-
Notifications
You must be signed in to change notification settings - Fork 13.9k
[TargetLowering][RISCV] Allow scalable non-simple EVTs to be split even if the element type isn't a legal scalar type. #144007
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.
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
Conversation
…en if the element type isn't a legal scalar type. This fixes an inconsistency in i64 vector handling between RV32 and RV64. Even if i64 isn't legal as a scalar, we can should still be able to split a large i64 vector to get down to a legal vector type. We only need to give up if we need to split a vscale x 1 vector.
@llvm/pr-subscribers-llvm-analysis Author: Craig Topper (topperc) ChangesThis fixes an inconsistency in i64 vector handling between RV32 and RV64. Even if i64 isn't legal as a scalar, we can should still be able to split a large i64 vector to get down to a legal vector type. We only need to give up if we need to split a vscale x 1 vector. Patch is 1.12 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/144007.diff 5 Files Affected:
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index 935afaf9dd550..b1afdc2a3ac39 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -1030,7 +1030,7 @@ TargetLoweringBase::getTypeConversion(LLVMContext &Context, EVT VT) const {
// If type is to be expanded, split the vector.
// <4 x i140> -> <2 x i140>
if (LK.first == TypeExpandInteger) {
- if (VT.getVectorElementCount().isScalable())
+ if (NumElts.isScalable() && NumElts.getKnownMinValue() == 1)
return LegalizeKind(TypeScalarizeScalableVector, EltVT);
return LegalizeKind(TypeSplitVector,
VT.getHalfNumVectorElementsVT(Context));
diff --git a/llvm/test/Analysis/CostModel/RISCV/cast-half.ll b/llvm/test/Analysis/CostModel/RISCV/cast-half.ll
index 244c42cc94ba0..971b14467c0f8 100644
--- a/llvm/test/Analysis/CostModel/RISCV/cast-half.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/cast-half.ll
@@ -74,7 +74,7 @@ define void @fptosi() {
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %nxv64f16_nxv64i8 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i8>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64f16_nxv64i16 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i16>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %nxv64f16_nxv64i32 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i32>
-; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 103 for instruction: %nxv64f16_nxv64i64 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i64>
+; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 102 for instruction: %nxv64f16_nxv64i64 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i64>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %nxv64f16_nxv64i1 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i1>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -147,7 +147,7 @@ define void @fptosi() {
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 59 for instruction: %nxv64f16_nxv64i8 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i8>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64f16_nxv64i16 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i16>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 66 for instruction: %nxv64f16_nxv64i32 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i32>
-; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 103 for instruction: %nxv64f16_nxv64i64 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i64>
+; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 102 for instruction: %nxv64f16_nxv64i64 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i64>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 83 for instruction: %nxv64f16_nxv64i1 = fptosi <vscale x 64 x half> undef to <vscale x 64 x i1>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -440,7 +440,7 @@ define void @fptoui() {
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %nxv64f16_nxv64i8 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i8>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64f16_nxv64i16 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i16>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %nxv64f16_nxv64i32 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i32>
-; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 103 for instruction: %nxv64f16_nxv64i64 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i64>
+; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 102 for instruction: %nxv64f16_nxv64i64 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i64>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %nxv64f16_nxv64i1 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i1>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -513,7 +513,7 @@ define void @fptoui() {
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 59 for instruction: %nxv64f16_nxv64i8 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i8>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64f16_nxv64i16 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i16>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 66 for instruction: %nxv64f16_nxv64i32 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i32>
-; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 103 for instruction: %nxv64f16_nxv64i64 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i64>
+; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 102 for instruction: %nxv64f16_nxv64i64 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i64>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 83 for instruction: %nxv64f16_nxv64i1 = fptoui <vscale x 64 x half> undef to <vscale x 64 x i1>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -806,7 +806,7 @@ define void @sitofp() {
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %nxv64i8_nxv64f16 = sitofp <vscale x 64 x i8> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64i16_nxv64f16 = sitofp <vscale x 64 x i16> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %nxv64i32_nxv64f16 = sitofp <vscale x 64 x i32> undef to <vscale x 64 x half>
-; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 55 for instruction: %nxv64i64_nxv64f16 = sitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
+; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 54 for instruction: %nxv64i64_nxv64f16 = sitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %nxv64i1_nxv64f16 = sitofp <vscale x 64 x i1> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -879,7 +879,7 @@ define void @sitofp() {
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 83 for instruction: %nxv64i8_nxv64f16 = sitofp <vscale x 64 x i8> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64i16_nxv64f16 = sitofp <vscale x 64 x i16> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %nxv64i32_nxv64f16 = sitofp <vscale x 64 x i32> undef to <vscale x 64 x half>
-; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 55 for instruction: %nxv64i64_nxv64f16 = sitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
+; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 54 for instruction: %nxv64i64_nxv64f16 = sitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 99 for instruction: %nxv64i1_nxv64f16 = sitofp <vscale x 64 x i1> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -1172,7 +1172,7 @@ define void @uitofp() {
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %nxv64i8_nxv64f16 = uitofp <vscale x 64 x i8> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64i16_nxv64f16 = uitofp <vscale x 64 x i16> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %nxv64i32_nxv64f16 = uitofp <vscale x 64 x i32> undef to <vscale x 64 x half>
-; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 55 for instruction: %nxv64i64_nxv64f16 = uitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
+; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 54 for instruction: %nxv64i64_nxv64f16 = uitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %nxv64i1_nxv64f16 = uitofp <vscale x 64 x i1> undef to <vscale x 64 x half>
; RV32ZVFH-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
@@ -1245,7 +1245,7 @@ define void @uitofp() {
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 83 for instruction: %nxv64i8_nxv64f16 = uitofp <vscale x 64 x i8> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %nxv64i16_nxv64f16 = uitofp <vscale x 64 x i16> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %nxv64i32_nxv64f16 = uitofp <vscale x 64 x i32> undef to <vscale x 64 x half>
-; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 55 for instruction: %nxv64i64_nxv64f16 = uitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
+; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 54 for instruction: %nxv64i64_nxv64f16 = uitofp <vscale x 64 x i64> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 99 for instruction: %nxv64i1_nxv64f16 = uitofp <vscale x 64 x i1> undef to <vscale x 64 x half>
; RV32ZVFHMIN-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
diff --git a/llvm/test/Analysis/CostModel/RISCV/cast.ll b/llvm/test/Analysis/CostModel/RISCV/cast.ll
index e498ccc733040..bdd8540a2c475 100644
--- a/llvm/test/Analysis/CostModel/RISCV/cast.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/cast.ll
@@ -3,651 +3,328 @@
; RUN: opt < %s -mtriple=riscv64 -mattr=+v,+zfh,+zvfh -passes="print<cost-model>" -cost-kind=throughput -intrinsic-cost-strategy=type-based-intrinsic-cost 2>&1 -disable-output | FileCheck %s --check-prefixes=CHECK,RV64
define void @sext() {
-; RV32-LABEL: 'sext'
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_v2i16 = sext <2 x i8> undef to <2 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_v2i32 = sext <2 x i8> undef to <2 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_v2i64 = sext <2 x i8> undef to <2 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_v2i32 = sext <2 x i16> undef to <2 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_v2i64 = sext <2 x i16> undef to <2 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i32_v2i64 = sext <2 x i32> undef to <2 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i1_v2i8 = sext <2 x i1> undef to <2 x i8>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i1_v2i16 = sext <2 x i1> undef to <2 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i1_v2i32 = sext <2 x i1> undef to <2 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i1_v2i64 = sext <2 x i1> undef to <2 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i8_v2i16 = call <2 x i16> @llvm.vp.sext.v2i16.v2i8(<2 x i8> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i8_v2i32 = call <2 x i32> @llvm.vp.sext.v2i32.v2i8(<2 x i8> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i8_v2i64 = call <2 x i64> @llvm.vp.sext.v2i64.v2i8(<2 x i8> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i16_v2i32 = call <2 x i32> @llvm.vp.sext.v2i32.v2i16(<2 x i16> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i16_v2i64 = call <2 x i64> @llvm.vp.sext.v2i64.v2i16(<2 x i16> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v2i32_v2i64 = call <2 x i64> @llvm.vp.sext.v2i64.v2i32(<2 x i32> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v2i1_v2i8 = call <2 x i8> @llvm.vp.sext.v2i8.v2i1(<2 x i1> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v2i1_v2i16 = call <2 x i16> @llvm.vp.sext.v2i16.v2i1(<2 x i1> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v2i1_v2i32 = call <2 x i32> @llvm.vp.sext.v2i32.v2i1(<2 x i1> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v2i1_v2i64 = call <2 x i64> @llvm.vp.sext.v2i64.v2i1(<2 x i1> undef, <2 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_v4i16 = sext <4 x i8> undef to <4 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_v4i32 = sext <4 x i8> undef to <4 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i8_v4i64 = sext <4 x i8> undef to <4 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_v4i32 = sext <4 x i16> undef to <4 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i16_v4i64 = sext <4 x i16> undef to <4 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i32_v4i64 = sext <4 x i32> undef to <4 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i1_v4i8 = sext <4 x i1> undef to <4 x i8>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i1_v4i16 = sext <4 x i1> undef to <4 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i1_v4i32 = sext <4 x i1> undef to <4 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v4i1_v4i64 = sext <4 x i1> undef to <4 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v4i8_v4i16 = call <4 x i16> @llvm.vp.sext.v4i16.v4i8(<4 x i8> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v4i8_v4i32 = call <4 x i32> @llvm.vp.sext.v4i32.v4i8(<4 x i8> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i8_v4i64 = call <4 x i64> @llvm.vp.sext.v4i64.v4i8(<4 x i8> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v4i16_v4i32 = call <4 x i32> @llvm.vp.sext.v4i32.v4i16(<4 x i16> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i16_v4i64 = call <4 x i64> @llvm.vp.sext.v4i64.v4i16(<4 x i16> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i32_v4i64 = call <4 x i64> @llvm.vp.sext.v4i64.v4i32(<4 x i32> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i1_v4i8 = call <4 x i8> @llvm.vp.sext.v4i8.v4i1(<4 x i1> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i1_v4i16 = call <4 x i16> @llvm.vp.sext.v4i16.v4i1(<4 x i1> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v4i1_v4i32 = call <4 x i32> @llvm.vp.sext.v4i32.v4i1(<4 x i1> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v4i1_v4i64 = call <4 x i64> @llvm.vp.sext.v4i64.v4i1(<4 x i1> undef, <4 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_v8i16 = sext <8 x i8> undef to <8 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i8_v8i32 = sext <8 x i8> undef to <8 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8i8_v8i64 = sext <8 x i8> undef to <8 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i16_v8i32 = sext <8 x i16> undef to <8 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8i16_v8i64 = sext <8 x i16> undef to <8 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8i32_v8i64 = sext <8 x i32> undef to <8 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i1_v8i8 = sext <8 x i1> undef to <8 x i8>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i1_v8i16 = sext <8 x i1> undef to <8 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8i1_v8i32 = sext <8 x i1> undef to <8 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v8i1_v8i64 = sext <8 x i1> undef to <8 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %vp_v8i8_v8i16 = call <8 x i16> @llvm.vp.sext.v8i16.v8i8(<8 x i8> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v8i8_v8i32 = call <8 x i32> @llvm.vp.sext.v8i32.v8i8(<8 x i8> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v8i8_v8i64 = call <8 x i64> @llvm.vp.sext.v8i64.v8i8(<8 x i8> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v8i16_v8i32 = call <8 x i32> @llvm.vp.sext.v8i32.v8i16(<8 x i16> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v8i16_v8i64 = call <8 x i64> @llvm.vp.sext.v8i64.v8i16(<8 x i16> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v8i32_v8i64 = call <8 x i64> @llvm.vp.sext.v8i64.v8i32(<8 x i32> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v8i1_v8i8 = call <8 x i8> @llvm.vp.sext.v8i8.v8i1(<8 x i1> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v8i1_v8i16 = call <8 x i16> @llvm.vp.sext.v8i16.v8i1(<8 x i1> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v8i1_v8i32 = call <8 x i32> @llvm.vp.sext.v8i32.v8i1(<8 x i1> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %vp_v8i1_v8i64 = call <8 x i64> @llvm.vp.sext.v8i64.v8i1(<8 x i1> undef, <8 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16i8_v16i16 = sext <16 x i8> undef to <16 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v16i8_v16i32 = sext <16 x i8> undef to <16 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16i8_v16i64 = sext <16 x i8> undef to <16 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v16i16_v16i32 = sext <16 x i16> undef to <16 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16i16_v16i64 = sext <16 x i16> undef to <16 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16i32_v16i64 = sext <16 x i32> undef to <16 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16i1_v16i8 = sext <16 x i1> undef to <16 x i8>
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v16i1_v16i16 = sext <16 x i1> undef to <16 x i16>
-; RV32-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16i1_v16i32 = sext <16 x i1> undef to <16 x i32>
-; RV32-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %v16i1_v16i64 = sext <16 x i1> undef to <16 x i64>
-; RV32-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %vp_v16i8_v16i16 = call <16 x i16> @llvm.vp.sext.v16i16.v16i8(<16 x i8> undef, <16 x i1> undef, i32 undef)
-; RV32-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %vp_v16i8_v16i3...
[truncated]
|
You can test this locally with the following command:git diff -U0 --pickaxe-regex -S '([^a-zA-Z0-9#_-]undef[^a-zA-Z0-9_-]|UndefValue::get)' 'HEAD~1' HEAD llvm/lib/CodeGen/TargetLoweringBase.cpp llvm/test/Analysis/CostModel/RISCV/cast-half.ll llvm/test/Analysis/CostModel/RISCV/cast.ll llvm/test/Analysis/CostModel/RISCV/cmp.ll llvm/test/CodeGen/RISCV/rvv/vsub-sdnode.ll The following files introduce new uses of undef:
Undef is now deprecated and should only be used in the rare cases where no replacement is possible. For example, a load of uninitialized memory yields In tests, avoid using For example, this is considered a bad practice: define void @fn() {
...
br i1 undef, ...
} Please use the following instead: define void @fn(i1 %cond) {
...
br i1 %cond, ...
} Please refer to the Undefined Behavior Manual for more information. |
Ping |
This fixes an inconsistency in i64 vector handling between RV32 and RV64. Even if i64 isn't legal as a scalar, we can should still be able to split a large i64 vector to get down to a legal vector type. We only need to give up if we need to split a vscale x 1 vector.