derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
415 lines
15 KiB
LLVM
415 lines
15 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
|
|
; RUN: opt < %s -passes=slsr,gvn -S | FileCheck %s
|
|
|
|
target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64-p:64:64:64-p1:32:32:32-p2:128:128:128:32"
|
|
|
|
; Index Delta
|
|
; Most of the original test cases in this file were optimized by Index-delta.
|
|
; After adding Base-delta and Stride-delta, most of the GEP test cases
|
|
; are optimized by Stride-delta now. The only case that GEP needs index-delta
|
|
; SLSR is to reuse address computation from a GEP with different pointee type.
|
|
; Once LLVM completely moves from typed GEP to PtrAdd, we can remove
|
|
; index-delta for GEP/PtrAdd.
|
|
|
|
define void @index_delta(ptr %input, i32 %c, i32 %b, i32 %n, float %r) {
|
|
; CHECK-LABEL: define void @index_delta(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i32 [[C:%.*]], i32 [[B:%.*]], i32 [[N:%.*]], float [[R:%.*]]) {
|
|
; CHECK-NEXT: [[ADD0:%.*]] = add nsw i32 [[B]], 1
|
|
; CHECK-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[ADD0]], [[N]]
|
|
; CHECK-NEXT: [[ADD1:%.*]] = add i32 [[MUL_1]], [[C]]
|
|
; CHECK-NEXT: [[OFFSET:%.*]] = sext i32 [[ADD1]] to i64
|
|
; CHECK-NEXT: [[GETELEM:%.*]] = getelementptr i8, ptr [[INPUT]], i64 [[OFFSET]]
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM]], align 4
|
|
; CHECK-NEXT: [[TMP:%.*]] = mul i64 [[OFFSET]], 3
|
|
; CHECK-NEXT: [[GETELEM_1:%.*]] = getelementptr inbounds i8, ptr [[GETELEM]], i64 [[TMP]]
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM_1]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
|
|
%add0 = add nsw i32 %b, 1
|
|
%mul.1 = mul nsw i32 %add0, %n
|
|
%add.1 = add i32 %mul.1, %c
|
|
%offset = sext i32 %add.1 to i64
|
|
%getElem = getelementptr i8, ptr %input, i64 %offset
|
|
store float %r, ptr %getElem, align 4
|
|
%getElem.1 = getelementptr inbounds float, ptr %input, i64 %offset
|
|
store float %r, ptr %getElem.1, align 4
|
|
|
|
ret void
|
|
}
|
|
|
|
; Stride Delta
|
|
|
|
; foo(input[0]);
|
|
; foo(input[s]);
|
|
; foo(input[s * 2]);
|
|
; =>
|
|
; p0 = &input[0];
|
|
; foo(*p);
|
|
; p1 = p0 + s;
|
|
; foo(*p1);
|
|
; p2 = p1 + s;
|
|
; foo(*p2);
|
|
define void @slsr_gep(ptr %input, i64 %s) {
|
|
; CHECK-LABEL: define void @slsr_gep(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i64 [[S:%.*]]) {
|
|
; CHECK-NEXT: call void @foo(ptr [[INPUT]])
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i32, ptr [[INPUT]], i64 [[S]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P1]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[S]], 2
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; v0 = input[0];
|
|
call void @foo(ptr %input)
|
|
|
|
; v1 = input[s];
|
|
%p1 = getelementptr inbounds i32, ptr %input, i64 %s
|
|
call void @foo(ptr %p1)
|
|
|
|
; v2 = input[s * 2];
|
|
%s2 = shl nsw i64 %s, 1
|
|
%p2 = getelementptr inbounds i32, ptr %input, i64 %s2
|
|
call void @foo(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
; foo(input[0]);
|
|
; foo(input[(long)s]);
|
|
; foo(input[(long)(s * 2)]);
|
|
; =>
|
|
; p0 = &input[0];
|
|
; foo(*p);
|
|
; p1 = p0 + (long)s;
|
|
; foo(*p1);
|
|
; p2 = p1 + (long)s;
|
|
; foo(*p2);
|
|
define void @slsr_gep_sext(ptr %input, i32 %s) {
|
|
; CHECK-LABEL: define void @slsr_gep_sext(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i32 [[S:%.*]]) {
|
|
; CHECK-NEXT: call void @foo(ptr [[INPUT]])
|
|
; CHECK-NEXT: [[T:%.*]] = sext i32 [[S]] to i64
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i32, ptr [[INPUT]], i64 [[T]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P1]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[T]], 2
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; v0 = input[0];
|
|
call void @foo(ptr %input)
|
|
|
|
; v1 = input[s];
|
|
%t = sext i32 %s to i64
|
|
%p1 = getelementptr inbounds i32, ptr %input, i64 %t
|
|
call void @foo(ptr %p1)
|
|
|
|
; v2 = input[s * 2];
|
|
%s2 = shl nsw i32 %s, 1
|
|
%t2 = sext i32 %s2 to i64
|
|
%p2 = getelementptr inbounds i32, ptr %input, i64 %t2
|
|
call void @foo(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
; int input[10][5];
|
|
; foo(input[s][t]);
|
|
; foo(input[s * 2][t]);
|
|
; foo(input[s * 3][t]);
|
|
; =>
|
|
; p0 = &input[s][t];
|
|
; foo(*p0);
|
|
; p1 = p0 + 5s;
|
|
; foo(*p1);
|
|
; p2 = p1 + 5s;
|
|
; foo(*p2);
|
|
define void @slsr_gep_2d(ptr %input, i64 %s, i64 %t) {
|
|
; CHECK-LABEL: define void @slsr_gep_2d(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i64 [[S:%.*]], i64 [[T:%.*]]) {
|
|
; CHECK-NEXT: [[P0:%.*]] = getelementptr [10 x [5 x i32]], ptr [[INPUT]], i64 0, i64 [[S]], i64 [[T]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P0]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[S]], 20
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i8, ptr [[P0]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P1]])
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; v0 = input[s][t];
|
|
%p0 = getelementptr inbounds [10 x [5 x i32]], ptr %input, i64 0, i64 %s, i64 %t
|
|
call void @foo(ptr %p0)
|
|
|
|
; v1 = input[s * 2][t];
|
|
%s2 = shl nsw i64 %s, 1
|
|
%p1 = getelementptr inbounds [10 x [5 x i32]], ptr %input, i64 0, i64 %s2, i64 %t
|
|
call void @foo(ptr %p1)
|
|
|
|
; v3 = input[s * 3][t];
|
|
%s3 = mul nsw i64 %s, 3
|
|
%p2 = getelementptr inbounds [10 x [5 x i32]], ptr %input, i64 0, i64 %s3, i64 %t
|
|
call void @foo(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
%struct.S = type <{ i64, i32 }>
|
|
|
|
; In this case, the bump
|
|
; = (char *)&input[s * 2][t].f1 - (char *)&input[s][t].f1
|
|
; = 60 * s
|
|
; which may not be divisible by typeof(input[s][t].f1) = 8. Therefore, we
|
|
; rewrite the candidates using byte offset instead of index offset as in
|
|
; @slsr_gep_2d.
|
|
define void @slsr_gep_uglygep(ptr %input, i64 %s, i64 %t) {
|
|
; CHECK-LABEL: define void @slsr_gep_uglygep(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i64 [[S:%.*]], i64 [[T:%.*]]) {
|
|
; CHECK-NEXT: [[P0:%.*]] = getelementptr [10 x [5 x [[STRUCT_S:%.*]]]], ptr [[INPUT]], i64 0, i64 [[S]], i64 [[T]], i32 0
|
|
; CHECK-NEXT: call void @bar(ptr [[P0]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[S]], 60
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i8, ptr [[P0]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @bar(ptr [[P1]])
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @bar(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; v0 = input[s][t].f1;
|
|
%p0 = getelementptr inbounds [10 x [5 x %struct.S]], ptr %input, i64 0, i64 %s, i64 %t, i32 0
|
|
call void @bar(ptr %p0)
|
|
|
|
; v1 = input[s * 2][t].f1;
|
|
%s2 = shl nsw i64 %s, 1
|
|
%p1 = getelementptr inbounds [10 x [5 x %struct.S]], ptr %input, i64 0, i64 %s2, i64 %t, i32 0
|
|
call void @bar(ptr %p1)
|
|
|
|
; v2 = input[s * 3][t].f1;
|
|
%s3 = mul nsw i64 %s, 3
|
|
%p2 = getelementptr inbounds [10 x [5 x %struct.S]], ptr %input, i64 0, i64 %s3, i64 %t, i32 0
|
|
call void @bar(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
define void @slsr_out_of_bounds_gep(ptr %input, i32 %s) {
|
|
; CHECK-LABEL: define void @slsr_out_of_bounds_gep(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i32 [[S:%.*]]) {
|
|
; CHECK-NEXT: call void @foo(ptr [[INPUT]])
|
|
; CHECK-NEXT: [[T:%.*]] = sext i32 [[S]] to i64
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i32, ptr [[INPUT]], i64 [[T]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P1]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[T]], 2
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr i8, ptr [[P1]], i64 [[TMP1]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; v0 = input[0];
|
|
call void @foo(ptr %input)
|
|
|
|
; v1 = input[(long)s];
|
|
%t = sext i32 %s to i64
|
|
%p1 = getelementptr i32, ptr %input, i64 %t
|
|
call void @foo(ptr %p1)
|
|
|
|
; v2 = input[(long)(s * 2)];
|
|
%s2 = shl nsw i32 %s, 1
|
|
%t2 = sext i32 %s2 to i64
|
|
%p2 = getelementptr i32, ptr %input, i64 %t2
|
|
call void @foo(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
define void @slsr_gep_128bit_index(ptr %input, i128 %s) {
|
|
; CHECK-LABEL: define void @slsr_gep_128bit_index(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i128 [[S:%.*]]) {
|
|
; CHECK-NEXT: call void @foo(ptr [[INPUT]])
|
|
; CHECK-NEXT: [[S125:%.*]] = shl nsw i128 [[S]], 125
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds i32, ptr [[INPUT]], i128 [[S125]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P1]])
|
|
; CHECK-NEXT: [[S126:%.*]] = shl nsw i128 [[S]], 126
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i32, ptr [[INPUT]], i128 [[S126]]
|
|
; CHECK-NEXT: call void @foo(ptr [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; p0 = &input[0]
|
|
call void @foo(ptr %input)
|
|
|
|
; p1 = &input[s << 125]
|
|
%s125 = shl nsw i128 %s, 125
|
|
%p1 = getelementptr inbounds i32, ptr %input, i128 %s125
|
|
call void @foo(ptr %p1)
|
|
|
|
; p2 = &input[s << 126]
|
|
%s126 = shl nsw i128 %s, 126
|
|
%p2 = getelementptr inbounds i32, ptr %input, i128 %s126
|
|
call void @foo(ptr %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
define void @slsr_gep_32bit_pointer(ptr addrspace(1) %input, i64 %s) {
|
|
; CHECK-LABEL: define void @slsr_gep_32bit_pointer(
|
|
; CHECK-SAME: ptr addrspace(1) [[INPUT:%.*]], i64 [[S:%.*]]) {
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[INPUT]], i64 [[S]]
|
|
; CHECK-NEXT: call void @baz(ptr addrspace(1) [[P1]])
|
|
; CHECK-NEXT: [[S2:%.*]] = mul nsw i64 [[S]], 2
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[INPUT]], i64 [[S2]]
|
|
; CHECK-NEXT: call void @baz(ptr addrspace(1) [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; p1 = &input[s]
|
|
%p1 = getelementptr inbounds i32, ptr addrspace(1) %input, i64 %s
|
|
call void @baz(ptr addrspace(1) %p1)
|
|
|
|
; p2 = &input[s * 2]
|
|
%s2 = mul nsw i64 %s, 2
|
|
%p2 = getelementptr inbounds i32, ptr addrspace(1) %input, i64 %s2
|
|
; %s2 is wider than the pointer size of addrspace(1), so do not factor it.
|
|
call void @baz(ptr addrspace(1) %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
define void @slsr_gep_fat_pointer(ptr addrspace(2) %input, i32 %s) {
|
|
; p1 = &input[s]
|
|
; CHECK-LABEL: define void @slsr_gep_fat_pointer(
|
|
; CHECK-SAME: ptr addrspace(2) [[INPUT:%.*]], i32 [[S:%.*]]) {
|
|
; CHECK-NEXT: [[P1:%.*]] = getelementptr i32, ptr addrspace(2) [[INPUT]], i32 [[S]]
|
|
; CHECK-NEXT: call void @baz2(ptr addrspace(2) [[P1]])
|
|
; CHECK-NEXT: [[TMP1:%.*]] = shl i32 [[S]], 2
|
|
; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i8, ptr addrspace(2) [[P1]], i32 [[TMP1]]
|
|
; CHECK-NEXT: call void @baz2(ptr addrspace(2) [[P2]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%p1 = getelementptr inbounds i32, ptr addrspace(2) %input, i32 %s
|
|
call void @baz2(ptr addrspace(2) %p1)
|
|
|
|
; p2 = &input[s * 2]
|
|
%s2 = mul nsw i32 %s, 2
|
|
%p2 = getelementptr inbounds i32, ptr addrspace(2) %input, i32 %s2
|
|
; Use index bitwidth, not pointer size (i128)
|
|
call void @baz2(ptr addrspace(2) %p2)
|
|
|
|
ret void
|
|
}
|
|
|
|
|
|
declare void @foo(ptr)
|
|
declare void @bar(ptr)
|
|
declare void @baz(ptr addrspace(1))
|
|
declare void @baz2(ptr addrspace(2))
|
|
|
|
define void @stride_const(ptr %input, i32 %c, i32 %b, i32 %n, float %r) {
|
|
; CHECK-LABEL: define void @stride_const(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i32 [[C:%.*]], i32 [[B:%.*]], i32 [[N:%.*]], float [[R:%.*]]) {
|
|
; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[B]], [[N]]
|
|
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[MUL]], [[C]]
|
|
; CHECK-NEXT: [[ADD_1:%.*]] = add i32 [[ADD]], [[N]]
|
|
; CHECK-NEXT: [[ADD_2:%.*]] = add i32 [[ADD_1]], [[N]]
|
|
; CHECK-NEXT: [[OFFSET:%.*]] = sext i32 [[ADD_2]] to i64
|
|
; CHECK-NEXT: [[GETELEM_1:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[OFFSET]]
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM_1]], align 4
|
|
; CHECK-NEXT: [[GETELEM_2:%.*]] = getelementptr i8, ptr [[GETELEM_1]], i64 16
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM_2]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
|
|
%mul = mul nsw i32 %b, %n
|
|
%add = add i32 %mul, %c
|
|
%add.1 = add i32 %add, %n
|
|
%add.2 = add i32 %add.1, %n
|
|
|
|
%offset = sext i32 %add.2 to i64
|
|
%1 = getelementptr float, ptr %input, i64 %offset
|
|
store float %r, ptr %1, align 4
|
|
|
|
%offset3 = add i64 %offset, 4
|
|
%2 = getelementptr float, ptr %input, i64 %offset3
|
|
store float %r, ptr %2, align 4
|
|
ret void
|
|
}
|
|
|
|
|
|
define void @stride_var(ptr %input, i32 %c, i32 %b, i32 %n, float %r) {
|
|
; CHECK-LABEL: define void @stride_var(
|
|
; CHECK-SAME: ptr [[INPUT:%.*]], i32 [[C:%.*]], i32 [[B:%.*]], i32 [[N:%.*]], float [[R:%.*]]) {
|
|
; CHECK-NEXT: [[ADD0:%.*]] = add nsw i32 [[B]], 1
|
|
; CHECK-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[ADD0]], [[N]]
|
|
; CHECK-NEXT: [[ADD1:%.*]] = add i32 [[MUL_1]], [[C]]
|
|
; CHECK-NEXT: [[I:%.*]] = sext i32 [[ADD1]] to i64
|
|
; CHECK-NEXT: [[GETELEM:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[I]]
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM]], align 4
|
|
; CHECK-NEXT: [[TMP1:%.*]] = sext i32 [[N]] to i64
|
|
; CHECK-NEXT: [[TMP2:%.*]] = shl i64 [[TMP1]], 2
|
|
; CHECK-NEXT: [[GETELEM_1:%.*]] = getelementptr inbounds i8, ptr [[GETELEM]], i64 [[TMP2]]
|
|
; CHECK-NEXT: store float [[R]], ptr [[GETELEM_1]], align 4
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
; Reuse getElem to compute getElem.1 and getElem.2 with variable offset n extracted from Stride
|
|
|
|
%add0 = add nsw i32 %b, 1
|
|
%mul.1 = mul nsw i32 %add0, %n
|
|
%add.1 = add i32 %mul.1, %c
|
|
%offset = sext i32 %add.1 to i64
|
|
%getElem = getelementptr float, ptr %input, i64 %offset
|
|
store float %r, ptr %getElem, align 4
|
|
|
|
%mul = mul nsw i32 %b, %n
|
|
%add = add nsw i32 %mul, %c
|
|
%add.11 = add nsw i32 %add, %n
|
|
%add.2 = add nsw i32 %add.11, %n
|
|
%offset1 = sext i32 %add.2 to i64
|
|
%getElem.1 = getelementptr inbounds float, ptr %input, i64 %offset1
|
|
store float %r, ptr %getElem.1, align 4
|
|
|
|
ret void
|
|
}
|
|
|
|
; Base Delta
|
|
|
|
%struct.B = type { i16 }
|
|
%struct.A = type { %struct.B, %struct.B }
|
|
|
|
define void @base_const(i32 %a, ptr %base, i16 %r) {
|
|
; Reuse getElem1 to compute getElem2
|
|
; CHECK-LABEL: define void @base_const(
|
|
; CHECK-SAME: i32 [[A:%.*]], ptr [[BASE:%.*]], i16 [[R:%.*]]) {
|
|
; CHECK-NEXT: [[TMP1:%.*]] = sext i32 [[A]] to i64
|
|
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], ptr [[BASE]], i64 [[TMP1]]
|
|
; CHECK-NEXT: store i16 [[R]], ptr [[GEP1]], align 2
|
|
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i8, ptr [[GEP1]], i64 2
|
|
; CHECK-NEXT: store i16 [[R]], ptr [[GEP2]], align 2
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
|
|
%1 = sext i32 %a to i64
|
|
%getElem1 = getelementptr inbounds %struct.A, ptr %base, i64 %1
|
|
store i16 %r, ptr %getElem1, align 2
|
|
%getElem2 = getelementptr inbounds %struct.A, ptr %base, i64 %1, i32 1
|
|
store i16 %r, ptr %getElem2, align 2
|
|
ret void
|
|
}
|
|
|
|
define void @base_var(i32 %a, ptr %base, i16 %r, i64 %n) {
|
|
; Reuse getElem1 to compute getElem2
|
|
; CHECK-LABEL: define void @base_var(
|
|
; CHECK-SAME: i32 [[A:%.*]], ptr [[BASE:%.*]], i16 [[R:%.*]], i64 [[N:%.*]]) {
|
|
; CHECK-NEXT: [[TMP1:%.*]] = sext i32 [[A]] to i64
|
|
; CHECK-NEXT: [[GETELEM1:%.*]] = getelementptr inbounds [[STRUCT_A:%.*]], ptr [[BASE]], i64 [[TMP1]]
|
|
; CHECK-NEXT: store i16 [[R]], ptr [[GETELEM1]], align 2
|
|
; CHECK-NEXT: [[GETELEM2:%.*]] = getelementptr inbounds i8, ptr [[GETELEM1]], i64 [[N]]
|
|
; CHECK-NEXT: store i16 [[R]], ptr [[GETELEM2]], align 2
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
|
|
%1 = sext i32 %a to i64
|
|
%base1 = getelementptr inbounds i8, ptr %base, i64 %n
|
|
%getElem1 = getelementptr inbounds %struct.A, ptr %base, i64 %1
|
|
store i16 %r, ptr %getElem1, align 2
|
|
%getElem2 = getelementptr inbounds %struct.A, ptr %base1, i64 %1
|
|
store i16 %r, ptr %getElem2, align 2
|
|
ret void
|
|
}
|