101e04c3fSmrg/* 201e04c3fSmrg * Copyright (C) 2016 Intel Corporation 301e04c3fSmrg * 401e04c3fSmrg * Permission is hereby granted, free of charge, to any person obtaining a 501e04c3fSmrg * copy of this software and associated documentation files (the "Software"), 601e04c3fSmrg * to deal in the Software without restriction, including without limitation 701e04c3fSmrg * the rights to use, copy, modify, merge, publish, distribute, sublicense, 801e04c3fSmrg * and/or sell copies of the Software, and to permit persons to whom the 901e04c3fSmrg * Software is furnished to do so, subject to the following conditions: 1001e04c3fSmrg * 1101e04c3fSmrg * The above copyright notice and this permission notice (including the next 1201e04c3fSmrg * paragraph) shall be included in all copies or substantial portions of the 1301e04c3fSmrg * Software. 1401e04c3fSmrg * 1501e04c3fSmrg * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 1601e04c3fSmrg * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 1701e04c3fSmrg * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 1801e04c3fSmrg * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 1901e04c3fSmrg * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 2001e04c3fSmrg * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 2101e04c3fSmrg * IN THE SOFTWARE. 2201e04c3fSmrg */ 2301e04c3fSmrg 247ec681f3Smrg#ifndef MESA_V3D_PACKET_HELPERS_H 257ec681f3Smrg#define MESA_V3D_PACKET_HELPERS_H 267ec681f3Smrg 2701e04c3fSmrg#include <stdio.h> 2801e04c3fSmrg#include <stdint.h> 2901e04c3fSmrg#include <stdbool.h> 3001e04c3fSmrg#include <assert.h> 3101e04c3fSmrg#include <math.h> 3201e04c3fSmrg#include "util/u_math.h" 3301e04c3fSmrg 3401e04c3fSmrg#ifdef HAVE_VALGRIND 3501e04c3fSmrg#include <valgrind.h> 3601e04c3fSmrg#include <memcheck.h> 3701e04c3fSmrg#define VG(x) x 3801e04c3fSmrg#ifndef NDEBUG 3901e04c3fSmrg#define __gen_validate_value(x) VALGRIND_CHECK_MEM_IS_DEFINED(&(x), sizeof(x)) 4001e04c3fSmrg#endif 4101e04c3fSmrg#else 427ec681f3Smrg#define VG(x) ((void)0) 4301e04c3fSmrg#endif 4401e04c3fSmrg 4501e04c3fSmrg#ifndef __gen_validate_value 4601e04c3fSmrg#define __gen_validate_value(x) 4701e04c3fSmrg#endif 4801e04c3fSmrg/* 4901e04c3fSmrg#ifndef __gen_address_type 5001e04c3fSmrg#error #define __gen_address_type before including this file 5101e04c3fSmrg#endif 5201e04c3fSmrg 5301e04c3fSmrg#ifndef __gen_user_data 5401e04c3fSmrg#error #define __gen_combine_address before including this file 5501e04c3fSmrg#endif 5601e04c3fSmrg*/ 5701e04c3fSmrgunion __gen_value { 5801e04c3fSmrg float f; 5901e04c3fSmrg uint32_t dw; 6001e04c3fSmrg}; 6101e04c3fSmrg 6201e04c3fSmrgstatic inline uint64_t 6301e04c3fSmrg__gen_mbo(uint32_t start, uint32_t end) 6401e04c3fSmrg{ 6501e04c3fSmrg return (~0ull >> (64 - (end - start + 1))) << start; 6601e04c3fSmrg} 6701e04c3fSmrg 6801e04c3fSmrgstatic inline uint64_t 6901e04c3fSmrg__gen_uint(uint64_t v, uint32_t start, uint32_t end) 7001e04c3fSmrg{ 7101e04c3fSmrg __gen_validate_value(v); 7201e04c3fSmrg 7301e04c3fSmrg#ifndef NDEBUG 7401e04c3fSmrg const int width = end - start + 1; 7501e04c3fSmrg if (width < 64) { 7601e04c3fSmrg const uint64_t max = (1ull << width) - 1; 7701e04c3fSmrg assert(v <= max); 7801e04c3fSmrg } 7901e04c3fSmrg#endif 8001e04c3fSmrg 8101e04c3fSmrg return v << start; 8201e04c3fSmrg} 8301e04c3fSmrg 8401e04c3fSmrgstatic inline uint64_t 8501e04c3fSmrg__gen_sint(int64_t v, uint32_t start, uint32_t end) 8601e04c3fSmrg{ 8701e04c3fSmrg const int width = end - start + 1; 8801e04c3fSmrg 8901e04c3fSmrg __gen_validate_value(v); 9001e04c3fSmrg 9101e04c3fSmrg#ifndef NDEBUG 9201e04c3fSmrg if (width < 64) { 9301e04c3fSmrg const int64_t max = (1ll << (width - 1)) - 1; 9401e04c3fSmrg const int64_t min = -(1ll << (width - 1)); 9501e04c3fSmrg assert(min <= v && v <= max); 9601e04c3fSmrg } 9701e04c3fSmrg#endif 9801e04c3fSmrg 9901e04c3fSmrg const uint64_t mask = ~0ull >> (64 - width); 10001e04c3fSmrg 10101e04c3fSmrg return (v & mask) << start; 10201e04c3fSmrg} 10301e04c3fSmrg 10401e04c3fSmrgstatic inline uint64_t 10501e04c3fSmrg__gen_offset(uint64_t v, uint32_t start, uint32_t end) 10601e04c3fSmrg{ 10701e04c3fSmrg __gen_validate_value(v); 10801e04c3fSmrg#ifndef NDEBUG 10901e04c3fSmrg uint64_t mask = (~0ull >> (64 - (end - start + 1))) << start; 11001e04c3fSmrg 11101e04c3fSmrg assert((v & ~mask) == 0); 11201e04c3fSmrg#endif 11301e04c3fSmrg 11401e04c3fSmrg return v; 11501e04c3fSmrg} 11601e04c3fSmrg 11701e04c3fSmrgstatic inline uint32_t 11801e04c3fSmrg__gen_float(float v) 11901e04c3fSmrg{ 12001e04c3fSmrg __gen_validate_value(v); 12101e04c3fSmrg return ((union __gen_value) { .f = (v) }).dw; 12201e04c3fSmrg} 12301e04c3fSmrg 12401e04c3fSmrgstatic inline uint64_t 12501e04c3fSmrg__gen_sfixed(float v, uint32_t start, uint32_t end, uint32_t fract_bits) 12601e04c3fSmrg{ 12701e04c3fSmrg __gen_validate_value(v); 12801e04c3fSmrg 12901e04c3fSmrg const float factor = (1 << fract_bits); 13001e04c3fSmrg 13101e04c3fSmrg#ifndef NDEBUG 13201e04c3fSmrg const float max = ((1 << (end - start)) - 1) / factor; 13301e04c3fSmrg const float min = -(1 << (end - start)) / factor; 13401e04c3fSmrg assert(min <= v && v <= max); 13501e04c3fSmrg#endif 13601e04c3fSmrg 13701e04c3fSmrg const int64_t int_val = llroundf(v * factor); 13801e04c3fSmrg const uint64_t mask = ~0ull >> (64 - (end - start + 1)); 13901e04c3fSmrg 14001e04c3fSmrg return (int_val & mask) << start; 14101e04c3fSmrg} 14201e04c3fSmrg 14301e04c3fSmrgstatic inline uint64_t 14401e04c3fSmrg__gen_ufixed(float v, uint32_t start, uint32_t end, uint32_t fract_bits) 14501e04c3fSmrg{ 14601e04c3fSmrg __gen_validate_value(v); 14701e04c3fSmrg 14801e04c3fSmrg const float factor = (1 << fract_bits); 14901e04c3fSmrg 15001e04c3fSmrg#ifndef NDEBUG 15101e04c3fSmrg const float max = ((1 << (end - start + 1)) - 1) / factor; 15201e04c3fSmrg const float min = 0.0f; 15301e04c3fSmrg assert(min <= v && v <= max); 15401e04c3fSmrg#endif 15501e04c3fSmrg 15601e04c3fSmrg const uint64_t uint_val = llroundf(v * factor); 15701e04c3fSmrg 15801e04c3fSmrg return uint_val << start; 15901e04c3fSmrg} 16001e04c3fSmrg 16101e04c3fSmrgstatic inline uint64_t 16201e04c3fSmrg__gen_unpack_uint(const uint8_t *restrict cl, uint32_t start, uint32_t end) 16301e04c3fSmrg{ 16401e04c3fSmrg uint64_t val = 0; 16501e04c3fSmrg const int width = end - start + 1; 16601e04c3fSmrg const uint32_t mask = (width == 32 ? ~0 : (1 << width) - 1 ); 16701e04c3fSmrg 1687ec681f3Smrg for (uint32_t byte = start / 8; byte <= end / 8; byte++) { 16901e04c3fSmrg val |= cl[byte] << ((byte - start / 8) * 8); 17001e04c3fSmrg } 17101e04c3fSmrg 17201e04c3fSmrg return (val >> (start % 8)) & mask; 17301e04c3fSmrg} 17401e04c3fSmrg 17501e04c3fSmrgstatic inline uint64_t 17601e04c3fSmrg__gen_unpack_sint(const uint8_t *restrict cl, uint32_t start, uint32_t end) 17701e04c3fSmrg{ 17801e04c3fSmrg int size = end - start + 1; 17901e04c3fSmrg int64_t val = __gen_unpack_uint(cl, start, end); 18001e04c3fSmrg 18101e04c3fSmrg /* Get the sign bit extended. */ 18201e04c3fSmrg return (val << (64 - size)) >> (64 - size); 18301e04c3fSmrg} 18401e04c3fSmrg 18501e04c3fSmrgstatic inline float 18601e04c3fSmrg__gen_unpack_sfixed(const uint8_t *restrict cl, uint32_t start, uint32_t end, 18701e04c3fSmrg uint32_t fractional_size) 18801e04c3fSmrg{ 18901e04c3fSmrg int32_t bits = __gen_unpack_sint(cl, start, end); 19001e04c3fSmrg return (float)bits / (1 << fractional_size); 19101e04c3fSmrg} 19201e04c3fSmrg 19301e04c3fSmrgstatic inline float 19401e04c3fSmrg__gen_unpack_ufixed(const uint8_t *restrict cl, uint32_t start, uint32_t end, 19501e04c3fSmrg uint32_t fractional_size) 19601e04c3fSmrg{ 19701e04c3fSmrg int32_t bits = __gen_unpack_uint(cl, start, end); 19801e04c3fSmrg return (float)bits / (1 << fractional_size); 19901e04c3fSmrg} 20001e04c3fSmrg 20101e04c3fSmrgstatic inline float 20201e04c3fSmrg__gen_unpack_float(const uint8_t *restrict cl, uint32_t start, uint32_t end) 20301e04c3fSmrg{ 20401e04c3fSmrg assert(start % 8 == 0); 20501e04c3fSmrg assert(end - start == 31); 20601e04c3fSmrg 20701e04c3fSmrg struct PACKED { float f; } *f = (void *)(cl + (start / 8)); 20801e04c3fSmrg 20901e04c3fSmrg return f->f; 21001e04c3fSmrg} 21101e04c3fSmrg 21201e04c3fSmrgstatic inline float 21301e04c3fSmrg__gen_unpack_f187(const uint8_t *restrict cl, uint32_t start, uint32_t end) 21401e04c3fSmrg{ 21501e04c3fSmrg assert(end - start == 15); 21601e04c3fSmrg uint32_t bits = __gen_unpack_uint(cl, start, end); 21701e04c3fSmrg return uif(bits << 16); 21801e04c3fSmrg} 21901e04c3fSmrg 2207ec681f3Smrg#endif 221