
tail call void undef(i32 %add129)
%inc = add nsw i32 %j.03, 1 ; <i32> [#uses=1]
br i1 undef, label %for.inc131, label %for.body123
for.inc131: ; preds = %for.body123, %for.body
%indvar.next = add i64 %indvar, 1 ; <i64> [#uses=1]
br i1 undef, label %for.end134, label %for.body
for.end134: ; preds = %for.inc131
ret void
}
LSR needs to remember inserted instructions even in postinc mode, because there could be multiple subexpressions within a single expansion which require insert point adjustment. This fixes PR7306. llvm-svn: 105510
2010-06-05 02:33:07 +02:00
; LSR needs to remember inserted instructions even in postinc mode, because
; there could be multiple subexpressions within a single expansion which
; require insert point adjustment.
; PR7306
define fastcc i32 @GetOptimum() nounwind {
bb:
br label %bb1
bb1: ; preds = %bb1, %bb
%t = phi i32 [ 0, %bb ], [ %t2, %bb1 ] ; <i32> [#uses=1]
%t2 = add i32 %t, undef ; <i32> [#uses=3]
br i1 undef, label %bb1, label %bb3
bb3: ; preds = %bb1
%t4 = add i32 undef, -1 ; <i32> [#uses=1]
br label %bb5
bb5: ; preds = %bb16, %bb3
%t6 = phi i32 [ %t17, %bb16 ], [ 0, %bb3 ] ; <i32> [#uses=3]
%t7 = add i32 undef, %t6 ; <i32> [#uses=2]
%t8 = add i32 %t4, %t6 ; <i32> [#uses=1]
br i1 undef, label %bb9, label %bb10
bb9: ; preds = %bb5
br label %bb10
bb10: ; preds = %bb9, %bb5
br i1 undef, label %bb11, label %bb16
bb11: ; preds = %bb10
%t12 = icmp ugt i32 %t7, %t2 ; <i1> [#uses=1]
%t13 = select i1 %t12, i32 %t2, i32 %t7 ; <i32> [#uses=1]
br label %bb14
bb14: ; preds = %bb11
store i32 %t13, i32* null
ret i32 %t8
bb16: ; preds = %bb10
%t17 = add i32 %t6, 1 ; <i32> [#uses=1]
br label %bb5
}
Reference in New Issue Copy Permalink