define i32 @test5(i32 %A) {
; CHECK: @test5
-; CHECK: ret i32 0
+; CHECK: ret i32 undef
%B = lshr i32 %A, 32 ;; shift all bits out
ret i32 %B
}
define i32 @test5a(i32 %A) {
; CHECK: @test5a
-; CHECK: ret i32 0
+; CHECK: ret i32 undef
%B = shl i32 %A, 32 ;; shift all bits out
ret i32 %B
}
+define i32 @test5b() {
+; CHECK: @test5b
+; CHECK: ret i32 -1
+ %B = ashr i32 undef, 2 ;; top two bits must be equal, so not undef
+ ret i32 %B
+}
+
+define i32 @test5b2(i32 %A) {
+; CHECK: @test5b2
+; CHECK: ret i32 -1
+ %B = ashr i32 undef, %A ;; top %A bits must be equal, so not undef
+ ret i32 %B
+}
+
define i32 @test6(i32 %A) {
; CHECK: @test6
; CHECK-NEXT: mul i32 %A, 6
; CHECK-NEXT: ret i32
- %B = shl i32 %A, 1 ;; convert to an mul instruction
- %C = mul i32 %B, 3
+ %B = shl i32 %A, 1 ;; convert to an mul instruction
+ %C = mul i32 %B, 3
+ ret i32 %C
+}
+
+define i32 @test6a(i32 %A) {
+; CHECK: @test6a
+; CHECK-NEXT: mul i32 %A, 6
+; CHECK-NEXT: ret i32
+ %B = mul i32 %A, 3
+ %C = shl i32 %B, 1 ;; convert to an mul instruction
ret i32 %C
}
ret i8 %C
}
+;; This transformation is deferred to DAGCombine:
;; (A >> 7) << 7 === A & 128
+;; The shl may be valuable to scalar evolution.
define i8 @test10(i8 %A) {
; CHECK: @test10
; CHECK-NEXT: and i8 %A, -128
ret i8 %C
}
+;; Allow the simplification when the lshr shift is exact.
+define i8 @test10a(i8 %A) {
+; CHECK: @test10a
+; CHECK-NEXT: ret i8 %A
+ %B = lshr exact i8 %A, 7
+ %C = shl i8 %B, 7
+ ret i8 %C
+}
+
+;; This transformation is deferred to DAGCombine:
;; (A >> 3) << 4 === (A & 0x1F) << 1
+;; The shl may be valuable to scalar evolution.
define i8 @test11(i8 %A) {
; CHECK: @test11
-; CHECK-NEXT: mul i8 %A, 6
-; CHECK-NEXT: and i8
+; CHECK: shl i8
; CHECK-NEXT: ret i8
%a = mul i8 %A, 3 ; <i8> [#uses=1]
%B = lshr i8 %a, 3 ; <i8> [#uses=1]
ret i8 %C
}
+;; Allow the simplification in InstCombine when the lshr shift is exact.
+define i8 @test11a(i8 %A) {
+; CHECK: @test11a
+; CHECK-NEXT: mul i8 %A, 6
+; CHECK-NEXT: ret i8
+ %a = mul i8 %A, 3
+ %B = lshr exact i8 %a, 3
+ %C = shl i8 %B, 4
+ ret i8 %C
+}
+
+;; This is deferred to DAGCombine unless %B is single-use.
;; (A >> 8) << 8 === A & -256
define i32 @test12(i32 %A) {
; CHECK: @test12
ret i32 %C
}
+;; This transformation is deferred to DAGCombine:
;; (A >> 3) << 4 === (A & -8) * 2
+;; The shl may be valuable to scalar evolution.
define i8 @test13(i8 %A) {
; CHECK: @test13
-; CHECK-NEXT: mul i8 %A, 6
-; CHECK-NEXT: and i8
+; CHECK: shl i8
; CHECK-NEXT: ret i8
%a = mul i8 %A, 3 ; <i8> [#uses=1]
%B = ashr i8 %a, 3 ; <i8> [#uses=1]
ret i8 %C
}
+define i8 @test13a(i8 %A) {
+; CHECK: @test13a
+; CHECK-NEXT: mul i8 %A, 6
+; CHECK-NEXT: ret i8
+ %a = mul i8 %A, 3
+ %B = ashr exact i8 %a, 3
+ %C = shl i8 %B, 4
+ ret i8 %C
+}
+
;; D = ((B | 1234) << 4) === ((B << 4)|(1234 << 4)
define i32 @test14(i32 %A) {
; CHECK: @test14
ret i64 %tmp46
; CHECK: @test37
-; CHECK: %tmp23 = shl i128 %tmp22, 32
+; CHECK: %tmp23 = shl nuw nsw i128 %tmp22, 32
; CHECK: %ins = or i128 %tmp23, %A
; CHECK: %tmp46 = trunc i128 %ins to i64
}
%tmp49 = lshr i8 %tmp48, 5
%tmp50 = mul i8 %tmp49, 64
%tmp51 = xor i8 %tmp50, %tmp5
-; CHECK: and i8 %0, 16
%tmp52 = and i8 %tmp51, -128
%tmp53 = lshr i8 %tmp52, 7
+; CHECK: lshr i8 %tmp51, 7
%tmp54 = mul i8 %tmp53, 16
+; CHECK: shl nuw nsw i8 %tmp53, 4
%tmp55 = xor i8 %tmp54, %tmp51
; CHECK: ret i8 %tmp551
ret i8 %tmp55
}
+
+; PR9809
+define i32 @test40(i32 %a, i32 %b) nounwind {
+ %shl1 = shl i32 1, %b
+ %shl2 = shl i32 %shl1, 2
+ %div = udiv i32 %a, %shl2
+ ret i32 %div
+; CHECK: @test40
+; CHECK-NEXT: add i32 %b, 2
+; CHECK-NEXT: lshr i32 %a
+; CHECK-NEXT: ret i32
+}
+
+define i32 @test41(i32 %a, i32 %b) nounwind {
+ %1 = shl i32 1, %b
+ %2 = shl i32 %1, 3
+ ret i32 %2
+; CHECK: @test41
+; CHECK-NEXT: shl i32 8, %b
+; CHECK-NEXT: ret i32
+}
+
+define i32 @test42(i32 %a, i32 %b) nounwind {
+ %div = lshr i32 4096, %b ; must be exact otherwise we'd divide by zero
+ %div2 = udiv i32 %a, %div
+ ret i32 %div2
+; CHECK: @test42
+; CHECK-NEXT: lshr exact i32 4096, %b
+}
+
+define i32 @test43(i32 %a, i32 %b) nounwind {
+ %div = shl i32 4096, %b ; must be exact otherwise we'd divide by zero
+ %div2 = udiv i32 %a, %div
+ ret i32 %div2
+; CHECK: @test43
+; CHECK-NEXT: add i32 %b, 12
+; CHECK-NEXT: lshr
+; CHECK-NEXT: ret
+}
+
+define i32 @test44(i32 %a) nounwind {
+ %y = shl nuw i32 %a, 1
+ %z = shl i32 %y, 4
+ ret i32 %z
+; CHECK: @test44
+; CHECK-NEXT: %y = shl i32 %a, 5
+; CHECK-NEXT: ret i32 %y
+}
+
+define i32 @test45(i32 %a) nounwind {
+ %y = lshr exact i32 %a, 1
+ %z = lshr i32 %y, 4
+ ret i32 %z
+; CHECK: @test45
+; CHECK-NEXT: %y = lshr i32 %a, 5
+; CHECK-NEXT: ret i32 %y
+}
+
+define i32 @test46(i32 %a) {
+ %y = ashr exact i32 %a, 3
+ %z = shl i32 %y, 1
+ ret i32 %z
+; CHECK: @test46
+; CHECK-NEXT: %z = ashr exact i32 %a, 2
+; CHECK-NEXT: ret i32 %z
+}
+
+define i32 @test47(i32 %a) {
+ %y = lshr exact i32 %a, 3
+ %z = shl i32 %y, 1
+ ret i32 %z
+; CHECK: @test47
+; CHECK-NEXT: %z = lshr exact i32 %a, 2
+; CHECK-NEXT: ret i32 %z
+}
+
+define i32 @test48(i32 %x) {
+ %A = lshr exact i32 %x, 1
+ %B = shl i32 %A, 3
+ ret i32 %B
+; CHECK: @test48
+; CHECK-NEXT: %B = shl i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
+
+define i32 @test49(i32 %x) {
+ %A = ashr exact i32 %x, 1
+ %B = shl i32 %A, 3
+ ret i32 %B
+; CHECK: @test49
+; CHECK-NEXT: %B = shl i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
+
+define i32 @test50(i32 %x) {
+ %A = shl nsw i32 %x, 1
+ %B = ashr i32 %A, 3
+ ret i32 %B
+; CHECK: @test50
+; CHECK-NEXT: %B = ashr i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
+
+define i32 @test51(i32 %x) {
+ %A = shl nuw i32 %x, 1
+ %B = lshr i32 %A, 3
+ ret i32 %B
+; CHECK: @test51
+; CHECK-NEXT: %B = lshr i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
+
+define i32 @test52(i32 %x) {
+ %A = shl nsw i32 %x, 3
+ %B = ashr i32 %A, 1
+ ret i32 %B
+; CHECK: @test52
+; CHECK-NEXT: %B = shl nsw i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
+
+define i32 @test53(i32 %x) {
+ %A = shl nuw i32 %x, 3
+ %B = lshr i32 %A, 1
+ ret i32 %B
+; CHECK: @test53
+; CHECK-NEXT: %B = shl nuw i32 %x, 2
+; CHECK-NEXT: ret i32 %B
+}
projects / oota-llvm.git / blobdiff