8256995: [vector] Improve broadcast operations

Co-authored-by: Paul Sandoz <psandoz@openjdk.org>
Co-authored-by: Sandhya Viswanathan <sviswanathan@openjdk.org>
Reviewed-by: kvn, vlivanov

Author: Paul Sandoz

src/hotspot/cpu/x86/x86.ad

@@ -3726,6 +3726,10 @@ instruct ReplB_reg(vec dst, rRegI src) %{
       assert(VM_Version::supports_avx512bw(), "required"); // 512-bit byte vectors assume AVX512BW
       int vlen_enc = vector_length_encoding(this);
       __ evpbroadcastb($dst$$XMMRegister, $src$$Register, vlen_enc);
+    } else if (VM_Version::supports_avx2()) {
+      int vlen_enc = vector_length_encoding(this);
+      __ movdl($dst$$XMMRegister, $src$$Register);
+      __ vpbroadcastb($dst$$XMMRegister, $dst$$XMMRegister, vlen_enc);
     } else {
       __ movdl($dst$$XMMRegister, $src$$Register);
       __ punpcklbw($dst$$XMMRegister, $dst$$XMMRegister);
@@ -3805,6 +3809,10 @@ instruct ReplS_reg(vec dst, rRegI src) %{
       assert(VM_Version::supports_avx512bw(), "required"); // 512-bit short vectors assume AVX512BW
       int vlen_enc = vector_length_encoding(this);
       __ evpbroadcastw($dst$$XMMRegister, $src$$Register, vlen_enc);
+    } else if (VM_Version::supports_avx2()) {
+      int vlen_enc = vector_length_encoding(this);
+      __ movdl($dst$$XMMRegister, $src$$Register);
+      __ vpbroadcastw($dst$$XMMRegister, $dst$$XMMRegister, vlen_enc);
     } else {
       __ movdl($dst$$XMMRegister, $src$$Register);
       __ pshuflw($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
@@ -3880,6 +3888,10 @@ instruct ReplI_reg(vec dst, rRegI src) %{
     if (vlen == 16 || VM_Version::supports_avx512vl()) { // AVX512VL for <512bit operands
       int vlen_enc = vector_length_encoding(this);
       __ evpbroadcastd($dst$$XMMRegister, $src$$Register, vlen_enc);
+    } else if (VM_Version::supports_avx2()) {
+      int vlen_enc = vector_length_encoding(this);
+      __ movdl($dst$$XMMRegister, $src$$Register);
+      __ vpbroadcastd($dst$$XMMRegister, $dst$$XMMRegister, vlen_enc);
     } else {
       __ movdl($dst$$XMMRegister, $src$$Register);
       __ pshufd($dst$$XMMRegister, $dst$$XMMRegister, 0x00);
@@ -3975,6 +3987,11 @@ instruct ReplL_reg(vec dst, rRegL src) %{
     } else if (vlen == 8 || VM_Version::supports_avx512vl()) { // AVX512VL for <512bit operands
       int vlen_enc = vector_length_encoding(this);
       __ evpbroadcastq($dst$$XMMRegister, $src$$Register, vlen_enc);
+    } else if (VM_Version::supports_avx2()) {
+      assert(vlen == 4, "sanity");
+      int vlen_enc = vector_length_encoding(this);
+      __ movdq($dst$$XMMRegister, $src$$Register);
+      __ vpbroadcastq($dst$$XMMRegister, $dst$$XMMRegister, vlen_enc);
     } else {
       assert(vlen == 4, "sanity");
       __ movdq($dst$$XMMRegister, $src$$Register);

src/jdk.incubator.vector/share/classes/jdk/incubator/vector/DoubleVector.java

@@ -400,7 +400,7 @@ static boolean doBinTest(int cond, double a, double b) {
     /*package-private*/
     @ForceInline
     static long toBits(double e) {
-        return  Double.doubleToLongBits(e);
+        return  Double.doubleToRawLongBits(e);
     }
 
     /*package-private*/

src/jdk.incubator.vector/share/classes/jdk/incubator/vector/FloatVector.java

@@ -400,7 +400,7 @@ static boolean doBinTest(int cond, float a, float b) {
     /*package-private*/
     @ForceInline
     static long toBits(float e) {
-        return  Float.floatToIntBits(e);
+        return  Float.floatToRawIntBits(e);
     }
 
     /*package-private*/

src/jdk.incubator.vector/share/classes/jdk/incubator/vector/X-Vector.java.template

@@ -404,7 +404,7 @@ public abstract class $abstractvectortype$ extends AbstractVector<$Boxtype$> {
     /*package-private*/
     @ForceInline
     static long toBits($type$ e) {
-        return {#if[FP]? $Type$.$type$To$Bitstype$Bits(e): e};
+        return {#if[FP]? $Type$.$type$ToRaw$Bitstype$Bits(e): e};
     }
 
     /*package-private*/