[RISCV] Generalize the (ADD (SLLI X, 32), X) special case in constant… #66931
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-backend-risc-v Changes… materialization. We don't have to limit ourselves to a shift amount of 32. We can support other shift amounts that make the upper 32 bits line up. Full diff: https://github.com/llvm/llvm-project/pull/66931.diff 3 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index e6d0346c45e8d54..0e3e211fbf73b85 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -195,28 +195,41 @@ static SDValue selectImm(SelectionDAG *CurDAG, const SDLoc &DL, const MVT VT,
RISCVMatInt::InstSeq Seq =
RISCVMatInt::generateInstSeq(Imm, Subtarget.getFeatureBits());
- // See if we can create this constant as (ADD (SLLI X, 32), X) where X is at
+ // See if we can create this constant as (ADD (SLLI X, C), X) where X is at
// worst an LUI+ADDIW. This will require an extra register, but avoids a
// constant pool.
// If we have Zba we can use (ADD_UW X, (SLLI X, 32)) to handle cases where
// low and high 32 bits are the same and bit 31 and 63 are set.
if (Seq.size() > 3) {
int64_t LoVal = SignExtend64<32>(Imm);
- int64_t HiVal = SignExtend64<32>(((uint64_t)Imm - (uint64_t)LoVal) >> 32);
- if (LoVal == HiVal ||
- (Subtarget.hasStdExtZba() && Lo_32(Imm) == Hi_32(Imm))) {
- RISCVMatInt::InstSeq SeqLo =
- RISCVMatInt::generateInstSeq(LoVal, Subtarget.getFeatureBits());
- if ((SeqLo.size() + 2) < Seq.size()) {
- SDValue Lo = selectImmSeq(CurDAG, DL, VT, SeqLo);
-
- SDValue SLLI = SDValue(
- CurDAG->getMachineNode(RISCV::SLLI, DL, VT, Lo,
- CurDAG->getTargetConstant(32, DL, VT)),
- 0);
- // Prefer ADD when possible.
- unsigned AddOpc = (LoVal == HiVal) ? RISCV::ADD : RISCV::ADD_UW;
- return SDValue(CurDAG->getMachineNode(AddOpc, DL, VT, Lo, SLLI), 0);
+ if (LoVal != 0) {
+ // Subtract the LoVal to emulate the effect of the final add.
+ uint64_t Tmp = (uint64_t)Imm - (uint64_t)LoVal;
+
+ // Use trailing zero counts to figure how far we need to shift LoVal to
+ // line up with the remaining constant.
+ unsigned TzLo = llvm::countr_zero((uint64_t)LoVal);
+ unsigned TzHi = llvm::countr_zero(Tmp);
+ assert(TzLo < 32 && TzHi >= 32);
+ unsigned ShiftAmt = TzHi - TzLo;
+
+ bool MatchedAdd = Tmp == ((uint64_t)LoVal << ShiftAmt);
+
+ if (MatchedAdd ||
+ (Subtarget.hasStdExtZba() && Lo_32(Imm) == Hi_32(Imm))) {
+ RISCVMatInt::InstSeq SeqLo =
+ RISCVMatInt::generateInstSeq(LoVal, Subtarget.getFeatureBits());
+ if ((SeqLo.size() + 2) < Seq.size()) {
+ SDValue Lo = selectImmSeq(CurDAG, DL, VT, SeqLo);
+
+ SDValue SLLI = SDValue(
+ CurDAG->getMachineNode(RISCV::SLLI, DL, VT, Lo,
+ CurDAG->getTargetConstant(32, DL, VT)),
+ 0);
+ // Prefer ADD when possible.
+ unsigned AddOpc = MatchedAdd ? RISCV::ADD : RISCV::ADD_UW;
+ return SDValue(CurDAG->getMachineNode(AddOpc, DL, VT, Lo, SLLI), 0);
+ }
}
}
}
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index f1cea6c6756f4fc..fef2a61b9250221 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -4954,24 +4954,38 @@ static SDValue lowerConstant(SDValue Op, SelectionDAG &DAG,
if (Seq.size() <= Subtarget.getMaxBuildIntsCost())
return Op;
- // Special case. See if we can build the constant as (ADD (SLLI X, 32), X) do
+ // Optimizations below are disabled for opt size. If we're optimizing for
+ // size, use a constant pool.
+ if (DAG.shouldOptForSize())
+ return SDValue();
+
+ // Special case. See if we can build the constant as (ADD (SLLI X, C), X) do
// that if it will avoid a constant pool.
// It will require an extra temporary register though.
// If we have Zba we can use (ADD_UW X, (SLLI X, 32)) to handle cases where
// low and high 32 bits are the same and bit 31 and 63 are set.
- if (!DAG.shouldOptForSize()) {
- int64_t LoVal = SignExtend64<32>(Imm);
- int64_t HiVal = SignExtend64<32>(((uint64_t)Imm - (uint64_t)LoVal) >> 32);
- if (LoVal == HiVal ||
- (Subtarget.hasStdExtZba() && Lo_32(Imm) == Hi_32(Imm))) {
- RISCVMatInt::InstSeq SeqLo =
- RISCVMatInt::generateInstSeq(LoVal, Subtarget.getFeatureBits());
- if ((SeqLo.size() + 2) <= Subtarget.getMaxBuildIntsCost())
- return Op;
- }
+ int64_t LoVal = SignExtend64<32>(Imm);
+ if (LoVal == 0)
+ return SDValue();
+
+ // Subtract the LoVal to emulate the effect of the final add.
+ uint64_t Tmp = (uint64_t)Imm - (uint64_t)LoVal;
+
+ // Use trailing zero counts to figure how far we need to shift LoVal to line
+ // up with the remaining constant.
+ unsigned TzLo = llvm::countr_zero((uint64_t)LoVal);
+ unsigned TzHi = llvm::countr_zero(Tmp);
+ assert(TzLo < 32 && TzHi >= 32);
+ unsigned ShiftAmt = TzHi - TzLo;
+
+ if (Tmp == ((uint64_t)LoVal << ShiftAmt) ||
+ (Subtarget.hasStdExtZba() && Lo_32(Imm) == Hi_32(Imm))) {
+ RISCVMatInt::InstSeq SeqLo =
+ RISCVMatInt::generateInstSeq(LoVal, Subtarget.getFeatureBits());
+ if ((SeqLo.size() + 2) <= Subtarget.getMaxBuildIntsCost())
+ return Op;
}
- // Expand to a constant pool using the default expansion code.
return SDValue();
}
diff --git a/llvm/test/CodeGen/RISCV/imm.ll b/llvm/test/CodeGen/RISCV/imm.ll
index 4f9cf1d947d5c35..205693847e64a40 100644
--- a/llvm/test/CodeGen/RISCV/imm.ll
+++ b/llvm/test/CodeGen/RISCV/imm.ll
@@ -1058,47 +1058,42 @@ define i64 @imm_end_xori_1() nounwind {
;
; RV64I-LABEL: imm_end_xori_1:
; RV64I: # %bb.0:
-; RV64I-NEXT: li a0, -1
-; RV64I-NEXT: slli a0, a0, 36
-; RV64I-NEXT: addi a0, a0, 1
-; RV64I-NEXT: slli a0, a0, 25
-; RV64I-NEXT: addi a0, a0, -1
+; RV64I-NEXT: lui a0, 8192
+; RV64I-NEXT: addiw a0, a0, -1
+; RV64I-NEXT: slli a1, a0, 32
+; RV64I-NEXT: add a0, a0, a1
; RV64I-NEXT: ret
;
; RV64IZBA-LABEL: imm_end_xori_1:
; RV64IZBA: # %bb.0:
-; RV64IZBA-NEXT: li a0, -1
-; RV64IZBA-NEXT: slli a0, a0, 36
-; RV64IZBA-NEXT: addi a0, a0, 1
-; RV64IZBA-NEXT: slli a0, a0, 25
-; RV64IZBA-NEXT: addi a0, a0, -1
+; RV64IZBA-NEXT: lui a0, 8192
+; RV64IZBA-NEXT: addiw a0, a0, -1
+; RV64IZBA-NEXT: slli a1, a0, 32
+; RV64IZBA-NEXT: add a0, a0, a1
; RV64IZBA-NEXT: ret
;
; RV64IZBB-LABEL: imm_end_xori_1:
; RV64IZBB: # %bb.0:
-; RV64IZBB-NEXT: li a0, -1
-; RV64IZBB-NEXT: slli a0, a0, 36
-; RV64IZBB-NEXT: addi a0, a0, 1
-; RV64IZBB-NEXT: slli a0, a0, 25
-; RV64IZBB-NEXT: addi a0, a0, -1
+; RV64IZBB-NEXT: lui a0, 8192
+; RV64IZBB-NEXT: addiw a0, a0, -1
+; RV64IZBB-NEXT: slli a1, a0, 32
+; RV64IZBB-NEXT: add a0, a0, a1
; RV64IZBB-NEXT: ret
;
; RV64IZBS-LABEL: imm_end_xori_1:
; RV64IZBS: # %bb.0:
-; RV64IZBS-NEXT: li a0, -1
-; RV64IZBS-NEXT: slli a0, a0, 36
-; RV64IZBS-NEXT: addi a0, a0, 1
-; RV64IZBS-NEXT: slli a0, a0, 25
-; RV64IZBS-NEXT: addi a0, a0, -1
+; RV64IZBS-NEXT: lui a0, 8192
+; RV64IZBS-NEXT: addiw a0, a0, -1
+; RV64IZBS-NEXT: slli a1, a0, 32
+; RV64IZBS-NEXT: add a0, a0, a1
; RV64IZBS-NEXT: ret
;
; RV64IXTHEADBB-LABEL: imm_end_xori_1:
; RV64IXTHEADBB: # %bb.0:
-; RV64IXTHEADBB-NEXT: li a0, -1
-; RV64IXTHEADBB-NEXT: slli a0, a0, 36
-; RV64IXTHEADBB-NEXT: addi a0, a0, 1
-; RV64IXTHEADBB-NEXT: slli a0, a0, 25
-; RV64IXTHEADBB-NEXT: addi a0, a0, -1
+; RV64IXTHEADBB-NEXT: lui a0, 8192
+; RV64IXTHEADBB-NEXT: addiw a0, a0, -1
+; RV64IXTHEADBB-NEXT: slli a1, a0, 32
+; RV64IXTHEADBB-NEXT: add a0, a0, a1
; RV64IXTHEADBB-NEXT: ret
ret i64 -2305843009180139521 ; 0xE000_0000_01FF_FFFF
}
@@ -1172,64 +1167,44 @@ define i64 @imm_2reg_1() nounwind {
; RV32I-NEXT: lui a1, 983040
; RV32I-NEXT: ret
;
-; RV64-NOPOOL-LABEL: imm_2reg_1:
-; RV64-NOPOOL: # %bb.0:
-; RV64-NOPOOL-NEXT: li a0, -1
-; RV64-NOPOOL-NEXT: slli a0, a0, 35
-; RV64-NOPOOL-NEXT: addi a0, a0, 9
-; RV64-NOPOOL-NEXT: slli a0, a0, 13
-; RV64-NOPOOL-NEXT: addi a0, a0, 837
-; RV64-NOPOOL-NEXT: slli a0, a0, 12
-; RV64-NOPOOL-NEXT: addi a0, a0, 1656
-; RV64-NOPOOL-NEXT: ret
-;
-; RV64I-POOL-LABEL: imm_2reg_1:
-; RV64I-POOL: # %bb.0:
-; RV64I-POOL-NEXT: lui a0, %hi(.LCPI27_0)
-; RV64I-POOL-NEXT: ld a0, %lo(.LCPI27_0)(a0)
-; RV64I-POOL-NEXT: ret
+; RV64I-LABEL: imm_2reg_1:
+; RV64I: # %bb.0:
+; RV64I-NEXT: lui a0, 74565
+; RV64I-NEXT: addiw a0, a0, 1656
+; RV64I-NEXT: slli a1, a0, 32
+; RV64I-NEXT: add a0, a0, a1
+; RV64I-NEXT: ret
;
; RV64IZBA-LABEL: imm_2reg_1:
; RV64IZBA: # %bb.0:
-; RV64IZBA-NEXT: li a0, -1
-; RV64IZBA-NEXT: slli a0, a0, 35
-; RV64IZBA-NEXT: addi a0, a0, 9
-; RV64IZBA-NEXT: slli a0, a0, 13
-; RV64IZBA-NEXT: addi a0, a0, 837
-; RV64IZBA-NEXT: slli a0, a0, 12
-; RV64IZBA-NEXT: addi a0, a0, 1656
+; RV64IZBA-NEXT: lui a0, 74565
+; RV64IZBA-NEXT: addiw a0, a0, 1656
+; RV64IZBA-NEXT: slli a1, a0, 32
+; RV64IZBA-NEXT: add a0, a0, a1
; RV64IZBA-NEXT: ret
;
; RV64IZBB-LABEL: imm_2reg_1:
; RV64IZBB: # %bb.0:
-; RV64IZBB-NEXT: li a0, -1
-; RV64IZBB-NEXT: slli a0, a0, 35
-; RV64IZBB-NEXT: addi a0, a0, 9
-; RV64IZBB-NEXT: slli a0, a0, 13
-; RV64IZBB-NEXT: addi a0, a0, 837
-; RV64IZBB-NEXT: slli a0, a0, 12
-; RV64IZBB-NEXT: addi a0, a0, 1656
+; RV64IZBB-NEXT: lui a0, 74565
+; RV64IZBB-NEXT: addiw a0, a0, 1656
+; RV64IZBB-NEXT: slli a1, a0, 32
+; RV64IZBB-NEXT: add a0, a0, a1
; RV64IZBB-NEXT: ret
;
; RV64IZBS-LABEL: imm_2reg_1:
; RV64IZBS: # %bb.0:
; RV64IZBS-NEXT: lui a0, 74565
; RV64IZBS-NEXT: addiw a0, a0, 1656
-; RV64IZBS-NEXT: bseti a0, a0, 60
-; RV64IZBS-NEXT: bseti a0, a0, 61
-; RV64IZBS-NEXT: bseti a0, a0, 62
-; RV64IZBS-NEXT: bseti a0, a0, 63
+; RV64IZBS-NEXT: slli a1, a0, 32
+; RV64IZBS-NEXT: add a0, a0, a1
; RV64IZBS-NEXT: ret
;
; RV64IXTHEADBB-LABEL: imm_2reg_1:
; RV64IXTHEADBB: # %bb.0:
-; RV64IXTHEADBB-NEXT: li a0, -1
-; RV64IXTHEADBB-NEXT: slli a0, a0, 35
-; RV64IXTHEADBB-NEXT: addi a0, a0, 9
-; RV64IXTHEADBB-NEXT: slli a0, a0, 13
-; RV64IXTHEADBB-NEXT: addi a0, a0, 837
-; RV64IXTHEADBB-NEXT: slli a0, a0, 12
-; RV64IXTHEADBB-NEXT: addi a0, a0, 1656
+; RV64IXTHEADBB-NEXT: lui a0, 74565
+; RV64IXTHEADBB-NEXT: addiw a0, a0, 1656
+; RV64IXTHEADBB-NEXT: slli a1, a0, 32
+; RV64IXTHEADBB-NEXT: add a0, a0, a1
; RV64IXTHEADBB-NEXT: ret
ret i64 -1152921504301427080 ; 0xF000_0000_1234_5678
}
@@ -1722,43 +1697,28 @@ define i64 @imm_neg_9223372034778874949() {
; RV32I-NEXT: lui a1, 524288
; RV32I-NEXT: ret
;
-; RV64-NOPOOL-LABEL: imm_neg_9223372034778874949:
-; RV64-NOPOOL: # %bb.0:
-; RV64-NOPOOL-NEXT: li a0, -1
-; RV64-NOPOOL-NEXT: slli a0, a0, 37
-; RV64-NOPOOL-NEXT: addi a0, a0, 31
-; RV64-NOPOOL-NEXT: slli a0, a0, 12
-; RV64-NOPOOL-NEXT: addi a0, a0, -273
-; RV64-NOPOOL-NEXT: slli a0, a0, 14
-; RV64-NOPOOL-NEXT: addi a0, a0, -1093
-; RV64-NOPOOL-NEXT: ret
-;
-; RV64I-POOL-LABEL: imm_neg_9223372034778874949:
-; RV64I-POOL: # %bb.0:
-; RV64I-POOL-NEXT: lui a0, %hi(.LCPI38_0)
-; RV64I-POOL-NEXT: ld a0, %lo(.LCPI38_0)(a0)
-; RV64I-POOL-NEXT: ret
+; RV64I-LABEL: imm_neg_9223372034778874949:
+; RV64I: # %bb.0:
+; RV64I-NEXT: lui a0, 506812
+; RV64I-NEXT: addiw a0, a0, -1093
+; RV64I-NEXT: slli a1, a0, 32
+; RV64I-NEXT: add a0, a0, a1
+; RV64I-NEXT: ret
;
; RV64IZBA-LABEL: imm_neg_9223372034778874949:
; RV64IZBA: # %bb.0:
-; RV64IZBA-NEXT: li a0, -1
-; RV64IZBA-NEXT: slli a0, a0, 37
-; RV64IZBA-NEXT: addi a0, a0, 31
-; RV64IZBA-NEXT: slli a0, a0, 12
-; RV64IZBA-NEXT: addi a0, a0, -273
-; RV64IZBA-NEXT: slli a0, a0, 14
-; RV64IZBA-NEXT: addi a0, a0, -1093
+; RV64IZBA-NEXT: lui a0, 506812
+; RV64IZBA-NEXT: addiw a0, a0, -1093
+; RV64IZBA-NEXT: slli a1, a0, 32
+; RV64IZBA-NEXT: add a0, a0, a1
; RV64IZBA-NEXT: ret
;
; RV64IZBB-LABEL: imm_neg_9223372034778874949:
; RV64IZBB: # %bb.0:
-; RV64IZBB-NEXT: li a0, -1
-; RV64IZBB-NEXT: slli a0, a0, 37
-; RV64IZBB-NEXT: addi a0, a0, 31
-; RV64IZBB-NEXT: slli a0, a0, 12
-; RV64IZBB-NEXT: addi a0, a0, -273
-; RV64IZBB-NEXT: slli a0, a0, 14
-; RV64IZBB-NEXT: addi a0, a0, -1093
+; RV64IZBB-NEXT: lui a0, 506812
+; RV64IZBB-NEXT: addiw a0, a0, -1093
+; RV64IZBB-NEXT: slli a1, a0, 32
+; RV64IZBB-NEXT: add a0, a0, a1
; RV64IZBB-NEXT: ret
;
; RV64IZBS-LABEL: imm_neg_9223372034778874949:
@@ -1770,13 +1730,10 @@ define i64 @imm_neg_9223372034778874949() {
;
; RV64IXTHEADBB-LABEL: imm_neg_9223372034778874949:
; RV64IXTHEADBB: # %bb.0:
-; RV64IXTHEADBB-NEXT: li a0, -1
-; RV64IXTHEADBB-NEXT: slli a0, a0, 37
-; RV64IXTHEADBB-NEXT: addi a0, a0, 31
-; RV64IXTHEADBB-NEXT: slli a0, a0, 12
-; RV64IXTHEADBB-NEXT: addi a0, a0, -273
-; RV64IXTHEADBB-NEXT: slli a0, a0, 14
-; RV64IXTHEADBB-NEXT: addi a0, a0, -1093
+; RV64IXTHEADBB-NEXT: lui a0, 506812
+; RV64IXTHEADBB-NEXT: addiw a0, a0, -1093
+; RV64IXTHEADBB-NEXT: slli a1, a0, 32
+; RV64IXTHEADBB-NEXT: add a0, a0, a1
; RV64IXTHEADBB-NEXT: ret
ret i64 -9223372034778874949 ; 0x800000007bbbbbbb
}
@@ -1955,29 +1912,26 @@ define i64 @imm_9223372034904144827() {
;
; RV64I-LABEL: imm_9223372034904144827:
; RV64I: # %bb.0:
-; RV64I-NEXT: lui a0, 1048343
-; RV64I-NEXT: addiw a0, a0, 1911
-; RV64I-NEXT: slli a0, a0, 12
-; RV64I-NEXT: addi a0, a0, 1911
-; RV64I-NEXT: srli a0, a0, 1
+; RV64I-NEXT: lui a0, 572348
+; RV64I-NEXT: addiw a0, a0, -1093
+; RV64I-NEXT: slli a1, a0, 32
+; RV64I-NEXT: add a0, a0, a1
; RV64I-NEXT: ret
;
; RV64IZBA-LABEL: imm_9223372034904144827:
; RV64IZBA: # %bb.0:
-; RV64IZBA-NEXT: lui a0, 1048343
-; RV64IZBA-NEXT: addiw a0, a0, 1911
-; RV64IZBA-NEXT: slli a0, a0, 12
-; RV64IZBA-NEXT: addi a0, a0, 1911
-; RV64IZBA-NEXT: srli a0, a0, 1
+; RV64IZBA-NEXT: lui a0, 572348
+; RV64IZBA-NEXT: addiw a0, a0, -1093
+; RV64IZBA-NEXT: slli a1, a0, 32
+; RV64IZBA-NEXT: add a0, a0, a1
; RV64IZBA-NEXT: ret
;
; RV64IZBB-LABEL: imm_9223372034904144827:
; RV64IZBB: # %bb.0:
-; RV64IZBB-NEXT: lui a0, 1048343
-; RV64IZBB-NEXT: addiw a0, a0, 1911
-; RV64IZBB-NEXT: slli a0, a0, 12
-; RV64IZBB-NEXT: addi a0, a0, 1911
-; RV64IZBB-NEXT: srli a0, a0, 1
+; RV64IZBB-NEXT: lui a0, 572348
+; RV64IZBB-NEXT: addiw a0, a0, -1093
+; RV64IZBB-NEXT: slli a1, a0, 32
+; RV64IZBB-NEXT: add a0, a0, a1
; RV64IZBB-NEXT: ret
;
; RV64IZBS-LABEL: imm_9223372034904144827:
@@ -1989,11 +1943,10 @@ define i64 @imm_9223372034904144827() {
;
; RV64IXTHEADBB-LABEL: imm_9223372034904144827:
; RV64IXTHEADBB: # %bb.0:
-; RV64IXTHEADBB-NEXT: lui a0, 1048343
-; RV64IXTHEADBB-NEXT: addiw a0, a0, 1911
-; RV64IXTHEADBB-NEXT: slli a0, a0, 12
-; RV64IXTHEADBB-NEXT: addi a0, a0, 1911
-; RV64IXTHEADBB-NEXT: srli a0, a0, 1
+; RV64IXTHEADBB-NEXT: lui a0, 572348
+; RV64IXTHEADBB-NEXT: addiw a0, a0, -1093
+; RV64IXTHEADBB-NEXT: slli a1, a0, 32
+; RV64IXTHEADBB-NEXT: add a0, a0, a1
; RV64IXTHEADBB-NEXT: ret
ret i64 9223372034904144827 ; 0x7fffffff8bbbbbbb
}
|
preames
requested changes
Sep 20, 2023
| if ((SeqLo.size() + 2) <= Subtarget.getMaxBuildIntsCost()) | ||
| return Op; | ||
| } | ||
| int64_t LoVal = SignExtend64<32>(Imm); |
There was a problem hiding this comment.
I really dislike the increasing duplication here.
Could we at least introduce a matchShiftedConstant helper? Have it return the shift amount and instseq for the value by reference?
There was a problem hiding this comment.
I took a shot at rolling this into the InstSeq abstraction. I posted it here: #67159
I'm not sold on this approach myself. I think it's probably only worth doing if we anticipate other cases where we want multiple live values in the same sequence.
|
Moving to draft while I work out issues and refactor. |
topperc
force-pushed
the
pr/constant-mat
branch
from
5ceb805
to
3e13faf
Compare
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
… materialization. We don't have to limit ourselves to a shift amount of 32. We can support other shift amounts that make the upper 32 bits line up.
topperc
force-pushed
the
pr/constant-mat
branch
from
3e13faf
to
1b42b8a
Compare
|
I refactored the duplicate code into a new function in RISCVMatInt. |
preames
approved these changes
Oct 2, 2023
| CurDAG->getMachineNode(RISCV::SLLI, DL, VT, Lo, | ||
| CurDAG->getTargetConstant(ShiftAmt, DL, VT)), | ||
| 0); | ||
| // Prefer ADD when possible. |
There was a problem hiding this comment.
Kill the oddly placed comment
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
… materialization.
We don't have to limit ourselves to a shift amount of 32. We can support other shift amounts that make the upper 32 bits line up.