我正在尝试为4x4矩阵乘法运行C++函数。 2天后,它终于可以使用了,但是效果不理想。
通常会将参数反馈给函数,然后返回以下行:
dataHeap2.set( new Uint8Array(data2.buffer) );
产生错误“未捕获的RangeError:源太大”
乍一看,它只是一个具有16个元素的普通Float32Array,但在查看其内部的缓冲区大小后,它似乎有所不同
console.log(data2.buffer.bufferLength);
结果不是期望的64个字节,而是庞大的数字,例如3342345。这是问题吗?我发现了一种通过手动复制值(如下所示)并将其返回的解决方法,然后问题消失了。不幸的是,这使我的代码比直接在缓冲区上运行要慢得多。
// bad solution - which works
for(var i = 0; i < 16; i++) {
dataTarget[i] = result[i];
}
希望今晚能找到一个更好的解决方案,并且希望继续使用此函数,因为它使用的是使用ASM.JS + SIMD的C++编译代码来对多个字符进行动画处理。在纯JavaScript中,它仍然太慢。这是整个功能。我很确定data2正在使用Emscripten的HEAP,并且我想避免它。
matrix4multiply = function(data, data2) {
// Import function from Emscripten generated file
var mmul_vec4 = Module.cwrap(
'mmul_vec4', 'number', ['number', 'number', 'number']
);
var dataTarget = new Float32Array(16);
// Get data byte size, allocate memory on Emscripten heap, and get pointer
var nDataBytes = dataTarget.length * dataTarget.BYTES_PER_ELEMENT;
// First matrix copy data to Emscripten heap
var dataPtr = Module._malloc(nDataBytes);
var dataPtr2 = Module._malloc(nDataBytes);
var dataPtr3 = Module._malloc(nDataBytes);
var dataHeap = new Uint8Array(Module.HEAPU8.buffer, dataPtr, nDataBytes);
dataHeap.set( new Uint8Array(data.buffer) );
// second matrix allocate and copy to emscripten's heap
var dataHeap2 = new Uint8Array(Module.HEAPU8.buffer, dataPtr2, nDataBytes);
dataHeap2.set( new Uint8Array(data2.buffer) );
// target heap
var targetHeap = new Uint8Array(Module.HEAPU8.buffer, dataPtr3, nDataBytes);
targetHeap.set( new Uint8Array(dataTarget.buffer) );
// Call the function by passing a number pointing to the byte location of
// the array of pointers on the Emscripten heap. Emscripten knows what to do!
mmul_vec4(dataHeap.byteOffset, dataHeap2.byteOffset, targetHeap.byteOffset);
// get result
var result = new Float32Array(targetHeap.buffer, targetHeap.byteOffset, 16);
// bad solution - which works
//for(var i = 0; i < 16; i++) {
// dataTarget[i] = result[i];
//}
// Free memory
Module._free(dataHeap.byteOffset);
Module._free(dataHeap2.byteOffset);
Module._free(targetHeap.byteOffset);
return result;
}
编辑:简化版本不必担心malloc等
new Uint8Array(Module.HEAPU8.buffer, this.dataPtr, 64).set( new Uint8Array(data.buffer, data.byteOffset, 64) );
// second matrix allocate and copy to emscripten's heap
new Uint8Array(Module.HEAPU8.buffer, this.dataPtr + 72, 64).set( new Uint8Array(data2.buffer, data2.byteOffset, 64) );
// multiply first two parameters and return in the last one
this.mmul_vec4(this.dataPtr, this.dataPtr + 72, this.dataPtr + 144);
// like that it works, unfotunately copying is being made here
return new Float32Array(Module.HEAPU8.buffer.slice(this.dataPtr + 144, this.dataPtr + 208));
// with this version (if uncommented) there's just white screen(but it looks like the game is working.
//return new Float32Array(Module.HEAPU8.buffer, this.dataPtr + 144, 16);
最佳答案
要使用头64个字节的data2
,请指定偏移量和长度。
dataHeap2.set( new Uint8Array(data2.buffer, data2.byteOffset, nDataBytes) );
关于javascript - Emscripten未捕获的RangeError : Source is too large,多个Float32Arrays,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/31550198/