maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/mul-constant-result.ll

1239 lines
39 KiB
LLVM
Raw Normal View History

This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown | FileCheck %s --check-prefix=X86
Recommit [MachineCombiner] Update instruction depths incrementally for large BBs. This version of the patch fixes an off-by-one error causing PR34596. We do not need to use std::next(BlockIter) when calling updateDepths, as BlockIter already points to the next element. Original commit message: > For large basic blocks with lots of combinable instructions, the > MachineTraceMetrics computations in MachineCombiner can dominate the compile > time, as computing the trace information is quadratic in the number of > instructions in a BB and it's relevant successors/predecessors. > In most cases, knowing the instruction depth should be enough to make > combination decisions. As we already iterate over all instructions in a basic > block, the instruction depth can be computed incrementally. This reduces the > cost of machine-combine drastically in cases where lots of instructions > are combined. The major drawback is that AFAIK, computing the critical path > length cannot be done incrementally. Therefore we only compute > instruction depths incrementally, for basic blocks with more > instructions than inc_threshold. The -machine-combiner-inc-threshold > option can be used to set the threshold and allows for easier > experimenting and checking if using incremental updates for all basic > blocks has any impact on the performance. > > Reviewers: sanjoy, Gerolf, MatzeB, efriedma, fhahn > > Reviewed By: fhahn > > Subscribers: kiranchandramohan, javed.absar, efriedma, llvm-commits > > Differential Revision: https://reviews.llvm.org/D36619 llvm-svn: 313751
2017-09-20 19:54:37 +08:00
; Incremental updates of the instruction depths should be enough for this test
; case.
; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=haswell -machine-combiner-inc-threshold=0| FileCheck %s --check-prefix=X64-HSW
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; Function Attrs: norecurse nounwind readnone uwtable
define i32 @mult(i32, i32) local_unnamed_addr #0 {
; X86-LABEL: mult:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X86: # %bb.0:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: pushl %esi
; X86-NEXT: .cfi_def_cfa_offset 8
; X86-NEXT: .cfi_offset %esi, -8
; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-NEXT: cmpl $1, %edx
; X86-NEXT: movl $1, %eax
; X86-NEXT: movl $1, %esi
; X86-NEXT: jg .LBB0_2
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X86-NEXT: # %bb.1:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: movl %edx, %esi
; X86-NEXT: .LBB0_2:
; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: testl %edx, %edx
; X86-NEXT: je .LBB0_4
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X86-NEXT: # %bb.3:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: movl %esi, %eax
; X86-NEXT: .LBB0_4:
; X86-NEXT: decl %ecx
; X86-NEXT: cmpl $31, %ecx
; X86-NEXT: ja .LBB0_39
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X86-NEXT: # %bb.5:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: jmpl *.LJTI0_0(,%ecx,4)
; X86-NEXT: .LBB0_6:
; X86-NEXT: addl %eax, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_39:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: xorl %eax, %eax
; X86-NEXT: .LBB0_40:
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_7:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_8:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: shll $2, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_9:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_10:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: addl %eax, %eax
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_11:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (,%eax,8), %ecx
; X86-NEXT: jmp .LBB0_12
; X86-NEXT: .LBB0_13:
; X86-NEXT: shll $3, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_14:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,8), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_15:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: addl %eax, %eax
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_16:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %ecx
; X86-NEXT: leal (%eax,%ecx,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_17:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: shll $2, %eax
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_18:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,2), %ecx
; X86-NEXT: leal (%eax,%ecx,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_19:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,2), %ecx
; X86-NEXT: jmp .LBB0_20
; X86-NEXT: .LBB0_21:
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_22:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: shll $4, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_23:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: shll $4, %ecx
; X86-NEXT: addl %ecx, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_24:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: addl %eax, %eax
; X86-NEXT: leal (%eax,%eax,8), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_25:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %ecx
; X86-NEXT: shll $2, %ecx
; X86-NEXT: jmp .LBB0_12
; X86-NEXT: .LBB0_26:
; X86-NEXT: shll $2, %eax
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_27:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %ecx
; X86-NEXT: leal (%eax,%ecx,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_28:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %ecx
; X86-NEXT: .LBB0_20:
; X86-NEXT: leal (%eax,%ecx,4), %ecx
; X86-NEXT: addl %ecx, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_29:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,2), %ecx
; X86-NEXT: shll $3, %ecx
; X86-NEXT: jmp .LBB0_12
; X86-NEXT: .LBB0_30:
; X86-NEXT: shll $3, %eax
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_31:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: leal (%eax,%eax,4), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_32:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,8), %ecx
; X86-NEXT: leal (%ecx,%ecx,2), %ecx
; X86-NEXT: jmp .LBB0_12
; X86-NEXT: .LBB0_33:
; X86-NEXT: leal (%eax,%eax,8), %eax
; X86-NEXT: leal (%eax,%eax,2), %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_34:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,8), %ecx
; X86-NEXT: leal (%ecx,%ecx,2), %ecx
; X86-NEXT: addl %ecx, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_35:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: leal (%eax,%eax,8), %ecx
; X86-NEXT: leal (%ecx,%ecx,2), %ecx
Revert "[X86] Improvement in CodeGen instruction selection for LEAs." This reverts r319543, due to ASan bot breakage. llvm-svn: 319591
2017-12-02 06:20:26 +08:00
; X86-NEXT: addl %eax, %ecx
; X86-NEXT: addl %ecx, %eax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_36:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: shll $5, %ecx
; X86-NEXT: subl %eax, %ecx
; X86-NEXT: jmp .LBB0_12
; X86-NEXT: .LBB0_37:
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: shll $5, %ecx
; X86-NEXT: .LBB0_12:
; X86-NEXT: subl %eax, %ecx
; X86-NEXT: movl %ecx, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
; X86-NEXT: .LBB0_38:
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: shll $5, %eax
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
;
; X64-HSW-LABEL: mult:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X64-HSW: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $edi killed $edi def $rdi
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: cmpl $1, %esi
; X64-HSW-NEXT: movl $1, %ecx
; X64-HSW-NEXT: movl %esi, %eax
; X64-HSW-NEXT: cmovgl %ecx, %eax
; X64-HSW-NEXT: testl %esi, %esi
; X64-HSW-NEXT: cmovel %ecx, %eax
; X64-HSW-NEXT: addl $-1, %edi
; X64-HSW-NEXT: cmpl $31, %edi
; X64-HSW-NEXT: ja .LBB0_36
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X64-HSW-NEXT: # %bb.1:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: jmpq *.LJTI0_0(,%rdi,8)
; X64-HSW-NEXT: .LBB0_2:
; X64-HSW-NEXT: addl %eax, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_36:
; X64-HSW-NEXT: xorl %eax, %eax
; X64-HSW-NEXT: .LBB0_37:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_3:
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_4:
; X64-HSW-NEXT: shll $2, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_5:
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_6:
; X64-HSW-NEXT: addl %eax, %eax
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_7:
; X64-HSW-NEXT: leal (,%rax,8), %ecx
; X64-HSW-NEXT: jmp .LBB0_8
; X64-HSW-NEXT: .LBB0_9:
; X64-HSW-NEXT: shll $3, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_10:
; X64-HSW-NEXT: leal (%rax,%rax,8), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_11:
; X64-HSW-NEXT: addl %eax, %eax
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_12:
; X64-HSW-NEXT: leal (%rax,%rax,4), %ecx
; X64-HSW-NEXT: leal (%rax,%rcx,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_13:
; X64-HSW-NEXT: shll $2, %eax
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_14:
; X64-HSW-NEXT: leal (%rax,%rax,2), %ecx
; X64-HSW-NEXT: leal (%rax,%rcx,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_15:
; X64-HSW-NEXT: leal (%rax,%rax,2), %ecx
; X64-HSW-NEXT: jmp .LBB0_16
; X64-HSW-NEXT: .LBB0_18:
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_19:
; X64-HSW-NEXT: shll $4, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_20:
; X64-HSW-NEXT: movl %eax, %ecx
; X64-HSW-NEXT: shll $4, %ecx
; X64-HSW-NEXT: jmp .LBB0_17
; X64-HSW-NEXT: .LBB0_21:
; X64-HSW-NEXT: addl %eax, %eax
; X64-HSW-NEXT: leal (%rax,%rax,8), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_22:
; X64-HSW-NEXT: leal (%rax,%rax,4), %ecx
; X64-HSW-NEXT: shll $2, %ecx
; X64-HSW-NEXT: jmp .LBB0_8
; X64-HSW-NEXT: .LBB0_23:
; X64-HSW-NEXT: shll $2, %eax
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_24:
; X64-HSW-NEXT: leal (%rax,%rax,4), %ecx
; X64-HSW-NEXT: leal (%rax,%rcx,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_25:
; X64-HSW-NEXT: leal (%rax,%rax,4), %ecx
; X64-HSW-NEXT: .LBB0_16:
; X64-HSW-NEXT: leal (%rax,%rcx,4), %ecx
; X64-HSW-NEXT: jmp .LBB0_17
; X64-HSW-NEXT: .LBB0_26:
; X64-HSW-NEXT: leal (%rax,%rax,2), %ecx
; X64-HSW-NEXT: shll $3, %ecx
; X64-HSW-NEXT: jmp .LBB0_8
; X64-HSW-NEXT: .LBB0_27:
; X64-HSW-NEXT: shll $3, %eax
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_28:
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
; X64-HSW-NEXT: leal (%rax,%rax,4), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_29:
; X64-HSW-NEXT: leal (%rax,%rax,8), %ecx
; X64-HSW-NEXT: leal (%rcx,%rcx,2), %ecx
; X64-HSW-NEXT: jmp .LBB0_8
; X64-HSW-NEXT: .LBB0_30:
; X64-HSW-NEXT: leal (%rax,%rax,8), %eax
; X64-HSW-NEXT: leal (%rax,%rax,2), %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_31:
; X64-HSW-NEXT: leal (%rax,%rax,8), %ecx
; X64-HSW-NEXT: leal (%rcx,%rcx,2), %ecx
Revert "[X86] Improvement in CodeGen instruction selection for LEAs." This reverts r319543, due to ASan bot breakage. llvm-svn: 319591
2017-12-02 06:20:26 +08:00
; X64-HSW-NEXT: jmp .LBB0_17
[X86] Improvement in CodeGen instruction selection for LEAs. Summary: 1/ Operand folding during complex pattern matching for LEAs has been extended, such that it promotes Scale to accommodate similar operand appearing in the DAG e.g. T1 = A + B T2 = T1 + 10 T3 = T2 + A For above DAG rooted at T3, X86AddressMode will now look like Base = B , Index = A , Scale = 2 , Disp = 10 2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs so that if there is an opportunity then complex LEAs (having 3 operands) could be factored out e.g. leal 1(%rax,%rcx,1), %rdx leal 1(%rax,%rcx,2), %rcx will be factored as following leal 1(%rax,%rcx,1), %rdx leal (%rdx,%rcx) , %edx 3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops, thus avoiding creation of any complex LEAs within a loop. 4/ Simplify LEA converts (lea (BASE,1,INDEX,0) --> add (BASE, INDEX) which offers better through put. PR32755 will be taken care of by this pathc. Previous patch revisions : r313343 , r314886 Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy, jbhateja Reviewed By: lsaba, RKSimon, jbhateja Subscribers: jmolloy, spatel, igorb, llvm-commits Differential Revision: https://reviews.llvm.org/D35014 llvm-svn: 319543
2017-12-01 22:07:38 +08:00
; X64-HSW-NEXT: .LBB0_32:
; X64-HSW-NEXT: leal (%rax,%rax,8), %ecx
; X64-HSW-NEXT: leal (%rcx,%rcx,2), %ecx
Revert "[X86] Improvement in CodeGen instruction selection for LEAs." This reverts r319543, due to ASan bot breakage. llvm-svn: 319591
2017-12-02 06:20:26 +08:00
; X64-HSW-NEXT: addl %eax, %ecx
; X64-HSW-NEXT: .LBB0_17:
; X64-HSW-NEXT: addl %eax, %ecx
; X64-HSW-NEXT: movl %ecx, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
[X86] Improvement in CodeGen instruction selection for LEAs. Summary: 1/ Operand folding during complex pattern matching for LEAs has been extended, such that it promotes Scale to accommodate similar operand appearing in the DAG e.g. T1 = A + B T2 = T1 + 10 T3 = T2 + A For above DAG rooted at T3, X86AddressMode will now look like Base = B , Index = A , Scale = 2 , Disp = 10 2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs so that if there is an opportunity then complex LEAs (having 3 operands) could be factored out e.g. leal 1(%rax,%rcx,1), %rdx leal 1(%rax,%rcx,2), %rcx will be factored as following leal 1(%rax,%rcx,1), %rdx leal (%rdx,%rcx) , %edx 3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops, thus avoiding creation of any complex LEAs within a loop. 4/ Simplify LEA converts (lea (BASE,1,INDEX,0) --> add (BASE, INDEX) which offers better through put. PR32755 will be taken care of by this pathc. Previous patch revisions : r313343 , r314886 Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy, jbhateja Reviewed By: lsaba, RKSimon, jbhateja Subscribers: jmolloy, spatel, igorb, llvm-commits Differential Revision: https://reviews.llvm.org/D35014 llvm-svn: 319543
2017-12-01 22:07:38 +08:00
; X64-HSW-NEXT: retq
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: .LBB0_33:
; X64-HSW-NEXT: movl %eax, %ecx
; X64-HSW-NEXT: shll $5, %ecx
; X64-HSW-NEXT: subl %eax, %ecx
; X64-HSW-NEXT: jmp .LBB0_8
; X64-HSW-NEXT: .LBB0_34:
; X64-HSW-NEXT: movl %eax, %ecx
; X64-HSW-NEXT: shll $5, %ecx
; X64-HSW-NEXT: .LBB0_8:
; X64-HSW-NEXT: subl %eax, %ecx
; X64-HSW-NEXT: movl %ecx, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
; X64-HSW-NEXT: .LBB0_35:
; X64-HSW-NEXT: shll $5, %eax
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; X64-HSW-NEXT: # kill: def $eax killed $eax killed $rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
%3 = icmp eq i32 %1, 0
%4 = icmp sgt i32 %1, 1
%5 = or i1 %3, %4
%6 = select i1 %5, i32 1, i32 %1
switch i32 %0, label %69 [
i32 1, label %70
i32 2, label %7
i32 3, label %9
i32 4, label %11
i32 5, label %13
i32 6, label %15
i32 7, label %17
i32 8, label %19
i32 9, label %21
i32 10, label %23
i32 11, label %25
i32 12, label %27
i32 13, label %29
i32 14, label %31
i32 15, label %33
i32 16, label %35
i32 17, label %37
i32 18, label %39
i32 19, label %41
i32 20, label %43
i32 21, label %45
i32 22, label %47
i32 23, label %49
i32 24, label %51
i32 25, label %53
i32 26, label %55
i32 27, label %57
i32 28, label %59
i32 29, label %61
i32 30, label %63
i32 31, label %65
i32 32, label %67
]
; <label>:7: ; preds = %2
%8 = shl nsw i32 %6, 1
br label %70
; <label>:9: ; preds = %2
%10 = mul nsw i32 %6, 3
br label %70
; <label>:11: ; preds = %2
%12 = shl nsw i32 %6, 2
br label %70
; <label>:13: ; preds = %2
%14 = mul nsw i32 %6, 5
br label %70
; <label>:15: ; preds = %2
%16 = mul nsw i32 %6, 6
br label %70
; <label>:17: ; preds = %2
%18 = mul nsw i32 %6, 7
br label %70
; <label>:19: ; preds = %2
%20 = shl nsw i32 %6, 3
br label %70
; <label>:21: ; preds = %2
%22 = mul nsw i32 %6, 9
br label %70
; <label>:23: ; preds = %2
%24 = mul nsw i32 %6, 10
br label %70
; <label>:25: ; preds = %2
%26 = mul nsw i32 %6, 11
br label %70
; <label>:27: ; preds = %2
%28 = mul nsw i32 %6, 12
br label %70
; <label>:29: ; preds = %2
%30 = mul nsw i32 %6, 13
br label %70
; <label>:31: ; preds = %2
%32 = mul nsw i32 %6, 14
br label %70
; <label>:33: ; preds = %2
%34 = mul nsw i32 %6, 15
br label %70
; <label>:35: ; preds = %2
%36 = shl nsw i32 %6, 4
br label %70
; <label>:37: ; preds = %2
%38 = mul nsw i32 %6, 17
br label %70
; <label>:39: ; preds = %2
%40 = mul nsw i32 %6, 18
br label %70
; <label>:41: ; preds = %2
%42 = mul nsw i32 %6, 19
br label %70
; <label>:43: ; preds = %2
%44 = mul nsw i32 %6, 20
br label %70
; <label>:45: ; preds = %2
%46 = mul nsw i32 %6, 21
br label %70
; <label>:47: ; preds = %2
%48 = mul nsw i32 %6, 22
br label %70
; <label>:49: ; preds = %2
%50 = mul nsw i32 %6, 23
br label %70
; <label>:51: ; preds = %2
%52 = mul nsw i32 %6, 24
br label %70
; <label>:53: ; preds = %2
%54 = mul nsw i32 %6, 25
br label %70
; <label>:55: ; preds = %2
%56 = mul nsw i32 %6, 26
br label %70
; <label>:57: ; preds = %2
%58 = mul nsw i32 %6, 27
br label %70
; <label>:59: ; preds = %2
%60 = mul nsw i32 %6, 28
br label %70
; <label>:61: ; preds = %2
%62 = mul nsw i32 %6, 29
br label %70
; <label>:63: ; preds = %2
%64 = mul nsw i32 %6, 30
br label %70
; <label>:65: ; preds = %2
%66 = mul nsw i32 %6, 31
br label %70
; <label>:67: ; preds = %2
%68 = shl nsw i32 %6, 5
br label %70
; <label>:69: ; preds = %2
br label %70
; <label>:70: ; preds = %2, %69, %67, %65, %63, %61, %59, %57, %55, %53, %51, %49, %47, %45, %43, %41, %39, %37, %35, %33, %31, %29, %27, %25, %23, %21, %19, %17, %15, %13, %11, %9, %7
%71 = phi i32 [ %8, %7 ], [ %10, %9 ], [ %12, %11 ], [ %14, %13 ], [ %16, %15 ], [ %18, %17 ], [ %20, %19 ], [ %22, %21 ], [ %24, %23 ], [ %26, %25 ], [ %28, %27 ], [ %30, %29 ], [ %32, %31 ], [ %34, %33 ], [ %36, %35 ], [ %38, %37 ], [ %40, %39 ], [ %42, %41 ], [ %44, %43 ], [ %46, %45 ], [ %48, %47 ], [ %50, %49 ], [ %52, %51 ], [ %54, %53 ], [ %56, %55 ], [ %58, %57 ], [ %60, %59 ], [ %62, %61 ], [ %64, %63 ], [ %66, %65 ], [ %68, %67 ], [ 0, %69 ], [ %6, %2 ]
ret i32 %71
}
; Function Attrs: norecurse nounwind readnone uwtable
define i32 @foo() local_unnamed_addr #0 {
; X86-LABEL: foo:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X86: # %bb.0:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: pushl %ebx
; X86-NEXT: .cfi_def_cfa_offset 8
; X86-NEXT: pushl %edi
; X86-NEXT: .cfi_def_cfa_offset 12
; X86-NEXT: pushl %esi
; X86-NEXT: .cfi_def_cfa_offset 16
; X86-NEXT: .cfi_offset %esi, -16
; X86-NEXT: .cfi_offset %edi, -12
; X86-NEXT: .cfi_offset %ebx, -8
; X86-NEXT: pushl $0
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $1
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %esi
; X86-NEXT: xorl $1, %esi
; X86-NEXT: pushl $1
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $2
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $2, %edi
; X86-NEXT: pushl $1
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $3
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $3, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $2
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $4
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $4, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $2
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $5
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $5, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $3
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $6
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $6, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $3
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $7
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $7, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $4
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $8
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $8, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $4
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $9
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $9, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $5
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $10
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $10, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $5
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $11
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $11, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $6
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $12
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $12, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $6
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $13
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $13, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $7
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $14
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $14, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $7
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $15
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $15, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $8
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $16
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $16, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $8
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $17
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $17, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $9
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $18
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $18, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $9
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $19
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $19, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $10
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $20
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $20, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $10
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $21
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $21, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $11
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $22
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $22, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $11
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $23
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $23, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $12
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $24
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $24, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $12
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $25
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $25, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $13
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $26
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $26, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $13
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $27
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $27, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $14
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $28
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $28, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $14
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $29
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $29, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: pushl $15
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $30
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %edi
; X86-NEXT: xorl $30, %edi
; X86-NEXT: orl %ebx, %edi
; X86-NEXT: pushl $15
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $31
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: xorl $31, %ebx
; X86-NEXT: orl %edi, %ebx
; X86-NEXT: orl %esi, %ebx
; X86-NEXT: pushl $16
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: pushl $32
; X86-NEXT: .cfi_adjust_cfa_offset 4
; X86-NEXT: calll mult
; X86-NEXT: addl $8, %esp
; X86-NEXT: .cfi_adjust_cfa_offset -8
; X86-NEXT: xorl $32, %eax
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X86-NEXT: xorl %ecx, %ecx
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: orl %ebx, %eax
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X86-NEXT: setne %cl
; X86-NEXT: negl %ecx
; X86-NEXT: movl %ecx, %eax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: popl %esi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 12
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: popl %edi
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: popl %ebx
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X86-NEXT: .cfi_def_cfa_offset 4
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X86-NEXT: retl
;
; X64-HSW-LABEL: foo:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; X64-HSW: # %bb.0:
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: pushq %rbp
; X64-HSW-NEXT: .cfi_def_cfa_offset 16
; X64-HSW-NEXT: pushq %r15
; X64-HSW-NEXT: .cfi_def_cfa_offset 24
; X64-HSW-NEXT: pushq %r14
; X64-HSW-NEXT: .cfi_def_cfa_offset 32
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X64-HSW-NEXT: pushq %rbx
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 40
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X64-HSW-NEXT: pushq %rax
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 48
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X64-HSW-NEXT: .cfi_offset %rbx, -40
[x86] revert r310208 to investigate test-suite failures (PR34105 / PR34097) llvm-svn: 310264
2017-08-07 23:47:48 +08:00
; X64-HSW-NEXT: .cfi_offset %r14, -32
; X64-HSW-NEXT: .cfi_offset %r15, -24
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: .cfi_offset %rbp, -16
; X64-HSW-NEXT: movl $1, %edi
; X64-HSW-NEXT: xorl %esi, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $1, %ebx
; X64-HSW-NEXT: movl $2, %edi
; X64-HSW-NEXT: movl $1, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $2, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $3, %edi
; X64-HSW-NEXT: movl $1, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r14d
; X64-HSW-NEXT: xorl $3, %r14d
; X64-HSW-NEXT: movl $4, %edi
; X64-HSW-NEXT: movl $2, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $4, %ebx
; X64-HSW-NEXT: orl %r14d, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $5, %edi
; X64-HSW-NEXT: movl $2, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r14d
; X64-HSW-NEXT: xorl $5, %r14d
; X64-HSW-NEXT: movl $6, %edi
; X64-HSW-NEXT: movl $3, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $6, %ebp
; X64-HSW-NEXT: orl %r14d, %ebp
; X64-HSW-NEXT: movl $7, %edi
; X64-HSW-NEXT: movl $3, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r14d
; X64-HSW-NEXT: xorl $7, %r14d
; X64-HSW-NEXT: orl %ebp, %r14d
; X64-HSW-NEXT: orl %ebx, %r14d
; X64-HSW-NEXT: movl $8, %edi
; X64-HSW-NEXT: movl $4, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $8, %ebx
; X64-HSW-NEXT: movl $9, %edi
; X64-HSW-NEXT: movl $4, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $9, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $10, %edi
; X64-HSW-NEXT: movl $5, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $10, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $11, %edi
; X64-HSW-NEXT: movl $5, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r15d
; X64-HSW-NEXT: xorl $11, %r15d
; X64-HSW-NEXT: orl %ebx, %r15d
; X64-HSW-NEXT: orl %r14d, %r15d
; X64-HSW-NEXT: movl $12, %edi
; X64-HSW-NEXT: movl $6, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $12, %ebx
; X64-HSW-NEXT: movl $13, %edi
; X64-HSW-NEXT: movl $6, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $13, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $14, %edi
; X64-HSW-NEXT: movl $7, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $14, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $15, %edi
; X64-HSW-NEXT: movl $7, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $15, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $16, %edi
; X64-HSW-NEXT: movl $8, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r14d
; X64-HSW-NEXT: xorl $16, %r14d
; X64-HSW-NEXT: orl %ebp, %r14d
; X64-HSW-NEXT: orl %r15d, %r14d
; X64-HSW-NEXT: movl $17, %edi
; X64-HSW-NEXT: movl $8, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $17, %ebp
; X64-HSW-NEXT: movl $18, %edi
; X64-HSW-NEXT: movl $9, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $18, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $19, %edi
; X64-HSW-NEXT: movl $9, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $19, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $20, %edi
; X64-HSW-NEXT: movl $10, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $20, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $21, %edi
; X64-HSW-NEXT: movl $10, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $21, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $22, %edi
; X64-HSW-NEXT: movl $11, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r15d
; X64-HSW-NEXT: xorl $22, %r15d
; X64-HSW-NEXT: orl %ebp, %r15d
; X64-HSW-NEXT: orl %r14d, %r15d
; X64-HSW-NEXT: movl $23, %edi
; X64-HSW-NEXT: movl $11, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $23, %ebp
; X64-HSW-NEXT: movl $24, %edi
; X64-HSW-NEXT: movl $12, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $24, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $25, %edi
; X64-HSW-NEXT: movl $12, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $25, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $26, %edi
; X64-HSW-NEXT: movl $13, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $26, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $27, %edi
; X64-HSW-NEXT: movl $13, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $27, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: movl $28, %edi
; X64-HSW-NEXT: movl $14, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $28, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $29, %edi
; X64-HSW-NEXT: movl $14, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebp
; X64-HSW-NEXT: xorl $29, %ebp
; X64-HSW-NEXT: orl %ebx, %ebp
; X64-HSW-NEXT: orl %r15d, %ebp
; X64-HSW-NEXT: movl $30, %edi
; X64-HSW-NEXT: movl $15, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %r14d
; X64-HSW-NEXT: xorl $30, %r14d
; X64-HSW-NEXT: movl $31, %edi
; X64-HSW-NEXT: movl $15, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: movl %eax, %ebx
; X64-HSW-NEXT: xorl $31, %ebx
; X64-HSW-NEXT: orl %r14d, %ebx
; X64-HSW-NEXT: orl %ebp, %ebx
; X64-HSW-NEXT: movl $32, %edi
; X64-HSW-NEXT: movl $16, %esi
; X64-HSW-NEXT: callq mult
; X64-HSW-NEXT: xorl $32, %eax
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X64-HSW-NEXT: xorl %ecx, %ecx
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: orl %ebx, %eax
[x86] use more shift or LEA for select-of-constants (2nd try) The previous rev (r310208) failed to account for overflow when subtracting the constants to see if they're suitable for shift/lea. This version add a check for that and more test were added in r310490. We can convert any select-of-constants to math ops: http://rise4fun.com/Alive/d7d For this patch, I'm enhancing an existing x86 transform that uses fake multiplies (they always become shl/lea) to avoid cmov or branching. The current code misses cases where we have a negative constant and a positive constant, so this is just trying to plug that hole. The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start with a select in IR, create a select DAG node, convert it into a sext, convert it back into a select, and then lower it to sext machine code. Some notes about the test diffs: 1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR. 2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a post-DAG problem though. 3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if that's a regression, but those would always be nearly equivalent. 4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs. 5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0. 6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again. 7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops. Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass. Differential Revision: https://reviews.llvm.org/D35340 llvm-svn: 310717
2017-08-11 23:44:14 +08:00
; X64-HSW-NEXT: setne %cl
; X64-HSW-NEXT: negl %ecx
; X64-HSW-NEXT: movl %ecx, %eax
; X64-HSW-NEXT: addq $8, %rsp
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 40
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: popq %rbx
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 32
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: popq %r14
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 24
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: popq %r15
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 16
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: popq %rbp
Correct dwarf unwind information in function epilogue This patch aims to provide correct dwarf unwind information in function epilogue for X86. It consists of two parts. The first part inserts CFI instructions that set appropriate cfa offset and cfa register in emitEpilogue() in X86FrameLowering. This part is X86 specific. The second part is platform independent and ensures that: * CFI instructions do not affect code generation (they are not counted as instructions when tail duplicating or tail merging) * Unwind information remains correct when a function is modified by different passes. This is done in a late pass by analyzing information about cfa offset and cfa register in BBs and inserting additional CFI directives where necessary. Added CFIInstrInserter pass: * analyzes each basic block to determine cfa offset and register are valid at its entry and exit * verifies that outgoing cfa offset and register of predecessor blocks match incoming values of their successors * inserts additional CFI directives at basic block beginning to correct the rule for calculating CFA Having CFI instructions in function epilogue can cause incorrect CFA calculation rule for some basic blocks. This can happen if, due to basic block reordering, or the existence of multiple epilogue blocks, some of the blocks have wrong cfa offset and register values set by the epilogue block above them. CFIInstrInserter is currently run only on X86, but can be used by any target that implements support for adding CFI instructions in epilogue. Patch by Violeta Vukobrat. Differential Revision: https://reviews.llvm.org/D42848 llvm-svn: 330706
2018-04-24 18:32:08 +08:00
; X64-HSW-NEXT: .cfi_def_cfa_offset 8
This patch closes PR28513: an optimization of multiplication by different constants. The initial patch was rejected: I fixed the issue and re-apply it. llvm-svn: 304972
2017-06-08 18:20:13 +08:00
; X64-HSW-NEXT: retq
%1 = tail call i32 @mult(i32 1, i32 0)
%2 = icmp ne i32 %1, 1
%3 = tail call i32 @mult(i32 2, i32 1)
%4 = icmp ne i32 %3, 2
%5 = or i1 %2, %4
%6 = tail call i32 @mult(i32 3, i32 1)
%7 = icmp ne i32 %6, 3
%8 = or i1 %5, %7
%9 = tail call i32 @mult(i32 4, i32 2)
%10 = icmp ne i32 %9, 4
%11 = or i1 %8, %10
%12 = tail call i32 @mult(i32 5, i32 2)
%13 = icmp ne i32 %12, 5
%14 = or i1 %11, %13
%15 = tail call i32 @mult(i32 6, i32 3)
%16 = icmp ne i32 %15, 6
%17 = or i1 %14, %16
%18 = tail call i32 @mult(i32 7, i32 3)
%19 = icmp ne i32 %18, 7
%20 = or i1 %17, %19
%21 = tail call i32 @mult(i32 8, i32 4)
%22 = icmp ne i32 %21, 8
%23 = or i1 %20, %22
%24 = tail call i32 @mult(i32 9, i32 4)
%25 = icmp ne i32 %24, 9
%26 = or i1 %23, %25
%27 = tail call i32 @mult(i32 10, i32 5)
%28 = icmp ne i32 %27, 10
%29 = or i1 %26, %28
%30 = tail call i32 @mult(i32 11, i32 5)
%31 = icmp ne i32 %30, 11
%32 = or i1 %29, %31
%33 = tail call i32 @mult(i32 12, i32 6)
%34 = icmp ne i32 %33, 12
%35 = or i1 %32, %34
%36 = tail call i32 @mult(i32 13, i32 6)
%37 = icmp ne i32 %36, 13
%38 = or i1 %35, %37
%39 = tail call i32 @mult(i32 14, i32 7)
%40 = icmp ne i32 %39, 14
%41 = or i1 %38, %40
%42 = tail call i32 @mult(i32 15, i32 7)
%43 = icmp ne i32 %42, 15
%44 = or i1 %41, %43
%45 = tail call i32 @mult(i32 16, i32 8)
%46 = icmp ne i32 %45, 16
%47 = or i1 %44, %46
%48 = tail call i32 @mult(i32 17, i32 8)
%49 = icmp ne i32 %48, 17
%50 = or i1 %47, %49
%51 = tail call i32 @mult(i32 18, i32 9)
%52 = icmp ne i32 %51, 18
%53 = or i1 %50, %52
%54 = tail call i32 @mult(i32 19, i32 9)
%55 = icmp ne i32 %54, 19
%56 = or i1 %53, %55
%57 = tail call i32 @mult(i32 20, i32 10)
%58 = icmp ne i32 %57, 20
%59 = or i1 %56, %58
%60 = tail call i32 @mult(i32 21, i32 10)
%61 = icmp ne i32 %60, 21
%62 = or i1 %59, %61
%63 = tail call i32 @mult(i32 22, i32 11)
%64 = icmp ne i32 %63, 22
%65 = or i1 %62, %64
%66 = tail call i32 @mult(i32 23, i32 11)
%67 = icmp ne i32 %66, 23
%68 = or i1 %65, %67
%69 = tail call i32 @mult(i32 24, i32 12)
%70 = icmp ne i32 %69, 24
%71 = or i1 %68, %70
%72 = tail call i32 @mult(i32 25, i32 12)
%73 = icmp ne i32 %72, 25
%74 = or i1 %71, %73
%75 = tail call i32 @mult(i32 26, i32 13)
%76 = icmp ne i32 %75, 26
%77 = or i1 %74, %76
%78 = tail call i32 @mult(i32 27, i32 13)
%79 = icmp ne i32 %78, 27
%80 = or i1 %77, %79
%81 = tail call i32 @mult(i32 28, i32 14)
%82 = icmp ne i32 %81, 28
%83 = or i1 %80, %82
%84 = tail call i32 @mult(i32 29, i32 14)
%85 = icmp ne i32 %84, 29
%86 = or i1 %83, %85
%87 = tail call i32 @mult(i32 30, i32 15)
%88 = icmp ne i32 %87, 30
%89 = or i1 %86, %88
%90 = tail call i32 @mult(i32 31, i32 15)
%91 = icmp ne i32 %90, 31
%92 = or i1 %89, %91
%93 = tail call i32 @mult(i32 32, i32 16)
%94 = icmp ne i32 %93, 32
%95 = or i1 %92, %94
%96 = sext i1 %95 to i32
ret i32 %96
}
attributes #0 = { norecurse nounwind readnone uwtable }