Skip to content

[RISCV][VLOPT] Add support for 11.9 min/max instructions #112198

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 1 commit into from
Oct 14, 2024
Merged

[RISCV][VLOPT] Add support for 11.9 min/max instructions #112198

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 8 additions & 1 deletion llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -535,7 +535,14 @@ static bool isSupportedInstr(const MachineInstr &MI) {
// 11.8 Vector Integer Compare Instructions
// FIXME: Add support for 11.8 instructions
// 11.9. Vector Integer Min/Max Instructions
// FIXME: Add support for 11.9 instructions
case RISCV::VMINU_VV:
case RISCV::VMINU_VX:
case RISCV::VMIN_VV:
case RISCV::VMIN_VX:
case RISCV::VMAXU_VV:
case RISCV::VMAXU_VX:
case RISCV::VMAX_VV:
case RISCV::VMAX_VX:
// 11.10. Vector Single-Width Integer Multiply Instructions
case RISCV::VMUL_VV:
case RISCV::VMUL_VX:
Expand Down
158 changes: 158 additions & 0 deletions llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -642,7 +642,165 @@ define <vscale x 4 x i16> @vnsrl_wi(<vscale x 4 x i32> %a, <vscale x 4 x i16> %b
ret <vscale x 4 x i16> %2
}

define <vscale x 4 x i32> @vminu_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
; NOVLOPT-LABEL: vminu_vv:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a1, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vminu.vv v8, v8, v10
; NOVLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v8, v10
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vminu_vv:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; VLOPT-NEXT: vminu.vv v8, v8, v10
; VLOPT-NEXT: vadd.vv v8, v8, v10
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vminu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vminu_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
; NOVLOPT-LABEL: vminu_vx:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a2, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vminu.vx v10, v8, a0
; NOVLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v10, v8
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vminu_vx:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; VLOPT-NEXT: vminu.vx v10, v8, a0
; VLOPT-NEXT: vadd.vv v8, v10, v8
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vminu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmin_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
; NOVLOPT-LABEL: vmin_vv:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a1, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmin.vv v8, v8, v10
; NOVLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v8, v10
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmin_vv:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; VLOPT-NEXT: vmin.vv v8, v8, v10
; VLOPT-NEXT: vadd.vv v8, v8, v10
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmin.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmin_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
; NOVLOPT-LABEL: vmin_vx:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a2, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmin.vx v10, v8, a0
; NOVLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v10, v8
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmin_vx:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; VLOPT-NEXT: vmin.vx v10, v8, a0
; VLOPT-NEXT: vadd.vv v8, v10, v8
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmin.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmaxu_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
; NOVLOPT-LABEL: vmaxu_vv:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a1, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmaxu.vv v8, v8, v10
; NOVLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v8, v10
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmaxu_vv:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; VLOPT-NEXT: vmaxu.vv v8, v8, v10
; VLOPT-NEXT: vadd.vv v8, v8, v10
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmaxu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmaxu_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
; NOVLOPT-LABEL: vmaxu_vx:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a2, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmaxu.vx v10, v8, a0
; NOVLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v10, v8
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmaxu_vx:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; VLOPT-NEXT: vmaxu.vx v10, v8, a0
; VLOPT-NEXT: vadd.vv v8, v10, v8
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmaxu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmax_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
; NOVLOPT-LABEL: vmax_vv:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a1, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmax.vv v8, v8, v10
; NOVLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v8, v10
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmax_vv:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a0, e32, m2, ta, ma
; VLOPT-NEXT: vmax.vv v8, v8, v10
; VLOPT-NEXT: vadd.vv v8, v8, v10
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmax.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmax_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
; NOVLOPT-LABEL: vmax_vx:
; NOVLOPT: # %bb.0:
; NOVLOPT-NEXT: vsetvli a2, zero, e32, m2, ta, ma
; NOVLOPT-NEXT: vmax.vx v10, v8, a0
; NOVLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; NOVLOPT-NEXT: vadd.vv v8, v10, v8
; NOVLOPT-NEXT: ret
;
; VLOPT-LABEL: vmax_vx:
; VLOPT: # %bb.0:
; VLOPT-NEXT: vsetvli zero, a1, e32, m2, ta, ma
; VLOPT-NEXT: vmax.vx v10, v8, a0
; VLOPT-NEXT: vadd.vv v8, v10, v8
; VLOPT-NEXT: ret
%1 = call <vscale x 4 x i32> @llvm.riscv.vmax.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
%2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
ret <vscale x 4 x i32> %2
}

define <vscale x 4 x i32> @vmul_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
; NOVLOPT-LABEL: vmul_vv:
Expand Down
Loading