[PATCH] D14743: [ARM] Don't pessimize i32 vselect.

[Charlie Turner via llvm-commits]

chatur01 created this revision.
chatur01 added reviewers: mzolotukhin, aschwaighofer, jmolloy, mcrosier.
chatur01 added a subscriber: llvm-commits.
chatur01 set the repository for this revision to rL LLVM.
Herald added subscribers: rengolin, aemerson.

The underlying issues surrounding codegen for 32-bit vselects have been resolved.  The pessimistic costs for 64-bit vselects remain due to the bad
scalarization that is still happening there.

I tested this on A57 in T32, A32 and A64 modes. I saw no regressions, and some improvements.

From my benchmarks, I saw these improvements in A57 (T32)
spec.cpu2000.ref.177_mesa   5.95%
lnt.SingleSource/Benchmarks/Shootout/strcat 	12.93%
lnt.MultiSource/Benchmarks/MiBench/telecomm-CRC32/telecomm-CRC32 	11.89% 	

I also measured A57 A32 and A9 T32 and found no performance regressions. I see much bigger wins in third-party benchmarks with this change.

Repository:
  rL LLVM

http://reviews.llvm.org/D14743

Files:
  lib/Target/ARM/ARMTargetTransformInfo.cpp
  test/CodeGen/ARM/vselect_imax.ll

Index: test/CodeGen/ARM/vselect_imax.ll
===================================================================
--- test/CodeGen/ARM/vselect_imax.ll
+++ test/CodeGen/ARM/vselect_imax.ll
@@ -10,8 +10,6 @@
     ret void
 }
 
-; We adjusted the cost model of the following selects. When we improve code
-; lowering we also need to adjust the cost.
 %T0_10 = type <16 x i16>
 %T1_10 = type <16 x i1>
 ; CHECK-LABEL: func_blend10:
@@ -23,7 +21,7 @@
 ; CHECK: vmin.s16
 ; CHECK: vmin.s16
 ; COST: func_blend10
-; COST: cost of 40 {{.*}} select
+; COST: cost of 2 {{.*}} select
   %r = select %T1_10 %c, %T0_10 %v0, %T0_10 %v1
   store %T0_10 %r, %T0_10* %storeaddr
   ret void
@@ -39,7 +37,7 @@
 ; CHECK: vmin.s32
 ; CHECK: vmin.s32
 ; COST: func_blend14
-; COST: cost of 41 {{.*}} select
+; COST: cost of 2 {{.*}} select
   %r = select %T1_14 %c, %T0_14 %v0, %T0_14 %v1
   store %T0_14 %r, %T0_14* %storeaddr
   ret void
@@ -55,11 +53,14 @@
   %v1 = load %T0_15, %T0_15* %loadaddr2
   %c = icmp slt %T0_15 %v0, %v1
 ; COST: func_blend15
-; COST: cost of 82 {{.*}} select
+; COST: cost of 4 {{.*}} select
   %r = select %T1_15 %c, %T0_15 %v0, %T0_15 %v1
   store %T0_15 %r, %T0_15* %storeaddr
   ret void
 }
+
+; We adjusted the cost model of the following selects. When we improve code
+; lowering we also need to adjust the cost.
 %T0_18 = type <4 x i64>
 %T1_18 = type <4 x i1>
 ; CHECK-LABEL: func_blend18:
Index: lib/Target/ARM/ARMTargetTransformInfo.cpp
===================================================================
--- lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -274,9 +274,6 @@
   if (ST->hasNEON() && ValTy->isVectorTy() && ISD == ISD::SELECT) {
     // Lowering of some vector selects is currently far from perfect.
     static const TypeConversionCostTblEntry NEONVectorSelectTbl[] = {
-      { ISD::SELECT, MVT::v16i1, MVT::v16i16, 2*16 + 1 + 3*1 + 4*1 },
-      { ISD::SELECT, MVT::v8i1, MVT::v8i32, 4*8 + 1*3 + 1*4 + 1*2 },
-      { ISD::SELECT, MVT::v16i1, MVT::v16i32, 4*16 + 1*6 + 1*8 + 1*4 },
       { ISD::SELECT, MVT::v4i1, MVT::v4i64, 4*4 + 1*2 + 1 },
       { ISD::SELECT, MVT::v8i1, MVT::v8i64, 50 },
       { ISD::SELECT, MVT::v16i1, MVT::v16i64, 100 }


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D14743.40292.patch
Type: text/x-patch
Size: 2221 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151117/91225165/attachment.bin>