Skip to content

[RISCV] Consider truncate semantics in performBUILD_VECTORCombine #81168

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 1 commit into from
Feb 8, 2024
Merged

[RISCV] Consider truncate semantics in performBUILD_VECTORCombine #81168

merged 1 commit into from
Feb 8, 2024

Conversation

preames
Copy link
Collaborator

@preames preames commented Feb 8, 2024

Fixes #80910.

Per the documentation in ISDOpcodes.h, for BUILD_VECTOR "The types of the operands must match the vector element type, except that integer types are allowed to be larger than the element type, in which case the operands are implicitly truncated."

This transform was assuming that the scalar operand type matched the result type. This resulted in essentially performing a truncate before a binop, instead of after. As demonstrated by the test case changes, this is often not legal.

@preames
[RISCV] Consider truncate semantics in performBUILD_VECTORCombine
984cc32 
Fixes llvm#80910.

Per the documentation in ISDOpcodes.h, for BUILD_VECTOR "The types of the
operands must match the vector element type, except that integer types
are allowed to be larger than the element type, in which case the
operands are implicitly truncated."

This transform was assuming that the scalar operand type matched the
result type.  This resulted in essentially performing a truncate
before a binop, instead of after.  As demonstrated by the test case
changes, this is often not legal.
@preames preames requested review from topperc and luke957 February 8, 2024 18:20
@llvmbot
Copy link
Member

llvmbot commented Feb 8, 2024

@llvm/pr-subscribers-backend-risc-v

Author: Philip Reames (preames)

Changes

Fixes #80910.

Per the documentation in ISDOpcodes.h, for BUILD_VECTOR "The types of the operands must match the vector element type, except that integer types are allowed to be larger than the element type, in which case the operands are implicitly truncated."

This transform was assuming that the scalar operand type matched the result type. This resulted in essentially performing a truncate before a binop, instead of after. As demonstrated by the test case changes, this is often not legal.

Patch is 27.15 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/81168.diff

4 Files Affected:

  • (modified) llvm/lib/Target/RISCV/RISCVISelLowering.cpp (+6)
  • (modified) llvm/test/CodeGen/RISCV/rvv/fixed-vectors-buildvec-of-binop.ll (+4-2)
  • (modified) llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect.ll (+362-151)
  • (modified) llvm/test/CodeGen/RISCV/urem-seteq-illegal-types.ll (+27-26)
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 27037f4d5c5c85..0799cc2e5901cb 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -14956,6 +14956,11 @@ static SDValue performBUILD_VECTORCombine(SDNode *N, SelectionDAG &DAG,
   if (!TLI.isOperationLegalOrCustom(Opcode, VT) || !TLI.isTypeLegal(VT))
     return SDValue();
 
+  // This BUILD_VECTOR involves an implicit truncation, and sinking
+  // truncates through binops is non-trivial.
+  if (N->op_begin()->getValueType() != VT.getVectorElementType())
+    return SDValue();
+
   SmallVector<SDValue> LHSOps;
   SmallVector<SDValue> RHSOps;
   for (SDValue Op : N->ops()) {
@@ -14983,6 +14988,7 @@ static SDValue performBUILD_VECTORCombine(SDNode *N, SelectionDAG &DAG,
     // have different LHS and RHS types.
     if (Op.getOperand(0).getValueType() != Op.getOperand(1).getValueType())
       return SDValue();
+
     RHSOps.push_back(Op.getOperand(1));
   }
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-buildvec-of-binop.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-buildvec-of-binop.ll
index e376688aca8a76..af7d7f7ae755bc 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-buildvec-of-binop.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-buildvec-of-binop.ll
@@ -589,7 +589,8 @@ define <8 x i32> @add_constant_rhs_8xi32_partial(<8 x i32> %vin, i32 %a, i32 %b,
   ret <8 x i32> %v3
 }
 
-; FIXME: This is currently showing a miscompile, we effectively
+; Here we can not pull the ashr through into the vector domain due to
+; the truncate semantics of the build_vector.  Doing so would
 ; truncate before the ashr instead of after it, so if %a or %b
 ; is e.g. UINT32_MAX+1 we get different result.
 define <2 x i32> @build_vec_of_trunc_op(i64 %a, i64 %b) {
@@ -608,10 +609,11 @@ define <2 x i32> @build_vec_of_trunc_op(i64 %a, i64 %b) {
 ;
 ; RV64-LABEL: build_vec_of_trunc_op:
 ; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    srli a0, a0, 1
+; RV64-NEXT:    srli a1, a1, 1
 ; RV64-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
 ; RV64-NEXT:    vmv.v.x v8, a0
 ; RV64-NEXT:    vslide1down.vx v8, v8, a1
-; RV64-NEXT:    vsrl.vi v8, v8, 1
 ; RV64-NEXT:    ret
 entry:
   %conv11.i = ashr i64 %a, 1
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect.ll
index cd47720e38216f..ead41b0717ff1c 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vselect.ll
@@ -3,30 +3,65 @@
 ; RUN: llc -mtriple=riscv64 -target-abi=lp64d -mattr=+v,+zfh,+zvfh,+f,+d -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,RV64
 
 define void @vselect_vv_v6i32(ptr %a, ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vv_v6i32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a2, 0(a2)
-; CHECK-NEXT:    vle32.v v8, (a1)
-; CHECK-NEXT:    srli a1, a2, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vmv.v.x v10, a2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 3
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 4
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a2, a2, 5
-; CHECK-NEXT:    vslide1down.vx v10, v10, a2
-; CHECK-NEXT:    vslidedown.vi v10, v10, 2
-; CHECK-NEXT:    vand.vi v10, v10, 1
-; CHECK-NEXT:    vmsne.vi v0, v10, 0
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
-; CHECK-NEXT:    vle32.v v8, (a0), v0.t
-; CHECK-NEXT:    vse32.v v8, (a3)
-; CHECK-NEXT:    ret
+; RV32-LABEL: vselect_vv_v6i32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    lbu a2, 0(a2)
+; RV32-NEXT:    vle32.v v8, (a1)
+; RV32-NEXT:    slli a1, a2, 30
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    andi a4, a2, 1
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vmv.v.x v10, a4
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 29
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 28
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 27
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    srli a2, a2, 5
+; RV32-NEXT:    vslide1down.vx v10, v10, a2
+; RV32-NEXT:    vslidedown.vi v10, v10, 2
+; RV32-NEXT:    vand.vi v10, v10, 1
+; RV32-NEXT:    vmsne.vi v0, v10, 0
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
+; RV32-NEXT:    vle32.v v8, (a0), v0.t
+; RV32-NEXT:    vse32.v v8, (a3)
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vselect_vv_v6i32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    lbu a2, 0(a2)
+; RV64-NEXT:    vle32.v v8, (a1)
+; RV64-NEXT:    slli a1, a2, 62
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    andi a4, a2, 1
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vmv.v.x v10, a4
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 61
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 60
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 59
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    srli a2, a2, 5
+; RV64-NEXT:    vslide1down.vx v10, v10, a2
+; RV64-NEXT:    vslidedown.vi v10, v10, 2
+; RV64-NEXT:    vand.vi v10, v10, 1
+; RV64-NEXT:    vmsne.vi v0, v10, 0
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
+; RV64-NEXT:    vle32.v v8, (a0), v0.t
+; RV64-NEXT:    vse32.v v8, (a3)
+; RV64-NEXT:    ret
   %va = load <6 x i32>, ptr %a
   %vb = load <6 x i32>, ptr %b
   %vcc = load <6 x i1>, ptr %cc
@@ -36,31 +71,67 @@ define void @vselect_vv_v6i32(ptr %a, ptr %b, ptr %cc, ptr %z) {
 }
 
 define void @vselect_vx_v6i32(i32 %a, ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vx_v6i32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a2, 0(a2)
-; CHECK-NEXT:    vle32.v v8, (a1)
-; CHECK-NEXT:    srli a1, a2, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vmv.v.x v10, a2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 3
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 4
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a2, a2, 5
-; CHECK-NEXT:    vslide1down.vx v10, v10, a2
-; CHECK-NEXT:    vslidedown.vi v10, v10, 2
-; CHECK-NEXT:    vand.vi v10, v10, 1
-; CHECK-NEXT:    vmsne.vi v0, v10, 0
-; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
-; CHECK-NEXT:    vmerge.vxm v8, v8, a0, v0
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    vse32.v v8, (a3)
-; CHECK-NEXT:    ret
+; RV32-LABEL: vselect_vx_v6i32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    lbu a2, 0(a2)
+; RV32-NEXT:    vle32.v v8, (a1)
+; RV32-NEXT:    slli a1, a2, 30
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    andi a4, a2, 1
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vmv.v.x v10, a4
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 29
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 28
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 27
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    srli a2, a2, 5
+; RV32-NEXT:    vslide1down.vx v10, v10, a2
+; RV32-NEXT:    vslidedown.vi v10, v10, 2
+; RV32-NEXT:    vand.vi v10, v10, 1
+; RV32-NEXT:    vmsne.vi v0, v10, 0
+; RV32-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV32-NEXT:    vmerge.vxm v8, v8, a0, v0
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    vse32.v v8, (a3)
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vselect_vx_v6i32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    lbu a2, 0(a2)
+; RV64-NEXT:    vle32.v v8, (a1)
+; RV64-NEXT:    slli a1, a2, 62
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    andi a4, a2, 1
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vmv.v.x v10, a4
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 61
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 60
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 59
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    srli a2, a2, 5
+; RV64-NEXT:    vslide1down.vx v10, v10, a2
+; RV64-NEXT:    vslidedown.vi v10, v10, 2
+; RV64-NEXT:    vand.vi v10, v10, 1
+; RV64-NEXT:    vmsne.vi v0, v10, 0
+; RV64-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV64-NEXT:    vmerge.vxm v8, v8, a0, v0
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    vse32.v v8, (a3)
+; RV64-NEXT:    ret
   %vb = load <6 x i32>, ptr %b
   %ahead = insertelement <6 x i32> poison, i32 %a, i32 0
   %va = shufflevector <6 x i32> %ahead, <6 x i32> poison, <6 x i32> zeroinitializer
@@ -71,31 +142,67 @@ define void @vselect_vx_v6i32(i32 %a, ptr %b, ptr %cc, ptr %z) {
 }
 
 define void @vselect_vi_v6i32(ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vi_v6i32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a1, 0(a1)
-; CHECK-NEXT:    vle32.v v8, (a0)
-; CHECK-NEXT:    srli a0, a1, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vmv.v.x v10, a1
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 3
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 4
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a1, a1, 5
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    vslidedown.vi v10, v10, 2
-; CHECK-NEXT:    vand.vi v10, v10, 1
-; CHECK-NEXT:    vmsne.vi v0, v10, 0
-; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
-; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    vse32.v v8, (a2)
-; CHECK-NEXT:    ret
+; RV32-LABEL: vselect_vi_v6i32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    lbu a1, 0(a1)
+; RV32-NEXT:    vle32.v v8, (a0)
+; RV32-NEXT:    slli a0, a1, 30
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    andi a3, a1, 1
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vmv.v.x v10, a3
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 29
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 28
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 27
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    srli a1, a1, 5
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    vslidedown.vi v10, v10, 2
+; RV32-NEXT:    vand.vi v10, v10, 1
+; RV32-NEXT:    vmsne.vi v0, v10, 0
+; RV32-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV32-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    vse32.v v8, (a2)
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vselect_vi_v6i32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    lbu a1, 0(a1)
+; RV64-NEXT:    vle32.v v8, (a0)
+; RV64-NEXT:    slli a0, a1, 62
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    andi a3, a1, 1
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vmv.v.x v10, a3
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 61
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 60
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 59
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    srli a1, a1, 5
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    vslidedown.vi v10, v10, 2
+; RV64-NEXT:    vand.vi v10, v10, 1
+; RV64-NEXT:    vmsne.vi v0, v10, 0
+; RV64-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV64-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    vse32.v v8, (a2)
+; RV64-NEXT:    ret
   %vb = load <6 x i32>, ptr %b
   %a = insertelement <6 x i32> poison, i32 -1, i32 0
   %va = shufflevector <6 x i32> %a, <6 x i32> poison, <6 x i32> zeroinitializer
@@ -107,30 +214,65 @@ define void @vselect_vi_v6i32(ptr %b, ptr %cc, ptr %z) {
 
 
 define void @vselect_vv_v6f32(ptr %a, ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vv_v6f32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a2, 0(a2)
-; CHECK-NEXT:    vle32.v v8, (a1)
-; CHECK-NEXT:    srli a1, a2, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vmv.v.x v10, a2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 3
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a1, a2, 4
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    srli a2, a2, 5
-; CHECK-NEXT:    vslide1down.vx v10, v10, a2
-; CHECK-NEXT:    vslidedown.vi v10, v10, 2
-; CHECK-NEXT:    vand.vi v10, v10, 1
-; CHECK-NEXT:    vmsne.vi v0, v10, 0
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
-; CHECK-NEXT:    vle32.v v8, (a0), v0.t
-; CHECK-NEXT:    vse32.v v8, (a3)
-; CHECK-NEXT:    ret
+; RV32-LABEL: vselect_vv_v6f32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    lbu a2, 0(a2)
+; RV32-NEXT:    vle32.v v8, (a1)
+; RV32-NEXT:    slli a1, a2, 30
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    andi a4, a2, 1
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vmv.v.x v10, a4
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 29
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 28
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    slli a1, a2, 27
+; RV32-NEXT:    srli a1, a1, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    srli a2, a2, 5
+; RV32-NEXT:    vslide1down.vx v10, v10, a2
+; RV32-NEXT:    vslidedown.vi v10, v10, 2
+; RV32-NEXT:    vand.vi v10, v10, 1
+; RV32-NEXT:    vmsne.vi v0, v10, 0
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
+; RV32-NEXT:    vle32.v v8, (a0), v0.t
+; RV32-NEXT:    vse32.v v8, (a3)
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vselect_vv_v6f32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    lbu a2, 0(a2)
+; RV64-NEXT:    vle32.v v8, (a1)
+; RV64-NEXT:    slli a1, a2, 62
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    andi a4, a2, 1
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vmv.v.x v10, a4
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 61
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 60
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    slli a1, a2, 59
+; RV64-NEXT:    srli a1, a1, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    srli a2, a2, 5
+; RV64-NEXT:    vslide1down.vx v10, v10, a2
+; RV64-NEXT:    vslidedown.vi v10, v10, 2
+; RV64-NEXT:    vand.vi v10, v10, 1
+; RV64-NEXT:    vmsne.vi v0, v10, 0
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, tu, mu
+; RV64-NEXT:    vle32.v v8, (a0), v0.t
+; RV64-NEXT:    vse32.v v8, (a3)
+; RV64-NEXT:    ret
   %va = load <6 x float>, ptr %a
   %vb = load <6 x float>, ptr %b
   %vcc = load <6 x i1>, ptr %cc
@@ -140,31 +282,67 @@ define void @vselect_vv_v6f32(ptr %a, ptr %b, ptr %cc, ptr %z) {
 }
 
 define void @vselect_vx_v6f32(float %a, ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vx_v6f32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a1, 0(a1)
-; CHECK-NEXT:    vle32.v v8, (a0)
-; CHECK-NEXT:    srli a0, a1, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vmv.v.x v10, a1
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 2
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 3
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a0, a1, 4
-; CHECK-NEXT:    vslide1down.vx v10, v10, a0
-; CHECK-NEXT:    srli a1, a1, 5
-; CHECK-NEXT:    vslide1down.vx v10, v10, a1
-; CHECK-NEXT:    vslidedown.vi v10, v10, 2
-; CHECK-NEXT:    vand.vi v10, v10, 1
-; CHECK-NEXT:    vmsne.vi v0, v10, 0
-; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
-; CHECK-NEXT:    vfmerge.vfm v8, v8, fa0, v0
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    vse32.v v8, (a2)
-; CHECK-NEXT:    ret
+; RV32-LABEL: vselect_vx_v6f32:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    lbu a1, 0(a1)
+; RV32-NEXT:    vle32.v v8, (a0)
+; RV32-NEXT:    slli a0, a1, 30
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    andi a3, a1, 1
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vmv.v.x v10, a3
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 29
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 28
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    slli a0, a1, 27
+; RV32-NEXT:    srli a0, a0, 31
+; RV32-NEXT:    vslide1down.vx v10, v10, a0
+; RV32-NEXT:    srli a1, a1, 5
+; RV32-NEXT:    vslide1down.vx v10, v10, a1
+; RV32-NEXT:    vslidedown.vi v10, v10, 2
+; RV32-NEXT:    vand.vi v10, v10, 1
+; RV32-NEXT:    vmsne.vi v0, v10, 0
+; RV32-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV32-NEXT:    vfmerge.vfm v8, v8, fa0, v0
+; RV32-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV32-NEXT:    vse32.v v8, (a2)
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vselect_vx_v6f32:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    lbu a1, 0(a1)
+; RV64-NEXT:    vle32.v v8, (a0)
+; RV64-NEXT:    slli a0, a1, 62
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    andi a3, a1, 1
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vmv.v.x v10, a3
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 61
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 60
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    slli a0, a1, 59
+; RV64-NEXT:    srli a0, a0, 63
+; RV64-NEXT:    vslide1down.vx v10, v10, a0
+; RV64-NEXT:    srli a1, a1, 5
+; RV64-NEXT:    vslide1down.vx v10, v10, a1
+; RV64-NEXT:    vslidedown.vi v10, v10, 2
+; RV64-NEXT:    vand.vi v10, v10, 1
+; RV64-NEXT:    vmsne.vi v0, v10, 0
+; RV64-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; RV64-NEXT:    vfmerge.vfm v8, v8, fa0, v0
+; RV64-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
+; RV64-NEXT:    vse32.v v8, (a2)
+; RV64-NEXT:    ret
   %vb = load <6 x float>, ptr %b
   %ahead = insertelement <6 x float> poison, float %a, i32 0
   %va = shufflevector <6 x float> %ahead, <6 x float> poison, <6 x i32> zeroinitializer
@@ -175,31 +353,67 @@ define void @vselect_vx_v6f32(float %a, ptr %b, ptr %cc, ptr %z) {
 }
 
 define void @vselect_vfpzero_v6f32(ptr %b, ptr %cc, ptr %z) {
-; CHECK-LABEL: vselect_vfpzero_v6f32:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 6, e32, m2, ta, ma
-; CHECK-NEXT:    lbu a1, 0(a1)
-; CHECK-NEXT:    vle32.v v8, (a0)
-; CHECK-NEXT:    srli a0, a1, 1
-; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; ...
[truncated]

topperc
topperc approved these changes Feb 8, 2024
View reviewed changes
Copy link
Collaborator

@topperc topperc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@preames preames merged commit d0f72f8 into llvm:main Feb 8, 2024
@preames preames deleted the pr-riscv-build-vector-truncate branch February 8, 2024 19:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@topperc topperc topperc approved these changes

@luke957 luke957 Awaiting requested review from luke957

Assignees
No one assigned
Labels
backend:RISC-V
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

RISCV64 vector miscompile at -O2
3 participants
@preames @llvmbot @topperc