46e5052b8e
INC/DEC is really a special case of a more generic issue. We should also turn leas into add reg/reg or add reg/imm regardless of the slow lea flags. This also supports LEA64_32 which has 64 bit input registers and 32 bit output registers. So we need to convert the 64 bit inputs to their 32 bit equivalents to check if they are equal to base reg. One thing to note, the original code preserved the kill flags by adding operands to the new instruction instead of using addReg. But I think tied operands aren't supposed to have the kill flag set. I dropped the kill flags, but I could probably try to preserve it in the add reg/reg case if we think its important. Not sure which operand its supposed to go on for the LEA64_32r instruction due to the super reg implicit uses. Though I'm also not sure those are needed since they were probably just created by an INSERT_SUBREG from a 32-bit input. Differential Revision: https://reviews.llvm.org/D61472 llvm-svn: 361691
577 lines
20 KiB
LLVM
577 lines
20 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -O3 -mtriple=x86_64-unknown-unknown -mcpu=core2 | FileCheck %s -check-prefix=X64
|
|
; RUN: llc < %s -O3 -mtriple=i686-unknown-unknown -mcpu=core2 | FileCheck %s -check-prefix=X32
|
|
|
|
; @simple is the most basic chain of address induction variables. Chaining
|
|
; saves at least one register and avoids complex addressing and setup
|
|
; code.
|
|
;
|
|
; %x * 4
|
|
; no other address computation in the preheader
|
|
; no complex address modes
|
|
;
|
|
; no expensive address computation in the preheader
|
|
; no complex address modes
|
|
|
|
define i32 @simple(i32* %a, i32* %b, i32 %x) nounwind {
|
|
; X64-LABEL: simple:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: movslq %edx, %rcx
|
|
; X64-NEXT: shlq $2, %rcx
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB0_1: # %loop
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: addl (%rdi), %eax
|
|
; X64-NEXT: leaq (%rdi,%rcx), %r8
|
|
; X64-NEXT: addl (%rdi,%rcx), %eax
|
|
; X64-NEXT: leaq (%r8,%rcx), %rdx
|
|
; X64-NEXT: addl (%rcx,%r8), %eax
|
|
; X64-NEXT: addl (%rcx,%rdx), %eax
|
|
; X64-NEXT: addq %rcx, %rdx
|
|
; X64-NEXT: addq %rcx, %rdx
|
|
; X64-NEXT: movq %rdx, %rdi
|
|
; X64-NEXT: cmpq %rsi, %rdx
|
|
; X64-NEXT: jne .LBB0_1
|
|
; X64-NEXT: # %bb.2: # %exit
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: simple:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %ebx
|
|
; X32-NEXT: pushl %edi
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: shll $2, %edx
|
|
; X32-NEXT: xorl %eax, %eax
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB0_1: # %loop
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: addl (%esi), %eax
|
|
; X32-NEXT: leal (%esi,%edx), %edi
|
|
; X32-NEXT: addl (%esi,%edx), %eax
|
|
; X32-NEXT: leal (%edi,%edx), %ebx
|
|
; X32-NEXT: addl (%edx,%edi), %eax
|
|
; X32-NEXT: addl (%edx,%ebx), %eax
|
|
; X32-NEXT: addl %edx, %ebx
|
|
; X32-NEXT: addl %edx, %ebx
|
|
; X32-NEXT: movl %ebx, %esi
|
|
; X32-NEXT: cmpl %ecx, %ebx
|
|
; X32-NEXT: jne .LBB0_1
|
|
; X32-NEXT: # %bb.2: # %exit
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: popl %edi
|
|
; X32-NEXT: popl %ebx
|
|
; X32-NEXT: retl
|
|
entry:
|
|
br label %loop
|
|
loop:
|
|
%iv = phi i32* [ %a, %entry ], [ %iv4, %loop ]
|
|
%s = phi i32 [ 0, %entry ], [ %s4, %loop ]
|
|
%v = load i32, i32* %iv
|
|
%iv1 = getelementptr inbounds i32, i32* %iv, i32 %x
|
|
%v1 = load i32, i32* %iv1
|
|
%iv2 = getelementptr inbounds i32, i32* %iv1, i32 %x
|
|
%v2 = load i32, i32* %iv2
|
|
%iv3 = getelementptr inbounds i32, i32* %iv2, i32 %x
|
|
%v3 = load i32, i32* %iv3
|
|
%s1 = add i32 %s, %v
|
|
%s2 = add i32 %s1, %v1
|
|
%s3 = add i32 %s2, %v2
|
|
%s4 = add i32 %s3, %v3
|
|
%iv4 = getelementptr inbounds i32, i32* %iv3, i32 %x
|
|
%cmp = icmp eq i32* %iv4, %b
|
|
br i1 %cmp, label %exit, label %loop
|
|
exit:
|
|
ret i32 %s4
|
|
}
|
|
|
|
; @user is not currently chained because the IV is live across memory ops.
|
|
;
|
|
; expensive address computation in the preheader
|
|
; complex address modes
|
|
define i32 @user(i32* %a, i32* %b, i32 %x) nounwind {
|
|
; X64-LABEL: user:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: movslq %edx, %rcx
|
|
; X64-NEXT: movq %rcx, %rdx
|
|
; X64-NEXT: shlq $4, %rdx
|
|
; X64-NEXT: leaq (,%rcx,4), %rax
|
|
; X64-NEXT: leaq (%rax,%rax,2), %r8
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB1_1: # %loop
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: addl (%rdi), %eax
|
|
; X64-NEXT: addl (%rdi,%rcx,4), %eax
|
|
; X64-NEXT: addl (%rdi,%rcx,8), %eax
|
|
; X64-NEXT: addl (%rdi,%r8), %eax
|
|
; X64-NEXT: movl %eax, (%rdi)
|
|
; X64-NEXT: addq %rdx, %rdi
|
|
; X64-NEXT: cmpq %rdi, %rsi
|
|
; X64-NEXT: jne .LBB1_1
|
|
; X64-NEXT: # %bb.2: # %exit
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: user:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %ebx
|
|
; X32-NEXT: pushl %edi
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
; X32-NEXT: movl %ecx, %edi
|
|
; X32-NEXT: shll $4, %edi
|
|
; X32-NEXT: leal (,%ecx,4), %eax
|
|
; X32-NEXT: leal (%eax,%eax,2), %ebx
|
|
; X32-NEXT: xorl %eax, %eax
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB1_1: # %loop
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: addl (%esi), %eax
|
|
; X32-NEXT: addl (%esi,%ecx,4), %eax
|
|
; X32-NEXT: addl (%esi,%ecx,8), %eax
|
|
; X32-NEXT: addl (%esi,%ebx), %eax
|
|
; X32-NEXT: movl %eax, (%esi)
|
|
; X32-NEXT: addl %edi, %esi
|
|
; X32-NEXT: cmpl %esi, %edx
|
|
; X32-NEXT: jne .LBB1_1
|
|
; X32-NEXT: # %bb.2: # %exit
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: popl %edi
|
|
; X32-NEXT: popl %ebx
|
|
; X32-NEXT: retl
|
|
entry:
|
|
br label %loop
|
|
loop:
|
|
%iv = phi i32* [ %a, %entry ], [ %iv4, %loop ]
|
|
%s = phi i32 [ 0, %entry ], [ %s4, %loop ]
|
|
%v = load i32, i32* %iv
|
|
%iv1 = getelementptr inbounds i32, i32* %iv, i32 %x
|
|
%v1 = load i32, i32* %iv1
|
|
%iv2 = getelementptr inbounds i32, i32* %iv1, i32 %x
|
|
%v2 = load i32, i32* %iv2
|
|
%iv3 = getelementptr inbounds i32, i32* %iv2, i32 %x
|
|
%v3 = load i32, i32* %iv3
|
|
%s1 = add i32 %s, %v
|
|
%s2 = add i32 %s1, %v1
|
|
%s3 = add i32 %s2, %v2
|
|
%s4 = add i32 %s3, %v3
|
|
%iv4 = getelementptr inbounds i32, i32* %iv3, i32 %x
|
|
store i32 %s4, i32* %iv
|
|
%cmp = icmp eq i32* %iv4, %b
|
|
br i1 %cmp, label %exit, label %loop
|
|
exit:
|
|
ret i32 %s4
|
|
}
|
|
|
|
; @extrastride is a slightly more interesting case of a single
|
|
; complete chain with multiple strides. The test case IR is what LSR
|
|
; used to do, and exactly what we don't want to do. LSR's new IV
|
|
; chaining feature should now undo the damage.
|
|
;
|
|
; We currently don't handle this on X64 because the sexts cause
|
|
; strange increment expressions like this:
|
|
; IV + ((sext i32 (2 * %s) to i64) + (-1 * (sext i32 %s to i64)))
|
|
;
|
|
; For x32, no spills in the preheader, no complex address modes, no reloads.
|
|
|
|
define void @extrastride(i8* nocapture %main, i32 %main_stride, i32* nocapture %res, i32 %x, i32 %y, i32 %z) nounwind {
|
|
; X64-LABEL: extrastride:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: pushq %rbp
|
|
; X64-NEXT: pushq %r14
|
|
; X64-NEXT: pushq %rbx
|
|
; X64-NEXT: # kill: def $ecx killed $ecx def $rcx
|
|
; X64-NEXT: # kill: def $esi killed $esi def $rsi
|
|
; X64-NEXT: testl %r9d, %r9d
|
|
; X64-NEXT: je .LBB2_3
|
|
; X64-NEXT: # %bb.1: # %for.body.lr.ph
|
|
; X64-NEXT: leal (%rsi,%rsi), %r14d
|
|
; X64-NEXT: leal (%rsi,%rsi,2), %ebx
|
|
; X64-NEXT: addl %esi, %ecx
|
|
; X64-NEXT: leal (,%rsi,4), %eax
|
|
; X64-NEXT: leal (%rcx,%rsi,4), %ebp
|
|
; X64-NEXT: movslq %eax, %r10
|
|
; X64-NEXT: movslq %ebx, %r11
|
|
; X64-NEXT: movslq %r14d, %rbx
|
|
; X64-NEXT: movslq %esi, %rsi
|
|
; X64-NEXT: movslq %r8d, %rcx
|
|
; X64-NEXT: shlq $2, %rcx
|
|
; X64-NEXT: movslq %ebp, %rax
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB2_2: # %for.body
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: movl (%rdi,%rsi), %ebp
|
|
; X64-NEXT: addl (%rdi), %ebp
|
|
; X64-NEXT: addl (%rdi,%rbx), %ebp
|
|
; X64-NEXT: addl (%rdi,%r11), %ebp
|
|
; X64-NEXT: addl (%rdi,%r10), %ebp
|
|
; X64-NEXT: movl %ebp, (%rdx)
|
|
; X64-NEXT: addq %rax, %rdi
|
|
; X64-NEXT: addq %rcx, %rdx
|
|
; X64-NEXT: decl %r9d
|
|
; X64-NEXT: jne .LBB2_2
|
|
; X64-NEXT: .LBB2_3: # %for.end
|
|
; X64-NEXT: popq %rbx
|
|
; X64-NEXT: popq %r14
|
|
; X64-NEXT: popq %rbp
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: extrastride:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %ebp
|
|
; X32-NEXT: pushl %ebx
|
|
; X32-NEXT: pushl %edi
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X32-NEXT: testl %eax, %eax
|
|
; X32-NEXT: je .LBB2_3
|
|
; X32-NEXT: # %bb.1: # %for.body.lr.ph
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ebx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edi
|
|
; X32-NEXT: addl %esi, %edi
|
|
; X32-NEXT: shll $2, %ecx
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB2_2: # %for.body
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: movl (%ebx,%esi), %ebp
|
|
; X32-NEXT: addl (%ebx), %ebp
|
|
; X32-NEXT: addl %esi, %ebx
|
|
; X32-NEXT: addl (%esi,%ebx), %ebp
|
|
; X32-NEXT: addl %esi, %ebx
|
|
; X32-NEXT: addl (%esi,%ebx), %ebp
|
|
; X32-NEXT: addl %esi, %ebx
|
|
; X32-NEXT: addl (%esi,%ebx), %ebp
|
|
; X32-NEXT: movl %ebp, (%edx)
|
|
; X32-NEXT: addl %esi, %ebx
|
|
; X32-NEXT: addl %edi, %ebx
|
|
; X32-NEXT: addl %ecx, %edx
|
|
; X32-NEXT: decl %eax
|
|
; X32-NEXT: jne .LBB2_2
|
|
; X32-NEXT: .LBB2_3: # %for.end
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: popl %edi
|
|
; X32-NEXT: popl %ebx
|
|
; X32-NEXT: popl %ebp
|
|
; X32-NEXT: retl
|
|
entry:
|
|
%cmp8 = icmp eq i32 %z, 0
|
|
br i1 %cmp8, label %for.end, label %for.body.lr.ph
|
|
|
|
for.body.lr.ph: ; preds = %entry
|
|
%add.ptr.sum = shl i32 %main_stride, 1 ; s*2
|
|
%add.ptr1.sum = add i32 %add.ptr.sum, %main_stride ; s*3
|
|
%add.ptr2.sum = add i32 %x, %main_stride ; s + x
|
|
%add.ptr4.sum = shl i32 %main_stride, 2 ; s*4
|
|
%add.ptr3.sum = add i32 %add.ptr2.sum, %add.ptr4.sum ; total IV stride = s*5+x
|
|
br label %for.body
|
|
|
|
for.body: ; preds = %for.body.lr.ph, %for.body
|
|
%main.addr.011 = phi i8* [ %main, %for.body.lr.ph ], [ %add.ptr6, %for.body ]
|
|
%i.010 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
|
|
%res.addr.09 = phi i32* [ %res, %for.body.lr.ph ], [ %add.ptr7, %for.body ]
|
|
%0 = bitcast i8* %main.addr.011 to i32*
|
|
%1 = load i32, i32* %0, align 4
|
|
%add.ptr = getelementptr inbounds i8, i8* %main.addr.011, i32 %main_stride
|
|
%2 = bitcast i8* %add.ptr to i32*
|
|
%3 = load i32, i32* %2, align 4
|
|
%add.ptr1 = getelementptr inbounds i8, i8* %main.addr.011, i32 %add.ptr.sum
|
|
%4 = bitcast i8* %add.ptr1 to i32*
|
|
%5 = load i32, i32* %4, align 4
|
|
%add.ptr2 = getelementptr inbounds i8, i8* %main.addr.011, i32 %add.ptr1.sum
|
|
%6 = bitcast i8* %add.ptr2 to i32*
|
|
%7 = load i32, i32* %6, align 4
|
|
%add.ptr3 = getelementptr inbounds i8, i8* %main.addr.011, i32 %add.ptr4.sum
|
|
%8 = bitcast i8* %add.ptr3 to i32*
|
|
%9 = load i32, i32* %8, align 4
|
|
%add = add i32 %3, %1
|
|
%add4 = add i32 %add, %5
|
|
%add5 = add i32 %add4, %7
|
|
%add6 = add i32 %add5, %9
|
|
store i32 %add6, i32* %res.addr.09, align 4
|
|
%add.ptr6 = getelementptr inbounds i8, i8* %main.addr.011, i32 %add.ptr3.sum
|
|
%add.ptr7 = getelementptr inbounds i32, i32* %res.addr.09, i32 %y
|
|
%inc = add i32 %i.010, 1
|
|
%cmp = icmp eq i32 %inc, %z
|
|
br i1 %cmp, label %for.end, label %for.body
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
ret void
|
|
}
|
|
|
|
; @foldedidx is an unrolled variant of this loop:
|
|
; for (unsigned long i = 0; i < len; i += s) {
|
|
; c[i] = a[i] + b[i];
|
|
; }
|
|
; where 's' can be folded into the addressing mode.
|
|
; Consequently, we should *not* form any chains.
|
|
|
|
define void @foldedidx(i8* nocapture %a, i8* nocapture %b, i8* nocapture %c) nounwind ssp {
|
|
; X64-LABEL: foldedidx:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: movl $3, %eax
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB3_1: # %for.body
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: movzbl -3(%rdi,%rax), %r8d
|
|
; X64-NEXT: movzbl -3(%rsi,%rax), %ecx
|
|
; X64-NEXT: addl %r8d, %ecx
|
|
; X64-NEXT: movb %cl, -3(%rdx,%rax)
|
|
; X64-NEXT: movzbl -2(%rdi,%rax), %r8d
|
|
; X64-NEXT: movzbl -2(%rsi,%rax), %ecx
|
|
; X64-NEXT: addl %r8d, %ecx
|
|
; X64-NEXT: movb %cl, -2(%rdx,%rax)
|
|
; X64-NEXT: movzbl -1(%rdi,%rax), %r8d
|
|
; X64-NEXT: movzbl -1(%rsi,%rax), %ecx
|
|
; X64-NEXT: addl %r8d, %ecx
|
|
; X64-NEXT: movb %cl, -1(%rdx,%rax)
|
|
; X64-NEXT: movzbl (%rdi,%rax), %r8d
|
|
; X64-NEXT: movzbl (%rsi,%rax), %ecx
|
|
; X64-NEXT: addl %r8d, %ecx
|
|
; X64-NEXT: movb %cl, (%rdx,%rax)
|
|
; X64-NEXT: addq $4, %rax
|
|
; X64-NEXT: cmpl $403, %eax # imm = 0x193
|
|
; X64-NEXT: jne .LBB3_1
|
|
; X64-NEXT: # %bb.2: # %for.end
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: foldedidx:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %ebx
|
|
; X32-NEXT: pushl %edi
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: movl $3, %eax
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB3_1: # %for.body
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: movzbl -3(%esi,%eax), %edi
|
|
; X32-NEXT: movzbl -3(%edx,%eax), %ebx
|
|
; X32-NEXT: addl %edi, %ebx
|
|
; X32-NEXT: movb %bl, -3(%ecx,%eax)
|
|
; X32-NEXT: movzbl -2(%esi,%eax), %edi
|
|
; X32-NEXT: movzbl -2(%edx,%eax), %ebx
|
|
; X32-NEXT: addl %edi, %ebx
|
|
; X32-NEXT: movb %bl, -2(%ecx,%eax)
|
|
; X32-NEXT: movzbl -1(%esi,%eax), %edi
|
|
; X32-NEXT: movzbl -1(%edx,%eax), %ebx
|
|
; X32-NEXT: addl %edi, %ebx
|
|
; X32-NEXT: movb %bl, -1(%ecx,%eax)
|
|
; X32-NEXT: movzbl (%esi,%eax), %edi
|
|
; X32-NEXT: movzbl (%edx,%eax), %ebx
|
|
; X32-NEXT: addl %edi, %ebx
|
|
; X32-NEXT: movb %bl, (%ecx,%eax)
|
|
; X32-NEXT: addl $4, %eax
|
|
; X32-NEXT: cmpl $403, %eax # imm = 0x193
|
|
; X32-NEXT: jne .LBB3_1
|
|
; X32-NEXT: # %bb.2: # %for.end
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: popl %edi
|
|
; X32-NEXT: popl %ebx
|
|
; X32-NEXT: retl
|
|
entry:
|
|
br label %for.body
|
|
|
|
for.body: ; preds = %for.body, %entry
|
|
%i.07 = phi i32 [ 0, %entry ], [ %inc.3, %for.body ]
|
|
%arrayidx = getelementptr inbounds i8, i8* %a, i32 %i.07
|
|
%0 = load i8, i8* %arrayidx, align 1
|
|
%conv5 = zext i8 %0 to i32
|
|
%arrayidx1 = getelementptr inbounds i8, i8* %b, i32 %i.07
|
|
%1 = load i8, i8* %arrayidx1, align 1
|
|
%conv26 = zext i8 %1 to i32
|
|
%add = add nsw i32 %conv26, %conv5
|
|
%conv3 = trunc i32 %add to i8
|
|
%arrayidx4 = getelementptr inbounds i8, i8* %c, i32 %i.07
|
|
store i8 %conv3, i8* %arrayidx4, align 1
|
|
%inc1 = or i32 %i.07, 1
|
|
%arrayidx.1 = getelementptr inbounds i8, i8* %a, i32 %inc1
|
|
%2 = load i8, i8* %arrayidx.1, align 1
|
|
%conv5.1 = zext i8 %2 to i32
|
|
%arrayidx1.1 = getelementptr inbounds i8, i8* %b, i32 %inc1
|
|
%3 = load i8, i8* %arrayidx1.1, align 1
|
|
%conv26.1 = zext i8 %3 to i32
|
|
%add.1 = add nsw i32 %conv26.1, %conv5.1
|
|
%conv3.1 = trunc i32 %add.1 to i8
|
|
%arrayidx4.1 = getelementptr inbounds i8, i8* %c, i32 %inc1
|
|
store i8 %conv3.1, i8* %arrayidx4.1, align 1
|
|
%inc.12 = or i32 %i.07, 2
|
|
%arrayidx.2 = getelementptr inbounds i8, i8* %a, i32 %inc.12
|
|
%4 = load i8, i8* %arrayidx.2, align 1
|
|
%conv5.2 = zext i8 %4 to i32
|
|
%arrayidx1.2 = getelementptr inbounds i8, i8* %b, i32 %inc.12
|
|
%5 = load i8, i8* %arrayidx1.2, align 1
|
|
%conv26.2 = zext i8 %5 to i32
|
|
%add.2 = add nsw i32 %conv26.2, %conv5.2
|
|
%conv3.2 = trunc i32 %add.2 to i8
|
|
%arrayidx4.2 = getelementptr inbounds i8, i8* %c, i32 %inc.12
|
|
store i8 %conv3.2, i8* %arrayidx4.2, align 1
|
|
%inc.23 = or i32 %i.07, 3
|
|
%arrayidx.3 = getelementptr inbounds i8, i8* %a, i32 %inc.23
|
|
%6 = load i8, i8* %arrayidx.3, align 1
|
|
%conv5.3 = zext i8 %6 to i32
|
|
%arrayidx1.3 = getelementptr inbounds i8, i8* %b, i32 %inc.23
|
|
%7 = load i8, i8* %arrayidx1.3, align 1
|
|
%conv26.3 = zext i8 %7 to i32
|
|
%add.3 = add nsw i32 %conv26.3, %conv5.3
|
|
%conv3.3 = trunc i32 %add.3 to i8
|
|
%arrayidx4.3 = getelementptr inbounds i8, i8* %c, i32 %inc.23
|
|
store i8 %conv3.3, i8* %arrayidx4.3, align 1
|
|
%inc.3 = add nsw i32 %i.07, 4
|
|
%exitcond.3 = icmp eq i32 %inc.3, 400
|
|
br i1 %exitcond.3, label %for.end, label %for.body
|
|
|
|
for.end: ; preds = %for.body
|
|
ret void
|
|
}
|
|
|
|
; @multioper tests instructions with multiple IV user operands. We
|
|
; should be able to chain them independent of each other.
|
|
|
|
define void @multioper(i32* %a, i32 %n) nounwind {
|
|
; X64-LABEL: multioper:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB4_1: # %for.body
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: movl %eax, (%rdi,%rax,4)
|
|
; X64-NEXT: leal 1(%rax), %ecx
|
|
; X64-NEXT: movl %ecx, 4(%rdi,%rax,4)
|
|
; X64-NEXT: leal 2(%rax), %ecx
|
|
; X64-NEXT: movl %ecx, 8(%rdi,%rax,4)
|
|
; X64-NEXT: leal 3(%rax), %ecx
|
|
; X64-NEXT: movl %ecx, 12(%rdi,%rax,4)
|
|
; X64-NEXT: addq $4, %rax
|
|
; X64-NEXT: cmpl %esi, %eax
|
|
; X64-NEXT: jl .LBB4_1
|
|
; X64-NEXT: # %bb.2: # %exit
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: multioper:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: xorl %eax, %eax
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB4_1: # %for.body
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: movl %eax, (%edx,%eax,4)
|
|
; X32-NEXT: leal 1(%eax), %esi
|
|
; X32-NEXT: movl %esi, 4(%edx,%eax,4)
|
|
; X32-NEXT: leal 2(%eax), %esi
|
|
; X32-NEXT: movl %esi, 8(%edx,%eax,4)
|
|
; X32-NEXT: leal 3(%eax), %esi
|
|
; X32-NEXT: movl %esi, 12(%edx,%eax,4)
|
|
; X32-NEXT: addl $4, %eax
|
|
; X32-NEXT: cmpl %ecx, %eax
|
|
; X32-NEXT: jl .LBB4_1
|
|
; X32-NEXT: # %bb.2: # %exit
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: retl
|
|
entry:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%p = phi i32* [ %p.next, %for.body ], [ %a, %entry ]
|
|
%i = phi i32 [ %inc4, %for.body ], [ 0, %entry ]
|
|
store i32 %i, i32* %p, align 4
|
|
%inc1 = or i32 %i, 1
|
|
%add.ptr.i1 = getelementptr inbounds i32, i32* %p, i32 1
|
|
store i32 %inc1, i32* %add.ptr.i1, align 4
|
|
%inc2 = add nsw i32 %i, 2
|
|
%add.ptr.i2 = getelementptr inbounds i32, i32* %p, i32 2
|
|
store i32 %inc2, i32* %add.ptr.i2, align 4
|
|
%inc3 = add nsw i32 %i, 3
|
|
%add.ptr.i3 = getelementptr inbounds i32, i32* %p, i32 3
|
|
store i32 %inc3, i32* %add.ptr.i3, align 4
|
|
%p.next = getelementptr inbounds i32, i32* %p, i32 4
|
|
%inc4 = add nsw i32 %i, 4
|
|
%cmp = icmp slt i32 %inc4, %n
|
|
br i1 %cmp, label %for.body, label %exit
|
|
|
|
exit:
|
|
ret void
|
|
}
|
|
|
|
; @testCmpZero has a ICmpZero LSR use that should not be hidden from
|
|
; LSR. Profitable chains should have more than one nonzero increment
|
|
; anyway.
|
|
|
|
define void @testCmpZero(i8* %src, i8* %dst, i32 %srcidx, i32 %dstidx, i32 %len) nounwind ssp {
|
|
; X64-LABEL: testCmpZero:
|
|
; X64: # %bb.0: # %entry
|
|
; X64-NEXT: movslq %edx, %rdx
|
|
; X64-NEXT: addq %rdx, %rdi
|
|
; X64-NEXT: movslq %ecx, %r9
|
|
; X64-NEXT: addq %rsi, %r9
|
|
; X64-NEXT: addl %edx, %r8d
|
|
; X64-NEXT: movslq %r8d, %rcx
|
|
; X64-NEXT: subq %rdx, %rcx
|
|
; X64-NEXT: xorl %edx, %edx
|
|
; X64-NEXT: .p2align 4, 0x90
|
|
; X64-NEXT: .LBB5_1: # %for.body82.us
|
|
; X64-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X64-NEXT: movzbl (%r9,%rdx,4), %eax
|
|
; X64-NEXT: movb %al, (%rdi,%rdx)
|
|
; X64-NEXT: incq %rdx
|
|
; X64-NEXT: cmpq %rdx, %rcx
|
|
; X64-NEXT: jne .LBB5_1
|
|
; X64-NEXT: # %bb.2: # %return
|
|
; X64-NEXT: retq
|
|
;
|
|
; X32-LABEL: testCmpZero:
|
|
; X32: # %bb.0: # %entry
|
|
; X32-NEXT: pushl %ebx
|
|
; X32-NEXT: pushl %esi
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: addl {{[0-9]+}}(%esp), %ecx
|
|
; X32-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: addl {{[0-9]+}}(%esp), %edx
|
|
; X32-NEXT: xorl %esi, %esi
|
|
; X32-NEXT: .p2align 4, 0x90
|
|
; X32-NEXT: .LBB5_1: # %for.body82.us
|
|
; X32-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; X32-NEXT: movzbl (%edx,%esi,4), %ebx
|
|
; X32-NEXT: movb %bl, (%ecx,%esi)
|
|
; X32-NEXT: incl %esi
|
|
; X32-NEXT: cmpl %esi, %eax
|
|
; X32-NEXT: jne .LBB5_1
|
|
; X32-NEXT: # %bb.2: # %return
|
|
; X32-NEXT: popl %esi
|
|
; X32-NEXT: popl %ebx
|
|
; X32-NEXT: retl
|
|
entry:
|
|
%dest0 = getelementptr inbounds i8, i8* %src, i32 %srcidx
|
|
%source0 = getelementptr inbounds i8, i8* %dst, i32 %dstidx
|
|
%add.ptr79.us.sum = add i32 %srcidx, %len
|
|
%lftr.limit = getelementptr i8, i8* %src, i32 %add.ptr79.us.sum
|
|
br label %for.body82.us
|
|
|
|
for.body82.us:
|
|
%dest = phi i8* [ %dest0, %entry ], [ %incdec.ptr91.us, %for.body82.us ]
|
|
%source = phi i8* [ %source0, %entry ], [ %add.ptr83.us, %for.body82.us ]
|
|
%0 = bitcast i8* %source to i32*
|
|
%1 = load i32, i32* %0, align 4
|
|
%trunc = trunc i32 %1 to i8
|
|
%add.ptr83.us = getelementptr inbounds i8, i8* %source, i32 4
|
|
%incdec.ptr91.us = getelementptr inbounds i8, i8* %dest, i32 1
|
|
store i8 %trunc, i8* %dest, align 1
|
|
%exitcond = icmp eq i8* %incdec.ptr91.us, %lftr.limit
|
|
br i1 %exitcond, label %return, label %for.body82.us
|
|
|
|
return:
|
|
ret void
|
|
}
|