[llvm] [LLVM][CodeGen][SME] Make FMA decision independent of streaming mode. (PR #222598)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Sep 10 03:41:15 PDT 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-aarch64
Author: Paul Walker (paulwalker-arm)
<details>
<summary>Changes</summary>
The relevant bfloat instructions are available in streaming mode and thus we can benefit from the same BFMLAL based lowering.
---
Patch is 52.56 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/222598.diff
2 Files Affected:
- (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+1-2)
- (modified) llvm/test/CodeGen/AArch64/sve-bf16-combines.ll (+655-320)
``````````diff
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index a3823f8552a3a..ffd90e68f070c 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -20369,8 +20369,7 @@ bool AArch64TargetLowering::isFMAFasterThanFMulAndFAdd(
case MVT::f64:
return Subtarget->hasFPARMv8();
case MVT::bf16:
- return VT.isScalableVector() && Subtarget->hasBF16() &&
- Subtarget->isNonStreamingSVEorSME2Available();
+ return VT.isScalableVector() && Subtarget->hasBF16();
default:
break;
}
diff --git a/llvm/test/CodeGen/AArch64/sve-bf16-combines.ll b/llvm/test/CodeGen/AArch64/sve-bf16-combines.ll
index 9a45f67aad946..eab91bde321b0 100644
--- a/llvm/test/CodeGen/AArch64/sve-bf16-combines.ll
+++ b/llvm/test/CodeGen/AArch64/sve-bf16-combines.ll
@@ -1,32 +1,52 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mattr=+sve,+bf16 < %s | FileCheck %s --check-prefixes=SVE
-; RUN: llc -mattr=+sve,+bf16,+sve-b16b16 < %s | FileCheck %s --check-prefixes=SVE-B16B16
+; RUN: llc -mattr=+sve,+bf16 < %s | FileCheck %s --check-prefixes=SVE,SVE-NONSTREAMING
+; RUN: llc -mattr=+sve,+bf16,+sve-b16b16 < %s | FileCheck %s --check-prefixes=SVE-B16B16,SVE-B16B16-NONSTREAMING
+
+; RUN: llc -mattr=+sme --force-streaming < %s | FileCheck %s --check-prefixes=SVE,SVE-STREAMING
+; RUN: llc -mattr=+sme2,+sve-b16b16 --force-streaming < %s | FileCheck %s --check-prefixes=SVE-B16B16,B16B16-STREAMING
target triple = "aarch64-unknown-linux-gnu"
define <vscale x 8 x bfloat> @fmla_nxv8bf16(<vscale x 8 x bfloat> %acc, <vscale x 8 x bfloat> %m1, <vscale x 8 x bfloat> %m2) {
-; SVE-LABEL: fmla_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: movi v3.2d, #0000000000000000
-; SVE-NEXT: uunpkhi z4.s, z2.h
-; SVE-NEXT: uunpkhi z5.s, z1.h
-; SVE-NEXT: uunpklo z2.s, z2.h
-; SVE-NEXT: uunpklo z1.s, z1.h
-; SVE-NEXT: ptrue p0.s
-; SVE-NEXT: zip2 z6.h, z3.h, z0.h
-; SVE-NEXT: zip1 z0.h, z3.h, z0.h
-; SVE-NEXT: bfmlalb z6.s, z5.h, z4.h
-; SVE-NEXT: bfmlalb z0.s, z1.h, z2.h
-; SVE-NEXT: bfcvt z1.h, p0/m, z6.s
-; SVE-NEXT: bfcvt z0.h, p0/m, z0.s
-; SVE-NEXT: uzp1 z0.h, z0.h, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fmla_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: movi v3.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: uunpkhi z4.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpkhi z5.s, z1.h
+; SVE-NONSTREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-NONSTREAMING-NEXT: ptrue p0.s
+; SVE-NONSTREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z0.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z6.s, z5.h, z4.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z0.s, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p0/m, z6.s
+; SVE-NONSTREAMING-NEXT: bfcvt z0.h, p0/m, z0.s
+; SVE-NONSTREAMING-NEXT: uzp1 z0.h, z0.h, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fmla_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: ptrue p0.h
; SVE-B16B16-NEXT: bfmla z0.h, p0/m, z1.h, z2.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fmla_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: mov z3.h, #0 // =0x0
+; SVE-STREAMING-NEXT: uunpkhi z4.s, z2.h
+; SVE-STREAMING-NEXT: uunpkhi z5.s, z1.h
+; SVE-STREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-STREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-STREAMING-NEXT: ptrue p0.s
+; SVE-STREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z0.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: bfmlalb z6.s, z5.h, z4.h
+; SVE-STREAMING-NEXT: bfmlalb z0.s, z1.h, z2.h
+; SVE-STREAMING-NEXT: bfcvt z1.h, p0/m, z6.s
+; SVE-STREAMING-NEXT: bfcvt z0.h, p0/m, z0.s
+; SVE-STREAMING-NEXT: uzp1 z0.h, z0.h, z1.h
+; SVE-STREAMING-NEXT: ret
%mul = fmul contract <vscale x 8 x bfloat> %m1, %m2
%res = fadd contract <vscale x 8 x bfloat> %acc, %mul
ret <vscale x 8 x bfloat> %res
@@ -73,30 +93,49 @@ define <vscale x 2 x bfloat> @fmla_nxv2bf16(<vscale x 2 x bfloat> %acc, <vscale
}
define <vscale x 8 x bfloat> @fmls_nxv8bf16(<vscale x 8 x bfloat> %acc, <vscale x 8 x bfloat> %m1, <vscale x 8 x bfloat> %m2) {
-; SVE-LABEL: fmls_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: ptrue p0.h
-; SVE-NEXT: movi v3.2d, #0000000000000000
-; SVE-NEXT: uunpkhi z5.s, z2.h
-; SVE-NEXT: uunpklo z2.s, z2.h
-; SVE-NEXT: fneg z1.h, p0/m, z1.h
-; SVE-NEXT: ptrue p0.s
-; SVE-NEXT: zip2 z6.h, z3.h, z0.h
-; SVE-NEXT: zip1 z0.h, z3.h, z0.h
-; SVE-NEXT: uunpkhi z4.s, z1.h
-; SVE-NEXT: uunpklo z1.s, z1.h
-; SVE-NEXT: bfmlalb z6.s, z4.h, z5.h
-; SVE-NEXT: bfmlalb z0.s, z1.h, z2.h
-; SVE-NEXT: bfcvt z1.h, p0/m, z6.s
-; SVE-NEXT: bfcvt z0.h, p0/m, z0.s
-; SVE-NEXT: uzp1 z0.h, z0.h, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fmls_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: ptrue p0.h
+; SVE-NONSTREAMING-NEXT: movi v3.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: uunpkhi z5.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-NONSTREAMING-NEXT: fneg z1.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ptrue p0.s
+; SVE-NONSTREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z0.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: uunpkhi z4.s, z1.h
+; SVE-NONSTREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z6.s, z4.h, z5.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z0.s, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p0/m, z6.s
+; SVE-NONSTREAMING-NEXT: bfcvt z0.h, p0/m, z0.s
+; SVE-NONSTREAMING-NEXT: uzp1 z0.h, z0.h, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fmls_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: ptrue p0.h
; SVE-B16B16-NEXT: bfmls z0.h, p0/m, z1.h, z2.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fmls_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: ptrue p0.h
+; SVE-STREAMING-NEXT: mov z3.h, #0 // =0x0
+; SVE-STREAMING-NEXT: uunpkhi z5.s, z2.h
+; SVE-STREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-STREAMING-NEXT: fneg z1.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ptrue p0.s
+; SVE-STREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z0.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: uunpkhi z4.s, z1.h
+; SVE-STREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-STREAMING-NEXT: bfmlalb z6.s, z4.h, z5.h
+; SVE-STREAMING-NEXT: bfmlalb z0.s, z1.h, z2.h
+; SVE-STREAMING-NEXT: bfcvt z1.h, p0/m, z6.s
+; SVE-STREAMING-NEXT: bfcvt z0.h, p0/m, z0.s
+; SVE-STREAMING-NEXT: uzp1 z0.h, z0.h, z1.h
+; SVE-STREAMING-NEXT: ret
%mul = fmul contract <vscale x 8 x bfloat> %m1, %m2
%res = fsub contract <vscale x 8 x bfloat> %acc, %mul
ret <vscale x 8 x bfloat> %res
@@ -145,28 +184,46 @@ define <vscale x 2 x bfloat> @fmls_nxv2bf16(<vscale x 2 x bfloat> %acc, <vscale
}
define <vscale x 8 x bfloat> @fmla_sel_nxv8bf16(<vscale x 8 x i1> %pred, <vscale x 8 x bfloat> %acc, <vscale x 8 x bfloat> %m1, <vscale x 8 x bfloat> %m2) {
-; SVE-LABEL: fmla_sel_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: movi v3.2d, #0000000000000000
-; SVE-NEXT: uunpkhi z4.s, z2.h
-; SVE-NEXT: uunpkhi z5.s, z1.h
-; SVE-NEXT: uunpklo z2.s, z2.h
-; SVE-NEXT: uunpklo z1.s, z1.h
-; SVE-NEXT: ptrue p1.s
-; SVE-NEXT: zip2 z6.h, z3.h, z0.h
-; SVE-NEXT: zip1 z3.h, z3.h, z0.h
-; SVE-NEXT: bfmlalb z6.s, z5.h, z4.h
-; SVE-NEXT: bfmlalb z3.s, z1.h, z2.h
-; SVE-NEXT: bfcvt z1.h, p1/m, z6.s
-; SVE-NEXT: bfcvt z2.h, p1/m, z3.s
-; SVE-NEXT: uzp1 z1.h, z2.h, z1.h
-; SVE-NEXT: mov z0.h, p0/m, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fmla_sel_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: movi v3.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: uunpkhi z4.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpkhi z5.s, z1.h
+; SVE-NONSTREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-NONSTREAMING-NEXT: ptrue p1.s
+; SVE-NONSTREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z3.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z6.s, z5.h, z4.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z3.s, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p1/m, z6.s
+; SVE-NONSTREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-NONSTREAMING-NEXT: uzp1 z1.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fmla_sel_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: bfmla z0.h, p0/m, z1.h, z2.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fmla_sel_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: mov z3.h, #0 // =0x0
+; SVE-STREAMING-NEXT: uunpkhi z4.s, z2.h
+; SVE-STREAMING-NEXT: uunpkhi z5.s, z1.h
+; SVE-STREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-STREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-STREAMING-NEXT: ptrue p1.s
+; SVE-STREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z3.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: bfmlalb z6.s, z5.h, z4.h
+; SVE-STREAMING-NEXT: bfmlalb z3.s, z1.h, z2.h
+; SVE-STREAMING-NEXT: bfcvt z1.h, p1/m, z6.s
+; SVE-STREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-STREAMING-NEXT: uzp1 z1.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ret
%mul = fmul contract <vscale x 8 x bfloat> %m1, %m2
%add = fadd contract <vscale x 8 x bfloat> %acc, %mul
%res = select <vscale x 8 x i1> %pred, <vscale x 8 x bfloat> %add, <vscale x 8 x bfloat> %acc
@@ -213,30 +270,50 @@ define <vscale x 2 x bfloat> @fmla_sel_nxv2bf16(<vscale x 2 x i1> %pred, <vscale
}
define <vscale x 8 x bfloat> @fmls_sel_nxv8bf16(<vscale x 8 x i1> %pred, <vscale x 8 x bfloat> %acc, <vscale x 8 x bfloat> %m1, <vscale x 8 x bfloat> %m2) {
-; SVE-LABEL: fmls_sel_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: ptrue p1.h
-; SVE-NEXT: movi v3.2d, #0000000000000000
-; SVE-NEXT: uunpkhi z5.s, z2.h
-; SVE-NEXT: uunpklo z2.s, z2.h
-; SVE-NEXT: fneg z1.h, p1/m, z1.h
-; SVE-NEXT: ptrue p1.s
-; SVE-NEXT: zip2 z6.h, z3.h, z0.h
-; SVE-NEXT: zip1 z3.h, z3.h, z0.h
-; SVE-NEXT: uunpkhi z4.s, z1.h
-; SVE-NEXT: uunpklo z1.s, z1.h
-; SVE-NEXT: bfmlalb z6.s, z4.h, z5.h
-; SVE-NEXT: bfmlalb z3.s, z1.h, z2.h
-; SVE-NEXT: bfcvt z1.h, p1/m, z6.s
-; SVE-NEXT: bfcvt z2.h, p1/m, z3.s
-; SVE-NEXT: uzp1 z1.h, z2.h, z1.h
-; SVE-NEXT: mov z0.h, p0/m, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fmls_sel_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: ptrue p1.h
+; SVE-NONSTREAMING-NEXT: movi v3.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: uunpkhi z5.s, z2.h
+; SVE-NONSTREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-NONSTREAMING-NEXT: fneg z1.h, p1/m, z1.h
+; SVE-NONSTREAMING-NEXT: ptrue p1.s
+; SVE-NONSTREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z3.h, z3.h, z0.h
+; SVE-NONSTREAMING-NEXT: uunpkhi z4.s, z1.h
+; SVE-NONSTREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z6.s, z4.h, z5.h
+; SVE-NONSTREAMING-NEXT: bfmlalb z3.s, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p1/m, z6.s
+; SVE-NONSTREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-NONSTREAMING-NEXT: uzp1 z1.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fmls_sel_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: bfmls z0.h, p0/m, z1.h, z2.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fmls_sel_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: ptrue p1.h
+; SVE-STREAMING-NEXT: mov z3.h, #0 // =0x0
+; SVE-STREAMING-NEXT: uunpkhi z5.s, z2.h
+; SVE-STREAMING-NEXT: uunpklo z2.s, z2.h
+; SVE-STREAMING-NEXT: fneg z1.h, p1/m, z1.h
+; SVE-STREAMING-NEXT: ptrue p1.s
+; SVE-STREAMING-NEXT: zip2 z6.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z3.h, z3.h, z0.h
+; SVE-STREAMING-NEXT: uunpkhi z4.s, z1.h
+; SVE-STREAMING-NEXT: uunpklo z1.s, z1.h
+; SVE-STREAMING-NEXT: bfmlalb z6.s, z4.h, z5.h
+; SVE-STREAMING-NEXT: bfmlalb z3.s, z1.h, z2.h
+; SVE-STREAMING-NEXT: bfcvt z1.h, p1/m, z6.s
+; SVE-STREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-STREAMING-NEXT: uzp1 z1.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ret
%mul = fmul contract <vscale x 8 x bfloat> %m1, %m2
%sub = fsub contract <vscale x 8 x bfloat> %acc, %mul
%res = select <vscale x 8 x i1> %pred, <vscale x 8 x bfloat> %sub, <vscale x 8 x bfloat> %acc
@@ -286,78 +363,126 @@ define <vscale x 2 x bfloat> @fmls_sel_nxv2bf16(<vscale x 2 x i1> %pred, <vscale
}
define <vscale x 8 x bfloat> @fadd_sel_nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x i1> %mask) {
-; SVE-LABEL: fadd_sel_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: movi v2.2d, #0000000000000000
-; SVE-NEXT: ptrue p1.s
-; SVE-NEXT: zip2 z3.h, z2.h, z1.h
-; SVE-NEXT: zip2 z4.h, z2.h, z0.h
-; SVE-NEXT: zip1 z1.h, z2.h, z1.h
-; SVE-NEXT: zip1 z2.h, z2.h, z0.h
-; SVE-NEXT: fadd z3.s, z4.s, z3.s
-; SVE-NEXT: fadd z1.s, z2.s, z1.s
-; SVE-NEXT: bfcvt z2.h, p1/m, z3.s
-; SVE-NEXT: bfcvt z1.h, p1/m, z1.s
-; SVE-NEXT: uzp1 z1.h, z1.h, z2.h
-; SVE-NEXT: mov z0.h, p0/m, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fadd_sel_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: movi v2.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: ptrue p1.s
+; SVE-NONSTREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: fadd z3.s, z4.s, z3.s
+; SVE-NONSTREAMING-NEXT: fadd z1.s, z2.s, z1.s
+; SVE-NONSTREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-NONSTREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fadd_sel_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: bfadd z0.h, p0/m, z0.h, z1.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fadd_sel_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: mov z2.h, #0 // =0x0
+; SVE-STREAMING-NEXT: ptrue p1.s
+; SVE-STREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: fadd z3.s, z4.s, z3.s
+; SVE-STREAMING-NEXT: fadd z1.s, z2.s, z1.s
+; SVE-STREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-STREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-STREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-STREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ret
%sel = select <vscale x 8 x i1> %mask, <vscale x 8 x bfloat> %b, <vscale x 8 x bfloat> zeroinitializer
%fadd = fadd nsz <vscale x 8 x bfloat> %a, %sel
ret <vscale x 8 x bfloat> %fadd
}
define <vscale x 8 x bfloat> @fsub_sel_nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x i1> %mask) {
-; SVE-LABEL: fsub_sel_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: movi v2.2d, #0000000000000000
-; SVE-NEXT: ptrue p1.s
-; SVE-NEXT: zip2 z3.h, z2.h, z1.h
-; SVE-NEXT: zip2 z4.h, z2.h, z0.h
-; SVE-NEXT: zip1 z1.h, z2.h, z1.h
-; SVE-NEXT: zip1 z2.h, z2.h, z0.h
-; SVE-NEXT: fsub z3.s, z4.s, z3.s
-; SVE-NEXT: fsub z1.s, z2.s, z1.s
-; SVE-NEXT: bfcvt z2.h, p1/m, z3.s
-; SVE-NEXT: bfcvt z1.h, p1/m, z1.s
-; SVE-NEXT: uzp1 z1.h, z1.h, z2.h
-; SVE-NEXT: mov z0.h, p0/m, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fsub_sel_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: movi v2.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: ptrue p1.s
+; SVE-NONSTREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: fsub z3.s, z4.s, z3.s
+; SVE-NONSTREAMING-NEXT: fsub z1.s, z2.s, z1.s
+; SVE-NONSTREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-NONSTREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fsub_sel_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: bfsub z0.h, p0/m, z0.h, z1.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fsub_sel_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: mov z2.h, #0 // =0x0
+; SVE-STREAMING-NEXT: ptrue p1.s
+; SVE-STREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: fsub z3.s, z4.s, z3.s
+; SVE-STREAMING-NEXT: fsub z1.s, z2.s, z1.s
+; SVE-STREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-STREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-STREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-STREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ret
%sel = select <vscale x 8 x i1> %mask, <vscale x 8 x bfloat> %b, <vscale x 8 x bfloat> zeroinitializer
%fsub = fsub <vscale x 8 x bfloat> %a, %sel
ret <vscale x 8 x bfloat> %fsub
}
define <vscale x 8 x bfloat> @fadd_sel_negzero_nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x i1> %mask) {
-; SVE-LABEL: fadd_sel_negzero_nxv8bf16:
-; SVE: // %bb.0:
-; SVE-NEXT: movi v2.2d, #0000000000000000
-; SVE-NEXT: ptrue p1.s
-; SVE-NEXT: zip2 z3.h, z2.h, z1.h
-; SVE-NEXT: zip2 z4.h, z2.h, z0.h
-; SVE-NEXT: zip1 z1.h, z2.h, z1.h
-; SVE-NEXT: zip1 z2.h, z2.h, z0.h
-; SVE-NEXT: fadd z3.s, z4.s, z3.s
-; SVE-NEXT: fadd z1.s, z2.s, z1.s
-; SVE-NEXT: bfcvt z2.h, p1/m, z3.s
-; SVE-NEXT: bfcvt z1.h, p1/m, z1.s
-; SVE-NEXT: uzp1 z1.h, z1.h, z2.h
-; SVE-NEXT: mov z0.h, p0/m, z1.h
-; SVE-NEXT: ret
+; SVE-NONSTREAMING-LABEL: fadd_sel_negzero_nxv8bf16:
+; SVE-NONSTREAMING: // %bb.0:
+; SVE-NONSTREAMING-NEXT: movi v2.2d, #0000000000000000
+; SVE-NONSTREAMING-NEXT: ptrue p1.s
+; SVE-NONSTREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-NONSTREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-NONSTREAMING-NEXT: fadd z3.s, z4.s, z3.s
+; SVE-NONSTREAMING-NEXT: fadd z1.s, z2.s, z1.s
+; SVE-NONSTREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-NONSTREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-NONSTREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-NONSTREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-NONSTREAMING-NEXT: ret
;
; SVE-B16B16-LABEL: fadd_sel_negzero_nxv8bf16:
; SVE-B16B16: // %bb.0:
; SVE-B16B16-NEXT: bfadd z0.h, p0/m, z0.h, z1.h
; SVE-B16B16-NEXT: ret
+;
+; SVE-STREAMING-LABEL: fadd_sel_negzero_nxv8bf16:
+; SVE-STREAMING: // %bb.0:
+; SVE-STREAMING-NEXT: mov z2.h, #0 // =0x0
+; SVE-STREAMING-NEXT: ptrue p1.s
+; SVE-STREAMING-NEXT: zip2 z3.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip2 z4.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: zip1 z1.h, z2.h, z1.h
+; SVE-STREAMING-NEXT: zip1 z2.h, z2.h, z0.h
+; SVE-STREAMING-NEXT: fadd z3.s, z4.s, z3.s
+; SVE-STREAMING-NEXT: fadd z1.s, z2.s, z1.s
+; SVE-STREAMING-NEXT: bfcvt z2.h, p1/m, z3.s
+; SVE-STREAMING-NEXT: bfcvt z1.h, p1/m, z1.s
+; SVE-STREAMING-NEXT: uzp1 z1.h, z1.h, z2.h
+; SVE-STREAMING-NEXT: mov z0.h, p0/m, z1.h
+; SVE-STREAMING-NEXT: ret
%nz = fneg <vscale x 8 x bfloat> zeroinitializer
%sel = select <vscale x 8 x i1> %mask, <vscale x 8 x bfloat> %b, <vscale x 8 x bfloat> %nz
%fadd = fadd <vscale x 8 x bfloat> %a, %sel
@@ -365,26 +490,42 @@ define <vscale x 8 x bfloat> @fadd_sel_negzero_nxv8bf16(<vscale x 8 x bfloat> %a
}
define <vscale x 8 x bfloat> @fsub_sel_negzero_nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x i1> %mask) {
-;...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/222598
More information about the llvm-commits
mailing list