Change Thumb2 jumptable codegen to one that uses two level jumps:
Before:
adr r12, #LJTI3_0_0
ldr pc, [r12, +r0, lsl #2]
LJTI3_0_0:
.long LBB3_24
.long LBB3_30
.long LBB3_31
.long LBB3_32
After:
adr r12, #LJTI3_0_0
add pc, r12, +r0, lsl #2
LJTI3_0_0:
b.w LBB3_24
b.w LBB3_30
b.w LBB3_31
b.w LBB3_32
This has several advantages.
1. This will make it easier to optimize this to a TBB / TBH instruction +
(smaller) table.
2. This eliminate the need for ugly asm printer hack to force the address
into thumb addresses (bit 0 is one).
3. Same codegen for pic and non-pic.
4. This eliminate the need to align the table so constantpool island pass
won't have to over-estimate the size.
Based on my calculation, the later is probably slightly faster as well since
ldr pc with shifter address is very slow. That is, it should be a win as long
as the HW implementation can do a reasonable job of branch predict the second
branch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77024 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index b03c308..0b61d4e 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -252,9 +252,11 @@
// ...likewise if it ends with a branch table followed by an unconditional
// branch. The branch folder can create these, and we must get rid of them for
// correctness of Thumb constant islands.
- if (((SecondLastOpc == getOpcode(ARMII::BR_JTr)) ||
- (SecondLastOpc == getOpcode(ARMII::BR_JTm)) ||
- (SecondLastOpc == getOpcode(ARMII::BR_JTadd))) &&
+ if ((SecondLastOpc == ARM::BR_JTr ||
+ SecondLastOpc == ARM::BR_JTm ||
+ SecondLastOpc == ARM::BR_JTadd ||
+ SecondLastOpc == ARM::tBR_JTr ||
+ SecondLastOpc == ARM::t2BR_JT) &&
(LastOpc == getOpcode(ARMII::B))) {
I = LastInst;
if (AllowModify)
@@ -451,9 +453,7 @@
case ARM::BR_JTr:
case ARM::BR_JTm:
case ARM::BR_JTadd:
- case ARM::t2BR_JTr:
- case ARM::t2BR_JTm:
- case ARM::t2BR_JTadd: {
+ case ARM::t2BR_JT: {
// These are jumptable branches, i.e. a branch followed by an inlined
// jumptable. The size is 4 + 4 * number of entries.
unsigned NumOps = TID.getNumOperands();