[AArch64][SVE] Predicated mul pattern

This is a simple predicated mul pattern, selecting from
select(p, mul(x, y), x).

Differential Revision: https://reviews.llvm.org/D149968

diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index b1bfdb39bcc1f0712c0dbe49836f0b459a804775..65ed502091e969fd011680457517828f3e05a581 100644 (file)
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -443,6 +443,7 @@ class fma_patfrags<SDPatternOperator intrinsic, SDPatternOperator add_zero, SDPa
 def AArch64fmla_m1 : fma_patfrags<int_aarch64_sve_fmla, AArch64fadd_p_nsz, AArch64fadd_p>;
 def AArch64fmls_m1 : fma_patfrags<int_aarch64_sve_fmls, AArch64fsub_p, AArch64fsub_p_nsz>;
 
+def AArch64mul_m1 : VSelectCommPredOrPassthruPatFrags<int_aarch64_sve_mul, AArch64mul_p>;
 def AArch64smax_m1 : VSelectCommPredOrPassthruPatFrags<int_aarch64_sve_smax, AArch64smax_p>;
 def AArch64umax_m1 : VSelectCommPredOrPassthruPatFrags<int_aarch64_sve_umax, AArch64umax_p>;
 def AArch64smin_m1 : VSelectCommPredOrPassthruPatFrags<int_aarch64_sve_smin, AArch64smin_p>;
@@ -529,7 +530,7 @@ let Predicates = [HasSVEorSME] in {
   defm UMIN_ZI   : sve_int_arith_imm1_unsigned<0b11, "umin", AArch64umin_p>;
 
   defm MUL_ZI     : sve_int_arith_imm2<"mul", AArch64mul_p>;
-  defm MUL_ZPmZ   : sve_int_bin_pred_arit_2<0b000, "mul",   "MUL_ZPZZ",   int_aarch64_sve_mul,   DestructiveBinaryComm>;
+  defm MUL_ZPmZ   : sve_int_bin_pred_arit_2<0b000, "mul",   "MUL_ZPZZ",   AArch64mul_m1,         DestructiveBinaryComm>;
   defm SMULH_ZPmZ : sve_int_bin_pred_arit_2<0b010, "smulh", "SMULH_ZPZZ", int_aarch64_sve_smulh, DestructiveBinaryComm>;
   defm UMULH_ZPmZ : sve_int_bin_pred_arit_2<0b011, "umulh", "UMULH_ZPZZ", int_aarch64_sve_umulh, DestructiveBinaryComm>;
 

diff --git a/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll b/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
index 0dcc5ce762ac2b1409a7f20886d5cbb5515683b8..f980378ef2db1dc499f18fe0d30b57c73352cddd 100644 (file)
--- a/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
@@ -125,9 +125,8 @@ define <vscale x 2 x i64> @mul_nxv2i64_x(<vscale x 2 x i64> %x, <vscale x 2 x i6
 ; CHECK-LABEL: mul_nxv2i64_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mul z1.d, z0.d, z1.d
 ; CHECK-NEXT:    cmpgt p0.d, p0/z, z2.d, #0
-; CHECK-NEXT:    mov z0.d, p0/m, z1.d
+; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 2 x i64> %n, zeroinitializer
@@ -140,9 +139,8 @@ define <vscale x 4 x i32> @mul_nxv4i32_x(<vscale x 4 x i32> %x, <vscale x 4 x i3
 ; CHECK-LABEL: mul_nxv4i32_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mul z1.s, z0.s, z1.s
 ; CHECK-NEXT:    cmpgt p0.s, p0/z, z2.s, #0
-; CHECK-NEXT:    mov z0.s, p0/m, z1.s
+; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 4 x i32> %n, zeroinitializer
@@ -155,9 +153,8 @@ define <vscale x 8 x i16> @mul_nxv8i16_x(<vscale x 8 x i16> %x, <vscale x 8 x i1
 ; CHECK-LABEL: mul_nxv8i16_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mul z1.h, z0.h, z1.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    mov z0.h, p0/m, z1.h
+; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 8 x i16> %n, zeroinitializer
@@ -170,9 +167,8 @@ define <vscale x 16 x i8> @mul_nxv16i8_x(<vscale x 16 x i8> %x, <vscale x 16 x i
 ; CHECK-LABEL: mul_nxv16i8_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mul z1.b, z0.b, z1.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
-; CHECK-NEXT:    mov z0.b, p0/m, z1.b
+; CHECK-NEXT:    mul z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 16 x i8> %n, zeroinitializer
@@ -1419,9 +1415,9 @@ define <vscale x 2 x i64> @mul_nxv2i64_y(<vscale x 2 x i64> %x, <vscale x 2 x i6
 ; CHECK-LABEL: mul_nxv2i64_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mul z0.d, z0.d, z1.d
 ; CHECK-NEXT:    cmpgt p0.d, p0/z, z2.d, #0
-; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    mul z1.d, p0/m, z1.d, z0.d
+; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 2 x i64> %n, zeroinitializer
@@ -1434,9 +1430,9 @@ define <vscale x 4 x i32> @mul_nxv4i32_y(<vscale x 4 x i32> %x, <vscale x 4 x i3
 ; CHECK-LABEL: mul_nxv4i32_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mul z0.s, z0.s, z1.s
 ; CHECK-NEXT:    cmpgt p0.s, p0/z, z2.s, #0
-; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    mul z1.s, p0/m, z1.s, z0.s
+; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 4 x i32> %n, zeroinitializer
@@ -1449,9 +1445,9 @@ define <vscale x 8 x i16> @mul_nxv8i16_y(<vscale x 8 x i16> %x, <vscale x 8 x i1
 ; CHECK-LABEL: mul_nxv8i16_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mul z0.h, z0.h, z1.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    mul z1.h, p0/m, z1.h, z0.h
+; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 8 x i16> %n, zeroinitializer
@@ -1464,9 +1460,9 @@ define <vscale x 16 x i8> @mul_nxv16i8_y(<vscale x 16 x i8> %x, <vscale x 16 x i
 ; CHECK-LABEL: mul_nxv16i8_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mul z0.b, z0.b, z1.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
-; CHECK-NEXT:    sel z0.b, p0, z0.b, z1.b
+; CHECK-NEXT:    mul z1.b, p0/m, z1.b, z0.b
+; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 16 x i8> %n, zeroinitializer