astojanov

#include <immintrin.h>

static inline __m256i avx_mm256_div_epi16_division(const __m256i &a_epi16, const __m256i &b_epi16) {
    //
    // Convert to two 32-bit integers
    //
    const __m256i a_hi_epi32       = _mm256_srai_epi32(a_epi16, 16);
    const __m256i a_lo_epi32_shift = _mm256_slli_epi32(a_epi16, 16);
    const __m256i a_lo_epi32       = _mm256_srai_epi32(a_lo_epi32_shift, 16);
    const __m256i b_hi_epi32       = _mm256_srai_epi32(b_epi16, 16);

astojanov

#include <immintrin.h>

//
// Performs division on 16-bit elements, using floating point division
//
static inline __m128i sse_mm_div_epi16_division(const __m128i &a_epi16, const __m128i &b_epi16) {

    const __m128i lo_mask = _mm_set1_epi32(0xFFFF);
    //
    // Convert to two 32-bit integers

astojanov

#include <immintrin.h>

#include <cstdint>
#include <iostream>
#include <iomanip>
#include <cassert>

//
// Transpose 8x8 registers
//

astojanov

Java HotSpot(TM) 64-Bit Server VM warning: printing of assembly code is enabled; turning on DebugNonSafepoints to gain additional output
CompilerOracle: print *JVector8.dot
Compiled method (c1)     807  864 %     3       ch.ethz.acl.ngen.precison.JVector8::dot @ 35 (134 bytes)
 total in heap  [0x000000010f6dd010,0x000000010f6ddbb0] = 2976
 relocation     [0x000000010f6dd138,0x000000010f6dd1b0] = 120
 main code      [0x000000010f6dd1c0,0x000000010f6dd640] = 1152
 stub code      [0x000000010f6dd640,0x000000010f6dd6d0] = 144
 oops           [0x000000010f6dd6d0,0x000000010f6dd6d8] = 8
 metadata       [0x000000010f6dd6d8,0x000000010f6dd6e0] = 8
 scopes data    [0x000000010f6dd6e0,0x000000010f6dd830] = 336

astojanov

// 1. package ch.ethz.acl.ngen.saxpy;
// 2. 
// 3. public class JSaxpy {
// 4.     public void apply(float[] a, float[] b, float s, int n){
// 5.         for (int i = 0; i < n; i += 1) {
// 6.             a[i] += b[i] * s;
// 7.         }
// 8.     }
// 9. }

astojanov

// Code being analyzed
//
// 1. package ch.ethz.acl.ngen.saxpy;
// 2.
// 3. public class JSaxpy {
// 4.    public void apply(int[] a, int[] b, int s, int n){
// 5.        for (int i = 0; i < n; i += 1) {
// 6.            a[i] += b[i] * s;
// 7.        }
// 8.    }

astojanov

install.packages("ggplot2")
library(ggplot2) 

getLegend <- function() { return(theme(legend.position="bottom")) } 
getTitle <- function(tmp) { return(ggtitle(tmp)) } 
getAxisTextElement <- function() { return(element_text(hjust=0.5, size=8, colour="black")) } 
getAxisTitleElement <- function() { return(element_text(hjust=0.5, size=10)) } 
getTitleElement <- function () { return(element_text(hjust=0, size=10)) } 
colorPalette <- c("#000000", "#585858", "#585858", "#585858", "#585858", "#585858", "#585858") 
shapePalette <- c(16, 23, 24, 25, 22) 

astojanov

IppsBigNumState* createBigNumState(int len, const Ipp32u* pData)  {
    int size;
    ippsBigNumGetSize(len, &size);
    IppsBigNumState* pBN = (IppsBigNumState*) ippMalloc(size);;
    ippsBigNumInit(len, pBN);
    if (pData != NULL) {
        ippsSet_BN(IppsBigNumPOS, len, pData, pBN);
    }
    return pBN;
}

astojanov

CPPFLAGS = -O3 -xHost -no-multibyte-chars
INCLUDE  = -I/opt/local/include/ -I/opt/intel/ipp/include
LIBS     = -L/opt/local/lib/ -lssl -lcrypto -L/opt/intel/ipp/lib -ipp=crypto

all: perf.h
	icpc $(INCLUDE) $(CFLAGS) -c main.cpp -o main.o
	icpc $(INCLUDE) $(CFLAGS) -c perf.cpp -o perf.o
	icpc $(LIBS) perf.o main.o -o rsatest

clean: