mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 11:13:28 +01:00
ca607f6065
llvm-svn: 364642
302 lines
9.3 KiB
LLVM
302 lines
9.3 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=i686-unknown-linux-gnu < %s | FileCheck %s --check-prefixes=CHECK,X86
|
|
; RUN: llc -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s --check-prefixes=CHECK,X64
|
|
|
|
;------------------------------------------------------------------------------;
|
|
; Odd divisors
|
|
;------------------------------------------------------------------------------;
|
|
|
|
; This tests the BuildREMEqFold optimization with UREM, i32, odd divisor, SETEQ.
|
|
; The corresponding pseudocode is:
|
|
; Q <- [N * multInv(5, 2^32)] <=> [N * 0xCCCCCCCD] <=> [N * (-858993459)]
|
|
; res <- [Q <= (2^32 - 1) / 5] <=> [Q <= 858993459] <=> [Q < 858993460]
|
|
define i32 @test_urem_odd(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_odd:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-858993459, {{[0-9]+}}(%esp), %ecx # imm = 0xCCCCCCCD
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $858993460, %ecx # imm = 0x33333334
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_odd:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-858993459, %edi, %ecx # imm = 0xCCCCCCCD
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $858993460, %ecx # imm = 0x33333334
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 5
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
define i32 @test_urem_odd_25(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_odd_25:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-1030792151, {{[0-9]+}}(%esp), %ecx # imm = 0xC28F5C29
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $171798692, %ecx # imm = 0xA3D70A4
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_odd_25:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-1030792151, %edi, %ecx # imm = 0xC28F5C29
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $171798692, %ecx # imm = 0xA3D70A4
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 25
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; This is like test_urem_odd, except the divisor has bit 30 set.
|
|
define i32 @test_urem_odd_bit30(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_odd_bit30:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $1789569707, {{[0-9]+}}(%esp), %ecx # imm = 0x6AAAAAAB
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $4, %ecx
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_odd_bit30:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $1789569707, %edi, %ecx # imm = 0x6AAAAAAB
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $4, %ecx
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 1073741827
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; This is like test_urem_odd, except the divisor has bit 31 set.
|
|
define i32 @test_urem_odd_bit31(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_odd_bit31:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $715827883, {{[0-9]+}}(%esp), %ecx # imm = 0x2AAAAAAB
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $2, %ecx
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_odd_bit31:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $715827883, %edi, %ecx # imm = 0x2AAAAAAB
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $2, %ecx
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 2147483651
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
;------------------------------------------------------------------------------;
|
|
; Even divisors
|
|
;------------------------------------------------------------------------------;
|
|
|
|
; This tests the BuildREMEqFold optimization with UREM, i16, even divisor, SETNE.
|
|
; In this case, D <=> 14 <=> 7 * 2^1, so D0 = 7 and K = 1.
|
|
; The corresponding pseudocode is:
|
|
; Q <- [N * multInv(D0, 2^16)] <=> [N * multInv(7, 2^16)] <=> [N * 28087]
|
|
; Q <- [Q >>rot K] <=> [Q >>rot 1]
|
|
; res <- ![Q <= (2^16 - 1) / 7] <=> ![Q <= 9362] <=> [Q > 9362]
|
|
define i16 @test_urem_even(i16 %X) nounwind {
|
|
; X86-LABEL: test_urem_even:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $28087, {{[0-9]+}}(%esp), %eax # imm = 0x6DB7
|
|
; X86-NEXT: rorw %ax
|
|
; X86-NEXT: movzwl %ax, %ecx
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $4681, %ecx # imm = 0x1249
|
|
; X86-NEXT: seta %al
|
|
; X86-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_even:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $28087, %edi, %eax # imm = 0x6DB7
|
|
; X64-NEXT: rorw %ax
|
|
; X64-NEXT: movzwl %ax, %ecx
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $4681, %ecx # imm = 0x1249
|
|
; X64-NEXT: seta %al
|
|
; X64-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-NEXT: retq
|
|
%urem = urem i16 %X, 14
|
|
%cmp = icmp ne i16 %urem, 0
|
|
%ret = zext i1 %cmp to i16
|
|
ret i16 %ret
|
|
}
|
|
|
|
define i32 @test_urem_even_100(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_even_100:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-1030792151, {{[0-9]+}}(%esp), %ecx # imm = 0xC28F5C29
|
|
; X86-NEXT: rorl $2, %ecx
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $42949673, %ecx # imm = 0x28F5C29
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_even_100:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-1030792151, %edi, %ecx # imm = 0xC28F5C29
|
|
; X64-NEXT: rorl $2, %ecx
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $42949673, %ecx # imm = 0x28F5C29
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 100
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; This is like test_urem_even, except the divisor has bit 30 set.
|
|
define i32 @test_urem_even_bit30(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_even_bit30:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-51622203, {{[0-9]+}}(%esp), %ecx # imm = 0xFCEC4EC5
|
|
; X86-NEXT: rorl $3, %ecx
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $4, %ecx
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_even_bit30:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-51622203, %edi, %ecx # imm = 0xFCEC4EC5
|
|
; X64-NEXT: rorl $3, %ecx
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $4, %ecx
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 1073741928
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; This is like test_urem_odd, except the divisor has bit 31 set.
|
|
define i32 @test_urem_even_bit31(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_even_bit31:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-1157956869, {{[0-9]+}}(%esp), %ecx # imm = 0xBAFAFAFB
|
|
; X86-NEXT: rorl %ecx
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $2, %ecx
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_even_bit31:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-1157956869, %edi, %ecx # imm = 0xBAFAFAFB
|
|
; X64-NEXT: rorl %ecx
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $2, %ecx
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 2147483750
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
;------------------------------------------------------------------------------;
|
|
; Special case
|
|
;------------------------------------------------------------------------------;
|
|
|
|
; 'NE' predicate is fine too.
|
|
define i32 @test_urem_odd_setne(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_odd_setne:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: imull $-858993459, {{[0-9]+}}(%esp), %ecx # imm = 0xCCCCCCCD
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $858993459, %ecx # imm = 0x33333333
|
|
; X86-NEXT: seta %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_odd_setne:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: imull $-858993459, %edi, %ecx # imm = 0xCCCCCCCD
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $858993459, %ecx # imm = 0x33333333
|
|
; X64-NEXT: seta %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 5
|
|
%cmp = icmp ne i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
;------------------------------------------------------------------------------;
|
|
; Negative tests
|
|
;------------------------------------------------------------------------------;
|
|
|
|
; The fold is invalid if divisor is 1.
|
|
define i32 @test_urem_one(i32 %X) nounwind {
|
|
; CHECK-LABEL: test_urem_one:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl $1, %eax
|
|
; CHECK-NEXT: ret{{[l|q]}}
|
|
%urem = urem i32 %X, 1
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; We can lower remainder of division by all-ones much better elsewhere.
|
|
define i32 @test_urem_allones(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_allones:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: xorl %ecx, %ecx
|
|
; X86-NEXT: subl {{[0-9]+}}(%esp), %ecx
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: cmpl $2, %ecx
|
|
; X86-NEXT: setb %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_allones:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: negl %edi
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: cmpl $2, %edi
|
|
; X64-NEXT: setb %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 4294967295
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|
|
|
|
; We can lower remainder of division by powers of two much better elsewhere.
|
|
define i32 @test_urem_pow2(i32 %X) nounwind {
|
|
; X86-LABEL: test_urem_pow2:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: xorl %eax, %eax
|
|
; X86-NEXT: testb $15, {{[0-9]+}}(%esp)
|
|
; X86-NEXT: sete %al
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: test_urem_pow2:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: xorl %eax, %eax
|
|
; X64-NEXT: testb $15, %dil
|
|
; X64-NEXT: sete %al
|
|
; X64-NEXT: retq
|
|
%urem = urem i32 %X, 16
|
|
%cmp = icmp eq i32 %urem, 0
|
|
%ret = zext i1 %cmp to i32
|
|
ret i32 %ret
|
|
}
|