[llvm] [RISCV][VLOPT] Add vector fp-conversion instruction to isSupportedInstr (PR #122033)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 7 17:12:14 PST 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-risc-v
Author: Michael Maitland (michaelmaitland)
<details>
<summary>Changes</summary>
When these instructions are marked nofpexcept, we can optimize them. There are some added toggles in the output, likley because other noexcept fp instructions are not part of isSupportedInstr yet. We may want to avoid marking an instruction as isSupported in the future if any of its FP users are missing nofpexcept to avoid added toggles. However, we seem to get some GPRs back as a result of this change, which may outweigh the cost of avoiding extra toggles.
The plan is to follow this patch up with added support for more FP instructions in the same way. The instructions in this patch are a natural starting point because they allow us to test with integer instructions which have good support already.
---
Patch is 256.36 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/122033.diff
17 Files Affected:
- (modified) llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp (+26)
- (modified) llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll (+9-9)
- (modified) llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll (+150-144)
- (modified) llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll (+150-144)
- (modified) llvm/test/CodeGen/RISCV/rvv/setcc-fp-vp.ll (+229-229)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfadd-vp.ll (+107-108)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfdiv-vp.ll (+103-104)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfmax-vp.ll (+37-37)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfmin-vp.ll (+37-37)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfmul-vp.ll (+103-104)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfptosi-vp-mask.ll (+4-4)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfptosi-vp.ll (+18-31)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfptoui-vp-mask.ll (+4-4)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfptoui-vp.ll (+18-31)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfsqrt-vp.ll (+34-32)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfsub-vp.ll (+103-104)
- (modified) llvm/test/CodeGen/RISCV/rvv/vl-opt.mir (+13)
``````````diff
diff --git a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
index 9ecb0453fb11cc..7b7c664e35e09c 100644
--- a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
@@ -962,6 +962,32 @@ static bool isSupportedInstr(const MachineInstr &MI) {
case RISCV::VIOTA_M:
case RISCV::VID_V:
return true;
+
+ // 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:
+ // 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:
+ // 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:
+ return MI.getFlags() & MachineInstr::NoFPExcept;
}
return false;
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll
index 15793eaada0783..66952cac8e00d3 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp.ll
@@ -1254,12 +1254,10 @@ define void @copysign_neg_trunc_v4f16_v4f32(ptr %x, ptr %y) {
define void @copysign_neg_trunc_v3f16_v3f32(ptr %x, ptr %y) {
; ZVFH-LABEL: copysign_neg_trunc_v3f16_v3f32:
; ZVFH: # %bb.0:
-; ZVFH-NEXT: vsetivli zero, 3, e32, m1, ta, ma
+; ZVFH-NEXT: vsetivli zero, 3, e16, mf2, ta, ma
; ZVFH-NEXT: vle32.v v8, (a1)
; ZVFH-NEXT: vle16.v v9, (a0)
-; ZVFH-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; ZVFH-NEXT: vfncvt.f.f.w v10, v8
-; ZVFH-NEXT: vsetivli zero, 3, e16, mf2, ta, ma
; ZVFH-NEXT: vfsgnjn.vv v8, v9, v10
; ZVFH-NEXT: vse16.v v8, (a0)
; ZVFH-NEXT: ret
@@ -1272,9 +1270,7 @@ define void @copysign_neg_trunc_v3f16_v3f32(ptr %x, ptr %y) {
; ZVFHMIN-NEXT: lui a1, 8
; ZVFHMIN-NEXT: addi a2, a1, -1
; ZVFHMIN-NEXT: vand.vx v8, v8, a2
-; ZVFHMIN-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vfncvt.f.f.w v10, v9
-; ZVFHMIN-NEXT: vsetivli zero, 3, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vxor.vx v9, v10, a1
; ZVFHMIN-NEXT: vand.vx v9, v9, a1
; ZVFHMIN-NEXT: vor.vv v8, v8, v9
@@ -4013,9 +4009,10 @@ define void @trunc_v6f16(ptr %x) {
; ZVFH-NEXT: vsetivli zero, 8, e16, m1, ta, ma
; ZVFH-NEXT: vfabs.v v9, v8
; ZVFH-NEXT: vmflt.vf v0, v9, fa5
+; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, ma
; ZVFH-NEXT: vfcvt.rtz.x.f.v v9, v8, v0.t
; ZVFH-NEXT: vfcvt.f.x.v v9, v9, v0.t
-; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, mu
+; ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, mu
; ZVFH-NEXT: vfsgnj.vv v8, v9, v8, v0.t
; ZVFH-NEXT: vse16.v v8, (a0)
; ZVFH-NEXT: ret
@@ -4197,10 +4194,11 @@ define void @ceil_v6f16(ptr %x) {
; ZVFH-NEXT: vfabs.v v9, v8
; ZVFH-NEXT: vmflt.vf v0, v9, fa5
; ZVFH-NEXT: fsrmi a1, 3
+; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, ma
; ZVFH-NEXT: vfcvt.x.f.v v9, v8, v0.t
; ZVFH-NEXT: fsrm a1
; ZVFH-NEXT: vfcvt.f.x.v v9, v9, v0.t
-; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, mu
+; ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, mu
; ZVFH-NEXT: vfsgnj.vv v8, v9, v8, v0.t
; ZVFH-NEXT: vse16.v v8, (a0)
; ZVFH-NEXT: ret
@@ -4388,10 +4386,11 @@ define void @floor_v6f16(ptr %x) {
; ZVFH-NEXT: vfabs.v v9, v8
; ZVFH-NEXT: vmflt.vf v0, v9, fa5
; ZVFH-NEXT: fsrmi a1, 2
+; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, ma
; ZVFH-NEXT: vfcvt.x.f.v v9, v8, v0.t
; ZVFH-NEXT: fsrm a1
; ZVFH-NEXT: vfcvt.f.x.v v9, v9, v0.t
-; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, mu
+; ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, mu
; ZVFH-NEXT: vfsgnj.vv v8, v9, v8, v0.t
; ZVFH-NEXT: vse16.v v8, (a0)
; ZVFH-NEXT: ret
@@ -4579,10 +4578,11 @@ define void @round_v6f16(ptr %x) {
; ZVFH-NEXT: vfabs.v v9, v8
; ZVFH-NEXT: vmflt.vf v0, v9, fa5
; ZVFH-NEXT: fsrmi a1, 4
+; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, ma
; ZVFH-NEXT: vfcvt.x.f.v v9, v8, v0.t
; ZVFH-NEXT: fsrm a1
; ZVFH-NEXT: vfcvt.f.x.v v9, v9, v0.t
-; ZVFH-NEXT: vsetivli zero, 6, e16, m1, ta, mu
+; ZVFH-NEXT: vsetvli zero, zero, e16, m1, ta, mu
; ZVFH-NEXT: vfsgnj.vv v8, v9, v8, v0.t
; ZVFH-NEXT: vse16.v v8, (a0)
; ZVFH-NEXT: ret
diff --git a/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
index 7649d60def111b..33fe73a097e32f 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
@@ -582,14 +582,14 @@ define <vscale x 1 x half> @vfmax_vv_nxv1f16(<vscale x 1 x half> %va, <vscale x
;
; ZVFHMIN-LABEL: vfmax_vv_nxv1f16:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vmv1r.v v10, v0
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v11, v11, v0.t
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v9
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v9, v11, v8, v0
; ZVFHMIN-NEXT: vmv1r.v v0, v10
; ZVFHMIN-NEXT: vmfeq.vv v0, v8, v8, v0.t
@@ -616,13 +616,13 @@ define <vscale x 1 x half> @vfmax_vv_nxv1f16_unmasked(<vscale x 1 x half> %va, <
;
; ZVFHMIN-LABEL: vfmax_vv_nxv1f16_unmasked:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v10, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v10, v10
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v9
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v9, v10, v8, v0
; ZVFHMIN-NEXT: vmfeq.vv v0, v8, v8
; ZVFHMIN-NEXT: vmerge.vvm v8, v8, v10, v0
@@ -652,14 +652,14 @@ define <vscale x 2 x half> @vfmax_vv_nxv2f16(<vscale x 2 x half> %va, <vscale x
;
; ZVFHMIN-LABEL: vfmax_vv_nxv2f16:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vmv1r.v v10, v0
; ZVFHMIN-NEXT: vfwcvt.f.f.v v11, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v11, v11, v0.t
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v9
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v9, v11, v8, v0
; ZVFHMIN-NEXT: vmv1r.v v0, v10
; ZVFHMIN-NEXT: vmfeq.vv v0, v8, v8, v0.t
@@ -686,13 +686,13 @@ define <vscale x 2 x half> @vfmax_vv_nxv2f16_unmasked(<vscale x 2 x half> %va, <
;
; ZVFHMIN-LABEL: vfmax_vv_nxv2f16_unmasked:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v10, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v10, v10
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, mf2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v8, v9
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v9, v10, v8, v0
; ZVFHMIN-NEXT: vmfeq.vv v0, v8, v8
; ZVFHMIN-NEXT: vmerge.vvm v8, v8, v10, v0
@@ -722,15 +722,15 @@ define <vscale x 4 x half> @vfmax_vv_nxv4f16(<vscale x 4 x half> %va, <vscale x
;
; ZVFHMIN-LABEL: vfmax_vv_nxv4f16:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m1, ta, ma
; ZVFHMIN-NEXT: vmv1r.v v10, v0
; ZVFHMIN-NEXT: vfwcvt.f.f.v v12, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v8, v12, v12, v0.t
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v14, v9
; ZVFHMIN-NEXT: vmv1r.v v0, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v16, v12, v14, v0
; ZVFHMIN-NEXT: vmv1r.v v0, v10
; ZVFHMIN-NEXT: vmfeq.vv v8, v14, v14, v0.t
@@ -758,13 +758,13 @@ define <vscale x 4 x half> @vfmax_vv_nxv4f16_unmasked(<vscale x 4 x half> %va, <
;
; ZVFHMIN-LABEL: vfmax_vv_nxv4f16_unmasked:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m1, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v10, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v10, v10
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v12, v9
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v8, v10, v12, v0
; ZVFHMIN-NEXT: vmfeq.vv v0, v12, v12
; ZVFHMIN-NEXT: vmerge.vvm v10, v12, v10, v0
@@ -796,15 +796,15 @@ define <vscale x 8 x half> @vfmax_vv_nxv8f16(<vscale x 8 x half> %va, <vscale x
;
; ZVFHMIN-LABEL: vfmax_vv_nxv8f16:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m2, ta, ma
; ZVFHMIN-NEXT: vmv1r.v v12, v0
; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v8, v16, v16, v0.t
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v20, v10
; ZVFHMIN-NEXT: vmv1r.v v0, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v24, v16, v20, v0
; ZVFHMIN-NEXT: vmv1r.v v0, v12
; ZVFHMIN-NEXT: vmfeq.vv v8, v20, v20, v0.t
@@ -832,13 +832,13 @@ define <vscale x 8 x half> @vfmax_vv_nxv8f16_unmasked(<vscale x 8 x half> %va, <
;
; ZVFHMIN-LABEL: vfmax_vv_nxv8f16_unmasked:
; ZVFHMIN: # %bb.0:
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v12, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v12, v12
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m2, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v10
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m4, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v8, v12, v16, v0
; ZVFHMIN-NEXT: vmfeq.vv v0, v16, v16
; ZVFHMIN-NEXT: vmerge.vvm v12, v16, v12, v0
@@ -876,15 +876,15 @@ define <vscale x 16 x half> @vfmax_vv_nxv16f16(<vscale x 16 x half> %va, <vscale
; ZVFHMIN-NEXT: slli a1, a1, 3
; ZVFHMIN-NEXT: sub sp, sp, a1
; ZVFHMIN-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; ZVFHMIN-NEXT: vmv1r.v v7, v0
; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v8, v24, v24, v0.t
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m4, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v12
; ZVFHMIN-NEXT: vmv1r.v v0, v8
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
; ZVFHMIN-NEXT: vmerge.vvm v8, v24, v16, v0
; ZVFHMIN-NEXT: addi a0, sp, 16
; ZVFHMIN-NEXT: vs8r.v v8, (a0) # Unknown-size Folded Spill
@@ -927,10 +927,10 @@ define <vscale x 16 x half> @vfmax_vv_nxv16f16_unmasked(<vscale x 16 x half> %va
; ZVFHMIN-NEXT: slli a1, a1, 3
; ZVFHMIN-NEXT: sub sp, sp, a1
; ZVFHMIN-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v8
; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v12
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
; ZVFHMIN-NEXT: vmfeq.vv v0, v16, v16
; ZVFHMIN-NEXT: vmfeq.vv v7, v24, v24
; ZVFHMIN-NEXT: vmerge.vvm v8, v16, v24, v0
@@ -995,64 +995,62 @@ define <vscale x 32 x half> @vfmax_vv_nxv32f16(<vscale x 32 x half> %va, <vscale
; ZVFHMIN-NEXT: add a1, a2, a1
; ZVFHMIN-NEXT: sub sp, sp, a1
; ZVFHMIN-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x21, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 33 * vlenb
-; ZVFHMIN-NEXT: csrr a2, vlenb
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m4, ta, ma
-; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v12
; ZVFHMIN-NEXT: csrr a1, vlenb
-; ZVFHMIN-NEXT: li a3, 24
-; ZVFHMIN-NEXT: mul a1, a1, a3
+; ZVFHMIN-NEXT: li a2, 25
+; ZVFHMIN-NEXT: mul a1, a1, a2
; ZVFHMIN-NEXT: add a1, sp, a1
; ZVFHMIN-NEXT: addi a1, a1, 16
-; ZVFHMIN-NEXT: vs8r.v v24, (a1) # Unknown-size Folded Spill
+; ZVFHMIN-NEXT: vs8r.v v16, (a1) # Unknown-size Folded Spill
+; ZVFHMIN-NEXT: vsetvli a1, zero, e8, mf2, ta, ma
+; ZVFHMIN-NEXT: vmv8r.v v16, v8
+; ZVFHMIN-NEXT: csrr a2, vlenb
; ZVFHMIN-NEXT: slli a1, a2, 1
; ZVFHMIN-NEXT: srli a2, a2, 2
; ZVFHMIN-NEXT: sub a3, a0, a1
; ZVFHMIN-NEXT: csrr a4, vlenb
-; ZVFHMIN-NEXT: slli a4, a4, 5
+; ZVFHMIN-NEXT: li a5, 24
+; ZVFHMIN-NEXT: mul a4, a4, a5
; ZVFHMIN-NEXT: add a4, sp, a4
; ZVFHMIN-NEXT: addi a4, a4, 16
; ZVFHMIN-NEXT: vs1r.v v0, (a4) # Unknown-size Folded Spill
-; ZVFHMIN-NEXT: vsetvli a4, zero, e8, mf2, ta, ma
-; ZVFHMIN-NEXT: vslidedown.vx v12, v0, a2
+; ZVFHMIN-NEXT: vslidedown.vx v0, v0, a2
+; ZVFHMIN-NEXT: addi a2, sp, 16
+; ZVFHMIN-NEXT: vs1r.v v0, (a2) # Unknown-size Folded Spill
; ZVFHMIN-NEXT: sltu a2, a0, a3
; ZVFHMIN-NEXT: addi a2, a2, -1
; ZVFHMIN-NEXT: and a2, a2, a3
-; ZVFHMIN-NEXT: vmv1r.v v0, v12
-; ZVFHMIN-NEXT: vsetvli zero, a2, e32, m8, ta, ma
-; ZVFHMIN-NEXT: vmfeq.vv v13, v24, v24, v0.t
-; ZVFHMIN-NEXT: vmv8r.v v0, v16
; ZVFHMIN-NEXT: csrr a3, vlenb
; ZVFHMIN-NEXT: slli a3, a3, 4
; ZVFHMIN-NEXT: add a3, sp, a3
; ZVFHMIN-NEXT: addi a3, a3, 16
-; ZVFHMIN-NEXT: vs8r.v v16, (a3) # Unknown-size Folded Spill
-; ZVFHMIN-NEXT: vsetvli a3, zero, e16, m4, ta, ma
-; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v4
-; ZVFHMIN-NEXT: vmv1r.v v0, v13
-; ZVFHMIN-NEXT: csrr a3, vlenb
-; ZVFHMIN-NEXT: li a4, 24
-; ZVFHMIN-NEXT: mul a3, a3, a4
-; ZVFHMIN-NEXT: add a3, sp, a3
-; ZVFHMIN-NEXT: addi a3, a3, 16
-; ZVFHMIN-NEXT: vl8r.v v24, (a3) # Unknown-size Folded Reload
-; ZVFHMIN-NEXT: vsetvli zero, a2, e32, m8, ta, ma
-; ZVFHMIN-NEXT: vmerge.vvm v24, v24, v16, v0
+; ZVFHMIN-NEXT: vs8r.v v8, (a3) # Unknown-size Folded Spill
+; ZVFHMIN-NEXT: vsetvli zero, a2, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v20
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vmfeq.vv v12, v24, v24, v0.t
; ZVFHMIN-NEXT: csrr a2, vlenb
-; ZVFHMIN-NEXT: slli a2, a2, 3
+; ZVFHMIN-NEXT: li a3, 25
+; ZVFHMIN-NEXT: mul a2, a2, a3
; ZVFHMIN-NEXT: add a2, sp, a2
; ZVFHMIN-NEXT: addi a2, a2, 16
-; ZVFHMIN-NEXT: vs8r.v v24, (a2) # Unknown-size Folded Spill
+; ZVFHMIN-NEXT: vl8r.v v0, (a2) # Unknown-size Folded Reload
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v4
; ZVFHMIN-NEXT: vmv1r.v v0, v12
-; ZVFHMIN-NEXT: vmfeq.vv v13, v16, v16, v0.t
-; ZVFHMIN-NEXT: vmv1r.v v0, v13
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vmerge.vvm v8, v24, v16, v0
; ZVFHMIN-NEXT: csrr a2, vlenb
-; ZVFHMIN-NEXT: li a3, 24
-; ZVFHMIN-NEXT: mul a2, a2, a3
+; ZVFHMIN-NEXT: slli a2, a2, 3
; ZVFHMIN-NEXT: add a2, sp, a2
; ZVFHMIN-NEXT: addi a2, a2, 16
-; ZVFHMIN-NEXT: vl8r.v v24, (a2) # Unknown-size Folded Reload
-; ZVFHMIN-NEXT: vmerge.vvm v16, v16, v24, v0
+; ZVFHMIN-NEXT: vs8r.v v8, (a2) # Unknown-size Folded Spill
+; ZVFHMIN-NEXT: addi a2, sp, 16
+; ZVFHMIN-NEXT: vl1r.v v8, (a2) # Unknown-size Folded Reload
+; ZVFHMIN-NEXT: vmv1r.v v0, v8
+; ZVFHMIN-NEXT: vmfeq.vv v12, v16, v16, v0.t
; ZVFHMIN-NEXT: vmv1r.v v0, v12
+; ZVFHMIN-NEXT: vmerge.vvm v16, v16, v24, v0
+; ZVFHMIN-NEXT: vmv1r.v v0, v8
; ZVFHMIN-NEXT: csrr a2, vlenb
; ZVFHMIN-NEXT: slli a2, a2, 3
; ZVFHMIN-NEXT: add a2, sp, a2
@@ -1070,35 +1068,43 @@ define <vscale x 32 x half> @vfmax_vv_nxv32f16(<vscale x 32 x half> %va, <vscale
; ZVFHMIN-NEXT: # %bb.1:
; ZVFHMIN-NEXT: mv a0, a1
; ZVFHMIN-NEXT: .LBB22_2:
-; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v8
-; ZVFHMIN-NEXT: csrr a1, vlenb
-; ZVFHMIN-NEXT: slli a1, a1, 5
-; ZVFHMIN-NEXT: add a1, sp, a1
-; ZVFHMIN-NEXT: addi a1, a1, 16
-; ZVFHMIN-NEXT: vl1r.v v0, (a1) # Unknown-size Folded Reload
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m8, ta, ma
-; ZVFHMIN-NEXT: vmfeq.vv v24, v16, v16, v0.t
-; ZVFHMIN-NEXT: vmv8r.v v8, v16
-; ZVFHMIN-NEXT: addi a1, sp, 16
-; ZVFHMIN-NEXT: vs8r.v v16, (a1) # Unknown-size Folded Spill
; ZVFHMIN-NEXT: csrr a1, vlenb
; ZVFHMIN-NEXT: slli a1, a1, 4
; ZVFHMIN-NEXT: add a1, sp, a1
; ZVFHMIN-NEXT: addi a1, a1, 16
-; ZVFHMIN-NEXT: vl8r.v v0, (a1) # Unknown-size Folded Reload
-; ZVFHMIN-NEXT: vsetvli a1, zero, e16, m4, ta, ma
-; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v0
-; ZVFHMIN-NEXT: vmv1r.v v0, v24
-; ZVFHMIN-NEXT: vsetvli zero, a0, e32, m8, ta, ma
-; ZVFHMIN-NEXT: vmerge.vvm v24, v8, v16, v0
+; ZVFHMIN-NEXT: vl8r.v v16, (a1) # Unknown-size Folded Reload
+; ZVFHMIN-NEXT: vsetvli zero, a0, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vfwcvt.f.f.v v24, v16
; ZVFHMIN-NEXT: csrr a0, vlenb
; ZVFHMIN-NEXT: li a1, 24
; ZVFHMIN-NEXT: mul a0, a0, a1
; ZVFHMIN-NEXT: add a0, sp, a0
; ZVFHMIN-NEXT: addi a0, a0, 16
+; ZVFHMIN-NEXT: vl1r.v v0, (a0) # Unknown-size Folded Reload
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vmfeq.vv v8, v24, v24, v0.t
+; ZVFHMIN-NEXT: addi a0, sp, 16
; ZVFHMIN-NEXT: vs8r.v v24, (a0) # Unknown-size Folded Spill
; ZVFHMIN-NEXT: csrr a0, vlenb
-; ZVFHMIN-NEXT: slli a0, a0, 5
+; ZVFHMIN-NEXT: li a1, 25
+; ZVFHMIN-NEXT: mul a0, a0, a1
+; ZVFHMIN-NEXT: add a0, sp, a0
+; ZVFHMIN-NEXT: addi a0, a0, 16
+; ZVFHMIN-NEXT: vl8r.v v0, (a0) # Unknown-size Folded Reload
+; ZVFHMIN-NEXT: vsetvli zero, zero, e16, m4, ta, ma
+; ZVFHMIN-NEXT: vfwcvt.f.f.v v16, v0
+; ZVFHMIN-NEXT: vmv1r.v v0, v8
+; ZVFHMIN-NEXT: vsetvli zero, zero, e32, m8, ta, ma
+; ZVFHMIN-NEXT: vmerge.vvm...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/122033
More information about the llvm-commits
mailing list