blob: 5f6a79d0b8344838a446b30c0b6077a15506f29a [file] [log] [blame]
//-------------------------------------------------------------------------------------------------------
// Copyright (C) Microsoft Corporation and contributors. All rights reserved.
// Licensed under the MIT license. See LICENSE.txt file in the project root for full license information.
//-------------------------------------------------------------------------------------------------------
this.WScript.LoadScriptFile("..\\UnitTestFramework\\SimdJsHelpers.js");
function asmModule(stdlib, imports, buffer) {
"use asm";
var i4 = stdlib.SIMD.Int32x4;
var i4check = i4.check;
var i4splat = i4.splat;
var i4fromFloat32x4 = i4.fromFloat32x4;
var i4fromFloat32x4Bits = i4.fromFloat32x4Bits;
//var i4abs = i4.abs;
var i4neg = i4.neg;
var i4add = i4.add;
var i4sub = i4.sub;
var i4mul = i4.mul;
var i4swizzle = i4.swizzle;
var i4shuffle = i4.shuffle;
var i4lessThan = i4.lessThan;
var i4equal = i4.equal;
var i4greaterThan = i4.greaterThan;
var i4select = i4.select;
var i4and = i4.and;
var i4or = i4.or;
var i4xor = i4.xor;
var i4not = i4.not;
var i4load = i4.load;
var i4load1 = i4.load1;
var i4load2 = i4.load2;
var i4load3 = i4.load3;
var i4store = i4.store
var i4store1 = i4.store1;
var i4store2 = i4.store2;
var i4store3 = i4.store3;
//var i4shiftLeftByScalar = i4.shiftLeftByScalar;
//var i4shiftRightByScalar = i4.shiftRightByScalar;
//var i4shiftRightArithmeticByScalar = i4.shiftRightArithmeticByScalar;
var f4 = stdlib.SIMD.Float32x4;
var f4check = f4.check;
var f4splat = f4.splat;
var f4fromInt32x4 = f4.fromInt32x4;
var f4fromInt32x4Bits = f4.fromInt32x4Bits;
var f4abs = f4.abs;
var f4neg = f4.neg;
var f4add = f4.add;
var f4sub = f4.sub;
var f4mul = f4.mul;
var f4div = f4.div;
var f4min = f4.min;
var f4max = f4.max;
var f4sqrt = f4.sqrt;
var f4swizzle = f4.swizzle;
var f4shuffle = f4.shuffle;
var f4lessThan = f4.lessThan;
var f4lessThanOrEqual = f4.lessThanOrEqual;
var f4equal = f4.equal;
var f4notEqual = f4.notEqual;
var f4greaterThan = f4.greaterThan;
var f4greaterThanOrEqual = f4.greaterThanOrEqual;
var f4select = f4.select;
var f4load = f4.load;
var f4load1 = f4.load1;
var f4load2 = f4.load2;
var f4load3 = f4.load3;
var f4store = f4.store;
var f4store1 = f4.store1;
var f4store2 = f4.store2;
var f4store3 = f4.store3;
var fround = stdlib.Math.fround;
var globImportF4 = f4check(imports.g1); // global var import
var globImportI4 = i4check(imports.g2); // global var import
var g1 = f4(-5033.2,-3401.0,665.34,32234.1); // global var initialized
var g2 = i4(1065353216, -1073741824, -1077936128, 1082130432); // global var initialized
var gval = 1234;
var gval2 = 1234.0;
var OFFSET_1 = 10;
var OFFSET_2 = 15;
var loopCOUNT = 10;
var Int8Heap = new stdlib.Int8Array (buffer);
var Uint8Heap = new stdlib.Uint8Array (buffer);
var Int16Heap = new stdlib.Int16Array(buffer);
var Uint16Heap = new stdlib.Uint16Array(buffer);
var Int32Heap = new stdlib.Int32Array(buffer);
var Uint32Heap = new stdlib.Uint32Array(buffer);
var Float32Heap = new stdlib.Float32Array(buffer);
// Merges count number of SIMD vectors from srcIdx1 and srcIdx2 to dstIdx
function merge(srcIdx1, srcIdx2, dstIdx, count)
{
srcIdx1 = srcIdx1 | 0;
srcIdx2 = srcIdx2 | 0;
dstIdx = dstIdx | 0;
count = count | 0;
var i = 0;
for (; (i|0) < ((2 * count)|0); i = (i + 1) |0)
{
if ((((i | 0) % 2)|0) == 0)
{
f4store(Int8Heap, dstIdx, f4load(Int8Heap, srcIdx1));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx1 = (srcIdx1 + 16) | 0;
}
else
{
f4store(Int8Heap, dstIdx, f4load(Int8Heap, srcIdx2));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx2 = (srcIdx2 + 16) | 0;
}
dstIdx = (dstIdx + 16) | 0;
}
return -1;
}
function merge1(srcIdx1, srcIdx2, dstIdx, count)
{
srcIdx1 = srcIdx1 | 0;
srcIdx2 = srcIdx2 | 0;
dstIdx = dstIdx | 0;
count = count | 0;
var i = 0;
for (; (i|0) < ((2 * count)|0); i = (i + 1) |0)
{
if ((((i | 0) % 2)|0) == 0)
{
i4store(Int8Heap, dstIdx, i4load(Int8Heap, srcIdx1));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx1 = (srcIdx1 + 16) | 0;
}
else
{
i4store(Int8Heap, dstIdx, i4load(Int8Heap, srcIdx2));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx2 = (srcIdx2 + 16) | 0;
}
dstIdx = (dstIdx + 16) | 0;
}
return -1;
}
/*
function merge2(srcIdx1, srcIdx2, dstIdx, count)
{
srcIdx1 = srcIdx1 | 0;
srcIdx2 = srcIdx2 | 0;
dstIdx = dstIdx | 0;
count = count | 0;
var i = 0;
for (; (i|0) < ((2 * count)|0); i = (i + 1) |0)
{
if ((((i | 0) % 2)|0) == 0)
{
d2store(Int8Heap, dstIdx, d2load(Int8Heap, srcIdx1));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx1 = (srcIdx1 + 16) | 0;
}
else
{
d2store(Int8Heap, dstIdx, d2load(Int8Heap, srcIdx2));
//f4store(Int8Heap, (dstIdx + 0), f4load(Int8Heap, (srcIdx1 + 0)));
srcIdx2 = (srcIdx2 + 16) | 0;
}
dstIdx = (dstIdx + 16) | 0;
}
return -1;
}
*/
return {merge:merge
, merge1:merge1
/*, merge2:merge2*/};
}
var buffer = new ArrayBuffer(0x10000); //16mb min 2^12
//Reset or flush the buffer
function initF32(buffer) {
var values = new Float32Array( buffer );
for( var i=0; i < values.length ; ++i ) {
values[i] = i * 10;
}
return values.length;
}
function initI32(buffer) {
var values = new Int32Array( buffer );
for( var i=0; i < values.length ; ++i ) {
values[i] = i * 10;
}
return values.length;
}
function validateBuffer(buffer, count)
{
var f4;
var data = [
SIMD.Float32x4(0.0,10.0,20.0,30.0),
SIMD.Float32x4(40.0,50.0,60.0,70.0),
SIMD.Float32x4(80.0,90.0,100.0,110.0),
SIMD.Float32x4(120.0,130.0,140.0,150.0),
SIMD.Float32x4(160.0,170.0,180.0,190.0),
SIMD.Float32x4(200.0,210.0,220.0,230.0),
SIMD.Float32x4(240.0,250.0,260.0,270.0),
SIMD.Float32x4(280.0,290.0,300.0,310.0),
SIMD.Float32x4(320.0,330.0,340.0,350.0),
SIMD.Float32x4(360.0,370.0,380.0,390.0),
SIMD.Float32x4(0.0,10.0,20.0,30.0),
SIMD.Float32x4(200.0,210.0,220.0,230.0),
SIMD.Float32x4(40.0,50.0,60.0,70.0),
SIMD.Float32x4(240.0,250.0,260.0,270.0),
SIMD.Float32x4(80.0,90.0,100.0,110.0),
SIMD.Float32x4(280.0,290.0,300.0,310.0),
SIMD.Float32x4(120.0,130.0,140.0,150.0),
SIMD.Float32x4(320.0,330.0,340.0,350.0),
SIMD.Float32x4(160.0,170.0,180.0,190.0),
SIMD.Float32x4(360.0,370.0,380.0,390.0),
SIMD.Float32x4(800.0,810.0,820.0,830.0),
SIMD.Float32x4(840.0,850.0,860.0,870.0),
SIMD.Float32x4(880.0,890.0,900.0,910.0),
SIMD.Float32x4(920.0,930.0,940.0,950.0),
SIMD.Float32x4(960.0,970.0,980.0,990.0)
];
for (var i = 0; i < count/* * 16*/; i += 4)
{
f4 = SIMD.Float32x4.load(buffer, i);
equalSimd(data[i/4], f4, SIMD.Float32x4, "validateBuffer");
}
}
function validateBuffer1(buffer, count)
{
var i4;
var data = [
SIMD.Int32x4(0,10,20,30),
SIMD.Int32x4(40,50,60,70),
SIMD.Int32x4(80,90,100,110),
SIMD.Int32x4(120,130,140,150),
SIMD.Int32x4(160,170,180,190),
SIMD.Int32x4(200,210,220,230),
SIMD.Int32x4(240,250,260,270),
SIMD.Int32x4(280,290,300,310),
SIMD.Int32x4(320,330,340,350),
SIMD.Int32x4(360,370,380,390),
SIMD.Int32x4(0,10,20,30),
SIMD.Int32x4(200,210,220,230),
SIMD.Int32x4(40,50,60,70),
SIMD.Int32x4(240,250,260,270),
SIMD.Int32x4(80,90,100,110),
SIMD.Int32x4(280,290,300,310),
SIMD.Int32x4(120,130,140,150),
SIMD.Int32x4(320,330,340,350),
SIMD.Int32x4(160,170,180,190),
SIMD.Int32x4(360,370,380,390),
SIMD.Int32x4(800,810,820,830),
SIMD.Int32x4(840,850,860,870),
SIMD.Int32x4(880,890,900,910),
SIMD.Int32x4(920,930,940,950),
SIMD.Int32x4(960,970,980,990)
];
for (var i = 0; i < count/* * 16*/; i += 4)
{
i4 = SIMD.Int32x4.load(buffer, i);
equalSimd(data[i/4], i4, SIMD.Int32x4, "validateBuffer1");
}
}
inputLength = initF32(buffer);
//Module initialization
var m = asmModule(this, {g0:initF32(buffer),g1:SIMD.Float32x4(9,9,9,9), g2:SIMD.Int32x4(1, 2, 3, 4)}, buffer);
var values = new Float32Array(buffer);
//Resetting the buffer.
initF32(buffer);
var ret = m.merge(0, 5 * 16, 10 * 16, 5);
validateBuffer(values, 4 * 25);
initI32(buffer);
var ret = m.merge1(0, 5 * 16, 10 * 16, 5);
validateBuffer1(values, 4 * 25);
print("PASS");