/* * Copyright (C) 2019-2020 Collabora, Ltd. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice (including the next * paragraph) shall be included in all copies or substantial portions of the * Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ #ifndef __BI_QUIRKS_H #define __BI_QUIRKS_H /* Model-specific quirks requiring compiler workarounds/etc. Quirks * may be errata requiring a workaround, or features. We're trying to be * quirk-positive here; quirky is the best! */ /* Whether this GPU lacks support for the preload mechanism. New GPUs can have * varyings and textures preloaded into the fragment shader to amortize the I/O * cost; early Bifrost models lacked this feature. */ #define BIFROST_NO_PRELOAD (1 << 0) /* Whether this GPU lacks support for fp32 transcendentals, requiring backend * lowering to low-precision lookup tables and polynomial approximation */ #define BIFROST_NO_FP32_TRANSCENDENTALS (1 << 1) /* Whether this GPU lacks support for the full form of the CLPER instruction. * These GPUs use a simple encoding of CLPER that does not support * inactive_result, subgroup_size, or lane_op. Using those features requires * lowering to additional ALU instructions. The encoding forces inactive_result * = zero, subgroup_size = subgroup4, and lane_op = none. */ #define BIFROST_LIMITED_CLPER (1 << 2) static inline unsigned bifrost_get_quirks(unsigned product_id) { switch (product_id >> 8) { case 0x60: return BIFROST_NO_PRELOAD | BIFROST_NO_FP32_TRANSCENDENTALS | BIFROST_LIMITED_CLPER; case 0x62: return BIFROST_NO_PRELOAD | BIFROST_LIMITED_CLPER; case 0x70: /* G31 */ return BIFROST_LIMITED_CLPER; case 0x71: case 0x72: case 0x73: case 0x74: return 0; case 0x90: case 0x91: case 0x92: case 0x93: case 0x94: case 0x95: return BIFROST_NO_PRELOAD; default: unreachable("Unknown Bifrost/Valhall GPU ID"); } } /* How many lanes per architectural warp (subgroup)? Used to lower divergent * indirects. */ static inline unsigned bifrost_lanes_per_warp(unsigned product_id) { switch (product_id >> 12) { case 6: return 4; case 7: return 8; case 9: return 16; default: unreachable("Invalid Bifrost/Valhall GPU major"); } } #endif