SIMD Instruction Set for OLC completed and tested.

[Johnnyg63]

Phase 1 release
Hi Team, Yes its back! and working as expected now:
There are a bunch of new SIMD instruction for the team to play with and enjoy. I have added a few extra methods for merging of sprites, have fun.

You can see a demo here I will work on updates here, and you can get a preview of things to come.

Phase 2 will have NEON(ARM) support, I may introduce threading (using different CPUs to execute SIMD), but it may need a lot of changes to the engine, so I will need to reach out to @OneLoneCoder before I do anything, it may not be straight forward and my goal is to ensure backward compatibly.

This document should explain everything, I am continuously updating it and will also upload drafts to Discord

OLC PGE How to implement SIMD.docx

Finally please give it a good test, please please please break it and report bugs to me.

CHANGES:

• #include <intrin.h>
  Public: Class Sprite

• Updated ~Sprite to clear any stored vector sprites

• Inline void: void setStoreSubSprites(bool bStore)

• Inline bool: bool getStoreSubSprites()

• Inline void: void setInsturctionSet(SIMD_INSTRUCTON_OPTION option)

• Inline int: SIMD_INSTRUCTON_OPTION getInsturctionSet()

• Inline: bool CheckForSSESupport()

• Inline: bool CheckForAVXSupport()

• Inline: bool CheckForAVX512Support()

• Inline: bool CheckForARMSupport() Placeholder not implemented yet

• olc::Sprite* Duplicate(olc::Sprite::Flip flip);

• olc::Sprite* Duplicate(uint32_t scale);

• olc::Sprite* DuplicateMerge(const olc::vi2d& vTargetPos, olc::Sprite* pTargetSprite, olc::Pixel p = olc::BLANK);

• olc::Sprite* Duplicate_SIMD();

• olc::Sprite* Duplicate_SIMD(olc::Sprite::Flip flip);

• olc::Sprite* Duplicate_SIMD(uint32_t scale); // John Galvin

• olc::Sprite* Duplicate_SIMD(const olc::vi2d& vPos, const olc::vi2d& vSize);

• olc::Sprite* DrawToTarget_SIMD(const olc::vi2d& vPos, olc::Sprite* pdrawTarget);

• olc::Sprite* DrawToTarget_SIMD(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint8_t flip);

• olc::Sprite* DrawToTarget_SIMD(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint32_t scale, uint8_t flip);

• olc::Sprite* DuplicateMerge_SIMD(const olc::vi2d& vTargetPos, olc::Sprite* pTargetSprite, olc::Pixel p = olc::BLANK);

• std::vector<std::tuple<olc::vi2d, olc::vi2d, uint32_t, uint8_t, Sprite*>> vecSubSprites;

  Private: Class Sprite

  • olc::Sprite* Duplicate_SSE(olc::Sprite::Flip flip)
  • olc::Sprite* Duplicate_AVX256(olc::Sprite::Flip flip)
  • olc::Sprite* Duplicate_AVX512(olc::Sprite::Flip flip)
  • olc::Sprite* Duplicate_SSE(uint32_t scale)
  • olc::Sprite* Duplicate_AVX256(uint32_t scale)
  • olc::Sprite* Duplicate_AVX512(uint32_t scale)
  • olc::Sprite* Duplicate_SSE(const olc::vi2d& vPos, const olc::vi2d& vSize)
  • olc::Sprite* Duplicate_AVX256(const olc::vi2d& vPos, const olc::vi2d& vSize)
  • olc::Sprite* Duplicate_AVX512(const olc::vi2d& vPos, const olc::vi2d& vSize)
  • olc::Sprite* Duplicate_SSE(const olc::vi2d& vPos, olc::Sprite* pdrawTarget)
  • olc::Sprite* Duplicate_AVX256(const olc::vi2d& vPos, olc::Sprite* pdrawTarget)
  • olc::Sprite* Duplicate_AVX512(const olc::vi2d& vPos, olc::Sprite* pdrawTarget)
  • olc::Sprite* Duplicate_SSE(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint8_t flip)
  • olc::Sprite* Duplicate_AVX256(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint8_t flip)
  • olc::Sprite* Duplicate_AVX512(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint8_t flip)
  • olc::Sprite* Duplicate_SSE(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint32_t scale, uint8_t flip)
  • olc::Sprite* Duplicate_AVX256(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint32_t scale, uint8_t flip)
  • olc::Sprite* Duplicate_AVX512(const olc::vi2d& vPos, olc::Sprite* pdrawTarget, uint32_t scale, uint8_t flip)
  • olc::Sprite* DuplicateMerge_SSE(const olc::vi2d& vTargetPos, olc::Sprite* pTargetSprite, std::vector vecPositions, olc::Pixel p)
  • olc::Sprite* DuplicateMerge_AVX256(const olc::vi2d& vTargetPos, olc::Sprite* pTargetSprite, std::vector vecPositions, olc::Pixel p)
  • olc::Sprite* DuplicateMerge_AVX512(const olc::vi2d& vTargetPos, olc::Sprite* pTargetSprite, std::vector vecPositions, olc::Pixel p)

  Public: Class PixelGameEngine

  • Inline enum: enum INSTRUCTON_OPTION
  • Inline void: void setInsturctionSet(INSTRUCTON_OPTION option)
  • Inline bool: int getInsturctionSet()
  • static bool CheckForSSESupport()
  • static bool CheckForAVXSupport()
  • static bool CheckForAVX512Support()
  • void Clear_SIMD(Pixel p = olc::BLANK)
  • void FillCircle_SIMD(olc::vi2d pos, int32_t radius, Pixel p = olc::WHITE)
  • void FillCircle_SIMD(int32_t x, int32_t y, int32_t radius, Pixel p = olc::WHITE)
  • void FillTriangle_SIMD(const olc::vi2d& pos1, const olc::vi2d& pos2, const olc::vi2d& pos3, Pixel p = olc::WHITE)
  • void FillTriangle_SIMD(int32_t x1, int32_t y1, int32_t x2, int32_t y2, int32_t x3, int32_t y3, Pixel p = olc::WHITE)
  • void FillRect_SIMD(const olc::vi2d& pos, const olc::vi2d& size, Pixel p = olc::WHITE)
  • void FillRect_SIMD(int32_t x, int32_t y, int32_t w, int32_t h, Pixel p = olc::WHITE)
  • void DrawSprite_SIMD(const olc::vi2d& pos, Sprite* sprite, uint32_t scale = 1, uint8_t flip = olc::Sprite::NONE)
  • void DrawSprite_SIMD(int32_t x, int32_t y, Sprite* sprite, uint32_t scale = 1, uint8_t flip = olc::Sprite::NONE)
  • void DrawPartialSprite_SIMD(const olc::vi2d& pos, Sprite* sprite, const olc::vi2d& sourcepos, const olc::vi2d& size, uint32_t scale = 1, uint8_t flip = olc::Sprite::NONE)
  • void DrawPartialSprite_SIMD(int32_t x, int32_t y, Sprite* sprite, int32_t ox, int32_t oy, int32_t w, int32_t h, uint32_t scale = 1, uint8_t flip = olc::Sprite::NONE)
  • void DrawMergeSprite_SIMD(const olc::vi2d& vPos, Sprite* pFromSprite, const olc::vi2d& vToSpritePos, Sprite* pToSprite, Pixel blendPixel = olc::BLANK, uint32_t scale = 1, olc::Sprite::Flip flip = olc::Sprite::NONE);
  • void DrawMergeSprite_SIMD(int32_t vPosx, int32_t vPosy, Sprite* pFromSprite, int32_t vToSpritePosx, int32_t vToSpritePosy, Sprite* pToSprite, Pixel blendPixel = olc::BLANK, uint32_t scale = 1, olc::Sprite::Flip flip = olc::Sprite::NONE);
  • void DrawMergeSprite(int32_t vPosx, int32_t vPosy, Sprite* pFromSprite, int32_t vToSpritePosx, int32_t vToSpritePosy, Sprite* pToSprite, Pixel blendPixel = olc::BLANK, uint32_t scale = 1, olc::Sprite::Flip flip = olc::Sprite::NONE);
  • void DrawMergeSprite(const olc::vi2d& vPos, Sprite* pFromSprite, const olc::vi2d& vToSpritePos, Sprite* pToSprite, Pixel blendPixel = olc::BLANK, uint32_t scale = 1, olc::Sprite::Flip flip = olc::Sprite::NONE);

  Private: Class PixelGameEngine

  • void Clear_SSE(int VecStartIndex, int VecEndIndex, Pixel p = olc::BLANK)
  • void Clear_AVX256(int VecStartIndex, int VecEndIndex, Pixel p = olc::BLANK)
  • void Clear_AVX512(int VecStartIndex, int VecEndIndex, Pixel p = olc::BLANK)
  • bool Draw_SIMD(int32_t sx, int32_t ex, int32_t ny, Pixel p = olc::WHITE)
  • void DrawFillLine_SSE(int sx, int ex, int ny, Pixel p = olc::WHITE)
  • void DrawFillLine_AVX256(int sx, int ex, int ny, Pixel p = olc::WHITE)
  • void DrawFillLine_AVX512(int sx, int ex, int ny, Pixel p = olc::WHITE)
  • void DrawFillLine_SIMD(int VecStartIndex, int VecEndIndex, int ny, Pixel p = olc::WHITE)