Skip to content

[BasicTTI] When costing a scalarized cast, use distinct src and dest types #109325

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 3 commits into from
Closed

[BasicTTI] When costing a scalarized cast, use distinct src and dest types #109325

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9ad1dfb
[BasicTTI] When costing a scalarized cast, use distinct src and dest …
preames Sep 19, 2024
bd5b63a
clang-format
preames Sep 19, 2024
fc36a5f
Update llvm/include/llvm/CodeGen/BasicTTIImpl.h
preames Sep 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions llvm/include/llvm/CodeGen/BasicTTIImpl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1186,8 +1186,10 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {

// Return the cost of multiple scalar invocation plus the cost of
// inserting and extracting the values.
return getScalarizationOverhead(DstVTy, /*Insert*/ true, /*Extract*/ true,
CostKind) +
return getScalarizationOverhead(SrcVTy, /*Insert*/ false,
/*Extract*/ true, CostKind) +
getScalarizationOverhead(DstVTy, /*Insert*/ true,
/*Extract*/ false, CostKind) +
Num * Cost;
}

Expand Down
128 changes: 64 additions & 64 deletions llvm/test/Analysis/CostModel/AArch64/cast.ll
View file
Open in desktop

Large diffs are not rendered by default.

24 changes: 12 additions & 12 deletions llvm/test/Analysis/CostModel/AArch64/fp-conversions-odd-vector-types.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ define <5 x i32> @fptoui_v5f32_to_v5i32(<5 x float> %in, ptr %dst) {

define <3 x i16> @fptoui_v3f32_to_v3i16(<3 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptoui_v3f32_to_v3i16'
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %conv = fptoui <3 x float> %in to <3 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %conv = fptoui <3 x float> %in to <3 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <3 x i16> %conv
;
%conv = fptoui <3 x float> %in to <3 x i16>
Expand All @@ -30,7 +30,7 @@ define <3 x i16> @fptoui_v3f32_to_v3i16(<3 x float> %in, ptr %dst) {

define <5 x i16> @fptoui_v5f32_to_v5i16(<5 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptoui_v5f32_to_v5i16'
; CHECK-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %conv = fptoui <5 x float> %in to <5 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %conv = fptoui <5 x float> %in to <5 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <5 x i16> %conv
;
%conv = fptoui <5 x float> %in to <5 x i16>
Expand All @@ -39,7 +39,7 @@ define <5 x i16> @fptoui_v5f32_to_v5i16(<5 x float> %in, ptr %dst) {

define <3 x i8> @fptoui_v3f32_to_v3i8(<3 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptoui_v3f32_to_v3i8'
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %conv = fptoui <3 x float> %in to <3 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %conv = fptoui <3 x float> %in to <3 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <3 x i8> %conv
;
%conv = fptoui <3 x float> %in to <3 x i8>
Expand All @@ -48,7 +48,7 @@ define <3 x i8> @fptoui_v3f32_to_v3i8(<3 x float> %in, ptr %dst) {

define <9 x i8> @fptoui_v9f32_to_v9i8(<9 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptoui_v9f32_to_v9i8'
; CHECK-NEXT: Cost Model: Found an estimated cost of 45 for instruction: %conv = fptoui <9 x float> %in to <9 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %conv = fptoui <9 x float> %in to <9 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <9 x i8> %conv
;
%conv = fptoui <9 x float> %in to <9 x i8>
Expand All @@ -75,7 +75,7 @@ define <6 x i32> @fptosi_v6f32_to_v6i32(<6 x float> %in, ptr %dst) {

define <3 x i16> @fptosi_v3f32_to_v3i16(<3 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptosi_v3f32_to_v3i16'
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %conv = fptosi <3 x float> %in to <3 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %conv = fptosi <3 x float> %in to <3 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <3 x i16> %conv
;
%conv = fptosi <3 x float> %in to <3 x i16>
Expand All @@ -84,7 +84,7 @@ define <3 x i16> @fptosi_v3f32_to_v3i16(<3 x float> %in, ptr %dst) {

define <6 x i16> @fptosi_v6f32_to_v6i16(<6 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptosi_v6f32_to_v6i16'
; CHECK-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %conv = fptosi <6 x float> %in to <6 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %conv = fptosi <6 x float> %in to <6 x i16>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <6 x i16> %conv
;
%conv = fptosi <6 x float> %in to <6 x i16>
Expand All @@ -93,7 +93,7 @@ define <6 x i16> @fptosi_v6f32_to_v6i16(<6 x float> %in, ptr %dst) {

define <3 x i8> @fptosi_v3f32_to_v3i8(<3 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptosi_v3f32_to_v3i8'
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %conv = fptosi <3 x float> %in to <3 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %conv = fptosi <3 x float> %in to <3 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <3 x i8> %conv
;
%conv = fptosi <3 x float> %in to <3 x i8>
Expand All @@ -102,7 +102,7 @@ define <3 x i8> @fptosi_v3f32_to_v3i8(<3 x float> %in, ptr %dst) {

define <5 x i8> @fptosi_v5f32_to_v5i8(<5 x float> %in, ptr %dst) {
; CHECK-LABEL: 'fptosi_v5f32_to_v5i8'
; CHECK-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %conv = fptosi <5 x float> %in to <5 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %conv = fptosi <5 x float> %in to <5 x i8>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <5 x i8> %conv
;
%conv = fptosi <5 x float> %in to <5 x i8>
Expand Down Expand Up @@ -138,7 +138,7 @@ define <3 x float> @uitofp_v3i16_to_v3f32(<3 x i16> %in, ptr %dst) {

define <7 x float> @uitofp_v7i16_to_v7f32(<7 x i16> %in, ptr %dst) {
; CHECK-LABEL: 'uitofp_v7i16_to_v7f32'
; CHECK-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %conv = uitofp <7 x i16> %in to <7 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 31 for instruction: %conv = uitofp <7 x i16> %in to <7 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <7 x float> %conv
;
%conv = uitofp <7 x i16> %in to <7 x float>
Expand All @@ -156,7 +156,7 @@ define <3 x float> @uitofp_v3i8_to_v3f32(<3 x i8> %in, ptr %dst) {

define <11 x float> @uitofp_v11i8_to_v11f32(<11 x i8> %in, ptr %dst) {
; CHECK-LABEL: 'uitofp_v11i8_to_v11f32'
; CHECK-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %conv = uitofp <11 x i8> %in to <11 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 49 for instruction: %conv = uitofp <11 x i8> %in to <11 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <11 x float> %conv
;
%conv = uitofp <11 x i8> %in to <11 x float>
Expand Down Expand Up @@ -192,7 +192,7 @@ define <3 x float> @sitofp_v3i16_to_v3f32(<3 x i16> %in, ptr %dst) {

define <6 x float> @sitofp_v6i16_to_v6f32(<6 x i16> %in, ptr %dst) {
; CHECK-LABEL: 'sitofp_v6i16_to_v6f32'
; CHECK-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %conv = sitofp <6 x i16> %in to <6 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %conv = sitofp <6 x i16> %in to <6 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <6 x float> %conv
;
%conv = sitofp <6 x i16> %in to <6 x float>
Expand All @@ -219,7 +219,7 @@ define <8 x float> @sitofp_v8i8_to_v8f32(<8 x i8> %in, ptr %dst) {

define <7 x float> @sitofp_v7i8_to_v7f32(<7 x i8> %in, ptr %dst) {
; CHECK-LABEL: 'sitofp_v7i8_to_v7f32'
; CHECK-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %conv = sitofp <7 x i8> %in to <7 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 31 for instruction: %conv = sitofp <7 x i8> %in to <7 x float>
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <7 x float> %conv
;
%conv = sitofp <7 x i8> %in to <7 x float>
Expand Down
Loading