Skip to content

[RISCV] Split fp rounding ops with zvfhmin nxv32f16 #108765

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 2 commits into from
Sep 18, 2024
Merged

[RISCV] Split fp rounding ops with zvfhmin nxv32f16 #108765

merged 2 commits into from
Sep 18, 2024

Conversation

lukel97
Copy link
Contributor

@lukel97 lukel97 commented Sep 15, 2024 •
edited
Loading

This adds zvfhmin test coverage for fceil, ffloor, fnearbyint, frint, fround and froundeven and splits them at nxv32f16 to avoid crashing, similarly to what we do for other nodes that we promote.

This also sets ftrunc to promote which was previously missing. We already promote the VP version of it, vp_froundtozero.
Marking it as promoted affects some of the cost model tests since they're no longer expanded.

@lukel97
[RISCV] Split fp rounding ops with zvfhmin nxv32f16
2f4d238 
This adds zvfhmin test coverage for fceil, ffloor, fnearbyint, frint, fround and froundeven and splits them at nxv32f16 to avoid crashing, similarly to what we do for other nodes that we promote.

This also sets ftrunc to promote which was previously missing. We already promote the VP version of it, vp_froundtozero.
@llvmbot
Copy link
Member

llvmbot commented Sep 15, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-analysis

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

Author: Luke Lau (lukel97)

Changes

This adds zvfhmin test coverage for fceil, ffloor, fnearbyint, frint, fround and froundeven and splits them at nxv32f16 to avoid crashing, similarly to what we do for other nodes that we promote.

This also sets ftrunc to promote which was previously missing. We already promote the VP version of it, vp_froundtozero.

Patch is 95.90 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/108765.diff

8 Files Affected:

  • (modified) llvm/lib/Target/RISCV/RISCVISelLowering.cpp (+11-6)
  • (modified) llvm/test/CodeGen/RISCV/rvv/fceil-sdnode.ll (+218-86)
  • (modified) llvm/test/CodeGen/RISCV/rvv/ffloor-sdnode.ll (+218-86)
  • (modified) llvm/test/CodeGen/RISCV/rvv/fnearbyint-sdnode.ll (+233-86)
  • (modified) llvm/test/CodeGen/RISCV/rvv/frint-sdnode.ll (+194-74)
  • (modified) llvm/test/CodeGen/RISCV/rvv/fround-sdnode.ll (+218-86)
  • (modified) llvm/test/CodeGen/RISCV/rvv/froundeven-sdnode.ll (+218-86)
  • (modified) llvm/test/CodeGen/RISCV/rvv/ftrunc-sdnode.ll (+194-74)
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 6f2dc710cb3d4d..7b1b35160cf460 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -938,12 +938,13 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
 
     // TODO: support more ops.
     static const unsigned ZvfhminPromoteOps[] = {
-        ISD::FMINNUM,     ISD::FMAXNUM,     ISD::FADD,         ISD::FSUB,
-        ISD::FMUL,        ISD::FMA,         ISD::FDIV,         ISD::FSQRT,
-        ISD::FCEIL,       ISD::FFLOOR,      ISD::FROUND,       ISD::FROUNDEVEN,
-        ISD::FRINT,       ISD::FNEARBYINT,  ISD::IS_FPCLASS,   ISD::SETCC,
-        ISD::FMAXIMUM,    ISD::FMINIMUM,    ISD::STRICT_FADD,  ISD::STRICT_FSUB,
-        ISD::STRICT_FMUL, ISD::STRICT_FDIV, ISD::STRICT_FSQRT, ISD::STRICT_FMA};
+        ISD::FMINNUM,     ISD::FMAXNUM,     ISD::FADD,        ISD::FSUB,
+        ISD::FMUL,        ISD::FMA,         ISD::FDIV,        ISD::FSQRT,
+        ISD::FCEIL,       ISD::FTRUNC,      ISD::FFLOOR,      ISD::FROUND,
+        ISD::FROUNDEVEN,  ISD::FRINT,       ISD::FNEARBYINT,  ISD::IS_FPCLASS,
+        ISD::SETCC,       ISD::FMAXIMUM,    ISD::FMINIMUM,    ISD::STRICT_FADD,
+        ISD::STRICT_FSUB, ISD::STRICT_FMUL, ISD::STRICT_FDIV, ISD::STRICT_FSQRT,
+        ISD::STRICT_FMA};
 
     // TODO: support more vp ops.
     static const unsigned ZvfhminPromoteVPOps[] = {ISD::VP_FADD,
@@ -6926,6 +6927,10 @@ SDValue RISCVTargetLowering::LowerOperation(SDValue Op,
   case ISD::FRINT:
   case ISD::FROUND:
   case ISD::FROUNDEVEN:
+    if (Op.getValueType() == MVT::nxv32f16 &&
+        (Subtarget.hasVInstructionsF16Minimal() &&
+         !Subtarget.hasVInstructionsF16()))
+      return SplitVectorOp(Op, DAG);
     return lowerFTRUNC_FCEIL_FFLOOR_FROUND(Op, DAG, Subtarget);
   case ISD::LRINT:
   case ISD::LLRINT:
diff --git a/llvm/test/CodeGen/RISCV/rvv/fceil-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/fceil-sdnode.ll
index 9efc3183f15a52..111d1d8e07d3bf 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fceil-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fceil-sdnode.ll
@@ -1,124 +1,256 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+zvfh,+v -target-abi=ilp32d \
-; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,ZVFH
 ; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+zvfh,+v -target-abi=lp64d \
-; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,ZVFH
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfhmin,+zvfhmin,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s \
+; RUN:     --check-prefixes=CHECK,ZVFHMIN
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfhmin,+zvfhmin,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s \
+; RUN:     --check-prefixes=CHECK,ZVFHMIN
 
 define <vscale x 1 x half> @ceil_nxv1f16(<vscale x 1 x half> %x) {
-; CHECK-LABEL: ceil_nxv1f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
-; CHECK-NEXT:    vfabs.v v9, v8
-; CHECK-NEXT:    vmflt.vf v0, v9, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv1f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI0_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; ZVFH-NEXT:    vfabs.v v9, v8
+; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv1f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v9
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v9
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 1 x half> @llvm.ceil.nxv1f16(<vscale x 1 x half> %x)
   ret <vscale x 1 x half> %a
 }
 declare <vscale x 1 x half> @llvm.ceil.nxv1f16(<vscale x 1 x half>)
 
 define <vscale x 2 x half> @ceil_nxv2f16(<vscale x 2 x half> %x) {
-; CHECK-LABEL: ceil_nxv2f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
-; CHECK-NEXT:    vfabs.v v9, v8
-; CHECK-NEXT:    vmflt.vf v0, v9, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv2f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI1_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; ZVFH-NEXT:    vfabs.v v9, v8
+; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv2f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v9
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v9
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 2 x half> @llvm.ceil.nxv2f16(<vscale x 2 x half> %x)
   ret <vscale x 2 x half> %a
 }
 declare <vscale x 2 x half> @llvm.ceil.nxv2f16(<vscale x 2 x half>)
 
 define <vscale x 4 x half> @ceil_nxv4f16(<vscale x 4 x half> %x) {
-; CHECK-LABEL: ceil_nxv4f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
-; CHECK-NEXT:    vfabs.v v9, v8
-; CHECK-NEXT:    vmflt.vf v0, v9, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv4f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI2_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; ZVFH-NEXT:    vfabs.v v9, v8
+; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv4f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v10, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v10
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v10, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v10, v8, v10, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v10
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 4 x half> @llvm.ceil.nxv4f16(<vscale x 4 x half> %x)
   ret <vscale x 4 x half> %a
 }
 declare <vscale x 4 x half> @llvm.ceil.nxv4f16(<vscale x 4 x half>)
 
 define <vscale x 8 x half> @ceil_nxv8f16(<vscale x 8 x half> %x) {
-; CHECK-LABEL: ceil_nxv8f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
-; CHECK-NEXT:    vfabs.v v10, v8
-; CHECK-NEXT:    vmflt.vf v0, v10, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv8f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI3_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; ZVFH-NEXT:    vfabs.v v10, v8
+; ZVFH-NEXT:    vmflt.vf v0, v10, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv8f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v12, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v12
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v12, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v12, v8, v12, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v12
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 8 x half> @llvm.ceil.nxv8f16(<vscale x 8 x half> %x)
   ret <vscale x 8 x half> %a
 }
 declare <vscale x 8 x half> @llvm.ceil.nxv8f16(<vscale x 8 x half>)
 
 define <vscale x 16 x half> @ceil_nxv16f16(<vscale x 16 x half> %x) {
-; CHECK-LABEL: ceil_nxv16f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
-; CHECK-NEXT:    vfabs.v v12, v8
-; CHECK-NEXT:    vmflt.vf v0, v12, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv16f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI4_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; ZVFH-NEXT:    vfabs.v v12, v8
+; ZVFH-NEXT:    vmflt.vf v0, v12, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv16f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v16, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v16
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v16, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v16, v8, v16, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v16
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 16 x half> @llvm.ceil.nxv16f16(<vscale x 16 x half> %x)
   ret <vscale x 16 x half> %a
 }
 declare <vscale x 16 x half> @llvm.ceil.nxv16f16(<vscale x 16 x half>)
 
 define <vscale x 32 x half> @ceil_nxv32f16(<vscale x 32 x half> %x) {
-; CHECK-LABEL: ceil_nxv32f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
-; CHECK-NEXT:    vfabs.v v16, v8
-; CHECK-NEXT:    vmflt.vf v0, v16, fa5
-; CHECK-NEXT:    fsrmi a0, 3
-; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: ceil_nxv32f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI5_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
+; ZVFH-NEXT:    vfabs.v v16, v8
+; ZVFH-NEXT:    vmflt.vf v0, v16, fa5
+; ZVFH-NEXT:    fsrmi a0, 3
+; ZVFH-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: ceil_nxv32f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v16, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v24, v16
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v24, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v24, v16, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v24, v24, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v16, v24, v16, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v16
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v16, v12
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v24, v16
+; ZVFHMIN-NEXT:    vmflt.vf v0, v24, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 3
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v24, v16, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v24, v24, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v16, v24, v16, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v12, v16
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 32 x half> @llvm.ceil.nxv32f16(<vscale x 32 x half> %x)
   ret <vscale x 32 x half> %a
 }
diff --git a/llvm/test/CodeGen/RISCV/rvv/ffloor-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/ffloor-sdnode.ll
index ec60b3ed3e0c88..97d84e91744038 100644
--- a/llvm/test/CodeGen/RISCV/rvv/ffloor-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/ffloor-sdnode.ll
@@ -1,124 +1,256 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+zvfh,+v -target-abi=ilp32d \
-; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,ZVFH
 ; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+zvfh,+v -target-abi=lp64d \
-; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,ZVFH
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfhmin,+zvfhmin,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s \
+; RUN:     --check-prefixes=CHECK,ZVFHMIN
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfhmin,+zvfhmin,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s \
+; RUN:     --check-prefixes=CHECK,ZVFHMIN
 
 define <vscale x 1 x half> @floor_nxv1f16(<vscale x 1 x half> %x) {
-; CHECK-LABEL: floor_nxv1f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
-; CHECK-NEXT:    vfabs.v v9, v8
-; CHECK-NEXT:    vmflt.vf v0, v9, fa5
-; CHECK-NEXT:    fsrmi a0, 2
-; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: floor_nxv1f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI0_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; ZVFH-NEXT:    vfabs.v v9, v8
+; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
+; ZVFH-NEXT:    fsrmi a0, 2
+; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: floor_nxv1f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v9
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 2
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; ZVFHMIN-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v9
+; ZVFHMIN-NEXT:    ret
   %a = call <vscale x 1 x half> @llvm.floor.nxv1f16(<vscale x 1 x half> %x)
   ret <vscale x 1 x half> %a
 }
 declare <vscale x 1 x half> @llvm.floor.nxv1f16(<vscale x 1 x half>)
 
 define <vscale x 2 x half> @floor_nxv2f16(<vscale x 2 x half> %x) {
-; CHECK-LABEL: floor_nxv2f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
-; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
-; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
-; CHECK-NEXT:    vfabs.v v9, v8
-; CHECK-NEXT:    vmflt.vf v0, v9, fa5
-; CHECK-NEXT:    fsrmi a0, 2
-; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
-; CHECK-NEXT:    fsrm a0
-; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
-; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: floor_nxv2f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    lui a0, %hi(.LCPI1_0)
+; ZVFH-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; ZVFH-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; ZVFH-NEXT:    vfabs.v v9, v8
+; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
+; ZVFH-NEXT:    fsrmi a0, 2
+; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; ZVFH-NEXT:    fsrm a0
+; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; ZVFH-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: floor_nxv2f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; ZVFHMIN-NEXT:    vfabs.v v8, v9
+; ZVFHMIN-NEXT:    lui a0, 307200
+; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
+; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
+; ZVFHMIN-NEXT:    fsrmi a0, 2
+; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
+; ZVFHMIN-NEXT:    fsrm a0
+; ...
[truncated]

@lukel97
Update fixed length vector tests
0350ec6 
Marking ftrunc as promoted means these are no longer expanded
@llvmbot llvmbot added the llvm:analysis Includes value tracking, cost tables and constant folding label Sep 16, 2024
@lukel97 lukel97 mentioned this pull request Sep 17, 2024
Merged
topperc
topperc approved these changes Sep 17, 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

@lukel97 lukel97 merged commit 8d7d4c2 into llvm:main Sep 18, 2024
9 checks passed
tmsri pushed a commit to tmsri/llvm-project that referenced this pull request Sep 19, 2024
@lukel97 @tmsri
[RISCV] Split fp rounding ops with zvfhmin nxv32f16 (llvm#108765)
aae38c1 
This adds zvfhmin test coverage for fceil, ffloor, fnearbyint, frint,
fround and froundeven and splits them at nxv32f16 to avoid crashing,
similarly to what we do for other nodes that we promote.

This also sets ftrunc to promote which was previously missing. We
already promote the VP version of it, vp_froundtozero.
Marking it as promoted affects some of the cost model tests since
they're no longer expanded.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jacquesguan jacquesguan jacquesguan approved these changes

@topperc topperc topperc approved these changes

@michaelmaitland michaelmaitland Awaiting requested review from michaelmaitland

@tclin914 tclin914 Awaiting requested review from tclin914

@wangpc-pp wangpc-pp Awaiting requested review from wangpc-pp

@mikhailramalho mikhailramalho Awaiting requested review from mikhailramalho

Assignees
No one assigned
Labels
backend:RISC-V llvm:analysis Includes value tracking, cost tables and constant folding
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@lukel97 @llvmbot @jacquesguan @topperc