[PATCH] D97470: [CostModel] Align the cost model for intrinsics for scalable/fixed-width vectors.
Sander de Smalen via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 31 07:02:21 PDT 2021
This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.
sdesmalen marked 4 inline comments as done.
Closed by commit rGb6d05297803d: [CostModel] Align the cost model for intrinsics for scalable/fixed-width… (authored by sdesmalen).
Changed prior to commit:
https://reviews.llvm.org/D97470?vs=327096&id=334430#toc
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D97470/new/
https://reviews.llvm.org/D97470
Files:
llvm/include/llvm/CodeGen/BasicTTIImpl.h
llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
llvm/test/Analysis/CostModel/AArch64/sve-math.ll
Index: llvm/test/Analysis/CostModel/AArch64/sve-math.ll
===================================================================
--- llvm/test/Analysis/CostModel/AArch64/sve-math.ll
+++ llvm/test/Analysis/CostModel/AArch64/sve-math.ll
@@ -29,7 +29,7 @@
define <vscale x 2 x double> @sqrt_v2f64(<vscale x 2 x double> %a) {
; THRU-LABEL: 'sqrt_v2f64'
-; THRU-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
+; THRU-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
; THRU-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <vscale x 2 x double> %r
;
; LATE-LABEL: 'sqrt_v2f64'
@@ -37,11 +37,11 @@
; LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <vscale x 2 x double> %r
;
; SIZE-LABEL: 'sqrt_v2f64'
-; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <vscale x 2 x double> %r
;
; SIZE_LATE-LABEL: 'sqrt_v2f64'
-; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
+; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %r = call <vscale x 2 x double> @llvm.sqrt.nxv2f64(<vscale x 2 x double> %a)
; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret <vscale x 2 x double> %r
;
%r = call <vscale x 2 x double> @llvm.sqrt.v2f64(<vscale x 2 x double> %a)
Index: llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
===================================================================
--- llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
+++ llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
@@ -105,8 +105,8 @@
define void @count_zeroes(<vscale x 4 x i32> %A) {
; CHECK-LABEL: 'count_zeroes'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %cttz = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %ctlz = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %cttz = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
%ctlz = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
%cttz = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> %A, i1 true)
ret void
Index: llvm/include/llvm/CodeGen/BasicTTIImpl.h
===================================================================
--- llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -1346,15 +1346,12 @@
return Cost;
}
}
- // TODO: Handle the remaining intrinsic with scalable vector type
- if (isa<ScalableVectorType>(RetTy))
- return BaseT::getIntrinsicInstrCost(ICA, CostKind);
// Assume that we need to scalarize this intrinsic.
// Compute the scalarization overhead based on Args for a vector
// intrinsic.
unsigned ScalarizationCost = std::numeric_limits<unsigned>::max();
- if (RetVF.isVector()) {
+ if (RetVF.isVector() && !RetVF.isScalable()) {
ScalarizationCost = 0;
if (!RetTy->isVoidTy())
ScalarizationCost +=
@@ -1399,6 +1396,12 @@
SmallVector<unsigned, 2> ISDs;
switch (IID) {
default: {
+ // Scalable vectors cannot be scalarized, so return Invalid.
+ if (isa<ScalableVectorType>(RetTy) || any_of(Tys, [](const Type *Ty) {
+ return isa<ScalableVectorType>(Ty);
+ }))
+ return InstructionCost::getInvalid();
+
// Assume that we need to scalarize this intrinsic.
InstructionCost ScalarizationCost = ScalarizationCostPassed;
unsigned ScalarCalls = 1;
@@ -1810,6 +1813,12 @@
// this will emit a costly libcall, adding call overhead and spills. Make it
// very expensive.
if (auto *RetVTy = dyn_cast<VectorType>(RetTy)) {
+ // Scalable vectors cannot be scalarized, so return Invalid.
+ if (isa<ScalableVectorType>(RetTy) || any_of(Tys, [](const Type *Ty) {
+ return isa<ScalableVectorType>(Ty);
+ }))
+ return InstructionCost::getInvalid();
+
unsigned ScalarizationCost = SkipScalarizationCost ?
ScalarizationCostPassed : getScalarizationOverhead(RetVTy, true, false);
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D97470.334430.patch
Type: text/x-patch
Size: 4960 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210331/2cdabfef/attachment.bin>
More information about the llvm-commits
mailing list