#include "ControlListUtil.h" #include "ConsecutivePoolAllocator.h" #include uint32_t divRoundUp(uint32_t n, uint32_t d) { return (((n) + (d) - 1) / (d)); } //move bits to offset, mask rest to 0 uint32_t moveBits(uint32_t d, uint32_t bits, uint32_t offset) { return (d << offset) & (~(~0 << bits) << offset); } uint32_t clHasEnoughSpace(ControlList* cl, uint32_t size) { assert(cl); assert(cl->CPA); uint32_t currSize = cl->nextFreeByteOffset - cl->offset; if(currSize + size < cl->numBlocks * cl->blockSize - 4) { return 1; //fits! } else { return 0; //need to reallocate } } void clInit(ControlList* cl, void* CPA, uint32_t offset, uint32_t blockSize) { assert(cl); assert(CPA); cl->offset = offset; cl->numBlocks = 1; cl->blockSize = blockSize; cl->nextFreeByteOffset = offset; cl->currMarkerOffset = -1; cl->CPA = CPA; } void clInsertNewCLMarker(ControlList* cl, ControlList* handlesCL, ControlList* shaderRecCL, uint32_t shaderRecCount, ControlList* uniformsCL) { //to be inserted when you'd insert tile binning mode config assert(cl); assert(cl->CPA); assert(handlesCL); assert(shaderRecCL); assert(uniformsCL); CLMarker marker = {}; marker.memGuard = 0xDDDDDDDD; marker.handlesBufOffset = handlesCL->offset; marker.shaderRecBufOffset = shaderRecCL->offset; marker.uniformsBufOffset = uniformsCL->offset; marker.nextMarkerOffset = -1; //close current marker if(cl->currMarkerOffset != -1 && !((CLMarker*)getCPAptrFromOffset(cl->CPA, cl->currMarkerOffset))->size) { clCloseCurrentMarker(cl, handlesCL, shaderRecCL, shaderRecCount, uniformsCL); } //if this is not the first marker if(cl->currMarkerOffset != -1) { CLMarker* currMarker = getCPAptrFromOffset(cl->CPA, cl->currMarkerOffset); marker.handlesBufOffset = currMarker->handlesBufOffset + currMarker->handlesSize; marker.shaderRecBufOffset = currMarker->shaderRecBufOffset + currMarker->shaderRecSize; marker.uniformsBufOffset = currMarker->uniformsBufOffset + currMarker->uniformsSize; marker.shaderRecCount = currMarker->shaderRecCount; //initialize with previous marker's data } *(CLMarker*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = marker; if(cl->currMarkerOffset != -1) { ((CLMarker*)getCPAptrFromOffset(cl->CPA, cl->currMarkerOffset))->nextMarkerOffset = cl->nextFreeByteOffset; } cl->currMarkerOffset = cl->nextFreeByteOffset; cl->nextFreeByteOffset += sizeof(CLMarker); } void clCloseCurrentMarker(ControlList* cl, ControlList* handlesCL, ControlList* shaderRecCL, uint32_t shaderRecCount, ControlList* uniformsCL) { assert(cl); assert(cl->CPA); assert(handlesCL); assert(shaderRecCL); assert(uniformsCL); CLMarker* currMarker = getCPAptrFromOffset(cl->CPA, cl->currMarkerOffset); currMarker->size = cl->nextFreeByteOffset - (cl->currMarkerOffset + sizeof(CLMarker)); currMarker->handlesSize = handlesCL->nextFreeByteOffset - currMarker->handlesBufOffset; currMarker->shaderRecSize = shaderRecCL->nextFreeByteOffset - currMarker->shaderRecBufOffset; currMarker->uniformsSize = uniformsCL->nextFreeByteOffset - currMarker->uniformsBufOffset; currMarker->shaderRecCount = shaderRecCount - currMarker->shaderRecCount; //update shader rec count to reflect added shader recs } void clInsertData(ControlList* cl, uint32_t size, uint8_t* data) { assert(cl); assert(cl->CPA); memcpy(getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset), data, size); cl->nextFreeByteOffset += size; } void clInsertUniformConstant(ControlList* cl, uint32_t data) { assert(cl); assert(cl->CPA); *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = data; cl->nextFreeByteOffset += 4; } void clInsertUniformXYScale(ControlList* cl, float data) { assert(cl); assert(cl->CPA); *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = data; cl->nextFreeByteOffset += 4; } void clInsertUniformZOffset(ControlList* cl, float data) { assert(cl); assert(cl->CPA); *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = data; cl->nextFreeByteOffset += 4; } void clInsertHalt(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_HALT_opcode; cl->nextFreeByteOffset++; } void clInsertNop(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_NOP_opcode; cl->nextFreeByteOffset++; } void clInsertFlush(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_FLUSH_opcode; cl->nextFreeByteOffset++; } void clInsertFlushAllState(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_FLUSH_ALL_STATE_opcode; cl->nextFreeByteOffset++; } void clInsertStartTileBinning(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_START_TILE_BINNING_opcode; cl->nextFreeByteOffset++; } void clInsertIncrementSemaphore(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_INCREMENT_SEMAPHORE_opcode; cl->nextFreeByteOffset++; } void clInsertWaitOnSemaphore(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_WAIT_ON_SEMAPHORE_opcode; cl->nextFreeByteOffset++; } //input: 2 cls (cl, handles cl) void clInsertBranch(ControlList* cls, ControlListAddress address) { assert(cls); assert(cls->CPA); *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = V3D21_BRANCH_opcode; cls->nextFreeByteOffset++; //TODO is this correct? //clEmitShaderRelocation(cls, &address); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = address.offset; cls->nextFreeByteOffset += 4; } //input: 2 cls (cl, handles cl) void clInsertBranchToSubList(ControlList* cls, ControlListAddress address) { assert(cls); assert(cls->CPA); *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = V3D21_BRANCH_TO_SUB_LIST_opcode; cls->nextFreeByteOffset++; //TODO is this correct? //clEmitShaderRelocation(cls, &address); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = address.offset; cls->nextFreeByteOffset += 4; } void clInsertReturnFromSubList(ControlList* cl) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_RETURN_FROM_SUB_LIST_opcode; cl->nextFreeByteOffset++; } /*void clInsertStoreMultiSampleResolvedTileColorBuffer(ControlList* cl) { assert(cl); assert(cl->buffer); assert(cl->nextFreeByte); *cl->nextFreeByte = V3D21_STORE_MULTI_SAMPLE_RESOLVED_TILE_COLOR_BUFFER_opcode; cl->nextFreeByte++; }*/ /*void clInsertStoreMultiSampleResolvedTileColorBufferAndEOF(ControlList* cl) { assert(cl); assert(cl->buffer); assert(cl->nextFreeByte); *cl->nextFreeByte = V3D21_STORE_MULTI_SAMPLE_RESOLVED_TILE_COLOR_BUFFER_AND_EOF_opcode; cl->nextFreeByte++; }*/ /* //input: 2 cls (cl, handles cl) void clInsertStoreFullResolutionTileBuffer(ControlList* cls, ControlListAddress address, uint32_t lastTile, //0/1 uint32_t disableClearOnWrite, //0/1 uint32_t disableZStencilBufferWrite, //0/1 uint32_t disableColorBufferWrite) //0/1 { assert(cls); assert(cls->buffer); assert(cls->nextFreeByte); *cls->nextFreeByte = V3D21_STORE_FULL_RESOLUTION_TILE_BUFFER_opcode; cls->nextFreeByte++; //is this correct? clEmitShaderRelocation(cls, &address); *(uint32_t*)cls->nextFreeByte = moveBits(disableColorBufferWrite, 1, 0) | moveBits(disableZStencilBufferWrite, 1, 1) | moveBits(disableClearOnWrite, 1, 2) | moveBits(lastTile, 1, 3) | moveBits(address.offset, 28, 4); cls->nextFreeByte += 4; } */ /* //input: 2 cls (cl, handles cl) void clInsertReLoadFullResolutionTileBuffer(ControlList* cls, ControlListAddress address, uint32_t disableZStencilBufferRead, //0/1 uint32_t disableColorBufferRead) //0/1 { assert(cls); assert(cls->buffer); assert(cls->nextFreeByte); *cls->nextFreeByte = V3D21_RE_LOAD_FULL_RESOLUTION_TILE_BUFFER_opcode; cls->nextFreeByte++; //is this correct? clEmitShaderRelocation(cls, &address); *(uint32_t*)cls->nextFreeByte = moveBits(disableColorBufferRead, 1, 0) | moveBits(disableZStencilBufferRead, 1, 1) | moveBits(address.offset, 28, 4); cls->nextFreeByte += 4; } */ /* //input: 2 cls (cl, handles cl) void clInsertStoreTileBufferGeneral(ControlList* cls, ControlListAddress address, uint32_t lastTileOfFrame, //0/1 uint32_t disableZStencilBufferDump, //0/1 uint32_t disableColorBufferDump, //0/1 uint32_t disableZStencilBufferClearOnStoreDump, //0/1 uint32_t disableColorBufferClearOnStoreDump, //0/1 uint32_t disableDoubleBufferSwap, //0/1 uint32_t pixelColorFormat, //0/1/2 RGBA8/BGR565dither/BGR565nodither uint32_t mode, //0/1/2 sample0/decimate4x/decimate16x uint32_t format, //0/1/2 raster/t/lt uint32_t bufferToStore) //0/1/2/3/5 none/color/zstencil/z/full { assert(cls); assert(cls->buffer); assert(cls->nextFreeByte); *cls->nextFreeByte = V3D21_STORE_TILE_BUFFER_GENERAL_opcode; cls->nextFreeByte++; //is this correct? *cls->nextFreeByte = moveBits(bufferToStore, 3, 0) | moveBits(format, 2, 4) | moveBits(mode, 2, 6); cls->nextFreeByte++; *cls->nextFreeByte = moveBits(pixelColorFormat, 2, 0) | moveBits(disableDoubleBufferSwap, 1, 4) | moveBits(disableColorBufferClearOnStoreDump, 1, 5) | moveBits(disableZStencilBufferClearOnStoreDump, 1, 6) | moveBits(1, 1, 7); //disable vg mask cls->nextFreeByte++; clEmitShaderRelocation(cls, &address); *(uint32_t*)cls->nextFreeByte = moveBits(disableColorBufferDump, 1, 0) | moveBits(disableZStencilBufferDump, 1, 1) | moveBits(1, 1, 2) | //disable vg mask moveBits(lastTileOfFrame, 1, 3) | moveBits(address.offset, 28, 4); cls->nextFreeByte += 4; } */ /* //input: 2 cls (cl, handles cl) void clInsertLoadTileBufferGeneral(ControlList* cls, ControlListAddress address, uint32_t disableZStencilBufferLoad, //0/1 uint32_t disableColorBufferLoad, //0/1 uint32_t pixelColorFormat, //0/1/2 RGBA8/BGR565dither/BGR565nodither uint32_t mode, //0/1/2 sample0/decimate4x/decimate16x uint32_t format, //0/1/2 raster/t/lt uint32_t bufferToLoad) //0/1/2/3/5 none/color/zstencil/z/full { assert(cls); assert(cls->buffer); assert(cls->nextFreeByte); *cls->nextFreeByte = V3D21_LOAD_TILE_BUFFER_GENERAL_opcode; cls->nextFreeByte++; //is this correct? *cls->nextFreeByte = moveBits(bufferToLoad, 3, 0) | moveBits(format, 2, 4); cls->nextFreeByte++; *cls->nextFreeByte = moveBits(pixelColorFormat, 2, 0); cls->nextFreeByte++; clEmitShaderRelocation(cls, &address); *(uint32_t*)cls->nextFreeByte = moveBits(disableColorBufferLoad, 1, 0) | moveBits(disableZStencilBufferLoad, 1, 1) | moveBits(1, 1, 2) | //disable vg mask moveBits(address.offset, 28, 4); cls->nextFreeByte += 4; } */ void clInsertIndexedPrimitiveList(ControlList* cl, uint32_t maxIndex, uint32_t indicesAddress, uint32_t length, uint32_t indexType, //0/1: 8 or 16 bit enum V3D21_Primitive primitiveMode) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_INDEXED_PRIMITIVE_LIST_opcode; cl->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(indexType, 4, 4) | moveBits(primitiveMode, 4, 0); cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = length; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = indicesAddress; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = maxIndex; cl->nextFreeByteOffset += 4; } void clInsertVertexArrayPrimitives(ControlList* cl, uint32_t firstVertexIndex, uint32_t length, enum V3D21_Primitive primitiveMode) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_VERTEX_ARRAY_PRIMITIVES_opcode; cl->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(primitiveMode, 8, 0); cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = length; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = firstVertexIndex; cl->nextFreeByteOffset += 4; } /*void clInsertPrimitiveListFormat(ControlList* cl, uint32_t dataType, //1/3: 16 or 32 bit uint32_t primitiveType) //0/1/2/3: point/line/tri/rhy { assert(cl); assert(cl->buffer); assert(cl->nextFreeByte); *cl->nextFreeByte = V3D21_PRIMITIVE_LIST_FORMAT_opcode; cl->nextFreeByte++; *cl->nextFreeByte = moveBits(dataType, 4, 4) | moveBits(primitiveType, 4, 0); cl->nextFreeByte++; }*/ void clInsertShaderState(ControlList* cl, uint32_t address, uint32_t extendedShaderRecord, //0/1: true/false uint32_t numberOfAttributeArrays) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_GL_SHADER_STATE_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(address, 28, 4) | moveBits(extendedShaderRecord, 1, 3) | moveBits(numberOfAttributeArrays, 3, 0); cl->nextFreeByteOffset += 4; } /* void clInsertClearColors(ControlList* cl, uint32_t clearStencil, uint32_t clearZ, //24 bit Z uint64_t clearColor) //2x RGBA8 or 1x RGBA16 { assert(cl); assert(cl->buffer); assert(cl->nextFreeByte); *cl->nextFreeByte = V3D21_CLEAR_COLORS_opcode; cl->nextFreeByte++; *(uint64_t*)cl->nextFreeByte = clearColor; cl->nextFreeByte += 8; *(uint32_t*)cl->nextFreeByte = clearZ; cl->nextFreeByte += 4; //24 bits for Z, 8 bit for vg mask (unused) *cl->nextFreeByte = clearStencil; cl->nextFreeByte++; } */ void clInsertConfigurationBits(ControlList* cl, uint32_t earlyZUpdatesEnable, //0/1 uint32_t earlyZEnable, //0/1 uint32_t zUpdatesEnable, //0/1 enum V3D21_Compare_Function depthTestFunction, uint32_t coverageReadMode, //0/1 clear/leave as is uint32_t coveragePipeSelect, //0/1 uint32_t coverageUpdateMode, //0/1/2/3 nonzero, odd, or, zero uint32_t coverageReadType, //0/1 4*8bit, 16 bit mask uint32_t rasterizerOversampleMode, //0/1/2 none, 4x, 16x uint32_t enableDepthOffset, //0/1 uint32_t clockwisePrimitives, //0/1 uint32_t enableReverseFacingPrimitive, //0/1 uint32_t enableForwardFacingPrimitive) //0/1 { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_CONFIGURATION_BITS_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(enableForwardFacingPrimitive, 1, 0) | moveBits(enableReverseFacingPrimitive, 1, 1) | moveBits(clockwisePrimitives, 1, 2) | moveBits(enableDepthOffset, 1, 3) | moveBits(coverageReadType, 1, 5) | moveBits(rasterizerOversampleMode, 2, 6) | moveBits(coveragePipeSelect, 1, 8) | moveBits(coverageUpdateMode, 2, 9) | moveBits(coverageReadMode, 1, 11) | moveBits(depthTestFunction, 3, 12) | moveBits(zUpdatesEnable, 1, 15) | moveBits(earlyZEnable, 1, 16) | moveBits(earlyZUpdatesEnable, 1, 17); cl->nextFreeByteOffset += 3; } void clInsertFlatShadeFlags(ControlList* cl, uint32_t flags) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_FLAT_SHADE_FLAGS_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = flags; cl->nextFreeByteOffset += 4; } void clInsertPointSize(ControlList* cl, float size) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_POINT_SIZE_opcode; cl->nextFreeByteOffset++; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = size; cl->nextFreeByteOffset += 4; } void clInsertLineWidth(ControlList* cl, float width) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_LINE_WIDTH_opcode; cl->nextFreeByteOffset++; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = width; cl->nextFreeByteOffset += 4; } void clInsertRHTXBoundary(ControlList* cl, uint32_t boundary) //sint16 { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_RHT_X_BOUNDARY_opcode; cl->nextFreeByteOffset++; *(uint16_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(boundary, 16, 0); cl->nextFreeByteOffset += 2; } uint32_t f32_to_f187(float f32) { uint32_t bits = *(uint32_t*)&f32; return bits >> 16; } void clInsertDepthOffset(ControlList* cl, float units, float factor) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_DEPTH_OFFSET_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(f32_to_f187(factor), 16, 0) | moveBits(f32_to_f187(units), 16, 16); cl->nextFreeByteOffset += 4; } void clInsertClipWindow(ControlList* cl, uint32_t width, //uint16 uint32_t height, //uint16 uint32_t bottomPixelCoord, //uint16 uint32_t leftPixelCoord) //uint16 { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_CLIP_WINDOW_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(leftPixelCoord, 16, 0) | moveBits(bottomPixelCoord, 16, 16); cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(width, 16, 0) | moveBits(height, 16, 16); cl->nextFreeByteOffset += 4; } uint16_t get16bitSignedFixedNumber(float x) { int32_t integerPart = roundf(x * 16.0f); return integerPart & 0xffff; } //viewport centre x/y coordinate void clInsertViewPortOffset(ControlList* cl, float x, float y ) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_VIEWPORT_OFFSET_opcode; cl->nextFreeByteOffset++; //expects 16 bit signed fixed point number with 4 fractional bits *(uint16_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = get16bitSignedFixedNumber(x); cl->nextFreeByteOffset += 2; *(uint16_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = get16bitSignedFixedNumber(y); cl->nextFreeByteOffset += 2; } void clInsertZMinMaxClippingPlanes(ControlList* cl, float minZw, float maxZw ) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_Z_MIN_AND_MAX_CLIPPING_PLANES_opcode; cl->nextFreeByteOffset++; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = minZw; cl->nextFreeByteOffset += 4; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = maxZw; cl->nextFreeByteOffset += 4; } void clInsertClipperXYScaling(ControlList* cl, float width, //half width in 1/16 of pixel float height //half height in 1/16 of pixel ) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_CLIPPER_XY_SCALING_opcode; cl->nextFreeByteOffset++; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = width; cl->nextFreeByteOffset += 4; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = height; cl->nextFreeByteOffset += 4; } void clInsertClipperZScaleOffset(ControlList* cl, float zOffset, //zc to zs float zScale //zc to zs ) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_CLIPPER_Z_SCALE_AND_OFFSET_opcode; cl->nextFreeByteOffset++; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = zScale; cl->nextFreeByteOffset += 4; *(float*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = zOffset; cl->nextFreeByteOffset += 4; } void clInsertTileBinningModeConfiguration(ControlList* cl, uint32_t doubleBufferInNonMsMode, //0/1 uint32_t tileAllocationBlockSize, //0/1/2/3 32/64/128/256 bytes uint32_t tileAllocationInitialBlockSize, //0/1/2/3 32/64/128/256 bytes uint32_t autoInitializeTileStateDataArray, //0/1 uint32_t tileBuffer64BitColorDepth, //0/1 uint32_t multisampleMode4x, //0/1 uint32_t widthInPixels, uint32_t heightInPixels, uint32_t tileStateDataArrayAddress, //16 byte aligned, size of 48 bytes * num tiles uint32_t tileAllocationMemorySize, uint32_t tileAllocationMemoryAddress ) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_TILE_BINNING_MODE_CONFIGURATION_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = tileAllocationMemoryAddress; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = tileAllocationMemorySize; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = tileStateDataArrayAddress; cl->nextFreeByteOffset += 4; uint32_t tileSizeW = 64; uint32_t tileSizeH = 64; if(multisampleMode4x) { tileSizeW >>= 1; tileSizeH >>= 1; } if(tileBuffer64BitColorDepth) { tileSizeH >>= 1; } uint32_t widthInTiles = divRoundUp(widthInPixels, tileSizeW); uint32_t heightInTiles = divRoundUp(heightInPixels, tileSizeH); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = widthInTiles; cl->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = heightInTiles; cl->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = moveBits(multisampleMode4x, 1, 0) | moveBits(tileBuffer64BitColorDepth, 1, 1) | moveBits(autoInitializeTileStateDataArray, 1, 2) | moveBits(tileAllocationInitialBlockSize, 2, 3) | moveBits(tileAllocationBlockSize, 2, 5) | moveBits(doubleBufferInNonMsMode, 1, 7); cl->nextFreeByteOffset++; } /* void clInsertTileRenderingModeConfiguration(ControlList* cls, ControlListAddress address, uint32_t doubleBufferInNonMsMode, //0/1 uint32_t earlyZEarlyCovDisable, //0/1 uint32_t earlyZUpdateDirection, //0/1 lt,le/gt,ge uint32_t selectCoverageMode, //0/1 uint32_t memoryFormat, //0/1/2 linear/t/lt uint32_t decimateMode, //0/1/2 0x/4x/16x uint32_t nonHDRFrameFormatColorFormat, //0/1/2 bgr565dithered/rgba8/bgr565nodither uint32_t tileBufferHDRMode, //0/1 uint32_t multisampleMode4x, //0/1 uint32_t widthPixels, uint32_t heightPixels) { assert(cls); assert(cls->buffer); assert(cls->nextFreeByte); *cls->nextFreeByte = V3D21_TILE_RENDERING_MODE_CONFIGURATION_opcode; cls->nextFreeByte++; //is this correct? clEmitShaderRelocation(cls, &address); *(uint32_t*)cls->nextFreeByte = address.offset; cls->nextFreeByte += 4; *(uint32_t*)cls->nextFreeByte = moveBits(widthPixels, 16, 0) | moveBits(heightPixels, 16, 16); cls->nextFreeByte += 4; *(uint16_t*)cls->nextFreeByte = moveBits(multisampleMode4x, 1, 0) | moveBits(tileBufferHDRMode, 1, 1) | moveBits(nonHDRFrameFormatColorFormat, 2, 2) | moveBits(decimateMode, 2, 4) | moveBits(memoryFormat, 2, 6) | moveBits(0, 1, 8) | //vg buffer enable moveBits(selectCoverageMode, 1, 9) | moveBits(earlyZUpdateDirection, 1, 10) | moveBits(earlyZEarlyCovDisable, 1, 11) | moveBits(doubleBufferInNonMsMode, 1, 12); cls->nextFreeByte += 2; } */ /* void clInsertTileCoordinates(ControlList* cl, uint32_t tileColumnNumber, //int8 uint32_t tileRowNumber) //int8 { assert(cl); assert(cl->buffer); assert(cl->nextFreeByte); *cl->nextFreeByte = V3D21_TILE_COORDINATES_opcode; cl->nextFreeByte++; *(uint16_t*)cl->nextFreeByte = moveBits(tileColumnNumber, 8, 0) | moveBits(tileRowNumber, 8, 8); cl->nextFreeByte += 2; } */ void clInsertGEMRelocations(ControlList* cl, uint32_t buffer0, uint32_t buffer1) { assert(cl); assert(cl->CPA); *(uint8_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = V3D21_GEM_RELOCATIONS_opcode; cl->nextFreeByteOffset++; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = buffer0; cl->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cl->CPA, cl->nextFreeByteOffset) = buffer1; cl->nextFreeByteOffset += 4; } //input: 2 cls (cl, handles cl) void clInsertShaderRecord(ControlList* cls, ControlList* relocCl, ControlList* handlesCl, uint32_t handlesOffset, uint32_t handlesSize, uint32_t fragmentShaderIsSingleThreaded, //0/1 uint32_t pointSizeIncludedInShadedVertexData, //0/1 uint32_t enableClipping, //0/1 uint32_t fragmentNumberOfUsedUniforms, uint32_t fragmentNumberOfVaryings, uint32_t fragmentUniformsAddress, ControlListAddress fragmentCodeAddress, uint32_t vertexNumberOfUsedUniforms, uint32_t vertexAttributeArraySelectBits, uint32_t vertexTotalAttributesSize, uint32_t vertexUniformsAddress, ControlListAddress vertexCodeAddress, uint32_t coordinateNumberOfUsedUniforms, uint32_t coordinateAttributeArraySelectBits, uint32_t coordinateTotalAttributesSize, uint32_t coordinateUniformsAddress, ControlListAddress coordinateCodeAddress) { assert(cls); assert(cls->CPA); *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = moveBits(fragmentShaderIsSingleThreaded, 1, 0) | moveBits(pointSizeIncludedInShadedVertexData, 1, 1) | moveBits(enableClipping, 1, 2); cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = 0; cls->nextFreeByteOffset++; *(uint16_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = moveBits(fragmentNumberOfUsedUniforms, 16, 0); cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) |= fragmentNumberOfVaryings; cls->nextFreeByteOffset++; clEmitShaderRelocation(relocCl, handlesCl, handlesOffset, handlesSize, &fragmentCodeAddress); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = fragmentCodeAddress.offset; cls->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = fragmentUniformsAddress; cls->nextFreeByteOffset += 4; *(uint16_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = moveBits(vertexNumberOfUsedUniforms, 16, 0); cls->nextFreeByteOffset += 2; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = vertexAttributeArraySelectBits; cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = vertexTotalAttributesSize; cls->nextFreeByteOffset++; clEmitShaderRelocation(relocCl, handlesCl, handlesOffset, handlesSize, &vertexCodeAddress); //wtf??? --> shader code will always have an offset of 0 so this is fine uint32_t offset = moveBits(vertexCodeAddress.offset, 32, 0) | moveBits(vertexUniformsAddress, 32, 0); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = offset; cls->nextFreeByteOffset += 4; cls->nextFreeByteOffset += 4; *(uint16_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = moveBits(coordinateNumberOfUsedUniforms, 16, 0); cls->nextFreeByteOffset += 2; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = coordinateAttributeArraySelectBits; cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = coordinateTotalAttributesSize; cls->nextFreeByteOffset++; clEmitShaderRelocation(relocCl, handlesCl, handlesOffset, handlesSize, &coordinateCodeAddress); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = coordinateCodeAddress.offset; cls->nextFreeByteOffset += 4; *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = coordinateUniformsAddress; cls->nextFreeByteOffset += 4; } //input: 2 cls (cl, handles cl) void clInsertAttributeRecord(ControlList* cls, ControlList* relocCl, ControlList* handlesCl, uint32_t handlesOffset, uint32_t handlesSize, ControlListAddress address, uint32_t sizeBytes, uint32_t stride, uint32_t vertexVPMOffset, uint32_t coordinateVPMOffset) { assert(cls); assert(cls->CPA); uint32_t sizeBytesMinusOne = sizeBytes - 1; clEmitShaderRelocation(relocCl, handlesCl, handlesOffset, handlesSize, &address); *(uint32_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = address.offset; cls->nextFreeByteOffset += 4; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = sizeBytesMinusOne; cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = stride; cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = vertexVPMOffset; cls->nextFreeByteOffset++; *(uint8_t*)getCPAptrFromOffset(cls->CPA, cls->nextFreeByteOffset) = coordinateVPMOffset; cls->nextFreeByteOffset++; } uint32_t clGetHandleIndex(ControlList* handlesCl, uint32_t handlesOffset, uint32_t handlesSize, uint32_t handle) { uint32_t c = 0; //if curr marker is closed already we need to work with the stored size uint32_t numHandles = (handlesSize ? handlesSize : (handlesCl->nextFreeByteOffset - handlesOffset)) / 4; for(; c < numHandles; ++c) { if(((uint32_t*)getCPAptrFromOffset(handlesCl->CPA, handlesOffset))[c] == handle) { //found return c; } } //write handle to handles cl *(uint32_t*)getCPAptrFromOffset(handlesCl->CPA, handlesCl->nextFreeByteOffset) = handle; handlesCl->nextFreeByteOffset += 4; return c; } //input: 2 cls (cl + handles cl) inline void clEmitShaderRelocation(ControlList* relocCl, ControlList* handlesCl, uint32_t handlesOffset, uint32_t handlesSize, const ControlListAddress* address) { assert(relocCl); assert(relocCl->CPA); assert(handlesCl); assert(handlesCl->CPA); assert(address); assert(address->handle); //store offset within handles in cl *(uint32_t*)getCPAptrFromOffset(relocCl->CPA, relocCl->nextFreeByteOffset) = clGetHandleIndex(handlesCl, handlesOffset, handlesSize, address->handle); relocCl->nextFreeByteOffset += 4; } inline void clDummyRelocation(ControlList* relocCl, const ControlListAddress* address) {}