Skip to content

[LoongArch] Load floating-point immediate using VLDI #101923

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 5 commits into from
Aug 9, 2024
Merged

[LoongArch] Load floating-point immediate using VLDI #101923

merged 5 commits into from
Aug 9, 2024

Conversation

heiher
Copy link
Member

@heiher heiher commented Aug 5, 2024

This commit uses the VLDI instruction to load some common floating-point constants when the LSX feature is enabled.

@heiher
[LoongArch] Load floating-point immediate using VLDI
751841a 
This commit uses the VLDI instruction to load some common floating-point
constants when the LSX feature is enabled.
@llvmbot
Copy link
Member

llvmbot commented Aug 5, 2024

@llvm/pr-subscribers-backend-loongarch

Author: hev (heiher)

Changes

This commit uses the VLDI instruction to load some common floating-point constants when the LSX feature is enabled.

Patch is 258.18 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/101923.diff

4 Files Affected:

  • (modified) llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp (+18-1)
  • (modified) llvm/lib/Target/LoongArch/LoongArchISelLowering.h (+2)
  • (modified) llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td (+31)
  • (modified) llvm/test/CodeGen/LoongArch/float-imm-vldi.ll (+1282-4864)
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index bdbb7ab0f5139..44f3c1496344e 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -5573,6 +5573,23 @@ SDValue LoongArchTargetLowering::LowerReturn(
   return DAG.getNode(LoongArchISD::RET, DL, MVT::Other, RetOps);
 }
 
+bool LoongArchTargetLowering::isFPImmVLDILegal(const APFloat &Imm,
+                                               EVT VT) const {
+  if (!Subtarget.hasExtLSX())
+    return false;
+
+  uint64_t val = Imm.bitcastToAPInt().getZExtValue();
+  if (VT == MVT::f32) {
+    uint64_t masked = val & 0x7e07ffff;
+    return (masked == 0x3e000000 || masked == 0x40000000);
+  } else if (VT == MVT::f64) {
+    uint64_t masked = val & 0x7fc0ffffffffffff;
+    return (masked == 0x3fc0000000000000 || masked == 0x4000000000000000);
+  }
+
+  return false;
+}
+
 bool LoongArchTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT,
                                            bool ForCodeSize) const {
   // TODO: Maybe need more checks here after vector extension is supported.
@@ -5580,7 +5597,7 @@ bool LoongArchTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT,
     return false;
   if (VT == MVT::f64 && !Subtarget.hasBasicD())
     return false;
-  return (Imm.isZero() || Imm.isExactlyValue(+1.0));
+  return (Imm.isZero() || Imm.isExactlyValue(1.0) || isFPImmVLDILegal(Imm, VT));
 }
 
 bool LoongArchTargetLowering::isCheapToSpeculateCttz(Type *) const {
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index d834a5d8587fd..9723789e919b1 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -260,6 +260,8 @@ class LoongArchTargetLowering : public TargetLowering {
   bool shouldAlignPointerArgs(CallInst *CI, unsigned &MinSize,
                               Align &PrefAlign) const override;
 
+  bool isFPImmVLDILegal(const APFloat &Imm, EVT VT) const;
+
 private:
   /// Target-specific function used to lower LoongArch calling conventions.
   typedef bool LoongArchCCAssignFn(const DataLayout &DL, LoongArchABI::ABI ABI,
diff --git a/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
index 0580683c3ce30..cac47e35afe25 100644
--- a/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
@@ -203,6 +203,29 @@ def to_valid_timm : SDNodeXForm<timm, [{
   return CurDAG->getTargetConstant(CN->getSExtValue(), SDLoc(N), Subtarget->getGRLenVT());
 }]>;
 
+// FP immediate of VLDI patterns.
+def f32imm_vldi : PatLeaf<(fpimm), [{
+  const auto &TLI =
+      *static_cast<const LoongArchTargetLowering*>(getTargetLowering());
+  return TLI.isFPImmVLDILegal(N->getValueAPF(), MVT::f32);
+}]>;
+def f64imm_vldi : PatLeaf<(fpimm), [{
+  const auto &TLI =
+      *static_cast<const LoongArchTargetLowering*>(getTargetLowering());
+  return TLI.isFPImmVLDILegal(N->getValueAPF(), MVT::f64);
+}]>;
+
+def to_f32imm_vldi : SDNodeXForm<fpimm, [{
+  uint64_t x = N->getValueAPF().bitcastToAPInt().getZExtValue();
+  x = (0b11011 << 8) | (((x >> 24) & 0xc0) ^ 0x40) | ((x >> 19) & 0x3f);
+  return CurDAG->getTargetConstant(SignExtend32<13>(x), SDLoc(N), MVT::i32);
+}]>;
+def to_f64imm_vldi : SDNodeXForm<fpimm, [{
+  uint64_t x = N->getValueAPF().bitcastToAPInt().getZExtValue();
+  x = (0b11100 << 8) | (((x >> 56) & 0xc0) ^ 0x40) | ((x >> 48) & 0x3f);
+  return CurDAG->getTargetConstant(SignExtend32<13>(x), SDLoc(N), MVT::i32);
+}]>;
+
 //===----------------------------------------------------------------------===//
 // Instruction class templates
 //===----------------------------------------------------------------------===//
@@ -663,7 +686,9 @@ def VMSKGEZ_B : LSX2R_VV<0x729c5000>;
 
 def VMSKNZ_B : LSX2R_VV<0x729c6000>;
 
+let isReMaterializable = 1, isAsCheapAsAMove = 1 in {
 def VLDI : LSX1RI13_VI<0x73e00000>;
+}
 
 def VAND_V : LSX3R_VVV<0x71260000>;
 def VOR_V : LSX3R_VVV<0x71268000>;
@@ -1910,6 +1935,12 @@ def : Pat<(v2i64 (fp_to_sint v2f64:$vj)), (VFTINTRZ_L_D v2f64:$vj)>;
 def : Pat<(v4i32 (fp_to_uint v4f32:$vj)), (VFTINTRZ_WU_S v4f32:$vj)>;
 def : Pat<(v2i64 (fp_to_uint v2f64:$vj)), (VFTINTRZ_LU_D v2f64:$vj)>;
 
+// Vector loads floating-point constants
+def : Pat<(f32 f32imm_vldi:$in),
+          (f32 (EXTRACT_SUBREG (VLDI (to_f32imm_vldi f32imm_vldi:$in)), sub_32))>;
+def : Pat<(f64 f64imm_vldi:$in),
+          (f64 (EXTRACT_SUBREG (VLDI (to_f64imm_vldi f64imm_vldi:$in)), sub_64))>;
+
 } // Predicates = [HasExtLSX]
 
 /// Intrinsic pattern
diff --git a/llvm/test/CodeGen/LoongArch/float-imm-vldi.ll b/llvm/test/CodeGen/LoongArch/float-imm-vldi.ll
index 0c6856bafd23d..551ab6ea44c66 100644
--- a/llvm/test/CodeGen/LoongArch/float-imm-vldi.ll
+++ b/llvm/test/CodeGen/LoongArch/float-imm-vldi.ll
@@ -1,6145 +1,2563 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc --mtriple=loongarch32 --mattr=+lsx < %s | FileCheck %s --check-prefix=LA32
-; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefix=LA64
+; RUN: llc --mtriple=loongarch32 --mattr=+lsx < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
 
 define dso_local { float, double } @test1() {
-; LA32-LABEL: test1:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI0_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI0_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI0_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI0_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test1:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI0_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI0_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI0_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI0_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test1:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1280
+; CHECK-NEXT:    vldi $vr1, -1024
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.0000000000, double 2.0000000000 }
 }
 
 define dso_local { float, double } @test2() {
-; LA32-LABEL: test2:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI1_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI1_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test2:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI1_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI1_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test2:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1279
+; CHECK-NEXT:    vldi $vr1, -1023
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.1250000000, double 2.1250000000 }
 }
 
 define dso_local { float, double } @test3() {
-; LA32-LABEL: test3:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI2_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI2_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI2_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI2_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test3:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI2_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI2_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI2_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI2_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test3:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1278
+; CHECK-NEXT:    vldi $vr1, -1022
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.2500000000, double 2.2500000000 }
 }
 
 define dso_local { float, double } @test4() {
-; LA32-LABEL: test4:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI3_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI3_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI3_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI3_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test4:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI3_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI3_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI3_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI3_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test4:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1277
+; CHECK-NEXT:    vldi $vr1, -1021
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.3750000000, double 2.3750000000 }
 }
 
 define dso_local { float, double } @test5() {
-; LA32-LABEL: test5:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI4_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI4_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI4_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI4_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test5:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI4_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI4_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI4_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI4_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test5:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1276
+; CHECK-NEXT:    vldi $vr1, -1020
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.5000000000, double 2.5000000000 }
 }
 
 define dso_local { float, double } @test6() {
-; LA32-LABEL: test6:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI5_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI5_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI5_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI5_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test6:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI5_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI5_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI5_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI5_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test6:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1275
+; CHECK-NEXT:    vldi $vr1, -1019
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.6250000000, double 2.6250000000 }
 }
 
 define dso_local { float, double } @test7() {
-; LA32-LABEL: test7:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI6_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI6_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI6_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI6_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test7:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI6_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI6_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI6_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI6_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test7:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1274
+; CHECK-NEXT:    vldi $vr1, -1018
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.7500000000, double 2.7500000000 }
 }
 
 define dso_local { float, double } @test8() {
-; LA32-LABEL: test8:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI7_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI7_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI7_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI7_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test8:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI7_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI7_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI7_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI7_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test8:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1273
+; CHECK-NEXT:    vldi $vr1, -1017
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 2.8750000000, double 2.8750000000 }
 }
 
 define dso_local { float, double } @test9() {
-; LA32-LABEL: test9:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI8_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI8_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI8_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI8_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test9:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI8_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI8_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI8_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI8_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test9:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1272
+; CHECK-NEXT:    vldi $vr1, -1016
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.0000000000, double 3.0000000000 }
 }
 
 define dso_local { float, double } @test10() {
-; LA32-LABEL: test10:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI9_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI9_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI9_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI9_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test10:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI9_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI9_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI9_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI9_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test10:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1271
+; CHECK-NEXT:    vldi $vr1, -1015
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.1250000000, double 3.1250000000 }
 }
 
 define dso_local { float, double } @test11() {
-; LA32-LABEL: test11:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI10_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI10_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI10_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI10_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test11:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI10_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI10_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI10_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI10_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test11:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1270
+; CHECK-NEXT:    vldi $vr1, -1014
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.2500000000, double 3.2500000000 }
 }
 
 define dso_local { float, double } @test12() {
-; LA32-LABEL: test12:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI11_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI11_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI11_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI11_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test12:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI11_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI11_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI11_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI11_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test12:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1269
+; CHECK-NEXT:    vldi $vr1, -1013
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.3750000000, double 3.3750000000 }
 }
 
 define dso_local { float, double } @test13() {
-; LA32-LABEL: test13:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI12_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI12_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI12_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI12_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test13:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI12_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI12_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI12_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI12_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test13:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1268
+; CHECK-NEXT:    vldi $vr1, -1012
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.5000000000, double 3.5000000000 }
 }
 
 define dso_local { float, double } @test14() {
-; LA32-LABEL: test14:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI13_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI13_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI13_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI13_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test14:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI13_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI13_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI13_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI13_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test14:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1267
+; CHECK-NEXT:    vldi $vr1, -1011
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.6250000000, double 3.6250000000 }
 }
 
 define dso_local { float, double } @test15() {
-; LA32-LABEL: test15:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI14_0)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI14_0)
-; LA32-NEXT:    fld.s $fa0, $a0, 0
-; LA32-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI14_1)
-; LA32-NEXT:    addi.w $a0, $a0, %pc_lo12(.LCPI14_1)
-; LA32-NEXT:    fld.d $fa1, $a0, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: test15:
-; LA64:       # %bb.0: # %entry
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI14_0)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI14_0)
-; LA64-NEXT:    fld.s $fa0, $a0, 0
-; LA64-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI14_1)
-; LA64-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI14_1)
-; LA64-NEXT:    fld.d $fa1, $a0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: test15:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vldi $vr0, -1266
+; CHECK-NEXT:    vldi $vr1, -1010
+; CHECK-NEXT:    ret
 entry:
   ret { float, double } { float 3.7500000000, double 3.7500000000 }
 }
 
 define dso_local { float, double } @test16() {
-; LA32-LABEL: test16:
-; LA32:       # %bb.0: # %entry
-; LA32-NEXT:    pcal...
[truncated]

@heiher heiher requested review from wangleiat and SixWeining August 7, 2024 07:44
heiher added 2 commits August 8, 2024 10:54
@heiher
Fix assertion failed
7e43b28 
llvm/include/llvm/ADT/APInt.h:1503: uint64_t llvm::APInt::getZExtValue() const:
 Assertion `getActiveBits() <= 64 && "Too many bits for uint64_t"' failed.
@heiher
Merge remote-tracking branch 'origin/main' into fpimm-vldi
de4f453
SixWeining
SixWeining reviewed Aug 9, 2024
View reviewed changes
@heiher heiher requested a review from SixWeining August 9, 2024 05:29
@heiher heiher merged commit dbae30d into llvm:main Aug 9, 2024
8 checks passed
@heiher heiher deleted the fpimm-vldi branch August 9, 2024 06:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@SixWeining SixWeining SixWeining approved these changes

@wangleiat wangleiat Awaiting requested review from wangleiat

Assignees
No one assigned
Labels
backend:loongarch
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@heiher @llvmbot @SixWeining