Skip to content

[RISCV][VLOPT] Add Vector FP instructions to getOperandInfo #121609

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

Merged
merged 5 commits into from
Jan 5, 2025
Merged

[RISCV][VLOPT] Add Vector FP instructions to getOperandInfo #121609

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
6c3c5c7
[RISCV][VLOPT] Add Vector FP instructions to getOperandInfo
michaelmaitland Jan 3, 2025
9de5551
fixup! clang-format
michaelmaitland Jan 3, 2025
e21886a
fixup! fix comment placement
michaelmaitland Jan 3, 2025
dc429af
fixup! move vfwmacc
michaelmaitland Jan 3, 2025
7511a1d
fixup! clang-format
michaelmaitland Jan 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
103 changes: 99 additions & 4 deletions llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -456,6 +456,49 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
case RISCV::VCOMPRESS_VM:
// Vector Element Index Instruction
case RISCV::VID_V:
// Vector Single-Width Floating-Point Add/Subtract Instructions
case RISCV::VFADD_VF:
case RISCV::VFADD_VV:
case RISCV::VFSUB_VF:
case RISCV::VFSUB_VV:
case RISCV::VFRSUB_VF:
// Vector Single-Width Floating-Point Multiply/Divide Instructions
case RISCV::VFMUL_VF:
case RISCV::VFMUL_VV:
case RISCV::VFDIV_VF:
case RISCV::VFDIV_VV:
case RISCV::VFRDIV_VF:
// Vector Floating-Point Square-Root Instruction
case RISCV::VFSQRT_V:
// Vector Floating-Point Reciprocal Square-Root Estimate Instruction
case RISCV::VFRSQRT7_V:
// Vector Floating-Point Reciprocal Estimate Instruction
case RISCV::VFREC7_V:
// Vector Floating-Point MIN/MAX Instructions
case RISCV::VFMIN_VF:
case RISCV::VFMIN_VV:
case RISCV::VFMAX_VF:
case RISCV::VFMAX_VV:
// Vector Floating-Point Sign-Injection Instructions
case RISCV::VFSGNJ_VF:
case RISCV::VFSGNJ_VV:
case RISCV::VFSGNJN_VV:
case RISCV::VFSGNJN_VF:
case RISCV::VFSGNJX_VF:
case RISCV::VFSGNJX_VV:
// Vector Floating-Point Classify Instruction
case RISCV::VFCLASS_V:
// Vector Floating-Point Move Instruction
case RISCV::VFMV_V_F:
// Single-Width Floating-Point/Integer Type-Convert Instructions
case RISCV::VFCVT_XU_F_V:
case RISCV::VFCVT_X_F_V:
case RISCV::VFCVT_RTZ_XU_F_V:
case RISCV::VFCVT_RTZ_X_F_V:
case RISCV::VFCVT_F_XU_V:
case RISCV::VFCVT_F_X_V:
// Vector Floating-Point Merge Instruction
case RISCV::VFMERGE_VFM:
return OperandInfo(MIVLMul, MILog2SEW);

// Vector Widening Integer Add/Subtract
Expand Down Expand Up @@ -488,7 +531,24 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
case RISCV::VWMACC_VX:
case RISCV::VWMACCSU_VV:
case RISCV::VWMACCSU_VX:
case RISCV::VWMACCUS_VX: {
case RISCV::VWMACCUS_VX:
// Vector Widening Floating-Point Add/Subtract Instructions
// Dest EEW=2*SEW and EMUL=2*LMUL. Source EEW=SEW and EMUL=LMUL.
case RISCV::VFWADD_VV:
case RISCV::VFWADD_VF:
case RISCV::VFWSUB_VV:
case RISCV::VFWSUB_VF:
// Vector Widening Floating-Point Multiply
case RISCV::VFWMUL_VF:
case RISCV::VFWMUL_VV:
// Widening Floating-Point/Integer Type-Convert Instructions
case RISCV::VFWCVT_XU_F_V:
case RISCV::VFWCVT_X_F_V:
case RISCV::VFWCVT_RTZ_XU_F_V:
case RISCV::VFWCVT_RTZ_X_F_V:
case RISCV::VFWCVT_F_XU_V:
case RISCV::VFWCVT_F_X_V:
case RISCV::VFWCVT_F_F_V: {
unsigned Log2EEW = IsMODef ? MILog2SEW + 1 : MILog2SEW;
RISCVII::VLMUL EMUL =
IsMODef ? RISCVVType::twoTimesVLMUL(MIVLMul) : MIVLMul;
Expand All @@ -503,7 +563,21 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
case RISCV::VWADD_WV:
case RISCV::VWADD_WX:
case RISCV::VWSUB_WV:
case RISCV::VWSUB_WX: {
case RISCV::VWSUB_WX:
// Vector Widening Floating-Point Add/Subtract Instructions
case RISCV::VFWADD_WF:
case RISCV::VFWADD_WV:
case RISCV::VFWSUB_WF:
case RISCV::VFWSUB_WV:
// Vector Widening Floating-Point Fused Multiply-Add Instructions
case RISCV::VFWMACC_VF:
case RISCV::VFWMACC_VV:
case RISCV::VFWNMACC_VF:
case RISCV::VFWNMACC_VV:
case RISCV::VFWMSAC_VF:
case RISCV::VFWMSAC_VV:
case RISCV::VFWNMSAC_VF:
case RISCV::VFWNMSAC_VV: {
bool IsOp1 = HasPassthru ? MO.getOperandNo() == 2 : MO.getOperandNo() == 1;
bool TwoTimes = IsMODef || IsOp1;
unsigned Log2EEW = TwoTimes ? MILog2SEW + 1 : MILog2SEW;
Expand Down Expand Up @@ -539,7 +613,16 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
case RISCV::VNCLIPU_WX:
case RISCV::VNCLIP_WI:
case RISCV::VNCLIP_WV:
case RISCV::VNCLIP_WX: {
case RISCV::VNCLIP_WX:
// Narrowing Floating-Point/Integer Type-Convert Instructions
case RISCV::VFNCVT_XU_F_W:
case RISCV::VFNCVT_X_F_W:
case RISCV::VFNCVT_RTZ_XU_F_W:
case RISCV::VFNCVT_RTZ_X_F_W:
case RISCV::VFNCVT_F_XU_W:
case RISCV::VFNCVT_F_X_W:
case RISCV::VFNCVT_F_F_W:
case RISCV::VFNCVT_ROD_F_F_W: {
bool IsOp1 = HasPassthru ? MO.getOperandNo() == 2 : MO.getOperandNo() == 1;
bool TwoTimes = IsOp1;
unsigned Log2EEW = TwoTimes ? MILog2SEW + 1 : MILog2SEW;
Expand Down Expand Up @@ -615,7 +698,19 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
case RISCV::VMADC_VI:
case RISCV::VMADC_VX:
case RISCV::VMSBC_VV:
case RISCV::VMSBC_VX: {
case RISCV::VMSBC_VX:
// 13.13. Vector Floating-Point Compare Instructions
// Dest EEW=1 and EMUL=(EEW/SEW)*LMUL. Source EEW=SEW EMUL=LMUL.
case RISCV::VMFEQ_VF:
case RISCV::VMFEQ_VV:
case RISCV::VMFNE_VF:
case RISCV::VMFNE_VV:
case RISCV::VMFLT_VF:
case RISCV::VMFLT_VV:
case RISCV::VMFLE_VF:
case RISCV::VMFLE_VV:
case RISCV::VMFGT_VF:
case RISCV::VMFGE_VF: {
if (IsMODef)
return OperandInfo(RISCVVType::getEMULEqualsEEWDivSEWTimesLMUL(0, MI), 0);
return OperandInfo(MIVLMul, MILog2SEW);
Expand Down
42 changes: 14 additions & 28 deletions llvm/test/CodeGen/RISCV/rvv/vfwmacc-vp.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -143,9 +143,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32(<vscale x 1 x half> %va, half %b,
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
Expand All @@ -170,9 +169,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32_commute(<vscale x 1 x half> %va,
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32_commute:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v11, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v11, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v10, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v11, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
Expand All @@ -198,9 +196,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32_unmasked(<vscale x 1 x half> %va,
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32_unmasked:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
Expand All @@ -225,9 +222,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32_tu(<vscale x 1 x half> %va, half
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32_tu:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, tu, mu
Expand All @@ -254,9 +250,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32_commute_tu(<vscale x 1 x half> %v
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32_commute_tu:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, tu, mu
Expand All @@ -283,9 +278,8 @@ define <vscale x 1 x float> @vfmacc_vf_nxv1f32_unmasked_tu(<vscale x 1 x half> %
; ZVFHMIN-LABEL: vfmacc_vf_nxv1f32_unmasked_tu:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, tu, ma
Expand Down Expand Up @@ -362,9 +356,8 @@ define <vscale x 2 x float> @vfmacc_vf_nxv2f32(<vscale x 2 x half> %va, half %b,
; ZVFHMIN-LABEL: vfmacc_vf_nxv2f32:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
Expand All @@ -389,9 +382,8 @@ define <vscale x 2 x float> @vfmacc_vf_nxv2f32_unmasked(<vscale x 2 x half> %va,
; ZVFHMIN-LABEL: vfmacc_vf_nxv2f32_unmasked:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v10, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v10
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
Expand Down Expand Up @@ -468,9 +460,8 @@ define <vscale x 4 x float> @vfmacc_vf_nxv4f32(<vscale x 4 x half> %va, half %b,
; ZVFHMIN-LABEL: vfmacc_vf_nxv4f32:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m1, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v12, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m1, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v12, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v14, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v12, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
Expand All @@ -495,9 +486,8 @@ define <vscale x 4 x float> @vfmacc_vf_nxv4f32_unmasked(<vscale x 4 x half> %va,
; ZVFHMIN-LABEL: vfmacc_vf_nxv4f32_unmasked:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m1, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v12, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m1, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v12, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v14, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v12
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
Expand Down Expand Up @@ -574,9 +564,8 @@ define <vscale x 8 x float> @vfmacc_vf_nxv8f32(<vscale x 8 x half> %va, half %b,
; ZVFHMIN-LABEL: vfmacc_vf_nxv8f32:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v16, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v16, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v20, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v16, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
Expand All @@ -601,9 +590,8 @@ define <vscale x 8 x float> @vfmacc_vf_nxv8f32_unmasked(<vscale x 8 x half> %va,
; ZVFHMIN-LABEL: vfmacc_vf_nxv8f32_unmasked:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v16, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m2, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v16, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v20, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v16
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
Expand Down Expand Up @@ -696,9 +684,8 @@ define <vscale x 16 x float> @vfmacc_vf_nxv16f32(<vscale x 16 x half> %va, half
; ZVFHMIN-LABEL: vfmacc_vf_nxv16f32:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v4, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v4, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v8, v0.t
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v4, v0.t
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
Expand All @@ -723,9 +710,8 @@ define <vscale x 16 x float> @vfmacc_vf_nxv16f32_unmasked(<vscale x 16 x half> %
; ZVFHMIN-LABEL: vfmacc_vf_nxv16f32_unmasked:
; ZVFHMIN: # %bb.0:
; ZVFHMIN-NEXT: fmv.x.h a1, fa0
; ZVFHMIN-NEXT: vsetvli a2, zero, e16, m4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v24, a1
; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; ZVFHMIN-NEXT: vmv.v.x v24, a1
; ZVFHMIN-NEXT: vfwcvt.f.f.v v0, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v24
; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
Expand Down
Loading
Loading