syoyo

isect4:
00000038	    b081	sub	sp, #4
0000003a	ed2d8b10	vstmdb	sp!, {d8-d15}
0000003e	    b088	sub	sp, #32
00000040	ed9deb28	vldr	d14, [sp, #160]
00000044	ed9dfb2a	vldr	d15, [sp, #168]
00000048	eddd2b2c	vldr	d18, [sp, #176]
0000004c	eddd3b2e	vldr	d19, [sp, #184]
00000050	ed9d6b54	vldr	d6, [sp, #336]
00000054	ed9d7b56	vldr	d7, [sp, #344]

syoyo

//
// SIMD Ray triangle intersection with ARM/NEON instruction.
//
// Compile:
//  CC=/Developer/Platforms/iPhoneOS.platform/Developer/usr/bin/arm-apple-darwin9-gcc-4.2.1
//  CFLAGS=-march=armv7-a -mtune=cortex-a8 -mfloat-abi=softfp -mfpu=neon
//  INCLUDES=-I/Developer/Platforms/iPhoneOS.platform/Developer/usr/lib/gcc/arm-apple-darwin9/4.2.1/include
//  OPTFLAGS=-O3
//
// Copyright 2009. Syoyo Fujita.

syoyo

// Benchmark divpd and approximate division using cvt/rcp + Newton method.
#include <stdio.h>
#include <sys/time.h>
#include <xmmintrin.h>

#define NLOOP 100000
#define ITER 256

__m128d outputs[ITER];
__m128d values[ITER];

syoyo

surface
mandel()
{
        float window_size = 512.0;

        float xx, yy;
        float x0, y0;

        float max_iter = 255;
        float iter = 0;

syoyo

surface
ambocc(float samples = 128)
{
        float occ = 0;

        normal nf = faceforward(normalize(N), I);

        gather("illuminance", P, nf, PI/2, samples, "distribution", "cosine") {
                    occ += 1;
        }

syoyo

FrameBegin 1
        Format 512 512 1
        PixelSamples 2 2
        PixelFilter "box" 1 1
        ShadingRate 0.5
        ShadingInterpolation "smooth"
        Display "ao.tiff" "file" "rgb"  # render image to buffer
        Projection "perspective" "fov" 22
        Translate 0 -0.5 8
        Rotate -40 1 0 0

syoyo

$ llvm-gcc -S -emit-llvm -fnested-functions main.c 

; ModuleID = 'main.c'
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
target triple = "i386-apple-darwin9.6"
	%struct.FRAME.muda = type { float }

define float @muda(float %x) nounwind {
entry:
	%x_addr = alloca float		; <float*> [#uses=2]

syoyo

/* main.c */
float muda(float x)
{
    float y;

    float bora (float z) {
        // extern float y;  NG in gcc
        return y + z;
    }

syoyo

##RenderMan RIB-Structure 1.0
version 3.03
Option "searchpath" "shader" ".:&"
Display "muda.tif" "file" "rgb"
Format 256 256 1
WorldBegin
Surface "bora" 
PointsPolygons [ 4 ] [ 0 3 2 1 ] "P" [ 0.500000 0.500000 0.500000 0.500000 -0.500000 0.500000 -0.500000 -0.500000 0.500000 -0.500000 0.500000 0.500000 ]
	"facevarying float s" [0.0 1.0 1.0 0.0]
	"facevarying float t" [0.0 0.0 1.0 1.0]

syoyo

##RenderMan RIB-Structure 1.0
version 3.03
Option "searchpath" "shader" ".:&"
Display "muda.tif" "file" "rgb"
Format 256 256 1
WorldBegin
Surface "bora" 
PointsPolygons [ 4 ] [ 0 3 2 1 ] "P" [ 0.500000 0.500000 0.500000 0.500000 -0.500000 0.500000 -0.500000 -0.500000 0.500000 -0.500000 0.500000 0.500000 ]
	"s" [0.0 1.0 1.0 0.0]
	"t" [0.0 0.0 1.0 1.0]