Skip to content

[RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and roundeven intrinsics. #108779

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
Sep 18, 2024
Merged

[RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and roundeven intrinsics. #108779

merged 1 commit into from
Sep 18, 2024

Conversation

topperc
Copy link
Collaborator

@topperc topperc commented Sep 15, 2024

No description provided.

@llvmbot
Copy link
Member

llvmbot commented Sep 15, 2024 •
edited
Loading

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

@llvm/pr-subscribers-llvm-globalisel

Author: Craig Topper (topperc)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/108779.diff

4 Files Affected:

  • (modified) llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp (+6)
  • (modified) llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp (+3-1)
  • (removed) llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-ceil-floor.mir (-98)
  • (added) llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-libcall.mir (+328)
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 3640b77ff4068c..00db5bc84e2916 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -485,6 +485,10 @@ static RTLIB::Libcall getRTLibDesc(unsigned Opcode, unsigned Size) {
     RTLIBCASE(RINT_F);
   case TargetOpcode::G_FNEARBYINT:
     RTLIBCASE(NEARBYINT_F);
+  case TargetOpcode::G_INTRINSIC_TRUNC:
+    RTLIBCASE(TRUNC_F);
+  case TargetOpcode::G_INTRINSIC_ROUND:
+    RTLIBCASE(ROUND_F);
   case TargetOpcode::G_INTRINSIC_ROUNDEVEN:
     RTLIBCASE(ROUNDEVEN_F);
   case TargetOpcode::G_INTRINSIC_LRINT:
@@ -1215,6 +1219,8 @@ LegalizerHelper::libcall(MachineInstr &MI, LostDebugLocObserver &LocObserver) {
   case TargetOpcode::G_FSQRT:
   case TargetOpcode::G_FRINT:
   case TargetOpcode::G_FNEARBYINT:
+  case TargetOpcode::G_INTRINSIC_TRUNC:
+  case TargetOpcode::G_INTRINSIC_ROUND:
   case TargetOpcode::G_INTRINSIC_ROUNDEVEN: {
     LLT LLTy = MRI.getType(MI.getOperand(0).getReg());
     unsigned Size = LLTy.getSizeInBits();
diff --git a/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp b/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
index 8de908d20d245c..c17a73b995c4e5 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
@@ -541,7 +541,9 @@ RISCVLegalizerInfo::RISCVLegalizerInfo(const RISCVSubtarget &ST)
 
   // FIXME: We can do custom inline expansion like SelectionDAG.
   // FIXME: Legal with Zfa.
-  getActionDefinitionsBuilder({G_FCEIL, G_FFLOOR})
+  getActionDefinitionsBuilder({G_FCEIL, G_FFLOOR, G_FRINT, G_FNEARBYINT,
+                               G_INTRINSIC_TRUNC, G_INTRINSIC_ROUND,
+                               G_INTRINSIC_ROUNDEVEN})
       .libcallFor({s32, s64});
 
   getActionDefinitionsBuilder(G_VASTART).customFor({p0});
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-ceil-floor.mir b/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-ceil-floor.mir
deleted file mode 100644
index 1e184bd0c1120f..00000000000000
--- a/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-ceil-floor.mir
+++ /dev/null
@@ -1,98 +0,0 @@
-# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=riscv32 -mattr=+d -run-pass=legalizer %s -o - \
-# RUN: | FileCheck %s
-# RUN: llc -mtriple=riscv64 -mattr=+d -run-pass=legalizer %s -o - \
-# RUN: | FileCheck %s
-
----
-name:            ceil_f32
-body:             |
-  bb.1:
-    liveins: $f10_f
-
-    ; CHECK-LABEL: name: ceil_f32
-    ; CHECK: liveins: $f10_f
-    ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
-    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
-    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &ceilf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
-    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
-    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
-    ; CHECK-NEXT: PseudoRET implicit $f10_f
-    %0:_(s32) = COPY $f10_f
-    %1:_(s32) = G_FCEIL %0
-    $f10_f = COPY %1(s32)
-    PseudoRET implicit $f10_f
-
-...
----
-name:            floor_f32
-body:             |
-  bb.1:
-    liveins: $f10_f
-
-    ; CHECK-LABEL: name: floor_f32
-    ; CHECK: liveins: $f10_f
-    ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
-    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
-    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &floorf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
-    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
-    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
-    ; CHECK-NEXT: PseudoRET implicit $f10_f
-    %0:_(s32) = COPY $f10_f
-    %1:_(s32) = G_FFLOOR %0
-    $f10_f = COPY %1(s32)
-    PseudoRET implicit $f10_f
-
-...
----
-name:            ceil_f64
-body:             |
-  bb.1:
-    liveins: $f10_d
-
-    ; CHECK-LABEL: name: ceil_f64
-    ; CHECK: liveins: $f10_d
-    ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
-    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
-    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &ceil, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
-    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
-    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
-    ; CHECK-NEXT: PseudoRET implicit $f10_d
-    %0:_(s64) = COPY $f10_d
-    %1:_(s64) = G_FCEIL %0
-    $f10_d = COPY %1(s64)
-    PseudoRET implicit $f10_d
-
-...
----
-name:            floor_f64
-body:             |
-  bb.1:
-    liveins: $f10_d
-
-    ; CHECK-LABEL: name: floor_f64
-    ; CHECK: liveins: $f10_d
-    ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
-    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
-    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &floor, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
-    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
-    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
-    ; CHECK-NEXT: PseudoRET implicit $f10_d
-    %0:_(s64) = COPY $f10_d
-    %1:_(s64) = G_FFLOOR %0
-    $f10_d = COPY %1(s64)
-    PseudoRET implicit $f10_d
-
-...
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-libcall.mir b/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-libcall.mir
new file mode 100644
index 00000000000000..3b4f6a065d9736
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-libcall.mir
@@ -0,0 +1,328 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=riscv32 -mattr=+d -run-pass=legalizer %s -o - \
+# RUN: | FileCheck %s
+# RUN: llc -mtriple=riscv64 -mattr=+d -run-pass=legalizer %s -o - \
+# RUN: | FileCheck %s
+
+---
+name:            ceil_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: ceil_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &ceilf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_FCEIL %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            floor_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: floor_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &floorf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_FFLOOR %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            trunc_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: trunc_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &truncf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_INTRINSIC_TRUNC %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            rint_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: rint_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &rintf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_FRINT %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            nearbyint_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: nearbyint_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &nearbyintf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_FNEARBYINT %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            round_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: round_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &roundf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_INTRINSIC_ROUND %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            roundeven_f32
+body:             |
+  bb.1:
+    liveins: $f10_f
+
+    ; CHECK-LABEL: name: roundeven_f32
+    ; CHECK: liveins: $f10_f
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_f = COPY [[COPY]](s32)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &roundevenf, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_f, implicit-def $f10_f
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $f10_f
+    ; CHECK-NEXT: $f10_f = COPY [[COPY1]](s32)
+    ; CHECK-NEXT: PseudoRET implicit $f10_f
+    %0:_(s32) = COPY $f10_f
+    %1:_(s32) = G_INTRINSIC_ROUNDEVEN %0
+    $f10_f = COPY %1(s32)
+    PseudoRET implicit $f10_f
+
+...
+---
+name:            ceil_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: ceil_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &ceil, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_FCEIL %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            floor_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: floor_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &floor, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_FFLOOR %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            trunc_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: trunc_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &trunc, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_INTRINSIC_TRUNC %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            rint_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: rint_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &rint, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_FRINT %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            nearbyint_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: nearbyint_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &nearbyint, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_FNEARBYINT %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            round_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: round_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &round, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_INTRINSIC_ROUND %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...
+---
+name:            roundeven_f64
+body:             |
+  bb.1:
+    liveins: $f10_d
+
+    ; CHECK-LABEL: name: roundeven_f64
+    ; CHECK: liveins: $f10_d
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: $f10_d = COPY [[COPY]](s64)
+    ; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &roundeven, csr_ilp32d_lp64d, implicit-def $x1, implicit $f10_d, implicit-def $f10_d
+    ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $x2, implicit $x2
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $f10_d
+    ; CHECK-NEXT: $f10_d = COPY [[COPY1]](s64)
+    ; CHECK-NEXT: PseudoRET implicit $f10_d
+    %0:_(s64) = COPY $f10_d
+    %1:_(s64) = G_INTRINSIC_ROUNDEVEN %0
+    $f10_d = COPY %1(s64)
+    PseudoRET implicit $f10_d
+
+...

@topperc topperc requested a review from tschuett September 16, 2024 20:56
@tschuett
Copy link

I cannot say much, but the title slightly differs from what you actually legalize, but so what. 4 vs. 5 and some missing fs.

@topperc topperc changed the title [RISCV][GISel] Use libcalls for rint, nearbyint, trunc, and round intrinsics. [RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and roundeven intrinsics. Sep 16, 2024
@topperc
Copy link
Collaborator Author

topperc commented Sep 16, 2024 •
edited
Loading

I cannot say much, but the title slightly differs from what you actually legalize, but so what. 4 vs. 5 and some missing fs.

I wrote them in terms of their C and IR names rather than the mishmash we have in G_ node names so I left off the 'f'.

arsenm
arsenm approved these changes Sep 18, 2024
View reviewed changes
llvm/test/CodeGen/RISCV/GlobalISel/legalizer/legalize-fp-libcall.mir
PseudoRET implicit $f10_d

...
---
Copy link
Contributor

Choose a reason for hiding this comment

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

I think these kind of tests should use IR tests. There's no weird legalizer edge cases being tested, and mir tests are harder to maintain

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Agreed. I'll commit this as is and then try to replace with an IR test as a followup.

@topperc topperc merged commit d5d1417 into llvm:main Sep 18, 2024
11 checks passed
@topperc topperc deleted the pr/gisel-libcall branch September 18, 2024 19:07
tmsri pushed a commit to tmsri/llvm-project that referenced this pull request Sep 19, 2024
@topperc @tmsri
[RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and ro…
566ac4a 
…undeven intrinsics. (llvm#108779)
PhilippvK pushed a commit to PhilippvK/llvm-project that referenced this pull request Oct 7, 2024
@topperc
[RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and ro…
44d52d9 
…undeven intrinsics. (llvm#108779)
PhilippvK pushed a commit to PhilippvK/llvm-project that referenced this pull request Jan 13, 2025
@topperc
[RISCV][GISel] Use libcalls for rint, nearbyint, trunc, round, and ro…
679c557 
…undeven intrinsics. (llvm#108779)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@arsenm arsenm arsenm approved these changes

@aemerson aemerson Awaiting requested review from aemerson

@michaelmaitland michaelmaitland Awaiting requested review from michaelmaitland

@tschuett tschuett Awaiting requested review from tschuett

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

Successfully merging this pull request may close these issues.
4 participants
@topperc @llvmbot @tschuett @arsenm