2019-04-17 12:52:47 +08:00
|
|
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
|
|
; RUN: opt < %s -instcombine -S | FileCheck %s
|
|
|
|
|
|
|
|
target datalayout = "e-p:64:64:64-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
|
|
|
|
|
|
|
|
define i32 @sub_constant(i32 %x) {
|
|
|
|
; CHECK-LABEL: @sub_constant(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add i32 [[X:%.*]], -42
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%r = sub i32 %x, 42
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
@g = global i32 0
|
|
|
|
|
|
|
|
define i32 @sub_constant_expression(i32 %x) {
|
|
|
|
; CHECK-LABEL: @sub_constant_expression(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = sub i32 [[X:%.*]], ptrtoint (i32* @g to i32)
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%r = sub i32 %x, ptrtoint (i32* @g to i32)
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @sub_constant_vec(<2 x i32> %x) {
|
|
|
|
; CHECK-LABEL: @sub_constant_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[X:%.*]], <i32 -42, i32 12>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%r = sub <2 x i32> %x, <i32 42, i32 -12>
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <3 x i33> @sub_constant_vec_weird_type(<3 x i33> %x) {
|
|
|
|
; CHECK-LABEL: @sub_constant_vec_weird_type(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <3 x i33> [[X:%.*]], <i33 42, i33 -42, i33 12>
|
|
|
|
; CHECK-NEXT: ret <3 x i33> [[R]]
|
|
|
|
;
|
|
|
|
%r = sub <3 x i33> %x, <i33 -42, i33 42, i33 -12>
|
|
|
|
ret <3 x i33> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <4 x i32> @sub_constant_expression_vec(<4 x i32> %x) {
|
|
|
|
; CHECK-LABEL: @sub_constant_expression_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = sub <4 x i32> [[X:%.*]], bitcast (i128 ptrtoint (i32* @g to i128) to <4 x i32>)
|
|
|
|
; CHECK-NEXT: ret <4 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%r = sub <4 x i32> %x, bitcast (i128 ptrtoint (i32* @g to i128) to <4 x i32>)
|
|
|
|
ret <4 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @neg_sub(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add i32 [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub i32 0, %x
|
|
|
|
%r = sub i32 %y, %neg
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @neg_nsw_sub(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add i32 [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw i32 0, %x
|
|
|
|
%r = sub i32 %y, %neg
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @neg_sub_nsw(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub_nsw(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add i32 [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub i32 0, %x
|
|
|
|
%r = sub nsw i32 %y, %neg
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @neg_nsw_sub_nsw(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub_nsw(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add nsw i32 [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw i32 0, %x
|
|
|
|
%r = sub nsw i32 %y, %neg
|
|
|
|
ret i32 %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_sub_vec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub <2 x i32> zeroinitializer, %x
|
|
|
|
%r = sub <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_nsw_sub_vec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw <2 x i32> zeroinitializer, %x
|
|
|
|
%r = sub <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_sub_nsw_vec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub_nsw_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub <2 x i32> zeroinitializer, %x
|
|
|
|
%r = sub nsw <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_nsw_sub_nsw_vec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub_nsw_vec(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add nsw <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw <2 x i32> zeroinitializer, %x
|
|
|
|
%r = sub nsw <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_sub_vec_undef(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub_vec_undef(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub <2 x i32> <i32 0, i32 undef>, %x
|
|
|
|
%r = sub <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_nsw_sub_vec_undef(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub_vec_undef(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw <2 x i32> <i32 undef, i32 0>, %x
|
|
|
|
%r = sub <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @neg_sub_nsw_vec_undef(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_sub_nsw_vec_undef(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub <2 x i32> <i32 undef, i32 0>, %x
|
|
|
|
%r = sub nsw <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
; This should not drop 'nsw'.
|
|
|
|
|
|
|
|
define <2 x i32> @neg_nsw_sub_nsw_vec_undef(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @neg_nsw_sub_nsw_vec_undef(
|
|
|
|
; CHECK-NEXT: [[R:%.*]] = add nsw <2 x i32> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[R]]
|
|
|
|
;
|
|
|
|
%neg = sub nsw <2 x i32> <i32 0, i32 undef>, %x
|
|
|
|
%r = sub nsw <2 x i32> %y, %neg
|
|
|
|
ret <2 x i32> %r
|
|
|
|
}
|
|
|
|
|
|
|
|
; (~X) - (~Y) --> Y - X
|
|
|
|
; Also, show that we can handle extra uses and vectors.
|
|
|
|
|
|
|
|
declare void @use8(i8)
|
|
|
|
|
|
|
|
define i8 @notnotsub(i8 %x, i8 %y) {
|
|
|
|
; CHECK-LABEL: @notnotsub(
|
|
|
|
; CHECK-NEXT: [[NX:%.*]] = xor i8 [[X:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[NY:%.*]] = xor i8 [[Y:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i8 [[Y]], [[X]]
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[NX]])
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[NY]])
|
|
|
|
; CHECK-NEXT: ret i8 [[SUB]]
|
|
|
|
;
|
|
|
|
%nx = xor i8 %x, -1
|
|
|
|
%ny = xor i8 %y, -1
|
|
|
|
%sub = sub i8 %nx, %ny
|
|
|
|
call void @use8(i8 %nx)
|
|
|
|
call void @use8(i8 %ny)
|
|
|
|
ret i8 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i8> @notnotsub_vec(<2 x i8> %x, <2 x i8> %y) {
|
|
|
|
; CHECK-LABEL: @notnotsub_vec(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub <2 x i8> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i8> [[SUB]]
|
|
|
|
;
|
|
|
|
%nx = xor <2 x i8> %x, <i8 -1, i8 -1>
|
|
|
|
%ny = xor <2 x i8> %y, <i8 -1, i8 -1>
|
|
|
|
%sub = sub <2 x i8> %nx, %ny
|
|
|
|
ret <2 x i8> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i8> @notnotsub_vec_undef_elts(<2 x i8> %x, <2 x i8> %y) {
|
|
|
|
; CHECK-LABEL: @notnotsub_vec_undef_elts(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub <2 x i8> [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i8> [[SUB]]
|
|
|
|
;
|
|
|
|
%nx = xor <2 x i8> %x, <i8 undef, i8 -1>
|
|
|
|
%ny = xor <2 x i8> %y, <i8 -1, i8 undef>
|
|
|
|
%sub = sub <2 x i8> %nx, %ny
|
|
|
|
ret <2 x i8> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test5(i32 %A, i32 %B, i32 %C) {
|
|
|
|
; CHECK-LABEL: @test5(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[D_NEG:%.*]] = sub i32 [[C:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[E:%.*]] = add i32 [[D_NEG]], [[A:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i32 [[E]]
|
|
|
|
;
|
|
|
|
%D = sub i32 %B, %C
|
|
|
|
%E = sub i32 %A, %D
|
|
|
|
ret i32 %E
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test6(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test6(
|
|
|
|
; CHECK-NEXT: [[B_NOT:%.*]] = xor i32 [[B:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = and i32 [[B_NOT]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[D]]
|
|
|
|
;
|
|
|
|
%C = and i32 %A, %B
|
|
|
|
%D = sub i32 %A, %C
|
|
|
|
ret i32 %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test6commuted(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test6commuted(
|
|
|
|
; CHECK-NEXT: [[B_NOT:%.*]] = xor i32 [[B:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = and i32 [[B_NOT]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[D]]
|
|
|
|
;
|
|
|
|
%C = and i32 %B, %A
|
|
|
|
%D = sub i32 %A, %C
|
|
|
|
ret i32 %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test7(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test7(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = xor i32 [[A:%.*]], -1
|
|
|
|
; CHECK-NEXT: ret i32 [[B]]
|
|
|
|
;
|
|
|
|
%B = sub i32 -1, %A
|
|
|
|
ret i32 %B
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test8(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test8(
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = shl i32 [[A:%.*]], 3
|
|
|
|
; CHECK-NEXT: ret i32 [[C]]
|
|
|
|
;
|
|
|
|
%B = mul i32 9, %A
|
|
|
|
%C = sub i32 %B, %A
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test9(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test9(
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = mul i32 [[A:%.*]], -2
|
|
|
|
; CHECK-NEXT: ret i32 [[C]]
|
|
|
|
;
|
|
|
|
%B = mul i32 3, %A
|
|
|
|
%C = sub i32 %A, %B
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i1 @test11(i8 %A, i8 %B) {
|
|
|
|
; CHECK-LABEL: @test11(
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = icmp ne i8 [[A:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret i1 [[D]]
|
|
|
|
;
|
|
|
|
%C = sub i8 %A, %B
|
|
|
|
%D = icmp ne i8 %C, 0
|
|
|
|
ret i1 %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i1> @test11vec(<2 x i8> %A, <2 x i8> %B) {
|
|
|
|
; CHECK-LABEL: @test11vec(
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = icmp ne <2 x i8> [[A:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i1> [[D]]
|
|
|
|
;
|
|
|
|
%C = sub <2 x i8> %A, %B
|
|
|
|
%D = icmp ne <2 x i8> %C, zeroinitializer
|
|
|
|
ret <2 x i1> %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test12(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test12(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[B_NEG:%.*]] = lshr i32 [[A:%.*]], 31
|
|
|
|
; CHECK-NEXT: ret i32 [[B_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = ashr i32 %A, 31
|
|
|
|
%C = sub i32 0, %B
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test13(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test13(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[B_NEG:%.*]] = ashr i32 [[A:%.*]], 31
|
|
|
|
; CHECK-NEXT: ret i32 [[B_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = lshr i32 %A, 31
|
|
|
|
%C = sub i32 0, %B
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test12vec(<2 x i32> %A) {
|
|
|
|
; CHECK-LABEL: @test12vec(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[B_NEG:%.*]] = lshr <2 x i32> [[A:%.*]], <i32 31, i32 31>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[B_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = ashr <2 x i32> %A, <i32 31, i32 31>
|
|
|
|
%C = sub <2 x i32> zeroinitializer, %B
|
|
|
|
ret <2 x i32> %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test13vec(<2 x i32> %A) {
|
|
|
|
; CHECK-LABEL: @test13vec(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[B_NEG:%.*]] = ashr <2 x i32> [[A:%.*]], <i32 31, i32 31>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[B_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = lshr <2 x i32> %A, <i32 31, i32 31>
|
|
|
|
%C = sub <2 x i32> zeroinitializer, %B
|
|
|
|
ret <2 x i32> %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test15(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test15(
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sub i32 0, [[A:%.*]]
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = srem i32 [[B:%.*]], [[C]]
|
|
|
|
; CHECK-NEXT: ret i32 [[D]]
|
|
|
|
;
|
|
|
|
%C = sub i32 0, %A
|
|
|
|
%D = srem i32 %B, %C
|
|
|
|
ret i32 %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test16(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test16(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[X_NEG:%.*]] = sdiv i32 [[A:%.*]], -1123
|
|
|
|
; CHECK-NEXT: ret i32 [[X_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%X = sdiv i32 %A, 1123
|
|
|
|
%Y = sub i32 0, %X
|
|
|
|
ret i32 %Y
|
|
|
|
}
|
|
|
|
|
|
|
|
; Can't fold subtract here because negation it might oveflow.
|
|
|
|
; PR3142
|
|
|
|
define i32 @test17(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test17(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = sub i32 0, [[A:%.*]]
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sdiv i32 [[B]], 1234
|
|
|
|
; CHECK-NEXT: ret i32 [[C]]
|
|
|
|
;
|
|
|
|
%B = sub i32 0, %A
|
|
|
|
%C = sdiv i32 %B, 1234
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test18(i64 %Y) {
|
|
|
|
; CHECK-LABEL: @test18(
|
|
|
|
; CHECK-NEXT: ret i64 0
|
|
|
|
;
|
|
|
|
%tmp.4 = shl i64 %Y, 2
|
|
|
|
%tmp.12 = shl i64 %Y, 2
|
|
|
|
%tmp.8 = sub i64 %tmp.4, %tmp.12
|
|
|
|
ret i64 %tmp.8
|
|
|
|
}
|
|
|
|
|
|
|
|
define i1 @test20(i32 %g, i32 %h) {
|
|
|
|
; CHECK-LABEL: @test20(
|
|
|
|
; CHECK-NEXT: [[TMP_4:%.*]] = icmp ne i32 [[H:%.*]], 0
|
|
|
|
; CHECK-NEXT: ret i1 [[TMP_4]]
|
|
|
|
;
|
|
|
|
%tmp.2 = sub i32 %g, %h
|
|
|
|
%tmp.4 = icmp ne i32 %tmp.2, %g
|
|
|
|
ret i1 %tmp.4
|
|
|
|
}
|
|
|
|
|
|
|
|
define i1 @test21(i32 %g, i32 %h) {
|
|
|
|
; CHECK-LABEL: @test21(
|
|
|
|
; CHECK-NEXT: [[TMP_4:%.*]] = icmp ne i32 [[H:%.*]], 0
|
|
|
|
; CHECK-NEXT: ret i1 [[TMP_4]]
|
|
|
|
;
|
|
|
|
%tmp.2 = sub i32 %g, %h
|
|
|
|
%tmp.4 = icmp ne i32 %tmp.2, %g
|
|
|
|
ret i1 %tmp.4
|
|
|
|
}
|
|
|
|
|
|
|
|
; PR2298
|
|
|
|
define zeroext i1 @test22(i32 %a, i32 %b) nounwind {
|
|
|
|
; CHECK-LABEL: @test22(
|
|
|
|
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[B:%.*]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret i1 [[TMP5]]
|
|
|
|
;
|
|
|
|
%tmp2 = sub i32 0, %a
|
|
|
|
%tmp4 = sub i32 0, %b
|
|
|
|
%tmp5 = icmp eq i32 %tmp2, %tmp4
|
|
|
|
ret i1 %tmp5
|
|
|
|
}
|
|
|
|
|
|
|
|
; rdar://7362831
|
|
|
|
define i32 @test23(i8* %P, i64 %A){
|
|
|
|
; CHECK-LABEL: @test23(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[A:%.*]] to i32
|
|
|
|
; CHECK-NEXT: ret i32 [[TMP1]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8* %P, i64 %A
|
|
|
|
%C = ptrtoint i8* %B to i64
|
|
|
|
%D = trunc i64 %C to i32
|
|
|
|
%E = ptrtoint i8* %P to i64
|
|
|
|
%F = trunc i64 %E to i32
|
|
|
|
%G = sub i32 %D, %F
|
|
|
|
ret i32 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i8 @test23_as1(i8 addrspace(1)* %P, i16 %A) {
|
|
|
|
; CHECK-LABEL: @test23_as1(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = trunc i16 [[A:%.*]] to i8
|
|
|
|
; CHECK-NEXT: ret i8 [[TMP1]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8 addrspace(1)* %P, i16 %A
|
|
|
|
%C = ptrtoint i8 addrspace(1)* %B to i16
|
|
|
|
%D = trunc i16 %C to i8
|
|
|
|
%E = ptrtoint i8 addrspace(1)* %P to i16
|
|
|
|
%F = trunc i16 %E to i8
|
|
|
|
%G = sub i8 %D, %F
|
|
|
|
ret i8 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test24(i8* %P, i64 %A){
|
|
|
|
; CHECK-LABEL: @test24(
|
|
|
|
; CHECK-NEXT: ret i64 [[A:%.*]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8* %P, i64 %A
|
|
|
|
%C = ptrtoint i8* %B to i64
|
|
|
|
%E = ptrtoint i8* %P to i64
|
|
|
|
%G = sub i64 %C, %E
|
|
|
|
ret i64 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i16 @test24_as1(i8 addrspace(1)* %P, i16 %A) {
|
|
|
|
; CHECK-LABEL: @test24_as1(
|
|
|
|
; CHECK-NEXT: ret i16 [[A:%.*]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8 addrspace(1)* %P, i16 %A
|
|
|
|
%C = ptrtoint i8 addrspace(1)* %B to i16
|
|
|
|
%E = ptrtoint i8 addrspace(1)* %P to i16
|
|
|
|
%G = sub i16 %C, %E
|
|
|
|
ret i16 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test24a(i8* %P, i64 %A){
|
|
|
|
; CHECK-LABEL: @test24a(
|
|
|
|
; CHECK-NEXT: [[DIFF_NEG:%.*]] = sub i64 0, [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[DIFF_NEG]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8* %P, i64 %A
|
|
|
|
%C = ptrtoint i8* %B to i64
|
|
|
|
%E = ptrtoint i8* %P to i64
|
|
|
|
%G = sub i64 %E, %C
|
|
|
|
ret i64 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i16 @test24a_as1(i8 addrspace(1)* %P, i16 %A) {
|
|
|
|
; CHECK-LABEL: @test24a_as1(
|
|
|
|
; CHECK-NEXT: [[DIFF_NEG:%.*]] = sub i16 0, [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret i16 [[DIFF_NEG]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds i8, i8 addrspace(1)* %P, i16 %A
|
|
|
|
%C = ptrtoint i8 addrspace(1)* %B to i16
|
|
|
|
%E = ptrtoint i8 addrspace(1)* %P to i16
|
|
|
|
%G = sub i16 %E, %C
|
|
|
|
ret i16 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
@Arr = external global [42 x i16]
|
|
|
|
|
|
|
|
define i64 @test24b(i8* %P, i64 %A){
|
|
|
|
; CHECK-LABEL: @test24b(
|
[Analysis] Don't assume that unsigned overflow can't happen in EmitGEPOffset (PR42699)
Summary:
Currently when computing a GEP offset using the function EmitGEPOffset
for the following instruction
getelementptr inbounds i32, i32* %p, i64 %offs
we get
mul nuw i64 %offs, 4
Unfortunately we cannot assume that unsigned wrapping won't happen
here because %offs is allowed to be negative.
Making such assumptions can lead to miscompilations: see the new test
test24_neg_offs in InstCombine/icmp.ll. Without the patch InstCombine
would generate the following comparison:
icmp eq i64 %offs, 4611686018427387902; 0x3ffffffffffffffe
Whereas the correct value to compare with is -2.
This patch replaces the NUW flag with NSW in the multiplication
instructions generated by EmitGEPOffset and adjusts the test suite.
https://bugs.llvm.org/show_bug.cgi?id=42699
Reviewers: chandlerc, craig.topper, ostannard, lebedev.ri, spatel, efriedma, nlopes, aqjune
Reviewed By: lebedev.ri
Subscribers: reames, lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68342
llvm-svn: 375089
2019-10-17 16:59:06 +08:00
|
|
|
; CHECK-NEXT: [[B_IDX:%.*]] = shl nsw i64 [[A:%.*]], 1
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i64 [[B_IDX]]
|
|
|
|
;
|
|
|
|
%B = getelementptr inbounds [42 x i16], [42 x i16]* @Arr, i64 0, i64 %A
|
|
|
|
%C = ptrtoint i16* %B to i64
|
|
|
|
%G = sub i64 %C, ptrtoint ([42 x i16]* @Arr to i64)
|
|
|
|
ret i64 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
define i64 @test25(i8* %P, i64 %A){
|
|
|
|
; CHECK-LABEL: @test25(
|
[Analysis] Don't assume that unsigned overflow can't happen in EmitGEPOffset (PR42699)
Summary:
Currently when computing a GEP offset using the function EmitGEPOffset
for the following instruction
getelementptr inbounds i32, i32* %p, i64 %offs
we get
mul nuw i64 %offs, 4
Unfortunately we cannot assume that unsigned wrapping won't happen
here because %offs is allowed to be negative.
Making such assumptions can lead to miscompilations: see the new test
test24_neg_offs in InstCombine/icmp.ll. Without the patch InstCombine
would generate the following comparison:
icmp eq i64 %offs, 4611686018427387902; 0x3ffffffffffffffe
Whereas the correct value to compare with is -2.
This patch replaces the NUW flag with NSW in the multiplication
instructions generated by EmitGEPOffset and adjusts the test suite.
https://bugs.llvm.org/show_bug.cgi?id=42699
Reviewers: chandlerc, craig.topper, ostannard, lebedev.ri, spatel, efriedma, nlopes, aqjune
Reviewed By: lebedev.ri
Subscribers: reames, lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68342
llvm-svn: 375089
2019-10-17 16:59:06 +08:00
|
|
|
; CHECK-NEXT: [[B_IDX:%.*]] = shl nsw i64 [[A:%.*]], 1
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add i64 [[B_IDX]], -84
|
|
|
|
; CHECK-NEXT: ret i64 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = getelementptr inbounds [42 x i16], [42 x i16]* @Arr, i64 0, i64 %A
|
|
|
|
%C = ptrtoint i16* %B to i64
|
|
|
|
%G = sub i64 %C, ptrtoint (i16* getelementptr ([42 x i16], [42 x i16]* @Arr, i64 1, i64 0) to i64)
|
|
|
|
ret i64 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
@Arr_as1 = external addrspace(1) global [42 x i16]
|
|
|
|
|
|
|
|
define i16 @test25_as1(i8 addrspace(1)* %P, i64 %A) {
|
|
|
|
; CHECK-LABEL: @test25_as1(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[A:%.*]] to i16
|
[Analysis] Don't assume that unsigned overflow can't happen in EmitGEPOffset (PR42699)
Summary:
Currently when computing a GEP offset using the function EmitGEPOffset
for the following instruction
getelementptr inbounds i32, i32* %p, i64 %offs
we get
mul nuw i64 %offs, 4
Unfortunately we cannot assume that unsigned wrapping won't happen
here because %offs is allowed to be negative.
Making such assumptions can lead to miscompilations: see the new test
test24_neg_offs in InstCombine/icmp.ll. Without the patch InstCombine
would generate the following comparison:
icmp eq i64 %offs, 4611686018427387902; 0x3ffffffffffffffe
Whereas the correct value to compare with is -2.
This patch replaces the NUW flag with NSW in the multiplication
instructions generated by EmitGEPOffset and adjusts the test suite.
https://bugs.llvm.org/show_bug.cgi?id=42699
Reviewers: chandlerc, craig.topper, ostannard, lebedev.ri, spatel, efriedma, nlopes, aqjune
Reviewed By: lebedev.ri
Subscribers: reames, lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68342
llvm-svn: 375089
2019-10-17 16:59:06 +08:00
|
|
|
; CHECK-NEXT: [[B_IDX:%.*]] = shl nsw i16 [[TMP1]], 1
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add i16 [[B_IDX]], -84
|
|
|
|
; CHECK-NEXT: ret i16 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%B = getelementptr inbounds [42 x i16], [42 x i16] addrspace(1)* @Arr_as1, i64 0, i64 %A
|
|
|
|
%C = ptrtoint i16 addrspace(1)* %B to i16
|
|
|
|
%G = sub i16 %C, ptrtoint (i16 addrspace(1)* getelementptr ([42 x i16], [42 x i16] addrspace(1)* @Arr_as1, i64 1, i64 0) to i16)
|
|
|
|
ret i16 %G
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test26(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test26(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i32 -3, [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SHL_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%shl = shl i32 3, %x
|
|
|
|
%neg = sub i32 0, %shl
|
|
|
|
ret i32 %neg
|
|
|
|
}
|
|
|
|
|
2020-01-18 23:54:25 +08:00
|
|
|
define i64 @test_neg_shl_sub(i64 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_sub(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[SUB_NEG:%.*]] = sub i64 [[B:%.*]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = shl i64 [[SUB_NEG]], 2
|
|
|
|
; CHECK-NEXT: ret i64 [[MUL_NEG]]
|
2020-01-18 23:54:25 +08:00
|
|
|
;
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
%mul = shl i64 %sub, 2
|
|
|
|
%neg = sub i64 0, %mul
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_shl_sub_extra_use1(i64 %a, i64 %b, i64* %p) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_sub_extra_use1(
|
2020-04-24 02:07:23 +08:00
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i64 [[A:%.*]], [[B:%.*]]
|
2020-01-18 23:54:25 +08:00
|
|
|
; CHECK-NEXT: store i64 [[SUB]], i64* [[P:%.*]], align 8
|
2020-04-24 02:07:23 +08:00
|
|
|
; CHECK-NEXT: [[MUL:%.*]] = shl i64 [[SUB]], 2
|
|
|
|
; CHECK-NEXT: [[NEG:%.*]] = sub i64 0, [[MUL]]
|
|
|
|
; CHECK-NEXT: ret i64 [[NEG]]
|
2020-01-18 23:54:25 +08:00
|
|
|
;
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
store i64 %sub, i64* %p
|
|
|
|
%mul = shl i64 %sub, 2
|
|
|
|
%neg = sub i64 0, %mul
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_shl_sub_extra_use2(i64 %a, i64 %b, i64* %p) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_sub_extra_use2(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i64 [[A:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[MUL:%.*]] = shl i64 [[SUB]], 2
|
|
|
|
; CHECK-NEXT: store i64 [[MUL]], i64* [[P:%.*]], align 8
|
|
|
|
; CHECK-NEXT: [[NEG:%.*]] = sub i64 0, [[MUL]]
|
|
|
|
; CHECK-NEXT: ret i64 [[NEG]]
|
|
|
|
;
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
%mul = shl i64 %sub, 2
|
|
|
|
store i64 %mul, i64* %p
|
|
|
|
%neg = sub i64 0, %mul
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
2020-01-24 04:12:41 +08:00
|
|
|
define i64 @test_neg_shl_div(i64 %a) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_div(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DIV_NEG:%.*]] = sdiv i64 [[A:%.*]], -3
|
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i64 [[DIV_NEG]], 2
|
|
|
|
; CHECK-NEXT: ret i64 [[SHL_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%div = sdiv i64 %a, 3
|
|
|
|
%shl = shl i64 %div, 2
|
|
|
|
%neg = sub i64 0, %shl
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_shl_zext_i1(i1 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_zext_i1(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[EXT_NEG:%.*]] = sext i1 [[A:%.*]] to i64
|
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i64 [[EXT_NEG]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[SHL_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%ext = zext i1 %a to i64
|
|
|
|
%shl = shl i64 %ext, %b
|
|
|
|
%neg = sub i64 0, %shl
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_shl_sext_i1(i1 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_shl_sext_i1(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[EXT_NEG:%.*]] = zext i1 [[A:%.*]] to i64
|
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i64 [[EXT_NEG]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[SHL_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%ext = sext i1 %a to i64
|
|
|
|
%shl = shl i64 %ext, %b
|
|
|
|
%neg = sub i64 0, %shl
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_zext_i1_extra_use(i1 %a, i64 %b, i64* %p) {
|
|
|
|
; CHECK-LABEL: @test_neg_zext_i1_extra_use(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[EXT_NEG:%.*]] = sext i1 [[A:%.*]] to i64
|
|
|
|
; CHECK-NEXT: [[EXT:%.*]] = zext i1 [[A]] to i64
|
2020-01-24 04:12:41 +08:00
|
|
|
; CHECK-NEXT: store i64 [[EXT]], i64* [[P:%.*]], align 8
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: ret i64 [[EXT_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%ext = zext i1 %a to i64
|
|
|
|
%neg = sub i64 0, %ext
|
|
|
|
store i64 %ext, i64* %p
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_sext_i1_extra_use(i1 %a, i64 %b, i64* %p) {
|
|
|
|
; CHECK-LABEL: @test_neg_sext_i1_extra_use(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[EXT_NEG:%.*]] = zext i1 [[A:%.*]] to i64
|
|
|
|
; CHECK-NEXT: [[EXT:%.*]] = sext i1 [[A]] to i64
|
2020-01-24 04:12:41 +08:00
|
|
|
; CHECK-NEXT: store i64 [[EXT]], i64* [[P:%.*]], align 8
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: ret i64 [[EXT_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%ext = sext i1 %a to i64
|
|
|
|
%neg = sub i64 0, %ext
|
|
|
|
store i64 %ext, i64* %p
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test_neg_trunc_shl_sub(i64 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_trunc_shl_sub(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[SUB_NEG:%.*]] = sub i64 [[B:%.*]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: [[SUB_TR_NEG:%.*]] = trunc i64 [[SUB_NEG]] to i32
|
|
|
|
; CHECK-NEXT: [[TRUNC_NEG:%.*]] = shl i32 [[SUB_TR_NEG]], 2
|
|
|
|
; CHECK-NEXT: ret i32 [[TRUNC_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
%shl = shl i64 %sub, 2
|
|
|
|
%trunc = trunc i64 %shl to i32
|
|
|
|
%neg = sub i32 0, %trunc
|
|
|
|
ret i32 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test_neg_trunc_shl_ashr(i64 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_trunc_shl_ashr(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[SHR_NEG:%.*]] = lshr i64 [[A:%.*]], 63
|
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i64 [[SHR_NEG]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[TRUNC_NEG:%.*]] = trunc i64 [[SHL_NEG]] to i32
|
|
|
|
; CHECK-NEXT: ret i32 [[TRUNC_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%shr = ashr i64 %a, 63
|
|
|
|
%shl = shl i64 %shr, %b
|
|
|
|
%trunc = trunc i64 %shl to i32
|
|
|
|
%neg = sub i32 0, %trunc
|
|
|
|
ret i32 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test_neg_trunc_shl_lshr(i64 %a, i64 %b) {
|
|
|
|
; CHECK-LABEL: @test_neg_trunc_shl_lshr(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[SHR_NEG:%.*]] = ashr i64 [[A:%.*]], 63
|
|
|
|
; CHECK-NEXT: [[SHL_NEG:%.*]] = shl i64 [[SHR_NEG]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[TRUNC_NEG:%.*]] = trunc i64 [[SHL_NEG]] to i32
|
|
|
|
; CHECK-NEXT: ret i32 [[TRUNC_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%shr = lshr i64 %a, 63
|
|
|
|
%shl = shl i64 %shr, %b
|
|
|
|
%trunc = trunc i64 %shl to i32
|
|
|
|
%neg = sub i32 0, %trunc
|
|
|
|
ret i32 %neg
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test_neg_mul_sub(i64 %a, i64 %b, i64 %c) {
|
|
|
|
; CHECK-LABEL: @test_neg_mul_sub(
|
2020-01-25 05:11:41 +08:00
|
|
|
; CHECK-NEXT: [[SUB_NEG:%.*]] = sub i64 [[B:%.*]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul i64 [[SUB_NEG]], [[C:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[MUL_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
%mul = mul i64 %sub, %c
|
|
|
|
%neg = sub i64 0, %mul
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
2020-01-25 05:11:41 +08:00
|
|
|
define i64 @test_neg_mul_sub_commuted(i64 %a, i64 %b, i64 %c) {
|
|
|
|
; CHECK-LABEL: @test_neg_mul_sub_commuted(
|
2020-01-24 04:12:41 +08:00
|
|
|
; CHECK-NEXT: [[COMPLEX:%.*]] = mul i64 [[C:%.*]], [[C]]
|
2020-01-25 05:11:41 +08:00
|
|
|
; CHECK-NEXT: [[SUB_NEG:%.*]] = sub i64 [[B:%.*]], [[A:%.*]]
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul i64 [[SUB_NEG]], [[COMPLEX]]
|
2020-01-25 05:11:41 +08:00
|
|
|
; CHECK-NEXT: ret i64 [[MUL_NEG]]
|
2020-01-24 04:12:41 +08:00
|
|
|
;
|
|
|
|
%complex = mul i64 %c, %c
|
|
|
|
%sub = sub i64 %a, %b
|
|
|
|
%mul = mul i64 %complex, %sub
|
|
|
|
%neg = sub i64 0, %mul
|
|
|
|
ret i64 %neg
|
|
|
|
}
|
|
|
|
|
2019-04-17 12:52:47 +08:00
|
|
|
define i32 @test27(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test27(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = shl i32 [[Y:%.*]], 3
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add i32 [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul i32 %y, -8
|
|
|
|
%sub = sub i32 %x, %mul
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27vec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27vec(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul <2 x i32> [[Y:%.*]], <i32 8, i32 6>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> %y, <i32 -8, i32 -6>
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27vecsplat(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27vecsplat(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = shl <2 x i32> [[Y:%.*]], <i32 3, i32 3>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> %y, <i32 -8, i32 -8>
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27vecmixed(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27vecmixed(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul <2 x i32> [[Y:%.*]], <i32 8, i32 -8>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> %y, <i32 -8, i32 8>
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test27commuted(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test27commuted(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = shl i32 [[Y:%.*]], 3
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add i32 [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul i32 -8, %y
|
|
|
|
%sub = sub i32 %x, %mul
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27commutedvec(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27commutedvec(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul <2 x i32> [[Y:%.*]], <i32 8, i32 6>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> <i32 -8, i32 -6>, %y
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27commutedvecsplat(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27commutedvecsplat(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = shl <2 x i32> [[Y:%.*]], <i32 3, i32 3>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> <i32 -8, i32 -8>, %y
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test27commutedvecmixed(<2 x i32> %x, <2 x i32> %y) {
|
|
|
|
; CHECK-LABEL: @test27commutedvecmixed(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[MUL_NEG:%.*]] = mul <2 x i32> [[Y:%.*]], <i32 8, i32 -8>
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i32> [[MUL_NEG]], [[X:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i32> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i32> <i32 -8, i32 8>, %y
|
|
|
|
%sub = sub <2 x i32> %x, %mul
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test28(i32 %x, i32 %y, i32 %z) {
|
|
|
|
; CHECK-LABEL: @test28(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[Z:%.*]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add i32 [[TMP1]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%neg = sub i32 0, %z
|
|
|
|
%mul = mul i32 %neg, %y
|
|
|
|
%sub = sub i32 %x, %mul
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test28commuted(i32 %x, i32 %y, i32 %z) {
|
|
|
|
; CHECK-LABEL: @test28commuted(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[Z:%.*]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add i32 [[TMP1]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%neg = sub i32 0, %z
|
|
|
|
%mul = mul i32 %y, %neg
|
|
|
|
%sub = sub i32 %x, %mul
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test29(i8* %foo, i64 %i, i64 %j) {
|
|
|
|
; CHECK-LABEL: @test29(
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = sub i64 [[I:%.*]], [[J:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%gep1 = getelementptr inbounds i8, i8* %foo, i64 %i
|
|
|
|
%gep2 = getelementptr inbounds i8, i8* %foo, i64 %j
|
|
|
|
%cast1 = ptrtoint i8* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test30(i8* %foo, i64 %i, i64 %j) {
|
|
|
|
; CHECK-LABEL: @test30(
|
[Analysis] Don't assume that unsigned overflow can't happen in EmitGEPOffset (PR42699)
Summary:
Currently when computing a GEP offset using the function EmitGEPOffset
for the following instruction
getelementptr inbounds i32, i32* %p, i64 %offs
we get
mul nuw i64 %offs, 4
Unfortunately we cannot assume that unsigned wrapping won't happen
here because %offs is allowed to be negative.
Making such assumptions can lead to miscompilations: see the new test
test24_neg_offs in InstCombine/icmp.ll. Without the patch InstCombine
would generate the following comparison:
icmp eq i64 %offs, 4611686018427387902; 0x3ffffffffffffffe
Whereas the correct value to compare with is -2.
This patch replaces the NUW flag with NSW in the multiplication
instructions generated by EmitGEPOffset and adjusts the test suite.
https://bugs.llvm.org/show_bug.cgi?id=42699
Reviewers: chandlerc, craig.topper, ostannard, lebedev.ri, spatel, efriedma, nlopes, aqjune
Reviewed By: lebedev.ri
Subscribers: reames, lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68342
llvm-svn: 375089
2019-10-17 16:59:06 +08:00
|
|
|
; CHECK-NEXT: [[GEP1_IDX:%.*]] = shl nsw i64 [[I:%.*]], 2
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = sub i64 [[GEP1_IDX]], [[J:%.*]]
|
|
|
|
; CHECK-NEXT: ret i64 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%bit = bitcast i8* %foo to i32*
|
|
|
|
%gep1 = getelementptr inbounds i32, i32* %bit, i64 %i
|
|
|
|
%gep2 = getelementptr inbounds i8, i8* %foo, i64 %j
|
|
|
|
%cast1 = ptrtoint i32* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i16 @test30_as1(i8 addrspace(1)* %foo, i16 %i, i16 %j) {
|
|
|
|
; CHECK-LABEL: @test30_as1(
|
[Analysis] Don't assume that unsigned overflow can't happen in EmitGEPOffset (PR42699)
Summary:
Currently when computing a GEP offset using the function EmitGEPOffset
for the following instruction
getelementptr inbounds i32, i32* %p, i64 %offs
we get
mul nuw i64 %offs, 4
Unfortunately we cannot assume that unsigned wrapping won't happen
here because %offs is allowed to be negative.
Making such assumptions can lead to miscompilations: see the new test
test24_neg_offs in InstCombine/icmp.ll. Without the patch InstCombine
would generate the following comparison:
icmp eq i64 %offs, 4611686018427387902; 0x3ffffffffffffffe
Whereas the correct value to compare with is -2.
This patch replaces the NUW flag with NSW in the multiplication
instructions generated by EmitGEPOffset and adjusts the test suite.
https://bugs.llvm.org/show_bug.cgi?id=42699
Reviewers: chandlerc, craig.topper, ostannard, lebedev.ri, spatel, efriedma, nlopes, aqjune
Reviewed By: lebedev.ri
Subscribers: reames, lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68342
llvm-svn: 375089
2019-10-17 16:59:06 +08:00
|
|
|
; CHECK-NEXT: [[GEP1_IDX:%.*]] = shl nsw i16 [[I:%.*]], 2
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = sub i16 [[GEP1_IDX]], [[J:%.*]]
|
|
|
|
; CHECK-NEXT: ret i16 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%bit = bitcast i8 addrspace(1)* %foo to i32 addrspace(1)*
|
|
|
|
%gep1 = getelementptr inbounds i32, i32 addrspace(1)* %bit, i16 %i
|
|
|
|
%gep2 = getelementptr inbounds i8, i8 addrspace(1)* %foo, i16 %j
|
|
|
|
%cast1 = ptrtoint i32 addrspace(1)* %gep1 to i16
|
|
|
|
%cast2 = ptrtoint i8 addrspace(1)* %gep2 to i16
|
|
|
|
%sub = sub i16 %cast1, %cast2
|
|
|
|
ret i16 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i64> @test31(<2 x i64> %A) {
|
|
|
|
; CHECK-LABEL: @test31(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add <2 x i64> [[A:%.*]], <i64 3, i64 4>
|
|
|
|
; CHECK-NEXT: ret <2 x i64> [[SUB]]
|
|
|
|
;
|
|
|
|
%xor = xor <2 x i64> %A, <i64 -1, i64 -1>
|
|
|
|
%sub = sub <2 x i64> <i64 2, i64 3>, %xor
|
|
|
|
ret <2 x i64> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i64> @test32(<2 x i64> %A) {
|
|
|
|
; CHECK-LABEL: @test32(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub <2 x i64> <i64 3, i64 4>, [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i64> [[SUB]]
|
|
|
|
;
|
|
|
|
%add = add <2 x i64> %A, <i64 -1, i64 -1>
|
|
|
|
%sub = sub <2 x i64> <i64 2, i64 3>, %add
|
|
|
|
ret <2 x i64> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i64> @test35(<2 x i64> %A) {
|
|
|
|
; CHECK-LABEL: @test35(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = mul <2 x i64> [[A:%.*]], <i64 -2, i64 -3>
|
|
|
|
; CHECK-NEXT: ret <2 x i64> [[SUB]]
|
|
|
|
;
|
|
|
|
%mul = mul <2 x i64> %A, <i64 3, i64 4>
|
|
|
|
%sub = sub <2 x i64> %A, %mul
|
|
|
|
ret <2 x i64> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i64> @test36(<2 x i64> %A) {
|
|
|
|
; CHECK-LABEL: @test36(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = mul <2 x i64> [[A:%.*]], <i64 7, i64 15>
|
|
|
|
; CHECK-NEXT: ret <2 x i64> [[SUB]]
|
|
|
|
;
|
|
|
|
%shl = shl <2 x i64> %A, <i64 3, i64 4>
|
|
|
|
%sub = sub <2 x i64> %shl, %A
|
|
|
|
ret <2 x i64> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test37(<2 x i32> %A) {
|
|
|
|
; CHECK-LABEL: @test37(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <2 x i32> [[A:%.*]], <i32 -2147483648, i32 -2147483648>
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DIV_NEG:%.*]] = sext <2 x i1> [[TMP1]] to <2 x i32>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[DIV_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%div = sdiv <2 x i32> %A, <i32 -2147483648, i32 -2147483648>
|
|
|
|
%sub = sub nsw <2 x i32> zeroinitializer, %div
|
|
|
|
ret <2 x i32> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test38(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test38(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[A:%.*]], -2147483648
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DIV_NEG:%.*]] = sext i1 [[TMP1]] to i32
|
|
|
|
; CHECK-NEXT: ret i32 [[DIV_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%div = sdiv i32 %A, -2147483648
|
|
|
|
%sub = sub nsw i32 0, %div
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i16 @test40(i16 %a, i16 %b) {
|
|
|
|
; CHECK-LABEL: @test40(
|
|
|
|
; CHECK-NEXT: [[ASHR:%.*]] = ashr i16 [[A:%.*]], 1
|
|
|
|
; CHECK-NEXT: [[ASHR1:%.*]] = ashr i16 [[B:%.*]], 1
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub nsw i16 [[ASHR]], [[ASHR1]]
|
|
|
|
; CHECK-NEXT: ret i16 [[SUB]]
|
|
|
|
;
|
|
|
|
%ashr = ashr i16 %a, 1
|
|
|
|
%ashr1 = ashr i16 %b, 1
|
|
|
|
%sub = sub i16 %ashr, %ashr1
|
|
|
|
ret i16 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test41(i16 %a, i16 %b) {
|
|
|
|
; CHECK-LABEL: @test41(
|
|
|
|
; CHECK-NEXT: [[CONV:%.*]] = sext i16 [[A:%.*]] to i32
|
|
|
|
; CHECK-NEXT: [[CONV1:%.*]] = sext i16 [[B:%.*]] to i32
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[CONV]], [[CONV1]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%conv = sext i16 %a to i32
|
|
|
|
%conv1 = sext i16 %b to i32
|
|
|
|
%sub = sub i32 %conv, %conv1
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i4 @test42(i4 %x, i4 %y) {
|
|
|
|
; CHECK-LABEL: @test42(
|
|
|
|
; CHECK-NEXT: [[A:%.*]] = and i4 [[Y:%.*]], 7
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = and i4 [[X:%.*]], 7
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sub nsw i4 [[A]], [[B]]
|
|
|
|
; CHECK-NEXT: ret i4 [[C]]
|
|
|
|
;
|
|
|
|
%a = and i4 %y, 7
|
|
|
|
%b = and i4 %x, 7
|
|
|
|
%c = sub i4 %a, %b
|
|
|
|
ret i4 %c
|
|
|
|
}
|
|
|
|
|
|
|
|
define i4 @test43(i4 %x, i4 %y) {
|
|
|
|
; CHECK-LABEL: @test43(
|
|
|
|
; CHECK-NEXT: [[A:%.*]] = or i4 [[X:%.*]], -8
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = and i4 [[Y:%.*]], 7
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sub nuw i4 [[A]], [[B]]
|
|
|
|
; CHECK-NEXT: ret i4 [[C]]
|
|
|
|
;
|
|
|
|
%a = or i4 %x, -8
|
|
|
|
%b = and i4 %y, 7
|
|
|
|
%c = sub i4 %a, %b
|
|
|
|
ret i4 %c
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test44(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test44(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = add nsw i32 [[X:%.*]], -32768
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%sub = sub nsw i32 %x, 32768
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test45(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test45(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = and i32 [[X:%.*]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%or = or i32 %x, %y
|
|
|
|
%xor = xor i32 %x, %y
|
|
|
|
%sub = sub i32 %or, %xor
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test45commuted(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test45commuted(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = and i32 [[Y:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%or = or i32 %x, %y
|
|
|
|
%xor = xor i32 %y, %x
|
|
|
|
%sub = sub i32 %or, %xor
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test46(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test46(
|
2019-05-03 23:14:19 +08:00
|
|
|
; CHECK-NEXT: [[X_NOT:%.*]] = xor i32 [[X:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = and i32 [[X_NOT]], [[Y:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%or = or i32 %x, %y
|
|
|
|
%sub = sub i32 %or, %x
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test46commuted(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @test46commuted(
|
2019-05-03 23:14:19 +08:00
|
|
|
; CHECK-NEXT: [[X_NOT:%.*]] = xor i32 [[X:%.*]], -1
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = and i32 [[X_NOT]], [[Y:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%or = or i32 %y, %x
|
|
|
|
%sub = sub i32 %or, %x
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test47(i1 %A, i32 %B, i32 %C, i32 %D) {
|
|
|
|
; CHECK-LABEL: @test47(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[D:%.*]], [[C:%.*]]
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = select i1 [[A:%.*]], i32 [[TMP1]], i32 0
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%sel0 = select i1 %A, i32 %D, i32 %B
|
|
|
|
%sel1 = select i1 %A, i32 %C, i32 %B
|
|
|
|
%sub = sub i32 %sel0, %sel1
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test48(i1 %A, i32 %B, i32 %C, i32 %D) {
|
|
|
|
; CHECK-LABEL: @test48(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[D:%.*]], [[C:%.*]]
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = select i1 [[A:%.*]], i32 0, i32 [[TMP1]]
|
|
|
|
; CHECK-NEXT: ret i32 [[SUB]]
|
|
|
|
;
|
|
|
|
%sel0 = select i1 %A, i32 %B, i32 %D
|
|
|
|
%sel1 = select i1 %A, i32 %B, i32 %C
|
|
|
|
%sub = sub i32 %sel0, %sel1
|
|
|
|
ret i32 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test49(i32 %X) {
|
|
|
|
; CHECK-LABEL: @test49(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i32 1, [[X:%.*]]
|
|
|
|
; CHECK-NEXT: [[RES:%.*]] = and i32 [[SUB]], 64
|
|
|
|
; CHECK-NEXT: ret i32 [[RES]]
|
|
|
|
;
|
|
|
|
%sub = sub i32 129, %X
|
|
|
|
%res = and i32 %sub, 64
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test50(i32 %X) {
|
|
|
|
; CHECK-LABEL: @test50(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i32 1, [[X:%.*]]
|
|
|
|
; CHECK-NEXT: [[RES:%.*]] = and i32 [[SUB]], 127
|
|
|
|
; CHECK-NEXT: ret i32 [[RES]]
|
|
|
|
;
|
|
|
|
%sub = sub i32 129, %X
|
|
|
|
%res = and i32 %sub, 127
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test51(i32 %X) {
|
|
|
|
; CHECK-LABEL: @test51(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i32 126, [[X:%.*]]
|
|
|
|
; CHECK-NEXT: [[RES:%.*]] = and i32 [[SUB]], 64
|
|
|
|
; CHECK-NEXT: ret i32 [[RES]]
|
|
|
|
;
|
|
|
|
%sub = sub i32 254, %X
|
|
|
|
%res = and i32 %sub, 64
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test52(i32 %X) {
|
|
|
|
; CHECK-LABEL: @test52(
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i32 126, [[X:%.*]]
|
|
|
|
; CHECK-NEXT: [[RES:%.*]] = and i32 [[SUB]], 127
|
|
|
|
; CHECK-NEXT: ret i32 [[RES]]
|
|
|
|
;
|
|
|
|
%sub = sub i32 254, %X
|
|
|
|
%res = and i32 %sub, 127
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i1> @test53(<2 x i1> %A, <2 x i1> %B) {
|
|
|
|
; CHECK-LABEL: @test53(
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = xor <2 x i1> [[B:%.*]], [[A:%.*]]
|
2019-04-17 12:52:47 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i1> [[SUB]]
|
|
|
|
;
|
|
|
|
%sub = sub <2 x i1> %A, %B
|
|
|
|
ret <2 x i1> %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test54(i1 %C) {
|
|
|
|
; CHECK-LABEL: @test54(
|
|
|
|
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], i32 -877, i32 113
|
|
|
|
; CHECK-NEXT: ret i32 [[V]]
|
|
|
|
;
|
|
|
|
%A = select i1 %C, i32 1000, i32 10
|
|
|
|
%V = sub i32 123, %A
|
|
|
|
ret i32 %V
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test54vec(i1 %C) {
|
|
|
|
; CHECK-LABEL: @test54vec(
|
|
|
|
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], <2 x i32> <i32 -877, i32 -877>, <2 x i32> <i32 113, i32 113>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[V]]
|
|
|
|
;
|
|
|
|
%A = select i1 %C, <2 x i32> <i32 1000, i32 1000>, <2 x i32> <i32 10, i32 10>
|
|
|
|
%V = sub <2 x i32> <i32 123, i32 123>, %A
|
|
|
|
ret <2 x i32> %V
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test54vec2(i1 %C) {
|
|
|
|
; CHECK-LABEL: @test54vec2(
|
|
|
|
; CHECK-NEXT: [[V:%.*]] = select i1 [[C:%.*]], <2 x i32> <i32 -877, i32 -2167>, <2 x i32> <i32 113, i32 303>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[V]]
|
|
|
|
;
|
|
|
|
%A = select i1 %C, <2 x i32> <i32 1000, i32 2500>, <2 x i32> <i32 10, i32 30>
|
|
|
|
%V = sub <2 x i32> <i32 123, i32 333>, %A
|
|
|
|
ret <2 x i32> %V
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test55(i1 %which) {
|
|
|
|
; CHECK-LABEL: @test55(
|
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
|
|
|
|
; CHECK: delay:
|
|
|
|
; CHECK-NEXT: br label [[FINAL]]
|
|
|
|
; CHECK: final:
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[A_NEG:%.*]] = phi i32 [ -877, [[ENTRY:%.*]] ], [ 113, [[DELAY]] ]
|
|
|
|
; CHECK-NEXT: ret i32 [[A_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
entry:
|
|
|
|
br i1 %which, label %final, label %delay
|
|
|
|
|
|
|
|
delay:
|
|
|
|
br label %final
|
|
|
|
|
|
|
|
final:
|
|
|
|
%A = phi i32 [ 1000, %entry ], [ 10, %delay ]
|
|
|
|
%value = sub i32 123, %A
|
|
|
|
ret i32 %value
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test55vec(i1 %which) {
|
|
|
|
; CHECK-LABEL: @test55vec(
|
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
|
|
|
|
; CHECK: delay:
|
|
|
|
; CHECK-NEXT: br label [[FINAL]]
|
|
|
|
; CHECK: final:
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[A_NEG:%.*]] = phi <2 x i32> [ <i32 -877, i32 -877>, [[ENTRY:%.*]] ], [ <i32 113, i32 113>, [[DELAY]] ]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[A_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
entry:
|
|
|
|
br i1 %which, label %final, label %delay
|
|
|
|
|
|
|
|
delay:
|
|
|
|
br label %final
|
|
|
|
|
|
|
|
final:
|
|
|
|
%A = phi <2 x i32> [ <i32 1000, i32 1000>, %entry ], [ <i32 10, i32 10>, %delay ]
|
|
|
|
%value = sub <2 x i32> <i32 123, i32 123>, %A
|
|
|
|
ret <2 x i32> %value
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test55vec2(i1 %which) {
|
|
|
|
; CHECK-LABEL: @test55vec2(
|
|
|
|
; CHECK-NEXT: entry:
|
|
|
|
; CHECK-NEXT: br i1 [[WHICH:%.*]], label [[FINAL:%.*]], label [[DELAY:%.*]]
|
|
|
|
; CHECK: delay:
|
|
|
|
; CHECK-NEXT: br label [[FINAL]]
|
|
|
|
; CHECK: final:
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[A_NEG:%.*]] = phi <2 x i32> [ <i32 -877, i32 -2167>, [[ENTRY:%.*]] ], [ <i32 113, i32 303>, [[DELAY]] ]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[A_NEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
entry:
|
|
|
|
br i1 %which, label %final, label %delay
|
|
|
|
|
|
|
|
delay:
|
|
|
|
br label %final
|
|
|
|
|
|
|
|
final:
|
|
|
|
%A = phi <2 x i32> [ <i32 1000, i32 2500>, %entry ], [ <i32 10, i32 30>, %delay ]
|
|
|
|
%value = sub <2 x i32> <i32 123, i32 333>, %A
|
|
|
|
ret <2 x i32> %value
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test56(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test56(
|
|
|
|
; CHECK-NEXT: [[Y:%.*]] = sub i32 0, [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[Y]]
|
|
|
|
;
|
|
|
|
%X = add i32 %A, %B
|
|
|
|
%Y = sub i32 %A, %X
|
|
|
|
ret i32 %Y
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test57(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test57(
|
|
|
|
; CHECK-NEXT: [[Y:%.*]] = sub i32 0, [[B:%.*]]
|
|
|
|
; CHECK-NEXT: ret i32 [[Y]]
|
|
|
|
;
|
|
|
|
%X = add i32 %B, %A
|
|
|
|
%Y = sub i32 %A, %X
|
|
|
|
ret i32 %Y
|
|
|
|
}
|
|
|
|
|
|
|
|
@dummy_global1 = external global i8*
|
|
|
|
@dummy_global2 = external global i8*
|
|
|
|
|
|
|
|
define i64 @test58([100 x [100 x i8]]* %foo, i64 %i, i64 %j) {
|
|
|
|
; Note the reassociate pass and another instcombine pass will further optimize this to
|
|
|
|
; "%sub = i64 %i, %j, ret i64 %sub"
|
|
|
|
; gep1 and gep2 have only one use
|
|
|
|
; CHECK-LABEL: @test58(
|
|
|
|
; CHECK-NEXT: [[GEP2_OFFS:%.*]] = add i64 [[J:%.*]], 4200
|
|
|
|
; CHECK-NEXT: [[GEP1_OFFS:%.*]] = add i64 [[I:%.*]], 4200
|
2020-01-24 04:13:57 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = sub i64 [[GEP1_OFFS]], [[GEP2_OFFS]]
|
|
|
|
; CHECK-NEXT: ret i64 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%gep1 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 %i
|
|
|
|
%gep2 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 %j
|
|
|
|
%cast1 = ptrtoint i8* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test59([100 x [100 x i8]]* %foo, i64 %i) {
|
|
|
|
; CHECK-LABEL: @test59(
|
|
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO:%.*]], i64 0, i64 42, i64 [[I:%.*]]
|
|
|
|
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO]], i64 0, i64 42, i64 0
|
|
|
|
; CHECK-NEXT: store i8* [[GEP1]], i8** @dummy_global1, align 8
|
|
|
|
; CHECK-NEXT: store i8* [[GEP2]], i8** @dummy_global2, align 8
|
|
|
|
; CHECK-NEXT: ret i64 [[I]]
|
|
|
|
;
|
|
|
|
; gep1 and gep2 have more than one uses
|
|
|
|
%gep1 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 %i
|
|
|
|
%gep2 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 0
|
|
|
|
%cast1 = ptrtoint i8* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
store i8* %gep1, i8** @dummy_global1
|
|
|
|
store i8* %gep2, i8** @dummy_global2
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test60([100 x [100 x i8]]* %foo, i64 %i, i64 %j) {
|
|
|
|
; CHECK-LABEL: @test60(
|
|
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO:%.*]], i64 0, i64 [[J:%.*]], i64 [[I:%.*]]
|
|
|
|
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO]], i64 0, i64 42, i64 0
|
|
|
|
; CHECK-NEXT: [[CAST1:%.*]] = ptrtoint i8* [[GEP1]] to i64
|
|
|
|
; CHECK-NEXT: [[CAST2:%.*]] = ptrtoint i8* [[GEP2]] to i64
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i64 [[CAST1]], [[CAST2]]
|
|
|
|
; CHECK-NEXT: store i8* [[GEP1]], i8** @dummy_global1, align 8
|
|
|
|
; CHECK-NEXT: ret i64 [[SUB]]
|
|
|
|
;
|
|
|
|
; gep1 has a non-constant index and more than one uses. Shouldn't duplicate the arithmetic.
|
|
|
|
%gep1 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 %j, i64 %i
|
|
|
|
%gep2 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 0
|
|
|
|
%cast1 = ptrtoint i8* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
store i8* %gep1, i8** @dummy_global1
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i64 @test61([100 x [100 x i8]]* %foo, i64 %i, i64 %j) {
|
|
|
|
; CHECK-LABEL: @test61(
|
|
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO:%.*]], i64 0, i64 42, i64 0
|
|
|
|
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* [[FOO]], i64 0, i64 [[J:%.*]], i64 [[I:%.*]]
|
|
|
|
; CHECK-NEXT: [[CAST1:%.*]] = ptrtoint i8* [[GEP1]] to i64
|
|
|
|
; CHECK-NEXT: [[CAST2:%.*]] = ptrtoint i8* [[GEP2]] to i64
|
|
|
|
; CHECK-NEXT: [[SUB:%.*]] = sub i64 [[CAST1]], [[CAST2]]
|
|
|
|
; CHECK-NEXT: store i8* [[GEP2]], i8** @dummy_global2, align 8
|
|
|
|
; CHECK-NEXT: ret i64 [[SUB]]
|
|
|
|
;
|
|
|
|
; gep2 has a non-constant index and more than one uses. Shouldn't duplicate the arithmetic.
|
|
|
|
%gep1 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 42, i64 0
|
|
|
|
%gep2 = getelementptr inbounds [100 x [100 x i8]], [100 x [100 x i8]]* %foo, i64 0, i64 %j, i64 %i
|
|
|
|
%cast1 = ptrtoint i8* %gep1 to i64
|
|
|
|
%cast2 = ptrtoint i8* %gep2 to i64
|
|
|
|
%sub = sub i64 %cast1, %cast2
|
|
|
|
store i8* %gep2, i8** @dummy_global2
|
|
|
|
ret i64 %sub
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test62(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test62(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = shl i32 [[A:%.*]], 1
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sub i32 2, [[B]]
|
|
|
|
; CHECK-NEXT: ret i32 [[C]]
|
|
|
|
;
|
|
|
|
%B = sub i32 1, %A
|
|
|
|
%C = shl i32 %B, 1
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test62vec(<2 x i32> %A) {
|
|
|
|
; CHECK-LABEL: @test62vec(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = shl <2 x i32> [[A:%.*]], <i32 1, i32 1>
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = sub <2 x i32> <i32 2, i32 2>, [[B]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[C]]
|
|
|
|
;
|
|
|
|
%B = sub <2 x i32> <i32 1, i32 1>, %A
|
|
|
|
%C = shl <2 x i32> %B, <i32 1, i32 1>
|
|
|
|
ret <2 x i32> %C
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test63(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test63(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = shl i32 [[A:%.*]], 1
|
|
|
|
; CHECK-NEXT: ret i32 [[B]]
|
|
|
|
;
|
|
|
|
%B = sub i32 1, %A
|
|
|
|
%C = shl i32 %B, 1
|
|
|
|
%D = sub i32 2, %C
|
|
|
|
ret i32 %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define <2 x i32> @test63vec(<2 x i32> %A) {
|
|
|
|
; CHECK-LABEL: @test63vec(
|
|
|
|
; CHECK-NEXT: [[B:%.*]] = shl <2 x i32> [[A:%.*]], <i32 1, i32 1>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[B]]
|
|
|
|
;
|
|
|
|
%B = sub <2 x i32> <i32 1, i32 1>, %A
|
|
|
|
%C = shl <2 x i32> %B, <i32 1, i32 1>
|
|
|
|
%D = sub <2 x i32> <i32 2, i32 2>, %C
|
|
|
|
ret <2 x i32> %D
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test64(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test64(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[X:%.*]], 255
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[X]], i32 255
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add nsw i32 [[TMP2]], 1
|
|
|
|
; CHECK-NEXT: ret i32 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor i32 %x, -1
|
|
|
|
%2 = icmp sgt i32 %1, -256
|
|
|
|
%3 = select i1 %2, i32 %1, i32 -256
|
|
|
|
%res = sub i32 0, %3
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test65(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test65(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[X:%.*]], -256
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[X]], i32 -256
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add i32 [[TMP2]], 1
|
|
|
|
; CHECK-NEXT: ret i32 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor i32 %x, -1
|
|
|
|
%2 = icmp slt i32 %1, 255
|
|
|
|
%3 = select i1 %2, i32 %1, i32 255
|
|
|
|
%res = sub i32 0, %3
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test66(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test66(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp ult i32 [[X:%.*]], -101
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[X]], i32 -101
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add nuw i32 [[TMP2]], 1
|
|
|
|
; CHECK-NEXT: ret i32 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor i32 %x, -1
|
|
|
|
%2 = icmp ugt i32 %1, 100
|
|
|
|
%3 = select i1 %2, i32 %1, i32 100
|
|
|
|
%res = sub i32 0, %3
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @test67(i32 %x) {
|
|
|
|
; CHECK-LABEL: @test67(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i32 [[X:%.*]], 100
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[X]], i32 100
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add i32 [[TMP2]], 1
|
|
|
|
; CHECK-NEXT: ret i32 [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor i32 %x, -1
|
|
|
|
%2 = icmp ult i32 %1, -101
|
|
|
|
%3 = select i1 %2, i32 %1, i32 -101
|
|
|
|
%res = sub i32 0, %3
|
|
|
|
ret i32 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check splat vectors too
|
|
|
|
define <2 x i32> @test68(<2 x i32> %x) {
|
|
|
|
; CHECK-LABEL: @test68(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt <2 x i32> [[X:%.*]], <i32 255, i32 255>
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select <2 x i1> [[TMP1]], <2 x i32> [[X]], <2 x i32> <i32 255, i32 255>
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add nsw <2 x i32> [[TMP2]], <i32 1, i32 1>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor <2 x i32> %x, <i32 -1, i32 -1>
|
|
|
|
%2 = icmp sgt <2 x i32> %1, <i32 -256, i32 -256>
|
|
|
|
%3 = select <2 x i1> %2, <2 x i32> %1, <2 x i32> <i32 -256, i32 -256>
|
|
|
|
%res = sub <2 x i32> zeroinitializer, %3
|
|
|
|
ret <2 x i32> %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; And non-splat constant vectors.
|
|
|
|
define <2 x i32> @test69(<2 x i32> %x) {
|
|
|
|
; CHECK-LABEL: @test69(
|
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt <2 x i32> [[X:%.*]], <i32 255, i32 127>
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = select <2 x i1> [[TMP1]], <2 x i32> [[X]], <2 x i32> <i32 255, i32 127>
|
[InstCombine] Negator - sink sinkable negations
Summary:
As we have discussed previously (e.g. in D63992 / D64090 / [[ https://bugs.llvm.org/show_bug.cgi?id=42457 | PR42457 ]]), `sub` instruction
can almost be considered non-canonical. While we do convert `sub %x, C` -> `add %x, -C`,
we sparsely do that for non-constants. But we should.
Here, i propose to interpret `sub %x, %y` as `add (sub 0, %y), %x` IFF the negation can be sinked into the `%y`
This has some potential to cause endless combine loops (either around PHI's, or if there are some opposite transforms).
For former there's `-instcombine-negator-max-depth` option to mitigate it, should this expose any such issues
For latter, if there are still any such opposing folds, we'd need to remove the colliding fold.
In any case, reproducers welcomed!
Reviewers: spatel, nikic, efriedma, xbolva00
Reviewed By: spatel
Subscribers: xbolva00, mgorny, hiraditya, reames, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68408
2020-04-22 02:24:36 +08:00
|
|
|
; CHECK-NEXT: [[DOTNEG:%.*]] = add <2 x i32> [[TMP2]], <i32 1, i32 1>
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[DOTNEG]]
|
2019-04-17 12:52:47 +08:00
|
|
|
;
|
|
|
|
%1 = xor <2 x i32> %x, <i32 -1, i32 -1>
|
|
|
|
%2 = icmp sgt <2 x i32> %1, <i32 -256, i32 -128>
|
|
|
|
%3 = select <2 x i1> %2, <2 x i32> %1, <2 x i32> <i32 -256, i32 -128>
|
|
|
|
%res = sub <2 x i32> zeroinitializer, %3
|
|
|
|
ret <2 x i32> %res
|
|
|
|
}
|
|
|
|
|
2019-05-07 22:14:29 +08:00
|
|
|
; Check (X | Y) - Y --> X & ~Y when Y is a constant
|
|
|
|
define i32 @test70(i32 %A) {
|
|
|
|
; CHECK-LABEL: @test70(
|
2019-05-08 03:36:41 +08:00
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[A:%.*]], -124
|
|
|
|
; CHECK-NEXT: ret i32 [[TMP1]]
|
2019-05-07 22:14:29 +08:00
|
|
|
;
|
|
|
|
%B = or i32 %A, 123
|
|
|
|
%C = sub i32 %B, 123
|
|
|
|
ret i32 %C
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check (X | Y) - Y --> (X | Y) ^ Y doesn't happen where (X | Y) has multiple uses
|
|
|
|
define i32 @test71(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test71(
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = or i32 [[A:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = sub i32 [[C]], [[B]]
|
|
|
|
; CHECK-NEXT: [[E:%.*]] = mul i32 [[C]], [[D]]
|
|
|
|
; CHECK-NEXT: ret i32 [[E]]
|
|
|
|
;
|
|
|
|
%C = or i32 %A, %B
|
|
|
|
%D = sub i32 %C, %B
|
|
|
|
%E = mul i32 %C, %D
|
|
|
|
ret i32 %E
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check (X | Y) - Y --> X & ~Y where X and Y are vectors
|
|
|
|
define <2 x i32> @test72(<2 x i32> %A, <2 x i32> %B) {
|
|
|
|
; CHECK-LABEL: @test72(
|
|
|
|
; CHECK-NEXT: [[B_NOT:%.*]] = xor <2 x i32> [[B:%.*]], <i32 -1, i32 -1>
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = and <2 x i32> [[B_NOT]], [[A:%.*]]
|
|
|
|
; CHECK-NEXT: ret <2 x i32> [[D]]
|
|
|
|
;
|
|
|
|
%C = or <2 x i32> %A, %B
|
|
|
|
%D = sub <2 x i32> %C, %B
|
|
|
|
ret <2 x i32> %D
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check reversing sub operands won't trigger (X | Y) - Y --> X & ~Y
|
|
|
|
define i32 @test73(i32 %A, i32 %B) {
|
|
|
|
; CHECK-LABEL: @test73(
|
|
|
|
; CHECK-NEXT: [[C:%.*]] = or i32 [[A:%.*]], [[B:%.*]]
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = sub i32 [[B]], [[C]]
|
|
|
|
; CHECK-NEXT: ret i32 [[D]]
|
|
|
|
;
|
|
|
|
%C = or i32 %A, %B
|
|
|
|
%D = sub i32 %B, %C
|
|
|
|
ret i32 %D
|
|
|
|
}
|
|
|
|
|
2019-04-17 12:52:47 +08:00
|
|
|
define i32 @nsw_inference1(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @nsw_inference1(
|
|
|
|
; CHECK-NEXT: [[X2:%.*]] = or i32 [[X:%.*]], 1024
|
|
|
|
; CHECK-NEXT: [[Y2:%.*]] = and i32 [[Y:%.*]], 1
|
|
|
|
; CHECK-NEXT: [[Z:%.*]] = sub nuw nsw i32 [[X2]], [[Y2]]
|
|
|
|
; CHECK-NEXT: ret i32 [[Z]]
|
|
|
|
;
|
|
|
|
%x2 = or i32 %x, 1024
|
|
|
|
%y2 = and i32 %y, 1
|
|
|
|
%z = sub i32 %x2, %y2
|
|
|
|
ret i32 %z
|
|
|
|
}
|
|
|
|
|
|
|
|
define i32 @nsw_inference2(i32 %x, i32 %y) {
|
|
|
|
; CHECK-LABEL: @nsw_inference2(
|
|
|
|
; CHECK-NEXT: [[X2:%.*]] = and i32 [[X:%.*]], -1025
|
|
|
|
; CHECK-NEXT: [[Y2:%.*]] = or i32 [[Y:%.*]], -2
|
|
|
|
; CHECK-NEXT: [[Z:%.*]] = sub nsw i32 [[X2]], [[Y2]]
|
|
|
|
; CHECK-NEXT: ret i32 [[Z]]
|
|
|
|
;
|
|
|
|
%x2 = and i32 %x, -1025
|
|
|
|
%y2 = or i32 %y, -2
|
|
|
|
%z = sub i32 %x2, %y2
|
|
|
|
ret i32 %z
|
|
|
|
}
|
2020-03-07 00:40:13 +08:00
|
|
|
|
|
|
|
define i8 @test74(i8 %x, i8 %y) {
|
|
|
|
; CHECK-LABEL: @test74(
|
|
|
|
; CHECK-NEXT: [[T0:%.*]] = and i8 [[X:%.*]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[T0]])
|
|
|
|
; CHECK-NEXT: [[T1:%.*]] = sub i8 [[X]], [[T0]]
|
|
|
|
; CHECK-NEXT: ret i8 [[T1]]
|
|
|
|
;
|
|
|
|
%t0 = and i8 %x, %y
|
|
|
|
call void @use8(i8 %t0)
|
|
|
|
%t1 = sub i8 %x, %t0
|
|
|
|
ret i8 %t1
|
|
|
|
}
|
|
|
|
|
|
|
|
define i8 @test75(i8 %x) {
|
|
|
|
; CHECK-LABEL: @test75(
|
|
|
|
; CHECK-NEXT: [[T0:%.*]] = and i8 [[X:%.*]], -8
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[T0]])
|
2020-03-07 02:39:07 +08:00
|
|
|
; CHECK-NEXT: [[T1:%.*]] = and i8 [[X]], 7
|
2020-03-07 00:40:13 +08:00
|
|
|
; CHECK-NEXT: ret i8 [[T1]]
|
|
|
|
;
|
|
|
|
%t0 = and i8 %x, -8
|
|
|
|
call void @use8(i8 %t0)
|
|
|
|
%t1 = sub i8 %x, %t0
|
|
|
|
ret i8 %t1
|
|
|
|
}
|
2020-05-27 01:27:16 +08:00
|
|
|
|
|
|
|
; ((w-x) + y) - z --> (w+y) - (x+z)
|
|
|
|
|
|
|
|
define i8 @sub_add_sub_reassoc(i8 %w, i8 %x, i8 %y, i8 %z) {
|
|
|
|
; CHECK-LABEL: @sub_add_sub_reassoc(
|
2020-05-27 02:32:57 +08:00
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = add i8 [[W:%.*]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[X:%.*]], [[Z:%.*]]
|
|
|
|
; CHECK-NEXT: [[S2:%.*]] = sub i8 [[TMP1]], [[TMP2]]
|
2020-05-27 01:27:16 +08:00
|
|
|
; CHECK-NEXT: ret i8 [[S2]]
|
|
|
|
;
|
|
|
|
%s1 = sub i8 %w, %x
|
|
|
|
%a = add i8 %s1, %y
|
|
|
|
%s2 = sub i8 %a, %z
|
|
|
|
ret i8 %s2
|
|
|
|
}
|
|
|
|
|
|
|
|
; vectors work too.
|
|
|
|
|
|
|
|
define <2 x i8> @sub_add_sub_reassoc_commute(<2 x i8> %w, <2 x i8> %x, <2 x i8> %y, <2 x i8> %z) {
|
|
|
|
; CHECK-LABEL: @sub_add_sub_reassoc_commute(
|
|
|
|
; CHECK-NEXT: [[D:%.*]] = sdiv <2 x i8> [[Y:%.*]], <i8 42, i8 -42>
|
2020-05-27 02:32:57 +08:00
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = add <2 x i8> [[D]], [[W:%.*]]
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i8> [[X:%.*]], [[Z:%.*]]
|
|
|
|
; CHECK-NEXT: [[S2:%.*]] = sub <2 x i8> [[TMP1]], [[TMP2]]
|
2020-05-27 01:27:16 +08:00
|
|
|
; CHECK-NEXT: ret <2 x i8> [[S2]]
|
|
|
|
;
|
|
|
|
%d = sdiv <2 x i8> %y, <i8 42, i8 -42> ; thwart complexity-based canonicalization
|
|
|
|
%s1 = sub <2 x i8> %w, %x
|
|
|
|
%a = add <2 x i8> %d, %s1
|
|
|
|
%s2 = sub <2 x i8> %a, %z
|
|
|
|
ret <2 x i8> %s2
|
|
|
|
}
|
|
|
|
|
|
|
|
; (v-w) + (x-y) - z --> (v+x) - (w+y+z)
|
|
|
|
|
|
|
|
define i8 @sub_add_sub_reassoc_twice(i8 %v, i8 %w, i8 %x, i8 %y, i8 %z) {
|
|
|
|
; CHECK-LABEL: @sub_add_sub_reassoc_twice(
|
2020-05-27 02:32:57 +08:00
|
|
|
; CHECK-NEXT: [[TMP1:%.*]] = add i8 [[W:%.*]], [[Z:%.*]]
|
|
|
|
; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[X:%.*]], [[V:%.*]]
|
|
|
|
; CHECK-NEXT: [[TMP3:%.*]] = add i8 [[TMP1]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: [[S3:%.*]] = sub i8 [[TMP2]], [[TMP3]]
|
2020-05-27 01:27:16 +08:00
|
|
|
; CHECK-NEXT: ret i8 [[S3]]
|
|
|
|
;
|
|
|
|
%s1 = sub i8 %v, %w
|
|
|
|
%s2 = sub i8 %x, %y
|
|
|
|
%a = add i8 %s1, %s2
|
|
|
|
%s3 = sub i8 %a, %z
|
|
|
|
ret i8 %s3
|
|
|
|
}
|
|
|
|
|
|
|
|
; negative test - uses
|
|
|
|
|
|
|
|
define i8 @sub_add_sub_reassoc_use1(i8 %w, i8 %x, i8 %y, i8 %z) {
|
|
|
|
; CHECK-LABEL: @sub_add_sub_reassoc_use1(
|
|
|
|
; CHECK-NEXT: [[S1:%.*]] = sub i8 [[W:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[S1]])
|
|
|
|
; CHECK-NEXT: [[A:%.*]] = add i8 [[S1]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: [[S2:%.*]] = sub i8 [[A]], [[Z:%.*]]
|
|
|
|
; CHECK-NEXT: ret i8 [[S2]]
|
|
|
|
;
|
|
|
|
%s1 = sub i8 %w, %x
|
|
|
|
call void @use8(i8 %s1)
|
|
|
|
%a = add i8 %s1, %y
|
|
|
|
%s2 = sub i8 %a, %z
|
|
|
|
ret i8 %s2
|
|
|
|
}
|
|
|
|
|
|
|
|
; negative test - uses
|
|
|
|
|
|
|
|
define i8 @sub_add_sub_reassoc_use2(i8 %w, i8 %x, i8 %y, i8 %z) {
|
|
|
|
; CHECK-LABEL: @sub_add_sub_reassoc_use2(
|
|
|
|
; CHECK-NEXT: [[S1:%.*]] = sub i8 [[W:%.*]], [[X:%.*]]
|
|
|
|
; CHECK-NEXT: [[A:%.*]] = add i8 [[S1]], [[Y:%.*]]
|
|
|
|
; CHECK-NEXT: call void @use8(i8 [[A]])
|
|
|
|
; CHECK-NEXT: [[S2:%.*]] = sub i8 [[A]], [[Z:%.*]]
|
|
|
|
; CHECK-NEXT: ret i8 [[S2]]
|
|
|
|
;
|
|
|
|
%s1 = sub i8 %w, %x
|
|
|
|
%a = add i8 %s1, %y
|
|
|
|
call void @use8(i8 %a)
|
|
|
|
%s2 = sub i8 %a, %z
|
|
|
|
ret i8 %s2
|
|
|
|
}
|