Learn iOS Game Optimization. Ultimate Guide

Learn iOS Game Optimization. Ultimate Guide

by Dmitriy Vovk

Want to achieve the same technology level? Welcome in!

Image is used without any permissions

GENERAL RECOMMENDATIONS

What you might know

• Batch, Batch, Batch!http://ce.u-sys.org/Veranstaltungen/Interaktive%20Computergraphik%20(Stamminger)/papers/BatchBatchBatch.pdf

• Render from one thread only• Avoid synchronizations:

1. glFlush/glFinish;2. Querying GL states;3. Accessing render targets;

http://ce.u-sys.org/Veranstaltungen/Interaktive%20Computergraphik%20(Stamminger)/papers/BatchBatchBatch.pdf




VERTEX DATA RECOMMENDATIONS

What you might know

• Pixel perfect HSR (Hidden Surface Removal),• But still needs to sort transparent geometry!• Avoid doing alpha test. Use alpha blend

instead

What you might not know

• HSR still requires vertices to be processed!• …thus don’t forget to cull your geometry on

CPU!• Prefer Stencil Test before Scissor.– Stencil test is performed in hardware on PowerVR

GPUs, thus resulting in dramatically increased performance.

– Stencil can be of any form in contrast to the rectangular Scissor


• Why no alpha test?! o Alpha test\discard requires fragment shader to run, before visibility for

current fragment can be determined. This will remove benefits of HSRo Even more! If shader code contains discard, than any geometry rendered

with this shader will suffer from alpha test drawbacks. Even if this key-word is under condition, USSE does assumes, that this condition may be hit.

o Move discard into separate shadero Draw opaque geometry, than alpha tested one and alpha blended in the

end

What you might know

• Bandwidth matters1. Use constant color per object, instead of per

vertex2. Simplify your models. Use smaller data types.3. Use indexed triangles or non-indexed triangle

strips4. Use VBO instead of client arrays5. Use VAO


• VAO implementation on at least iOS 4.0 did harmed your performance

• VBOs are allocated at 4KB page size multiples. Be aware of that. Large amount of small VBOs can defragment and waste you memory.


• Updating your VBO data each frame:1. glBufferSubData, that updates big part of the original

data do harm performance. Try not to update buffer, that is used now

2. glBufferData, that will completely overwrite original data is OK. Old data will be orphaned by driver and storage for new one will be allocated

3. glMapBuffer with triple buffered VBO is preferred way to update your data• EXT_map_buffer_range (iOS 6 only), when you need to

update only a subset of a buffer object.

What you might not knowint bufferID = 0; //initializationfor (int i = 0; i < 3; ++i)// only allocate data for 3 vbo, do not upload it{

glBindBuffer(vertexBuffer[i]);glBufferData(GL_ARRAY_BUFFER, 0, 0, GL_DYNAMIC_DRAW);

}//...glBindBuffer(GL_ARRAY_BUFFER, vertexBuffer[bufferID]);void* ptr = glMapBufferOES(GL_ARRAY_BUFFER, GL_WRITE_ONLY_OES);//update data hereglUnmapBufferOES(GL_ARRAY_BUFFER);++bufferID;if (bufferID == 3)//cycling through 3 buffers{

bufferID = 0;}


• This scheme will give you the best performance possible – no blocking CPU by GPU (or vice versa), no redundant memcpy operations, lower CPU load, but extra memory is used (note, that you will need no extra temporal buffer to store your data before sending it to VBO).

update(1), draw(1), gpuworking(................) update(2), draw(2), gpuworking(................)

update(3), draw(3), gpuworking(................)


• Float type is native to GPU• …that means any other type will be converted

to float by USSE• …resulting in few additional cycles• Thus it’s your choice in tradeoff between

bandwidth\storage and additional cycles

What you might know

• Use interleaved vertex data– Align each vertex attribute by 4 bytes boundaries


• Why you have to do this?!– You don’t. Driver can do this instead of you– …resulting in slower performance.


• PowerVR SGX 5XT GPU series have a cache for last 12 vertex indices. Optimize your indexed geometry for this cache size.

• Take a look at optimizers, that use Tom Forsyth’s algorithm http://home.comcast.net/~tom_forsyth/papers/fast_vert_cache_opt.html

http://home.comcast.net/~tom_forsyth/papers/fast_vert_cache_opt.html



What you might know

• Split your vertex data into two parts:1. Static VBO - the one, that never will be changed2. Dynamic VBO – the one, that needs to be

updated frequently• Split your vertex data into few VBOs, when

few meshes share the same set of attributes

TEXTURE DATA RECOMMENDATIONS

What you might know

• Bandwidth matters1. Use lower precision formats i.e. RGB5652. Use PVRTC compressed textures3. Use atlases4. Use mipmaps. They improve texture cache

efficiency and quality.


• iOS OpenGL ES drivers from 4.0 version prior to 6.0 has a bug, that will ALWAYS reserve memory for mipmaps, regardless, whether you requested to create them, or not. And you don’t need mip maps for 2D graphics.

• …but there are one workaround – make your textures NPOT (non-power of two).


• NPOT textures works only with the GL_CLAMP_TO_EDGE warp mode

• POT are preferable, they gives you the best performance possible

• Use NPOT textures with dimensions multiple to 32 pixels for best performance

• Driver will pad data of your NPOT texture to match the size of the closes POT values.


• Why do I have to use PVRTC? It looks ugly!1. PVRTC provides great compression, resulting in

smaller texture size, improved cache, saved bandwidth and decreased power consumption

2. PVRTC stores pixel data in GPU’s native order i.e BGRA, instead of RGBA


• BGRA vs RGBA1. RGBA:• Requires pixel data to be shuffled by driver into

BGRA• Has options for RGB422, RGB565, RGBA4444,

RGBA5551

2. BGRA:• Stores data in GPU’s native order• Has option only for BGRA8888 for upload and

BGRA888, BGRA5551, BGRA4444 for ReadPixels


• Prefer OES_texture_half_float instead of OES_texture_float• Texture reads read only 32 bits per texel, thus RGBA float

texture will result in 4 texture reads

What you might know

• Prefer multitexturing instead of multiple passes

• Configure texture parameters before feeding image data to driver


• Texture uploading to the GPU is a mess!• Usual way to do this:

1. Load texture to temporal buffer in RAM2. Feed this buffer to glTexImage2D3. Draw!

• Looks simple and fast, right?


• …NO!void* buf = malloc(TEXTURE_SIZE); //4mb for RGBA8 1024x1024 textureLoadTexture(textureName);glBindTexture(GL_TEXTURE_2D, textureID);glTexImage2D(GL_TEXTURE_2D, 0, 4, 1024, 1024, 0, GL_RGBA, GL_UNSIGNED_BYTE, &buf);// buf is copied into internal buffer, created by driver (that's obvious)free(buf); // because buffer can be freed immediately after glTexImage2DglDrawElements(GL_TRIANGLES, 6, GL_UNSIGNED_BYTE, 0);// driver will do some additional work to fully upload texture first time it is actually used!

• Textures are finally uploaded only when they are used first time. So draw them off screen immediately after glTexImage2D

• A lot of redundant work!


• Jedi way to upload textures:void* ptr = mmap(NULL, TEXTURE_SIZE, PROT_READ, MAP_PRIVATE, fileHandle, 0); //file mappingglBindTexture(GL_TEXTURE_2D, textureID);glTexImage2D(GL_TEXTURE_2D, 0, 4, 1024, 1024, 0, GL_RGBA, GL_UNSIGNED_BYTE, ptr);glDrawElements(GL_TRIANGLES, 6, GL_UNSIGNED_BYTE, 0);// driver will do some additional work to fully upload texture first time it is actually used!munmap(ptr, TEXTURE_SIZE);

• File mapping does not copy your file data into RAM! It does load file data page by page, when it’s accessed.

• Thus we eliminated one redundant copy, dramatically decreased texture upload time and decreased memory fragmentation


• Always use glClear at the beginning of the frame…

• … and EXT_discard_framebuffer at the end.• PVR GPU series have a fast on chip depth

buffer for each tile. If you forget to clear\discard depth buffer, it will be uploaded from HW to SW

SHADERS BEST PRACTICES

What you might know

• Be wise with precision hints• Avoid branching• Eliminate loops• Do not use discard. Place discard instruction

as early, as possible to avoid useless computations


• Code inside of dynamic branch (it’s condition is evaluated against value calculated in shader) will be executed anyway and than it will be orphaned if condition is false


• highp – represents 32 bit floating point value• mediump – represents 16 bit floating point

value in range of [-65520, 65520]• lowp – 10 bit fixed point values in range of [-2,

2] with step of 1/256• Try to give the same precision to all you

operands, because conversion takes some time


• highp values are calculated on a scalar processor on USSE1 only:highp vec4 v1, v2; highp float s1, s2; // Bad v2 = (v1 * s1) * s2;//scalar processor executes v1 * s1 – 4 operations, and than this result is multiplied by s2 on //a scalar processor again – 4 additional operations// Good v2 = v1 * (s1 * s2);//s1 * s2 – 1 operation on a scalar processor; result * v1 – 4 operations on a scalar processor

HARDWARE FEATURES

What you might know

• Typical CPU found in iOS devices:1. ARMv7 architecture2. Cortex A8\Cortex A9\Custom Apple cores3. 600 – 1300 MHz4. 1-2 cores5. Thumb-2 instructions set


• ARMv7 has no hardware support for integer division

• VFPv3 FPU\VFPv4 on Apple A6 (rumored)• NEON SIMD engine• Unaligned access is done in software on Cortex

A8. That means a hundred times slower• Cortex A8 is in-order CPU. Cortex A9+ are out of

order


• Cortex A9 core has full VFPv3 FPU, while Cortex A8 has a VFPLite. That means, that float operations take 1 cycle on A9 and 10 cycles on A8!


• NEON – 16 registers, 128 bit wide each. Supports operations on 8, 16, 32 and 64 bits integers and 32 bits float values

• NEON can be used for:– Software geometry instancing;– Skinning on ES 1.1;– As a general vertex processor;– Other, typical, applications for SIMD.


• USSE1 architecture is scalar, NEON is vector by nature. Move your vertex processing to CPU from GPU to speedup calculations*

• ???????• PROFIT!!!111

• *NOTE. That doesn’t apply to USSE2 hardware


• There are 3 ways to use NEON engine in your code:1. Intrinsics

1.1 GLKMath

2. Handwritten NEON assembly3. Autovectorization. Add –mllvm –vectorize –

mllvm –bb-vectorize-aligned-only to Other C Flags in project settings and you are ready to go.


• Intrinsics:


• Assembly:


• Summary:

• Intrinsics got me 25% speedup over assembly. Let’s see the code!

• Note that speed of intrinsics code vary from compiler to compiler.

Running time, ms CPU usage, %

Intrinsics 2764 19

Assembly 3664 20

FPU 6209 25-28

FPU autovectorized 5028 22-24


• Intrinsics advantages over assembly:–Higher level code;–No need to manage registers;– You can vectorize basic blocks and build

solution for every new problem with this blocks. In contrast to assembly – you have to solve each new problem from scratch;


• Assembly advantages over intrinsics:–Code generated from intrinsics vary from

compiler to compiler and can give you really big difference in speed. Assembly code always will be the same.


__attribute__((always_inline)) void Matrix4ByVec4(const float32x4x4_t* __restrict__ mat, const float32x4_t* __restrict__ vec, float32x4_t* __restrict__ result){ (*result) = vmulq_n_f32((*mat).val[0], (*vec)[0]); (*result) = vmlaq_n_f32((*result), (*mat).val[1], (*vec)[1]); (*result) = vmlaq_n_f32((*result), (*mat).val[2], (*vec)[2]); (*result) = vmlaq_n_f32((*result), (*mat).val[3], (*vec)[3]);}

What you might not know__attribute__((always_inline)) void Matrix4ByMatrix4(const float32x4x4_t* __restrict__ m1, const float32x4x4_t* __restrict__ m2, float32x4x4_t* __restrict__ r){#ifdef INTRINSICS (*r).val[0] = vmulq_n_f32((*m1).val[0], vgetq_lane_f32((*m2).val[0], 0)); (*r).val[1] = vmulq_n_f32((*m1).val[0], vgetq_lane_f32((*m2).val[1], 0)); (*r).val[2] = vmulq_n_f32((*m1).val[0], vgetq_lane_f32((*m2).val[2], 0)); (*r).val[3] = vmulq_n_f32((*m1).val[0], vgetq_lane_f32((*m2).val[3], 0)); (*r).val[0] = vmlaq_n_f32((*r).val[0], (*m1).val[1], vgetq_lane_f32((*m2).val[0], 1)); (*r).val[1] = vmlaq_n_f32((*r).val[1], (*m1).val[1], vgetq_lane_f32((*m2).val[1], 1)); (*r).val[2] = vmlaq_n_f32((*r).val[2], (*m1).val[1], vgetq_lane_f32((*m2).val[2], 1)); (*r).val[3] = vmlaq_n_f32((*r).val[3], (*m1).val[1], vgetq_lane_f32((*m2).val[3], 1)); (*r).val[0] = vmlaq_n_f32((*r).val[0], (*m1).val[2], vgetq_lane_f32((*m2).val[0], 2)); (*r).val[1] = vmlaq_n_f32((*r).val[1], (*m1).val[2], vgetq_lane_f32((*m2).val[1], 2)); (*r).val[2] = vmlaq_n_f32((*r).val[2], (*m1).val[2], vgetq_lane_f32((*m2).val[2], 2)); (*r).val[3] = vmlaq_n_f32((*r).val[3], (*m1).val[2], vgetq_lane_f32((*m2).val[3], 2)); (*r).val[0] = vmlaq_n_f32((*r).val[0], (*m1).val[3], vgetq_lane_f32((*m2).val[0], 3)); (*r).val[1] = vmlaq_n_f32((*r).val[1], (*m1).val[3], vgetq_lane_f32((*m2).val[1], 3)); (*r).val[2] = vmlaq_n_f32((*r).val[2], (*m1).val[3], vgetq_lane_f32((*m2).val[2], 3)); (*r).val[3] = vmlaq_n_f32((*r).val[3], (*m1).val[3], vgetq_lane_f32((*m2).val[3], 3));}

What you might not know __asm__ volatile ( "vldmia %6, { q0-q3 } \n\t" "vldmia %0, { q8-q11 }\n\t" "vmul.f32 q12, q8, d0[0]\n\t" "vmul.f32 q13, q8, d2[0]\n\t" "vmul.f32 q14, q8, d4[0]\n\t" "vmul.f32 q15, q8, d6[0]\n\t" "vmla.f32 q12, q9, d0[1]\n\t" "vmla.f32 q13, q9, d2[1]\n\t" "vmla.f32 q14, q9, d4[1]\n\t" "vmla.f32 q15, q9, d6[1]\n\t" "vmla.f32 q12, q10, d1[0]\n\t" "vmla.f32 q13, q10, d3[0]\n\t" "vmla.f32 q14, q10, d5[0]\n\t" "vmla.f32 q15, q10, d7[0]\n\t" "vmla.f32 q12, q11, d1[1]\n\t" "vmla.f32 q13, q11, d3[1]\n\t" "vmla.f32 q14, q11, d5[1]\n\t" "vmla.f32 q15, q11, d7[1]\n\t" "vldmia %1, { q0-q3 } \n\t" "vmul.f32 q8, q12, d0[0]\n\t" "vmul.f32 q9, q12, d2[0]\n\t"

"vmul.f32 q10, q12, d4[0]\n\t" "vmul.f32 q11, q12, d6[0]\n\t" "vmla.f32 q8, q13, d0[1]\n\t" "vmla.f32 q8, q14, d1[0]\n\t" "vmla.f32 q8, q15, d1[1]\n\t" "vmla.f32 q9, q13, d2[1]\n\t" "vmla.f32 q9, q14, d3[0]\n\t" "vmla.f32 q9, q15, d3[1]\n\t" "vmla.f32 q10, q13, d4[1]\n\t" "vmla.f32 q10, q14, d5[0]\n\t" "vmla.f32 q10, q15, d5[1]\n\t" "vmla.f32 q11, q13, d6[1]\n\t" "vmla.f32 q11, q14, d7[0]\n\t" "vmla.f32 q11, q15, d7[1]\n\t" "vstmia %2, { q8 }\n\t" "vstmia %3, { q9 }\n\t" "vstmia %4, { q10 }\n\t" "vstmia %5, { q11 }" : : "r" (proj), "r" (squareVertices), "r" (v1), "r" (v2), "r" (v3), "r" (v4), "r" (modelView) : "memory", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15" );


• For detailed explanation on intrinsics\assembly see: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0491e/CIHJBEFE.html

http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0491e/CIHJBEFE.html



Contact me

http://www.linkedin.com/in/dvovk/http://nukecode.blogspot.com/

mailto:[email protected]




Learn iOS Game Optimization. Ultimate Guide

Technology

vbo data

original data

vertex data recommendations

data ext

old data

interleaved vertex data

texture data recommendations

oes update data