/* * Copyright © 2017 Broadcom * * 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. */ /** @file vc4_tiling_lt.c * * Helper functions from vc4_tiling.c that will be compiled for using NEON * assembly or not. * * If VC4_BUILD_NEON is set, then the functions will be suffixed with _neon. * They will only use NEON assembly if __ARM_ARCH is also set, to keep the x86 * sim build working. */ #include #include "pipe/p_state.h" #include "vc4_tiling.h" #ifdef VC4_BUILD_NEON #define NEON_TAG(x) x ## _neon #else #define NEON_TAG(x) x ## _base #endif static inline uint32_t align_down(uint32_t val, uint32_t align) { return val & ~(align - 1); } /** Returns the stride in bytes of a 64-byte microtile. */ static uint32_t vc4_utile_stride(int cpp) { switch (cpp) { case 1: return 8; case 2: case 4: case 8: return 16; default: unreachable("bad cpp"); } } static void vc4_load_utile(void *cpu, void *gpu, uint32_t cpu_stride, uint32_t cpp) { uint32_t gpu_stride = vc4_utile_stride(cpp); #if defined(VC4_BUILD_NEON) && defined(PIPE_ARCH_ARM) if (gpu_stride == 8) { __asm__ volatile ( /* Load from the GPU in one shot, no interleave, to * d0-d7. */ "vldm %0, {q0, q1, q2, q3}\n" /* Store each 8-byte line to cpu-side destination, * incrementing it by the stride each time. */ "vst1.8 d0, [%1], %2\n" "vst1.8 d1, [%1], %2\n" "vst1.8 d2, [%1], %2\n" "vst1.8 d3, [%1], %2\n" "vst1.8 d4, [%1], %2\n" "vst1.8 d5, [%1], %2\n" "vst1.8 d6, [%1], %2\n" "vst1.8 d7, [%1]\n" : : "r"(gpu), "r"(cpu), "r"(cpu_stride) : "q0", "q1", "q2", "q3"); } else { assert(gpu_stride == 16); __asm__ volatile ( /* Load from the GPU in one shot, no interleave, to * d0-d7. */ "vldm %0, {q0, q1, q2, q3};\n" /* Store each 16-byte line in 2 parts to the cpu-side * destination. (vld1 can only store one d-register * at a time). */ "vst1.8 d0, [%1], %3\n" "vst1.8 d1, [%2], %3\n" "vst1.8 d2, [%1], %3\n" "vst1.8 d3, [%2], %3\n" "vst1.8 d4, [%1], %3\n" "vst1.8 d5, [%2], %3\n" "vst1.8 d6, [%1]\n" "vst1.8 d7, [%2]\n" : : "r"(gpu), "r"(cpu), "r"(cpu + 8), "r"(cpu_stride) : "q0", "q1", "q2", "q3"); } #elif defined (PIPE_ARCH_AARCH64) if (gpu_stride == 8) { __asm__ volatile ( /* Load from the GPU in one shot, no interleave, to * d0-d7. */ "ld1 {v0.2d, v1.2d, v2.2d, v3.2d}, [%0]\n" /* Store each 8-byte line to cpu-side destination, * incrementing it by the stride each time. */ "st1 {v0.D}[0], [%1], %2\n" "st1 {v0.D}[1], [%1], %2\n" "st1 {v1.D}[0], [%1], %2\n" "st1 {v1.D}[1], [%1], %2\n" "st1 {v2.D}[0], [%1], %2\n" "st1 {v2.D}[1], [%1], %2\n" "st1 {v3.D}[0], [%1], %2\n" "st1 {v3.D}[1], [%1]\n" : : "r"(gpu), "r"(cpu), "r"(cpu_stride) : "v0", "v1", "v2", "v3"); } else { assert(gpu_stride == 16); __asm__ volatile ( /* Load from the GPU in one shot, no interleave, to * d0-d7. */ "ld1 {v0.2d, v1.2d, v2.2d, v3.2d}, [%0]\n" /* Store each 16-byte line in 2 parts to the cpu-side * destination. (vld1 can only store one d-register * at a time). */ "st1 {v0.D}[0], [%1], %3\n" "st1 {v0.D}[1], [%2], %3\n" "st1 {v1.D}[0], [%1], %3\n" "st1 {v1.D}[1], [%2], %3\n" "st1 {v2.D}[0], [%1], %3\n" "st1 {v2.D}[1], [%2], %3\n" "st1 {v3.D}[0], [%1]\n" "st1 {v3.D}[1], [%2]\n" : : "r"(gpu), "r"(cpu), "r"(cpu + 8), "r"(cpu_stride) : "v0", "v1", "v2", "v3"); } #else for (uint32_t gpu_offset = 0; gpu_offset < 64; gpu_offset += gpu_stride) { memcpy(cpu, gpu + gpu_offset, gpu_stride); cpu += cpu_stride; } #endif } static void vc4_store_utile(void *gpu, void *cpu, uint32_t cpu_stride, uint32_t cpp) { uint32_t gpu_stride = vc4_utile_stride(cpp); #if defined(VC4_BUILD_NEON) && defined(PIPE_ARCH_ARM) if (gpu_stride == 8) { __asm__ volatile ( /* Load each 8-byte line from cpu-side source, * incrementing it by the stride each time. */ "vld1.8 d0, [%1], %2\n" "vld1.8 d1, [%1], %2\n" "vld1.8 d2, [%1], %2\n" "vld1.8 d3, [%1], %2\n" "vld1.8 d4, [%1], %2\n" "vld1.8 d5, [%1], %2\n" "vld1.8 d6, [%1], %2\n" "vld1.8 d7, [%1]\n" /* Load from the GPU in one shot, no interleave, to * d0-d7. */ "vstm %0, {q0, q1, q2, q3}\n" : : "r"(gpu), "r"(cpu), "r"(cpu_stride) : "q0", "q1", "q2", "q3"); } else { assert(gpu_stride == 16); __asm__ volatile ( /* Load each 16-byte line in 2 parts from the cpu-side * destination. (vld1 can only store one d-register * at a time). */ "vld1.8 d0, [%1], %3\n" "vld1.8 d1, [%2], %3\n" "vld1.8 d2, [%1], %3\n" "vld1.8 d3, [%2], %3\n" "vld1.8 d4, [%1], %3\n" "vld1.8 d5, [%2], %3\n" "vld1.8 d6, [%1]\n" "vld1.8 d7, [%2]\n" /* Store to the GPU in one shot, no interleave. */ "vstm %0, {q0, q1, q2, q3}\n" : : "r"(gpu), "r"(cpu), "r"(cpu + 8), "r"(cpu_stride) : "q0", "q1", "q2", "q3"); } #elif defined (PIPE_ARCH_AARCH64) if (gpu_stride == 8) { __asm__ volatile ( /* Load each 8-byte line from cpu-side source, * incrementing it by the stride each time. */ "ld1 {v0.D}[0], [%1], %2\n" "ld1 {v0.D}[1], [%1], %2\n" "ld1 {v1.D}[0], [%1], %2\n" "ld1 {v1.D}[1], [%1], %2\n" "ld1 {v2.D}[0], [%1], %2\n" "ld1 {v2.D}[1], [%1], %2\n" "ld1 {v3.D}[0], [%1], %2\n" "ld1 {v3.D}[1], [%1]\n" /* Store to the GPU in one shot, no interleave. */ "st1 {v0.2d, v1.2d, v2.2d, v3.2d}, [%0]\n" : : "r"(gpu), "r"(cpu), "r"(cpu_stride) : "v0", "v1", "v2", "v3"); } else { assert(gpu_stride == 16); __asm__ volatile ( /* Load each 16-byte line in 2 parts from the cpu-side * destination. (vld1 can only store one d-register * at a time). */ "ld1 {v0.D}[0], [%1], %3\n" "ld1 {v0.D}[1], [%2], %3\n" "ld1 {v1.D}[0], [%1], %3\n" "ld1 {v1.D}[1], [%2], %3\n" "ld1 {v2.D}[0], [%1], %3\n" "ld1 {v2.D}[1], [%2], %3\n" "ld1 {v3.D}[0], [%1]\n" "ld1 {v3.D}[1], [%2]\n" /* Store to the GPU in one shot, no interleave. */ "st1 {v0.2d, v1.2d, v2.2d, v3.2d}, [%0]\n" : : "r"(gpu), "r"(cpu), "r"(cpu + 8), "r"(cpu_stride) : "v0", "v1", "v2", "v3"); } #else for (uint32_t gpu_offset = 0; gpu_offset < 64; gpu_offset += gpu_stride) { memcpy(gpu + gpu_offset, cpu, gpu_stride); cpu += cpu_stride; } #endif } /** * Returns the X value into the address bits for LT tiling. * * The LT tile load/stores rely on the X bits not intersecting with the Y * bits. Because of this, we have to choose to put the utile index within the * LT tile into one of the two values, and we do so in swizzle_lt_x() to make * NPOT handling easier. */ static uint32_t swizzle_lt_x(int x, int cpp) { switch (cpp) { case 1: /* 8x8 inside of 4x4 */ return ((x & 0x7) << (0 - 0) | (x & ~0x7) << (6 - 3)); case 2: /* 8x4 inside of 4x4 */ return ((x & 0x7) << (1 - 0) | (x & ~0x7) << (6 - 3)); case 4: /* 4x4 inside of 4x4 */ return ((x & 0x3) << (2 - 0) | (x & ~0x3) << (6 - 2)); case 8: /* 2x4 inside of 4x4 */ return ((x & 0x1) << (3 - 0) | (x & ~0x1) << (6 - 1)); default: unreachable("bad cpp"); } } /** * Returns the Y value into the address bits for LT tiling. * * The LT tile load/stores rely on the X bits not intersecting with the Y * bits. */ static uint32_t swizzle_lt_y(int y, int cpp) { switch (cpp) { case 1: /* 8x8 inside of 4x4 */ return ((y & 0x7) << 3); case 2: /* 8x4 inside of 4x4 */ return ((y & 0x3) << 4); case 4: /* 4x4 inside of 4x4 */ return ((y & 0x3) << 4); case 8: /* 2x4 inside of 4x4 */ return ((y & 0x3) << 4); default: unreachable("bad cpp"); } } /** * Helper for loading or storing to an LT image, where the box is aligned * to utiles. * * This just breaks the box down into calls to the fast * vc4_load_utile/vc4_store_utile helpers. */ static inline void vc4_lt_image_aligned(void *gpu, uint32_t gpu_stride, void *cpu, uint32_t cpu_stride, int cpp, const struct pipe_box *box, bool to_cpu) { uint32_t utile_w = vc4_utile_width(cpp); uint32_t utile_h = vc4_utile_height(cpp); uint32_t xstart = box->x; uint32_t ystart = box->y; for (uint32_t y = 0; y < box->height; y += utile_h) { for (uint32_t x = 0; x < box->width; x += utile_w) { void *gpu_tile = gpu + ((ystart + y) * gpu_stride + (xstart + x) * 64 / utile_w); if (to_cpu) { vc4_load_utile(cpu + (cpu_stride * y + x * cpp), gpu_tile, cpu_stride, cpp); } else { vc4_store_utile(gpu_tile, cpu + (cpu_stride * y + x * cpp), cpu_stride, cpp); } } } } /** * Helper for loading or storing to an LT image, where the box is not aligned * to utiles. * * This walks through the raster-order data, copying to/from the corresponding * tiled pixel. This means we don't get write-combining on stores, but the * loop is very few CPU instructions since the memcpy will be inlined. */ static inline void vc4_lt_image_unaligned(void *gpu, uint32_t gpu_stride, void *cpu, uint32_t cpu_stride, int cpp, const struct pipe_box *box, bool to_cpu) { /* These are the address bits for the start of the box, split out into * x/y so that they can be incremented separately in their loops. */ uint32_t offs_x0 = swizzle_lt_x(box->x, cpp); uint32_t offs_y = swizzle_lt_y(box->y, cpp); /* The *_mask values are "what bits of the address are from x or y" */ uint32_t x_mask = swizzle_lt_x(~0, cpp); uint32_t y_mask = swizzle_lt_y(~0, cpp); uint32_t incr_y = swizzle_lt_x(gpu_stride / cpp, cpp); assert(!(x_mask & y_mask)); offs_x0 += incr_y * (box->y / vc4_utile_height(cpp)); for (uint32_t y = 0; y < box->height; y++) { void *gpu_row = gpu + offs_y; uint32_t offs_x = offs_x0; for (uint32_t x = 0; x < box->width; x++) { /* Use a memcpy here to move a pixel's worth of data. * We're relying on this function to be inlined, so * this will get expanded into the appropriate 1, 2, * or 4-byte move. */ if (to_cpu) { memcpy(cpu + x * cpp, gpu_row + offs_x, cpp); } else { memcpy(gpu_row + offs_x, cpu + x * cpp, cpp); } /* This math trick with x_mask increments offs_x by 1 * in x. */ offs_x = (offs_x - x_mask) & x_mask; } offs_y = (offs_y - y_mask) & y_mask; /* When offs_y wraps (we hit the end of the utile), we * increment offs_x0 by effectively the utile stride. */ if (!offs_y) offs_x0 += incr_y; cpu += cpu_stride; } } /** * General LT image load/store helper. */ static inline void vc4_lt_image_helper(void *gpu, uint32_t gpu_stride, void *cpu, uint32_t cpu_stride, int cpp, const struct pipe_box *box, bool to_cpu) { if (box->x & (vc4_utile_width(cpp) - 1) || box->y & (vc4_utile_height(cpp) - 1) || box->width & (vc4_utile_width(cpp) - 1) || box->height & (vc4_utile_height(cpp) - 1)) { vc4_lt_image_unaligned(gpu, gpu_stride, cpu, cpu_stride, cpp, box, to_cpu); } else { vc4_lt_image_aligned(gpu, gpu_stride, cpu, cpu_stride, cpp, box, to_cpu); } } static inline void vc4_lt_image_cpp_helper(void *gpu, uint32_t gpu_stride, void *cpu, uint32_t cpu_stride, int cpp, const struct pipe_box *box, bool to_cpu) { switch (cpp) { case 1: vc4_lt_image_helper(gpu, gpu_stride, cpu, cpu_stride, 1, box, to_cpu); break; case 2: vc4_lt_image_helper(gpu, gpu_stride, cpu, cpu_stride, 2, box, to_cpu); break; case 4: vc4_lt_image_helper(gpu, gpu_stride, cpu, cpu_stride, 4, box, to_cpu); break; case 8: vc4_lt_image_helper(gpu, gpu_stride, cpu, cpu_stride, 8, box, to_cpu); break; default: unreachable("bad cpp"); } } void NEON_TAG(vc4_load_lt_image)(void *dst, uint32_t dst_stride, void *src, uint32_t src_stride, int cpp, const struct pipe_box *box) { vc4_lt_image_cpp_helper(src, src_stride, dst, dst_stride, cpp, box, true); } void NEON_TAG(vc4_store_lt_image)(void *dst, uint32_t dst_stride, void *src, uint32_t src_stride, int cpp, const struct pipe_box *box) { vc4_lt_image_cpp_helper(dst, dst_stride, src, src_stride, cpp, box, false); }