Skip to content

Commit

Permalink
Add DAGCombiner load combine tests with non-zero offset
Browse files Browse the repository at this point in the history 
This is separated from https://reviews.llvm.org/D29394 review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@294185 91177308-0d34-0410-b5e6-96231b3b80d8
  • Loading branch information
@arpilipe
arpilipe committed Feb 6, 2017
1 parent b0a111f commit cf5b8bc
Show file tree
Hide file tree
Showing 5 changed files with 805 additions and 3 deletions.
140 changes: 140 additions & 0 deletions test/CodeGen/AArch64/load-combine-big-endian.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -191,3 +191,143 @@ define i64 @load_i64_by_i8(i64* %arg) {
%tmp37 = or i64 %tmp33, %tmp36
ret i64 %tmp37
}

; i8* p; // p[1] is 4 byte aligned
; (i32) p[1] | ((i32) p[2] << 8) | ((i32) p[3] << 16) | ((i32) p[4] << 24)
define i32 @load_i32_by_i8_nonzero_offset(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_nonzero_offset:
; CHECK: ldrb w8, [x0, #1]
; CHECK-NEXT: ldrb w9, [x0, #2]
; CHECK-NEXT: ldrb w10, [x0, #3]
; CHECK-NEXT: ldrb w11, [x0, #4]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 1
%tmp2 = load i8, i8* %tmp1, align 4
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 2
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 3
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 4
%tmp15 = load i8, i8* %tmp14, align 1
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[-4] is 4 byte aligned
; (i32) p[-4] | ((i32) p[-3] << 8) | ((i32) p[-2] << 16) | ((i32) p[-1] << 24)
define i32 @load_i32_by_i8_neg_offset(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_neg_offset:
; CHECK: ldurb w8, [x0, #-4]
; CHECK-NEXT: ldurb w9, [x0, #-3]
; CHECK-NEXT: ldurb w10, [x0, #-2]
; CHECK-NEXT: ldurb w11, [x0, #-1]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 -4
%tmp2 = load i8, i8* %tmp1, align 4
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 -3
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 -2
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 -1
%tmp15 = load i8, i8* %tmp14, align 1
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[1] is 4 byte aligned
; (i32) p[4] | ((i32) p[3] << 8) | ((i32) p[2] << 16) | ((i32) p[1] << 24)
define i32 @load_i32_by_i8_nonzero_offset_bswap(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_nonzero_offset_bswap:
; CHECK: ldrb w8, [x0, #4]
; CHECK-NEXT: ldrb w9, [x0, #3]
; CHECK-NEXT: ldrb w10, [x0, #2]
; CHECK-NEXT: ldrb w11, [x0, #1]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 4
%tmp2 = load i8, i8* %tmp1, align 1
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 3
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 2
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 1
%tmp15 = load i8, i8* %tmp14, align 4
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[-4] is 4 byte aligned
; (i32) p[-1] | ((i32) p[-2] << 8) | ((i32) p[-3] << 16) | ((i32) p[-4] << 24)
define i32 @load_i32_by_i8_neg_offset_bswap(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_neg_offset_bswap:
; CHECK: ldurb w8, [x0, #-1]
; CHECK-NEXT: ldurb w9, [x0, #-2]
; CHECK-NEXT: ldurb w10, [x0, #-3]
; CHECK-NEXT: ldurb w11, [x0, #-4]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 -1
%tmp2 = load i8, i8* %tmp1, align 1
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 -2
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 -3
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 -4
%tmp15 = load i8, i8* %tmp14, align 4
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}
140 changes: 140 additions & 0 deletions test/CodeGen/AArch64/load-combine.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -178,3 +178,143 @@ define i64 @load_i64_by_i8_bswap(i64* %arg) {
%tmp37 = or i64 %tmp33, %tmp36
ret i64 %tmp37
}

; i8* p; // p[1] is 4 byte aligned
; (i32) p[1] | ((i32) p[2] << 8) | ((i32) p[3] << 16) | ((i32) p[4] << 24)
define i32 @load_i32_by_i8_nonzero_offset(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_nonzero_offset:
; CHECK: ldrb w8, [x0, #1]
; CHECK-NEXT: ldrb w9, [x0, #2]
; CHECK-NEXT: ldrb w10, [x0, #3]
; CHECK-NEXT: ldrb w11, [x0, #4]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 1
%tmp2 = load i8, i8* %tmp1, align 4
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 2
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 3
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 4
%tmp15 = load i8, i8* %tmp14, align 1
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[-4] is 4 byte aligned
; (i32) p[-4] | ((i32) p[-3] << 8) | ((i32) p[-2] << 16) | ((i32) p[-1] << 24)
define i32 @load_i32_by_i8_neg_offset(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_neg_offset:
; CHECK: ldurb w8, [x0, #-4]
; CHECK-NEXT: ldurb w9, [x0, #-3]
; CHECK-NEXT: ldurb w10, [x0, #-2]
; CHECK-NEXT: ldurb w11, [x0, #-1]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 -4
%tmp2 = load i8, i8* %tmp1, align 4
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 -3
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 -2
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 -1
%tmp15 = load i8, i8* %tmp14, align 1
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[1] is 4 byte aligned
; (i32) p[4] | ((i32) p[3] << 8) | ((i32) p[2] << 16) | ((i32) p[1] << 24)
define i32 @load_i32_by_i8_nonzero_offset_bswap(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_nonzero_offset_bswap:
; CHECK: ldrb w8, [x0, #4]
; CHECK-NEXT: ldrb w9, [x0, #3]
; CHECK-NEXT: ldrb w10, [x0, #2]
; CHECK-NEXT: ldrb w11, [x0, #1]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 4
%tmp2 = load i8, i8* %tmp1, align 1
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 3
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 2
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 1
%tmp15 = load i8, i8* %tmp14, align 4
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}

; i8* p; // p[-4] is 4 byte aligned
; (i32) p[-1] | ((i32) p[-2] << 8) | ((i32) p[-3] << 16) | ((i32) p[-4] << 24)
define i32 @load_i32_by_i8_neg_offset_bswap(i32* %arg) {
; CHECK-LABEL: load_i32_by_i8_neg_offset_bswap:
; CHECK: ldurb w8, [x0, #-1]
; CHECK-NEXT: ldurb w9, [x0, #-2]
; CHECK-NEXT: ldurb w10, [x0, #-3]
; CHECK-NEXT: ldurb w11, [x0, #-4]
; CHECK-NEXT: bfi w8, w9, #8, #8
; CHECK-NEXT: bfi w8, w10, #16, #8
; CHECK-NEXT: bfi w8, w11, #24, #8
; CHECK-NEXT: mov w0, w8
; CHECK-NEXT: ret
%tmp = bitcast i32* %arg to i8*
%tmp1 = getelementptr inbounds i8, i8* %tmp, i32 -1
%tmp2 = load i8, i8* %tmp1, align 1
%tmp3 = zext i8 %tmp2 to i32
%tmp4 = getelementptr inbounds i8, i8* %tmp, i32 -2
%tmp5 = load i8, i8* %tmp4, align 1
%tmp6 = zext i8 %tmp5 to i32
%tmp7 = shl nuw nsw i32 %tmp6, 8
%tmp8 = or i32 %tmp7, %tmp3
%tmp9 = getelementptr inbounds i8, i8* %tmp, i32 -3
%tmp10 = load i8, i8* %tmp9, align 1
%tmp11 = zext i8 %tmp10 to i32
%tmp12 = shl nuw nsw i32 %tmp11, 16
%tmp13 = or i32 %tmp8, %tmp12
%tmp14 = getelementptr inbounds i8, i8* %tmp, i32 -4
%tmp15 = load i8, i8* %tmp14, align 4
%tmp16 = zext i8 %tmp15 to i32
%tmp17 = shl nuw nsw i32 %tmp16, 24
%tmp18 = or i32 %tmp13, %tmp17
ret i32 %tmp18
}
Loading

0 comments on commit cf5b8bc

Please sign in to comment.