diff options
Diffstat (limited to 'src/compute')
55 files changed, 343 insertions, 343 deletions
diff --git a/src/compute/color/color.c b/src/compute/color/color.c index 3e4b1c8c10..929754639d 100644 --- a/src/compute/color/color.c +++ b/src/compute/color/color.c @@ -104,8 +104,8 @@ color_linear_to_srgb_rgb_f32(float rgb[3]) // // -static -float +static +float lerpf(float const a, float const b, float const t) { return fmaf(t, b, fmaf(-t, a, a)); @@ -115,8 +115,8 @@ lerpf(float const a, float const b, float const t) // // -void -color_linear_lerp_rgba_f32(float rgba_m[4], +void +color_linear_lerp_rgba_f32(float rgba_m[4], float const rgba_a[4], float const rgba_b[4], float const t) diff --git a/src/compute/color/color.h b/src/compute/color/color.h index 36f1293b20..452ed190e5 100644 --- a/src/compute/color/color.h +++ b/src/compute/color/color.h @@ -59,7 +59,7 @@ void color_linear_to_srgb_rgb_f32(float rgb[3]); // // -void color_linear_lerp_rgba_f32(float rgba_m[4], +void color_linear_lerp_rgba_f32(float rgba_m[4], float const rgba_a[4], float const rgba_b[4], float const t); diff --git a/src/compute/common/cl/assert_cl.c b/src/compute/common/cl/assert_cl.c index 91e5620c5f..5d420586b3 100644 --- a/src/compute/common/cl/assert_cl.c +++ b/src/compute/common/cl/assert_cl.c @@ -147,7 +147,7 @@ assert_cl(cl_int const code, char const * const file, int const line, bool const // void -cl_get_event_info(cl_event event, +cl_get_event_info(cl_event event, cl_int * const status, cl_command_type * const type) { @@ -169,7 +169,7 @@ cl_get_event_info(cl_event event, } -char const * +char const * cl_get_event_command_status_string(cl_int const status) { switch (status) @@ -177,14 +177,14 @@ cl_get_event_command_status_string(cl_int const status) CL_VAL_TO_STRING(CL_QUEUED); CL_VAL_TO_STRING(CL_SUBMITTED); CL_VAL_TO_STRING(CL_RUNNING); - CL_VAL_TO_STRING(CL_COMPLETE); + CL_VAL_TO_STRING(CL_COMPLETE); default: return "UNKNOWN COMMAND STATUS"; - } + } } -char const * +char const * cl_get_event_command_type_string(cl_command_type const type) { switch (type) diff --git a/src/compute/common/cl/assert_cl.h b/src/compute/common/cl/assert_cl.h index 5481f3850f..517ada8d37 100644 --- a/src/compute/common/cl/assert_cl.h +++ b/src/compute/common/cl/assert_cl.h @@ -19,13 +19,13 @@ // // -char const * +char const * cl_get_error_string(cl_int const err); -cl_int -assert_cl(cl_int const code, - char const * const file, - int const line, +cl_int +assert_cl(cl_int const code, + char const * const file, + int const line, bool const abort); // @@ -40,14 +40,14 @@ assert_cl(cl_int const code, // void -cl_get_event_info(cl_event event, +cl_get_event_info(cl_event event, cl_int * const status, cl_command_type * const type); -char const * +char const * cl_get_event_command_status_string(cl_int const status); -char const * +char const * cl_get_event_command_type_string(cl_command_type const type); // diff --git a/src/compute/common/cl/find_cl.c b/src/compute/common/cl/find_cl.c index 43b26d1d01..a04d9ebd69 100644 --- a/src/compute/common/cl/find_cl.c +++ b/src/compute/common/cl/find_cl.c @@ -42,7 +42,7 @@ clFindIdsByName(char const * const target_platform_substring, // get number of platforms // cl_uint platform_count; - + cl(GetPlatformIDs(0,NULL,&platform_count)); cl_platform_id * const platform_ids = ALLOCA(sizeof(*platform_ids) * platform_count); @@ -94,13 +94,13 @@ clFindIdsByName(char const * const target_platform_substring, &device_count); cl_device_id * const device_ids = ALLOCA(sizeof(*device_ids) * device_count); - + cl_err = clGetDeviceIDs(platform_ids[ii], CL_DEVICE_TYPE_ALL, device_count, device_ids, NULL); - + if (cl_err != CL_DEVICE_NOT_FOUND) cl_ok(cl_err); @@ -135,7 +135,7 @@ clFindIdsByName(char const * const target_platform_substring, driver_version_size, driver_version, NULL)); - + if (!match_device && match_platform && (strstr(device_name,target_device_substring) != NULL)) { match_device = true; @@ -148,7 +148,7 @@ clFindIdsByName(char const * const target_platform_substring, if (matched_device_name_size >= 1) matched_device_name[matched_device_name_size-1] = 0; - if (matched_device_name_size > 1) + if (matched_device_name_size > 1) { bytes = MIN_MACRO(device_name_size,matched_device_name_size-1); @@ -202,7 +202,7 @@ clFindIdsByName(char const * const target_platform_substring, } return CL_SUCCESS; -} +} // // diff --git a/src/compute/common/util.c b/src/compute/common/util.c index 69710125a3..eb05d91a9f 100644 --- a/src/compute/common/util.c +++ b/src/compute/common/util.c @@ -6,7 +6,7 @@ * */ -#include <intrin.h> +#include <intrin.h> // // diff --git a/src/compute/hs/cl/bench/sort.cpp b/src/compute/hs/cl/bench/sort.cpp index 861c498ae4..13e0af556e 100644 --- a/src/compute/hs/cl/bench/sort.cpp +++ b/src/compute/hs/cl/bench/sort.cpp @@ -84,11 +84,11 @@ hs_cpu_sort_u32(uint32_t * a, uint32_t const count) } extern "C" -char const * +char const * hs_cpu_sort_u64(uint64_t * a, uint32_t const count) { #if defined ( HS_USE_PARALLEL_SORT ) - std::sort(std::execution::par_unseq,a,a+count); + std::sort(std::execution::par_unseq,a,a+count); return "std::sort(std::execution::par_unseq)()"; #elif defined ( HS_USE_STD_SORT ) std::sort(a,a+count); diff --git a/src/compute/hs/cl/gen9/hs_cl.h b/src/compute/hs/cl/gen9/hs_cl.h index a33b2b7b93..4926a14fb3 100644 --- a/src/compute/hs/cl/gen9/hs_cl.h +++ b/src/compute/hs/cl/gen9/hs_cl.h @@ -1,26 +1,26 @@ -// -// Copyright 2016 Google Inc. -// -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. -// - -#ifndef HS_CL_ONCE -#define HS_CL_ONCE - -#define HS_LANES_PER_WARP_LOG2 3 -#define HS_LANES_PER_WARP (1 << HS_LANES_PER_WARP_LOG2) -#define HS_BS_WARPS 16 -#define HS_BS_WARPS_LOG2_RU 4 -#define HS_BC_WARPS_LOG2_MAX 4 -#define HS_FM_BLOCKS_LOG2_MIN 1 -#define HS_HM_BLOCKS_LOG2_MIN 1 -#define HS_KEYS_PER_LANE 16 -#define HS_REG_LAST(c) c##16 -#define HS_KEY_WORDS 2 -#define HS_KEY_TYPE ulong -#define HS_EMPTY - +// +// Copyright 2016 Google Inc. +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. +// + +#ifndef HS_CL_ONCE +#define HS_CL_ONCE + +#define HS_LANES_PER_WARP_LOG2 3 +#define HS_LANES_PER_WARP (1 << HS_LANES_PER_WARP_LOG2) +#define HS_BS_WARPS 16 +#define HS_BS_WARPS_LOG2_RU 4 +#define HS_BC_WARPS_LOG2_MAX 4 +#define HS_FM_BLOCKS_LOG2_MIN 1 +#define HS_HM_BLOCKS_LOG2_MIN 1 +#define HS_KEYS_PER_LANE 16 +#define HS_REG_LAST(c) c##16 +#define HS_KEY_WORDS 2 +#define HS_KEY_TYPE ulong +#define HS_EMPTY + #define HS_SLAB_ROWS() \ HS_SLAB_ROW( 1, 0 ) \ HS_SLAB_ROW( 2, 1 ) \ @@ -39,7 +39,7 @@ HS_SLAB_ROW( 15, 14 ) \ HS_SLAB_ROW( 16, 15 ) \ HS_EMPTY - + #define HS_TRANSPOSE_SLAB() \ HS_TRANSPOSE_STAGE( 1 ) \ HS_TRANSPOSE_STAGE( 2 ) \ @@ -85,38 +85,38 @@ HS_TRANSPOSE_REMAP( u, 15, 14 ) \ HS_TRANSPOSE_REMAP( u, 16, 16 ) \ HS_EMPTY - -#define HS_FM_BLOCKS_LOG2_1 0 -#define HS_FM_BLOCKS_LOG2_2 1 -#define HS_FM_BLOCKS_LOG2_3 2 -#define HS_FM_BLOCKS_LOG2_4 3 -#define HS_FM_BLOCKS_LOG2_5 4 -#define HS_FM_BLOCKS_LOG2_6 5 -#define HS_HM_BLOCKS_LOG2_5 0 -#define HS_FM_BLOCKS_LOG2_7 6 -#define HS_HM_BLOCKS_LOG2_6 1 -#define HS_FM_BLOCKS_LOG2_8 7 -#define HS_HM_BLOCKS_LOG2_7 2 -#define HS_FM_BLOCKS_LOG2_9 8 -#define HS_HM_BLOCKS_LOG2_8 3 -#define HS_FM_BLOCKS_LOG2_10 9 -#define HS_HM_BLOCKS_LOG2_9 4 -#define HS_FM_BLOCKS_LOG2_11 10 -#define HS_HM_BLOCKS_LOG2_10 5 -#define HS_FM_BLOCKS_LOG2_12 11 -#define HS_HM_BLOCKS_LOG2_11 6 -#define HS_FM_BLOCKS_LOG2_13 12 -#define HS_HM_BLOCKS_LOG2_12 7 -#define HS_FM_BLOCKS_LOG2_14 13 -#define HS_HM_BLOCKS_LOG2_13 8 -#define HS_FM_BLOCKS_LOG2_15 14 -#define HS_HM_BLOCKS_LOG2_14 9 -#define HS_FM_BLOCKS_LOG2_16 15 -#define HS_HM_BLOCKS_LOG2_15 10 - -#endif - -// -// -// - + +#define HS_FM_BLOCKS_LOG2_1 0 +#define HS_FM_BLOCKS_LOG2_2 1 +#define HS_FM_BLOCKS_LOG2_3 2 +#define HS_FM_BLOCKS_LOG2_4 3 +#define HS_FM_BLOCKS_LOG2_5 4 +#define HS_FM_BLOCKS_LOG2_6 5 +#define HS_HM_BLOCKS_LOG2_5 0 +#define HS_FM_BLOCKS_LOG2_7 6 +#define HS_HM_BLOCKS_LOG2_6 1 +#define HS_FM_BLOCKS_LOG2_8 7 +#define HS_HM_BLOCKS_LOG2_7 2 +#define HS_FM_BLOCKS_LOG2_9 8 +#define HS_HM_BLOCKS_LOG2_8 3 +#define HS_FM_BLOCKS_LOG2_10 9 +#define HS_HM_BLOCKS_LOG2_9 4 +#define HS_FM_BLOCKS_LOG2_11 10 +#define HS_HM_BLOCKS_LOG2_10 5 +#define HS_FM_BLOCKS_LOG2_12 11 +#define HS_HM_BLOCKS_LOG2_11 6 +#define HS_FM_BLOCKS_LOG2_13 12 +#define HS_HM_BLOCKS_LOG2_12 7 +#define HS_FM_BLOCKS_LOG2_14 13 +#define HS_HM_BLOCKS_LOG2_13 8 +#define HS_FM_BLOCKS_LOG2_15 14 +#define HS_HM_BLOCKS_LOG2_14 9 +#define HS_FM_BLOCKS_LOG2_16 15 +#define HS_HM_BLOCKS_LOG2_15 10 + +#endif + +// +// +// + diff --git a/src/compute/hs/cl/hs_cl_launcher.c b/src/compute/hs/cl/hs_cl_launcher.c index b0b9e4091b..f8a87f1dde 100644 --- a/src/compute/hs/cl/hs_cl_launcher.c +++ b/src/compute/hs/cl/hs_cl_launcher.c @@ -774,7 +774,7 @@ hs_fm_launcher(struct hs_state * const state, // how many full-sized scaled flip-merge spans are there? state->fm.full = state->bx.ru / full_span_warps; state->fm.frac = 0; - + // initialize down_warps *down_warps = state->fm.full * full_span_warps; @@ -787,10 +787,10 @@ hs_fm_launcher(struct hs_state * const state, uint32_t const frac_rem = span_rem - half_span_warps; uint32_t const frac_rem_pow2 = pow2_ru_u32(frac_rem); - if (frac_rem_pow2 >= half_span_warps) + if (frac_rem_pow2 >= half_span_warps) { *down_warps += full_span_warps; - state->fm.full += 1; + state->fm.full += 1; } else { @@ -895,7 +895,7 @@ hs_pad(uint32_t const count, void hs_sort(cl_command_queue cq, // out-of-order cq - cl_mem vin, + cl_mem vin, cl_mem vout, uint32_t const count, uint32_t const count_padded_in, @@ -1019,7 +1019,7 @@ hs_sort(cl_command_queue cq, // out-of-order cq } } - if (linearize) + if (linearize) { // launch linearize; hs_transpose_launcher(&state,cq); @@ -1113,7 +1113,7 @@ hs_create(cl_context context, HS_CREATE_KERNELS(bc_kernels); HS_CREATE_KERNELS(fm_kernels); HS_CREATE_KERNELS(hm_kernels); - HS_CREATE_KERNELS(transpose_kernels); + HS_CREATE_KERNELS(transpose_kernels); cl(ReleaseProgram(program)); } diff --git a/src/compute/hs/gen/main.c b/src/compute/hs/gen/main.c index a79c34c08e..42f4518bfd 100644 --- a/src/compute/hs/gen/main.c +++ b/src/compute/hs/gen/main.c @@ -882,7 +882,7 @@ hsg_bs_flip_merge_level(struct hsg_op * ops, uint32_t const level, uint32_t const s_pairs) { - // + // // Note there are a number of ways to flip merge these warps. There // is a magic number in the merge structure that indicates which // warp to activate as well as what network size to invoke. @@ -917,7 +917,7 @@ hsg_bs_flip_merge_level(struct hsg_op * ops, uint32_t s_rows = s_pairs * 2; uint32_t base = 0; - + while (s_rows > 0) { uint32_t active = merge->warps; @@ -934,7 +934,7 @@ hsg_bs_flip_merge_level(struct hsg_op * ops, // how many equal number of rows to merge? uint32_t loops = s_rows / active; - // decrement + // decrement s_rows -= loops * active; for (uint32_t ss=0; ss<loops; ss++) @@ -998,7 +998,7 @@ hsg_bs_flip_merge(struct hsg_op * ops, struct hsg_merge const * const merge) { uint32_t const count = merge->levels[level].count; - if (count == 0) + if (count == 0) continue; uint32_t const r_mid = hsg_config.thread.regs/2 + 1; diff --git a/src/compute/hs/gen/target_cuda_sm3x.c b/src/compute/hs/gen/target_cuda_sm3x.c index 436fc5f50d..6369aa33b0 100644 --- a/src/compute/hs/gen/target_cuda_sm3x.c +++ b/src/compute/hs/gen/target_cuda_sm3x.c @@ -51,7 +51,7 @@ hsg_target_cuda_sm3x(struct hsg_file * const files, uint32_t const bc_max = msb_idx_u32(pow2_rd_u32(merge[0].warps)); fprintf(files[HSG_FILE_TYPE_HEADER].file, - "// \n" + "// \n" "// Copyright 2016 Google Inc. \n" "// \n" "// Use of this source code is governed by a BSD-style \n" @@ -86,7 +86,7 @@ hsg_target_cuda_sm3x(struct hsg_file * const files, files[HSG_FILE_TYPE_SOURCE].name); fprintf(files[HSG_FILE_TYPE_SOURCE].file, - "// \n" + "// \n" "// Copyright 2016 Google Inc. \n" "// \n" "// Use of this source code is governed by a BSD-style \n" @@ -461,12 +461,12 @@ hsg_target_cuda_sm3x(struct hsg_file * const files, break; case HSG_OP_TYPE_CMP_FLIP: - fprintf(files[HSG_FILE_TYPE_SOURCE].file, + fprintf(files[HSG_FILE_TYPE_SOURCE].file, "HS_CMP_FLIP(r%-3u,r%-3u,r%-3u)\n",ops->a,ops->b,ops->c); break; case HSG_OP_TYPE_CMP_HALF: - fprintf(files[HSG_FILE_TYPE_SOURCE].file, + fprintf(files[HSG_FILE_TYPE_SOURCE].file, "HS_CMP_HALF(r%-3u,r%-3u)\n",ops->a,ops->b); break; diff --git a/src/compute/hs/gen/target_igp_genx.c b/src/compute/hs/gen/target_igp_genx.c index f321975ba0..3d0f2bc1b8 100644 --- a/src/compute/hs/gen/target_igp_genx.c +++ b/src/compute/hs/gen/target_igp_genx.c @@ -21,15 +21,15 @@ // // -static +static char hsg_transpose_reg_prefix(uint32_t const cols_log2) { return 'a' + (('r' + cols_log2 - 'a') % 26); } -static -void +static +void hsg_transpose_blend(uint32_t const cols_log2, uint32_t const row_ll, // lower-left uint32_t const row_ur, // upper-right @@ -41,10 +41,10 @@ hsg_transpose_blend(uint32_t const cols_log2, hsg_transpose_reg_prefix(cols_log2-1), hsg_transpose_reg_prefix(cols_log2), cols_log2,row_ll+1,row_ur+1); -} +} -static -void +static +void hsg_transpose_remap(uint32_t const row_from, uint32_t const row_to, FILE * file) @@ -54,7 +54,7 @@ hsg_transpose_remap(uint32_t const row_from, " HS_TRANSPOSE_REMAP( %c, %3u, %3u ) \\\n", hsg_transpose_reg_prefix(msb_idx_u32(hsg_config.warp.lanes)), row_from+1,row_to+1); -} +} // // @@ -89,7 +89,7 @@ hsg_target_igp_genx(struct hsg_file * const files, uint32_t const warp_lanes_log2 = msb_idx_u32(hsg_config.warp.lanes); fprintf(files[HSG_FILE_TYPE_HEADER].file, - "// \n" + "// \n" "// Copyright 2016 Google Inc. \n" "// \n" "// Use of this source code is governed by a BSD-style \n" @@ -125,7 +125,7 @@ hsg_target_igp_genx(struct hsg_file * const files, fprintf(files[HSG_FILE_TYPE_HEADER].file, "#define HS_SLAB_ROWS() \\\n"); - + for (uint32_t ii=1; ii<=hsg_config.thread.regs; ii++) fprintf(files[HSG_FILE_TYPE_HEADER].file, " HS_SLAB_ROW( %3u, %3u ) \\\n",ii,ii-1); @@ -147,13 +147,13 @@ hsg_target_igp_genx(struct hsg_file * const files, files[HSG_FILE_TYPE_HEADER].file, hsg_transpose_blend, hsg_transpose_remap); - + fprintf(files[HSG_FILE_TYPE_HEADER].file, " HS_EMPTY\n" " \n"); fprintf(files[HSG_FILE_TYPE_SOURCE].file, - "// \n" + "// \n" "// Copyright 2016 Google Inc. \n" "// \n" "// Use of this source code is governed by a BSD-style \n" @@ -215,11 +215,11 @@ hsg_target_igp_genx(struct hsg_file * const files, "HS_TRANSPOSE_SLAB()\n"); } break; - + case HSG_OP_TYPE_BS_KERNEL_PROTO: { struct hsg_merge const * const m = merge + ops->a; - + uint32_t const tpb = m->warps * hsg_config.warp.lanes; uint32_t const bs = pow2_ru_u32(m->warps); uint32_t const msb = msb_idx_u32(bs); @@ -236,7 +236,7 @@ hsg_target_igp_genx(struct hsg_file * const files, msb); } break; - + case HSG_OP_TYPE_BS_KERNEL_PREAMBLE: { fprintf(files[HSG_FILE_TYPE_SOURCE].file, @@ -466,12 +466,12 @@ hsg_target_igp_genx(struct hsg_file * const files, break; case HSG_OP_TYPE_CMP_FLIP: - fprintf(files[HSG_FILE_TYPE_SOURCE].file, + fprintf(files[HSG_FILE_TYPE_SOURCE].file, "HS_CMP_FLIP(%-3u,r%-3u,r%-3u)\n",ops->a,ops->b,ops->c); break; case HSG_OP_TYPE_CMP_HALF: - fprintf(files[HSG_FILE_TYPE_SOURCE].file, + fprintf(files[HSG_FILE_TYPE_SOURCE].file, "HS_CMP_HALF(%-3u,r%-3u)\n",ops->a,ops->b); break; @@ -523,7 +523,7 @@ hsg_target_igp_genx(struct hsg_file * const files, ops->c, ops->a); break; - + case HSG_OP_TYPE_BS_REG_SHARED_LOAD_LEFT: fprintf(files[HSG_FILE_TYPE_SOURCE].file, "HS_KEY_TYPE r%u_%-3u = (shared.m + smem_l_idx)[%u];\n", @@ -598,7 +598,7 @@ hsg_target_igp_genx(struct hsg_file * const files, #endif } break; - + case HSG_OP_TYPE_BC_MERGE_H_PREAMBLE: { struct hsg_merge const * const m = merge + ops->a; @@ -613,7 +613,7 @@ hsg_target_igp_genx(struct hsg_file * const files, } break; - + case HSG_OP_TYPE_BX_MERGE_H_PRED: fprintf(files[HSG_FILE_TYPE_SOURCE].file, "if (get_sub_group_id() < %u)\n", diff --git a/src/compute/hs/gen/transpose.c b/src/compute/hs/gen/transpose.c index f99e966ff7..de15c62631 100644 --- a/src/compute/hs/gen/transpose.c +++ b/src/compute/hs/gen/transpose.c @@ -33,7 +33,7 @@ hsg_transpose(uint32_t const cols_log2, { // get mapping array uint32_t * map_curr = ALLOCA(rows * sizeof(*map_curr)); - uint32_t * map_next = ALLOCA(rows * sizeof(*map_next)); + uint32_t * map_next = ALLOCA(rows * sizeof(*map_next)); // init the mapping array for (uint32_t ii=0; ii<rows; ii++) @@ -53,7 +53,7 @@ hsg_transpose(uint32_t const cols_log2, { for (uint32_t jj=0; jj<rows; jj++) { - if (map_curr[jj] == stay) + if (map_curr[jj] == stay) { map_next[jj] = stay; map_next[ii] = stay + (rows << (cc-1)); @@ -88,7 +88,7 @@ hsg_transpose(uint32_t const cols_log2, static uint32_t cols; // implicit on SIMD/GPU static -void +void hsg_debug_remap(uint32_t const row_from, uint32_t const row_to, uint32_t * const r) @@ -99,9 +99,9 @@ hsg_debug_remap(uint32_t const row_from, } static -void +void hsg_debug_blend(uint32_t const cols_log2, - uint32_t const row_ll, // lower-left + uint32_t const row_ll, // lower-left uint32_t const row_ur, // upper-right uint32_t * m) { diff --git a/src/compute/skc/allocator_host.c b/src/compute/skc/allocator_host.c index dbdcba0f51..0d7cbca9e4 100644 --- a/src/compute/skc/allocator_host.c +++ b/src/compute/skc/allocator_host.c @@ -33,7 +33,7 @@ // void * -skc_runtime_host_perm_alloc(struct skc_runtime * const runtime, +skc_runtime_host_perm_alloc(struct skc_runtime * const runtime, skc_mem_flags_e const flags, size_t const size) { @@ -42,7 +42,7 @@ skc_runtime_host_perm_alloc(struct skc_runtime * const runtime, } void -skc_runtime_host_perm_free(struct skc_runtime * const runtime, +skc_runtime_host_perm_free(struct skc_runtime * const runtime, void * const mem) { SKC_ALIGNED_FREE(mem); @@ -65,11 +65,11 @@ skc_runtime_host_temp_alloc(struct skc_runtime * const runtime, if (subbuf_size != NULL) *subbuf_size = 0; - + return NULL; } - return runtime->allocator.host.temp.extent + + return runtime->allocator.host.temp.extent + skc_suballocator_subbuf_alloc(&runtime->allocator.host.temp.suballocator, runtime->scheduler, size,subbuf_id,subbuf_size); @@ -101,7 +101,7 @@ skc_allocator_host_create(struct skc_runtime * const runtime) SKC_RUNTIME_HOST_CACHELINE_SIZE, runtime->config->suballocator.host.size); - runtime->allocator.host.temp.extent = + runtime->allocator.host.temp.extent = skc_runtime_host_perm_alloc(runtime, SKC_MEM_FLAGS_READ_WRITE, runtime->config->suballocator.host.size); diff --git a/src/compute/skc/context.c b/src/compute/skc/context.c index 8066dc2b9b..59c7956fd5 100644 --- a/src/compute/skc/context.c +++ b/src/compute/skc/context.c @@ -28,7 +28,7 @@ // skc_err -skc_context_create_cl(skc_context_t * context, +skc_context_create_cl(skc_context_t * context, cl_context context_cl, cl_device_id device_id_cl) { diff --git a/src/compute/skc/context.h b/src/compute/skc/context.h index 65da835c9c..1dc026151e 100644 --- a/src/compute/skc/context.h +++ b/src/compute/skc/context.h @@ -49,7 +49,7 @@ struct skc_context skc_err (* path_flush )(struct skc_runtime * const runtime, skc_path_t const * paths, uint32_t count); - + // // // diff --git a/src/compute/skc/extent_ring.c b/src/compute/skc/extent_ring.c index ca489441ab..ecb41e6fcd 100644 --- a/src/compute/skc/extent_ring.c +++ b/src/compute/skc/extent_ring.c @@ -99,17 +99,17 @@ skc_extent_ring_snap_alloc(struct skc_runtime * const runtime, { skc_subbuf_id_t id; - struct skc_extent_ring_snap * snap = + struct skc_extent_ring_snap * snap = skc_runtime_host_temp_alloc(runtime, SKC_MEM_FLAGS_READ_WRITE, sizeof(*snap),&id,NULL); // save the id snap->id = id; - + // back point to parent snap->ring = ring; snap->next = NULL; - + // save the inner boundaries of the ring to the snapshot snap->reads = ring->inner.reads; snap->writes = ring->inner.reads = ring->inner.writes; @@ -123,7 +123,7 @@ skc_extent_ring_snap_alloc(struct skc_runtime * const runtime, ring->head = snap; ring->last = snap; } - else + else { ring->last->next = snap; ring->last = snap; @@ -156,7 +156,7 @@ skc_extent_ring_snap_free(struct skc_runtime * const runtime, do { // increment read counter ring->outer.reads = curr->writes; - + struct skc_extent_ring_snap * const next = curr->next; skc_runtime_host_temp_free(runtime,curr,curr->id); @@ -164,7 +164,7 @@ skc_extent_ring_snap_free(struct skc_runtime * const runtime, curr = next; // this was the last snap... - if (curr == NULL) + if (curr == NULL) { ring->last = NULL; break; diff --git a/src/compute/skc/grid.c b/src/compute/skc/grid.c index fb5a073a16..3c5ebcdf4d 100644 --- a/src/compute/skc/grid.c +++ b/src/compute/skc/grid.c @@ -113,7 +113,7 @@ struct skc_grid struct skc_grid_pfn_name waiting; // optional - if defined, typically used to yank the grid away from host struct skc_grid_pfn_name execute; // optional - starts execution of waiting grid struct skc_grid_pfn_name dispose; // optional - invoked when grid is complete - + struct { skc_uint words[SKC_GRID_SIZE_WORDS]; // 0:inactive, 1:active skc_uint count; @@ -535,7 +535,7 @@ skc_grid_clear_dependency(skc_grid_t const after, skc_uint const before) after->before.count -= 1; - if ((after->before.count == 0) && ((after->state == SKC_GRID_STATE_WAITING) || + if ((after->before.count == 0) && ((after->state == SKC_GRID_STATE_WAITING) || (after->state == SKC_GRID_STATE_FORCED))) { // schedule grid for execution diff --git a/src/compute/skc/handle.h b/src/compute/skc/handle.h index 3751b653c9..34a3d7ab21 100644 --- a/src/compute/skc/handle.h +++ b/src/compute/skc/handle.h @@ -39,7 +39,7 @@ typedef enum skc_typed_handle_type_e SKC_TYPED_HANDLE_TYPE_IS_PATH = 0x40000000, SKC_TYPED_HANDLE_TYPE_IS_RASTER = 0x80000000 } skc_typed_handle_type_e; - + typedef skc_uint skc_typed_handle_t; typedef skc_uint skc_handle_t; diff --git a/src/compute/skc/interop.h b/src/compute/skc/interop.h index 555f3c41c4..775da8623f 100644 --- a/src/compute/skc/interop.h +++ b/src/compute/skc/interop.h @@ -52,7 +52,7 @@ void skc_interop_blit(struct skc_interop * interop); void -skc_interop_get_size(struct skc_interop * interop, +skc_interop_get_size(struct skc_interop * interop, uint32_t * width, uint32_t * height); // diff --git a/src/compute/skc/main.c b/src/compute/skc/main.c index fe37324229..067d24c773 100644 --- a/src/compute/skc/main.c +++ b/src/compute/skc/main.c @@ -194,7 +194,7 @@ main(int argc, char** argv) // rasterize, render and reclaim svg until escape // skc_pipeline_start_at_e pipeline_start_at_base = SKC_PIPELINE_START_AT_DEFINE_PATHS; - skc_pipeline_start_at_e pipeline_start_at_loop = SKC_PIPELINE_START_AT_DEFINE_PATHS; + skc_pipeline_start_at_e pipeline_start_at_loop = SKC_PIPELINE_START_AT_DEFINE_PATHS; skc_path_t * paths; skc_raster_t * rasters; @@ -269,7 +269,7 @@ main(int argc, char** argv) // do we only want to run part of the pipeline? if ((key >= SKC_PIPELINE_START_AT_DEFINE_PATHS) && (key <= SKC_PIPELINE_START_AT_RENDER)) pipeline_start_at_base = key; - + // valid for a loop pipeline_start_at_loop = pipeline_start_at_base; diff --git a/src/compute/skc/path_builder.h b/src/compute/skc/path_builder.h index c8e36dfbdf..a956475f49 100644 --- a/src/compute/skc/path_builder.h +++ b/src/compute/skc/path_builder.h @@ -55,10 +55,10 @@ struct skc_coords_rem_count_cubic struct skc_path_builder { struct skc_context * context; - + struct skc_path_builder_impl * impl; - void (* begin )(struct skc_path_builder_impl * const impl); + void (* begin )(struct skc_path_builder_impl * const impl); void (* end )(struct skc_path_builder_impl * const impl, skc_path_t * const path); void (* new_line )(struct skc_path_builder_impl * const impl); void (* new_quad )(struct skc_path_builder_impl * const impl); diff --git a/src/compute/skc/platforms/cl_12/allocator_device_cl.c b/src/compute/skc/platforms/cl_12/allocator_device_cl.c index 90ae26eb71..8003504706 100644 --- a/src/compute/skc/platforms/cl_12/allocator_device_cl.c +++ b/src/compute/skc/platforms/cl_12/allocator_device_cl.c @@ -57,7 +57,7 @@ skc_runtime_device_temp_alloc(struct skc_runtime * const runtime, if (subbuf_size != NULL) *subbuf_size = 0; - + return NULL; } @@ -83,7 +83,7 @@ skc_runtime_device_temp_alloc(struct skc_runtime * const runtime, void -skc_runtime_device_temp_free(struct skc_runtime * const runtime, +skc_runtime_device_temp_free(struct skc_runtime * const runtime, cl_mem const mem, skc_subbuf_id_t const subbuf_id) { @@ -92,7 +92,7 @@ skc_runtime_device_temp_free(struct skc_runtime * const runtime, skc_suballocator_subbuf_free(&runtime->allocator.device.temp.suballocator,subbuf_id); - cl(ReleaseMemObject(mem)); + cl(ReleaseMemObject(mem)); } // @@ -116,7 +116,7 @@ skc_allocator_device_create(struct skc_runtime * const runtime) cl_mem_flags const flags = CL_MEM_READ_WRITE; #endif - runtime->allocator.device.temp.extent = + runtime->allocator.device.temp.extent = skc_runtime_device_perm_alloc(runtime, flags, runtime->config->suballocator.device.size); diff --git a/src/compute/skc/platforms/cl_12/allocator_device_cl.h b/src/compute/skc/platforms/cl_12/allocator_device_cl.h index 67d4e41398..08c4518a6a 100644 --- a/src/compute/skc/platforms/cl_12/allocator_device_cl.h +++ b/src/compute/skc/platforms/cl_12/allocator_device_cl.h @@ -28,7 +28,7 @@ struct skc_allocator_device { #if 0 struct { - + } perm; #endif diff --git a/src/compute/skc/platforms/cl_12/composition_cl_12.c b/src/compute/skc/platforms/cl_12/composition_cl_12.c index 7853564636..5db86762f3 100644 --- a/src/compute/skc/platforms/cl_12/composition_cl_12.c +++ b/src/compute/skc/platforms/cl_12/composition_cl_12.c @@ -118,7 +118,7 @@ struct skc_composition_place static void -skc_composition_unseal_block(struct skc_composition_impl * const impl, +skc_composition_unseal_block(struct skc_composition_impl * const impl, skc_bool const block); // @@ -140,7 +140,7 @@ skc_composition_pfn_release(struct skc_composition_impl * const impl) skc_composition_unseal_block(impl,true); // block struct skc_runtime * const runtime = impl->runtime; - + // free host composition skc_runtime_host_perm_free(runtime,impl->composition); @@ -149,7 +149,7 @@ skc_composition_pfn_release(struct skc_composition_impl * const impl) // release kernels cl(ReleaseKernel(impl->kernels.place)); - cl(ReleaseKernel(impl->kernels.segment)); + cl(ReleaseKernel(impl->kernels.segment)); // release extents skc_extent_phw1g_tdrNs_free(runtime,&impl->cmds.extent); @@ -158,7 +158,7 @@ skc_composition_pfn_release(struct skc_composition_impl * const impl) skc_extent_pdrw_free (runtime,&impl->keys); skc_extent_pdrw_free (runtime,&impl->offsets); - + // free composition impl skc_runtime_host_perm_free(runtime,impl); } @@ -204,7 +204,7 @@ void skc_composition_place_read_cb(cl_event event, cl_int status, skc_grid_t const grid) { SKC_CL_CB(status); - + struct skc_composition_place * const place = skc_grid_get_data(grid); struct skc_composition_impl * const impl = place->impl; struct skc_runtime * const runtime = impl->runtime; @@ -288,7 +288,7 @@ skc_composition_snap(struct skc_composition_impl * const impl) // set grid data skc_grid_set_data(impl->grids.place,place); - + // acquire command queue place->cq = skc_runtime_acquire_cq_in_order(impl->runtime); @@ -297,7 +297,7 @@ skc_composition_snap(struct skc_composition_impl * const impl) // make a snapshot skc_extent_phw1g_tdrNs_snap_init(impl->runtime,&impl->cmds.ring,&place->cmds); - + // unmap the snapshot (could be a copy) skc_extent_phw1g_tdrNs_snap_alloc(impl->runtime, &impl->cmds.extent, @@ -382,7 +382,7 @@ skc_composition_sort_grid_pfn_execute(skc_grid_t const grid) { struct skc_composition_impl * const impl = skc_grid_get_data(grid); - // we should be sealing + // we should be sealing assert(impl->state == SKC_COMPOSITION_STATE_SEALING); struct skc_place_atomics * const atomics = impl->atomics.hr; @@ -422,7 +422,7 @@ skc_composition_sort_grid_pfn_execute(skc_grid_t const grid) // next stage needs to know number of key segments skc_extent_phr_pdrw_read(&impl->atomics,impl->cq,&complete); - + // register a callback cl(SetEventCallback(complete,CL_COMPLETE,skc_composition_sort_execute_cb,impl)); cl(ReleaseEvent(complete)); @@ -456,7 +456,7 @@ skc_composition_raster_release(struct skc_composition_impl * const impl) static void -skc_composition_unseal_block(struct skc_composition_impl * const impl, +skc_composition_unseal_block(struct skc_composition_impl * const impl, skc_bool const block) { // return if already unsealed @@ -526,7 +526,7 @@ skc_composition_reset(struct skc_composition_impl * const impl) static void -skc_composition_unseal_block_reset(struct skc_composition_impl * const impl, +skc_composition_unseal_block_reset(struct skc_composition_impl * const impl, skc_bool const block, skc_bool const reset) { @@ -590,7 +590,7 @@ skc_composition_pfn_place(struct skc_composition_impl * const impl, SKC_TYPED_HANDLE_TYPE_IS_RASTER, rasters, count); - if (err) + if (err) return err; skc_runtime_handle_device_retain(impl->runtime,rasters,count); @@ -614,7 +614,7 @@ skc_composition_pfn_place(struct skc_composition_impl * const impl, do { skc_uint rem; - // find out how much room is left in then ring's snap + // find out how much room is left in then ring's snap // if the place ring is full -- let it drain SKC_SCHEDULER_WAIT_WHILE(impl->runtime->scheduler,(rem = skc_extent_ring_wip_rem(&impl->cmds.ring)) == 0); diff --git a/src/compute/skc/platforms/cl_12/composition_cl_12.h b/src/compute/skc/platforms/cl_12/composition_cl_12.h index 81a544fdee..a6368d2bad 100644 --- a/src/compute/skc/platforms/cl_12/composition_cl_12.h +++ b/src/compute/skc/platforms/cl_12/composition_cl_12.h @@ -38,15 +38,15 @@ typedef enum skc_composition_state_e { // struct skc_composition_impl -{ +{ struct skc_composition * composition; struct skc_runtime * runtime; - + SKC_ASSERT_STATE_DECLARE(skc_composition_state_e); skc_int lock_count; // wip renders - struct { + struct { skc_grid_t place; skc_grid_t sort; } grids; diff --git a/src/compute/skc/platforms/cl_12/config_cl.h b/src/compute/skc/platforms/cl_12/config_cl.h index ac5cd76710..deef2db791 100644 --- a/src/compute/skc/platforms/cl_12/config_cl.h +++ b/src/compute/skc/platforms/cl_12/config_cl.h @@ -27,10 +27,10 @@ struct skc_config } host; // alignment determined by compiler struct { skc_uint size; - skc_uint subbufs; + skc_uint subbufs; } device; // alignment determined by device } suballocator; - + struct { skc_uint size; } scheduler; @@ -58,7 +58,7 @@ struct skc_config struct { skc_uint size; // a large fraction of block pool size skc_uint width; // determines number of launched reclamation subgroups - skc_uint recs; // how many in-flight width-subgroup reclamation grids + skc_uint recs; // how many in-flight width-subgroup reclamation grids } handle_pool; struct { diff --git a/src/compute/skc/platforms/cl_12/cq_pool_cl.c b/src/compute/skc/platforms/cl_12/cq_pool_cl.c index 2e37937b69..8d1537dc40 100644 --- a/src/compute/skc/platforms/cl_12/cq_pool_cl.c +++ b/src/compute/skc/platforms/cl_12/cq_pool_cl.c @@ -46,7 +46,7 @@ static cl_command_queue skc_runtime_cl_12_create_cq(struct skc_runtime * const runtime, struct skc_cq_pool * const pool) - + { cl_command_queue cq; @@ -59,7 +59,7 @@ skc_runtime_cl_12_create_cq(struct skc_runtime * const runtime, cq = clCreateCommandQueue(runtime->cl.context, runtime->cl.device_id, pool->cq_props, - &cl_err); cl_ok(cl_err); + &cl_err); cl_ok(cl_err); #else if (runtime_cl->version.major < 2) { @@ -71,7 +71,7 @@ skc_runtime_cl_12_create_cq(struct skc_runtime * const runtime, cq = clCreateCommandQueue(runtime_cl->context, runtime_cl->device_id, (cl_command_queue_properties)type, - &cl_err); cl_ok(cl_err); + &cl_err); cl_ok(cl_err); } else { @@ -135,7 +135,7 @@ skc_cq_pool_dispose(struct skc_runtime * const runtime, // // -static +static void skc_cq_pool_write(struct skc_cq_pool * const pool, cl_command_queue cq) @@ -174,7 +174,7 @@ skc_cq_pool_expand(struct skc_runtime * const runtime, // // -static +static cl_command_queue skc_cq_pool_read(struct skc_runtime * const runtime, struct skc_cq_pool * const pool) @@ -199,7 +199,7 @@ skc_runtime_acquire_cq_in_order(struct skc_runtime * const runtime) } void -skc_runtime_release_cq_in_order(struct skc_runtime * const runtime, +skc_runtime_release_cq_in_order(struct skc_runtime * const runtime, cl_command_queue cq) { skc_cq_pool_write(&runtime->cq_pool,cq); diff --git a/src/compute/skc/platforms/cl_12/cq_pool_cl.h b/src/compute/skc/platforms/cl_12/cq_pool_cl.h index c614600e19..b0953c053a 100644 --- a/src/compute/skc/platforms/cl_12/cq_pool_cl.h +++ b/src/compute/skc/platforms/cl_12/cq_pool_cl.h @@ -14,7 +14,7 @@ // Why we need to wrap command queue creation: // // - command queue creation is expensive -// +// // - the CL 1.2 function is deprecated in 2.0 // diff --git a/src/compute/skc/platforms/cl_12/device_cl_12.h b/src/compute/skc/platforms/cl_12/device_cl_12.h index ef574958b3..11dc208224 100644 --- a/src/compute/skc/platforms/cl_12/device_cl_12.h +++ b/src/compute/skc/platforms/cl_12/device_cl_12.h @@ -33,7 +33,7 @@ typedef enum skc_device_kernel_id { SKC_DEVICE_KERNEL_ID_FILLS_EXPAND, - SKC_DEVICE_KERNEL_ID_RASTERIZE_ALL, + SKC_DEVICE_KERNEL_ID_RASTERIZE_ALL, SKC_DEVICE_KERNEL_ID_RASTERIZE_LINES, SKC_DEVICE_KERNEL_ID_RASTERIZE_QUADS, SKC_DEVICE_KERNEL_ID_RASTERIZE_CUBICS, @@ -69,12 +69,12 @@ void skc_device_dispose(struct skc_runtime * const runtime); -// +// // multi-threading/context/device requires multiple kernel instances // cl_kernel -skc_device_acquire_kernel(struct skc_device * const device, +skc_device_acquire_kernel(struct skc_device * const device, skc_device_kernel_id const type); void @@ -85,13 +85,13 @@ skc_device_release_kernel(struct skc_device * const device, // grid shape can vary greatly by target platform // void -skc_device_enqueue_kernel(struct skc_device * const device, +skc_device_enqueue_kernel(struct skc_device * const device, skc_device_kernel_id const type, cl_command_queue cq, cl_kernel kernel, - size_t const work_size, + size_t const work_size, cl_uint num_events_in_wait_list, - cl_event const * const event_wait_list, + cl_event const * const event_wait_list, cl_event * const event); // diff --git a/src/compute/skc/platforms/cl_12/export_cl_12.h b/src/compute/skc/platforms/cl_12/export_cl_12.h index e577282791..23ff2343e6 100644 --- a/src/compute/skc/platforms/cl_12/export_cl_12.h +++ b/src/compute/skc/platforms/cl_12/export_cl_12.h @@ -15,7 +15,7 @@ #include "skc.h" // -// +// // skc_err @@ -23,7 +23,7 @@ skc_path_builder_cl_12_create(struct skc_context * const context, struct skc_path_builder * * const path_builder); // -// +// // skc_err diff --git a/src/compute/skc/platforms/cl_12/extent_cl_12.c b/src/compute/skc/platforms/cl_12/extent_cl_12.c index 73676d8063..e145d979c2 100644 --- a/src/compute/skc/platforms/cl_12/extent_cl_12.c +++ b/src/compute/skc/platforms/cl_12/extent_cl_12.c @@ -127,7 +127,7 @@ skc_extent_phr_pdrw_read(struct skc_extent_phr_pdrw * const extent, if (extent->size == 0) return; - cl(EnqueueReadBuffer(cq, + cl(EnqueueReadBuffer(cq, extent->drw, CL_FALSE, 0, @@ -191,7 +191,7 @@ skc_extent_thr_tdrw_read(struct skc_extent_thr_tdrw * const extent, if (extent->size == 0) return; - cl(EnqueueReadBuffer(cq, + cl(EnqueueReadBuffer(cq, extent->drw, CL_FALSE, 0, diff --git a/src/compute/skc/platforms/cl_12/extent_cl_12.h b/src/compute/skc/platforms/cl_12/extent_cl_12.h index 47ba951bb3..1801c9c57f 100644 --- a/src/compute/skc/platforms/cl_12/extent_cl_12.h +++ b/src/compute/skc/platforms/cl_12/extent_cl_12.h @@ -40,7 +40,7 @@ // DURABLE R/W HOST EXTENT -- STANDARD CACHED MEMORY // -struct skc_extent_phrw +struct skc_extent_phrw { void * hrw; }; @@ -58,7 +58,7 @@ skc_extent_phrw_free(struct skc_runtime * const runtime, // DURABLE R/W DEVICE EXTENT -- ALLOCATED FROM DEVICE HEAP // -struct skc_extent_pdrw +struct skc_extent_pdrw { cl_mem drw; }; @@ -76,7 +76,7 @@ skc_extent_pdrw_free(struct skc_runtime * const runtime, // EPHEMERAL DEVICE R/W EXTENT -- ALLOCATED QUICKLY FROM A MANAGED RING // -struct skc_extent_tdrw +struct skc_extent_tdrw { size_t size; cl_mem drw; diff --git a/src/compute/skc/platforms/cl_12/handle_pool_cl_12.h b/src/compute/skc/platforms/cl_12/handle_pool_cl_12.h index 4fefae3552..455d9384a1 100644 --- a/src/compute/skc/platforms/cl_12/handle_pool_cl_12.h +++ b/src/compute/skc/platforms/cl_12/handle_pool_cl_12.h @@ -88,7 +88,7 @@ union skc_handle_reclaim_rec { // ELEMENT 0 struct skc_runtime * runtime; - + // ELEMENT 1 struct { skc_uint rem; // # of available records diff --git a/src/compute/skc/platforms/cl_12/interop/interop_glfw.c b/src/compute/skc/platforms/cl_12/interop/interop_glfw.c index a5c0bfcb57..8f94100552 100644 --- a/src/compute/skc/platforms/cl_12/interop/interop_glfw.c +++ b/src/compute/skc/platforms/cl_12/interop/interop_glfw.c @@ -132,7 +132,7 @@ void skc_interop_key_callback(GLFWwindow * window, int key, int scancode, int action, int mods) { struct skc_interop * interop = glfwGetWindowUserPointer(window); - + if (action == GLFW_RELEASE) return; @@ -182,7 +182,7 @@ void skc_interop_window_size_callback(GLFWwindow * window, int width, int height) { struct skc_interop * interop = glfwGetWindowUserPointer(window); - + interop->width = width; interop->height = height; interop->is_resized = true; @@ -217,7 +217,7 @@ skc_interop_scroll_callback(GLFWwindow * window, double xoffset, double yoffset) if (!ctrl) return; - struct skc_interop * interop = glfwGetWindowUserPointer(window); + struct skc_interop * interop = glfwGetWindowUserPointer(window); skc_interop_scale(interop,yoffset); @@ -242,7 +242,7 @@ static void skc_interop_cursor_position_callback(GLFWwindow * window, double x, double y) { - + int const state = glfwGetMouseButton(window,GLFW_MOUSE_BUTTON_LEFT); static bool is_mouse_dragging = false; @@ -253,8 +253,8 @@ skc_interop_cursor_position_callback(GLFWwindow * window, double x, double y) if (state == GLFW_PRESS) { - struct skc_interop * interop = glfwGetWindowUserPointer(window); - + struct skc_interop * interop = glfwGetWindowUserPointer(window); + if (is_mouse_dragging) { const bool ctrl = @@ -294,7 +294,7 @@ skc_interop_cursor_position_callback(GLFWwindow * window, double x, double y) mx - x_prev, my - y_prev); } - + interop->is_transform = true; } else @@ -417,7 +417,7 @@ skc_interop_create() // save back pointer glfwSetWindowUserPointer(interop->window,interop); - + glfwMakeContextCurrent(interop->window); // set up GLAD @@ -519,7 +519,7 @@ skc_interop_transform(struct skc_interop * interop, // spinner... if (interop->is_spinning) interop->rotate_theta = fmodf(interop->rotate_theta + SKC_ROTATE_STEP,(float)(M_PI*2.0)); - + // always rotate and scale around surface center point skc_transform_stack_push_rotate_scale_xy(ts, interop->rotate_theta, @@ -673,7 +673,7 @@ skc_interop_poll(struct skc_interop * interop, int * key) bool const is_transform = interop->is_transform || interop->is_spinning; interop->is_transform = false; - + return is_transform; } @@ -736,7 +736,7 @@ skc_interop_should_exit(struct skc_interop * interop) // void -skc_interop_get_size(struct skc_interop * interop, +skc_interop_get_size(struct skc_interop * interop, uint32_t * width, uint32_t * height) { diff --git a/src/compute/skc/platforms/cl_12/kernels/devices/gen9/device_cl_12.c b/src/compute/skc/platforms/cl_12/kernels/devices/gen9/device_cl_12.c index d7f10e38bf..0be97235f3 100644 --- a/src/compute/skc/platforms/cl_12/kernels/devices/gen9/device_cl_12.c +++ b/src/compute/skc/platforms/cl_12/kernels/devices/gen9/device_cl_12.c @@ -90,7 +90,7 @@ // FIXME -- move these to log2 values where appropriate // -static +static struct skc_config const config = { .suballocator = { @@ -102,7 +102,7 @@ struct skc_config const config = .size = 128 * 1024 * 1024, .subbufs = 1024 // must be <= (1 << (8 * sizeof(skc_subbuf_id_t))) } - }, + }, .scheduler = { .size = 4096 // 128 // FIXME -- this is just for testing -- way too big -- schedulees should bring their own state @@ -382,7 +382,7 @@ skc_device_shaper_paths_copy(size_t const work_size, work_local [0] = SKC_PATHS_COPY_SUBGROUP_SIZE; return work_local; -#else +#else return NULL; // let runtime figure out local work size #endif } @@ -659,7 +659,7 @@ struct skc_device { // // FIXME -- an OpenCL 2.1+ device would clone these kernels in a - // multithreaded system. + // multithreaded system. // // Not having the ability to clone kernels (yet set their sticky // args) was an oversight in previous versions of OpenCL. diff --git a/src/compute/skc/platforms/cl_12/path_builder_cl_12.c b/src/compute/skc/platforms/cl_12/path_builder_cl_12.c index e915dffada..e0a9d029ee 100644 --- a/src/compute/skc/platforms/cl_12/path_builder_cl_12.c +++ b/src/compute/skc/platforms/cl_12/path_builder_cl_12.c @@ -256,7 +256,7 @@ struct skc_path_builder_impl static union skc_ringdex_expand -skc_ringdex_expand(struct skc_path_builder_impl * const impl, +skc_ringdex_expand(struct skc_path_builder_impl * const impl, skc_ringdex_t const ringdex) { return (union skc_ringdex_expand){ @@ -322,7 +322,7 @@ skc_ringdex_prev_is_equal(struct skc_path_builder_impl * const impl) static skc_uint -skc_ringdex_dont_map_last(struct skc_path_builder_impl * const impl, +skc_ringdex_dont_map_last(struct skc_path_builder_impl * const impl, skc_uint const to_block) { // no blocks acquired OR this is last block in subbuf @@ -509,7 +509,7 @@ skc_path_builder_impl_unmap(struct skc_path_builder_impl * const impl, #if 0 fprintf(stderr,"unmap: [%2u,%2u)\n",from,to); #endif - + while (from != to) // 'to' might be out of range { // bring 'from' back in range @@ -647,9 +647,9 @@ void skc_path_builder_paths_copy_cb(cl_event event, cl_int status, skc_grid_t grid) { SKC_CL_CB(status); - + struct skc_release_record * const release = skc_grid_get_data(grid); - + SKC_SCHEDULER_SCHEDULE(release->impl->runtime->scheduler,skc_path_builder_complete,grid); } @@ -802,7 +802,7 @@ skc_path_builder_grid_pfn_waiting(skc_grid_t const grid) // if there was no wip then round up to the next subbuf // skc_ringdex_wip_to_subbuf_inc(impl); - + // // update prev/curr with with incremented wip // diff --git a/src/compute/skc/platforms/cl_12/raster_builder_cl_12.c b/src/compute/skc/platforms/cl_12/raster_builder_cl_12.c index 33992cbdfb..d84b92bfd7 100644 --- a/src/compute/skc/platforms/cl_12/raster_builder_cl_12.c +++ b/src/compute/skc/platforms/cl_12/raster_builder_cl_12.c @@ -329,7 +329,7 @@ skc_raster_builder_pfn_release(struct skc_raster_builder_impl * const impl) // // -static +static void skc_raster_builder_rasters_release(struct skc_runtime * const runtime, skc_raster_t const * const rasters, @@ -344,12 +344,12 @@ skc_raster_builder_rasters_release(struct skc_runtime * const runtime, skc_grid_deps_unmap(runtime->deps,rasters_from,count_from); skc_runtime_raster_device_release(runtime,rasters_from,count_from); - } + } else // from > to implies wrap { skc_raster_t const * rasters_lo = rasters + from; skc_uint count_lo = size - from; - + skc_grid_deps_unmap(runtime->deps,rasters_lo,count_lo); skc_runtime_raster_device_release(runtime,rasters_lo,count_lo); @@ -358,7 +358,7 @@ skc_raster_builder_rasters_release(struct skc_runtime * const runtime, } } -static +static void skc_raster_builder_paths_release(struct skc_runtime * const runtime, struct skc_extent_phrwg_thr1s_snap * const snap) @@ -401,7 +401,7 @@ skc_raster_builder_cohort_grid_pfn_dispose(skc_grid_t const grid) // snap alloc the paths -- this host snap simply sets up pointers skc_extent_phrwg_thr1s_snap_alloc(runtime,&impl->path_ids,&cohort->path_ids); - + // unmap and release raster ids skc_raster_builder_paths_release(runtime,&cohort->path_ids); @@ -418,7 +418,7 @@ skc_raster_builder_cohort_grid_pfn_dispose(skc_grid_t const grid) // unmap and release raster ids skc_raster_builder_rasters_release(runtime,impl->raster_ids.hrw,size,from,to); - // release cohort's remaining allocated resources + // release cohort's remaining allocated resources skc_extent_phrwg_tdrNs_snap_free(runtime,&cohort->raster_ids); skc_runtime_release_cq_in_order(runtime,cohort->cq); skc_extent_thr_tdrw_free(runtime,&cohort->atomics); @@ -458,7 +458,7 @@ skc_raster_cohort_prefix_release(skc_grid_t const grid) // FIXME -- note that pfn_dispose can be accomplished here // release the grid - skc_grid_complete(grid); + skc_grid_complete(grid); } static @@ -466,7 +466,7 @@ void skc_raster_cohort_prefix_cb(cl_event event, cl_int status, skc_grid_t const grid) { SKC_CL_CB(status); - + struct skc_raster_cohort * const cohort = skc_grid_get_data(grid); struct skc_scheduler * const scheduler = cohort->impl->runtime->scheduler; @@ -599,7 +599,7 @@ skc_raster_cohort_sort_prefix(skc_grid_t const grid) // DELETE ALL THIS WHEN READY // -#if 0 +#if 0 // // // @@ -625,7 +625,7 @@ skc_raster_cohort_sort_prefix(skc_grid_t const grid) // // some counters // - union skc_raster_cohort_meta_in meta_in = { + union skc_raster_cohort_meta_in meta_in = { .blocks = 0, .offset = 0, .pk = 0, @@ -665,10 +665,10 @@ skc_raster_cohort_sort_prefix(skc_grid_t const grid) // store back to buffer metas->inout[curr.cohort].in = meta_in; - + // update meta_in meta_in.blocks = 0; - meta_in.offset = ii; + meta_in.offset = ii; meta_in.pk = 0; meta_in.rk = 0; } @@ -699,7 +699,7 @@ skc_raster_cohort_sort_prefix(skc_grid_t const grid) // store back to buffer metas->inout[curr.cohort].in = meta_in; - + // unmap skc_extent_thrw_tdrw_unmap(&cohort->keys, keys, @@ -795,7 +795,7 @@ void skc_raster_cohort_rasterize_cb(cl_event event, cl_int status, skc_grid_t const grid) { SKC_CL_CB(status); - + struct skc_raster_cohort * const cohort = skc_grid_get_data(grid); // as quickly as possible, enqueue next stage in pipeline to context command scheduler @@ -1079,7 +1079,7 @@ skc_raster_builder_cohort_grid_pfn_waiting(skc_grid_t const grid) // allocate the ephemeral/temp cohort skc_subbuf_id_t id; - struct skc_raster_cohort * const cohort = + struct skc_raster_cohort * const cohort = skc_runtime_host_temp_alloc(runtime, SKC_MEM_FLAGS_READ_WRITE, sizeof(*cohort), @@ -1145,7 +1145,7 @@ skc_raster_builder_cohort_create(struct skc_raster_builder_impl * const impl) static skc_err -skc_raster_builder_pfn_add(struct skc_raster_builder_impl * const impl, +skc_raster_builder_pfn_add(struct skc_raster_builder_impl * const impl, skc_path_t const * paths, skc_uint count) { @@ -1329,7 +1329,7 @@ skc_raster_builder_cl_12_create(struct skc_context * const context, // acquire kernels // impl->kernels.fills_expand = skc_device_acquire_kernel(runtime->device, SKC_DEVICE_KERNEL_ID_FILLS_EXPAND); - impl->kernels.rasterize_all = skc_device_acquire_kernel(runtime->device, SKC_DEVICE_KERNEL_ID_RASTERIZE_ALL); + impl->kernels.rasterize_all = skc_device_acquire_kernel(runtime->device, SKC_DEVICE_KERNEL_ID_RASTERIZE_ALL); #if 0 impl->kernels.rasterize_lines = skc_device_acquire_kernel(runtime->device, SKC_DEVICE_KERNEL_ID_RASTERIZE_LINES); diff --git a/src/compute/skc/platforms/cl_12/runtime_cl_12.c b/src/compute/skc/platforms/cl_12/runtime_cl_12.c index 81e1e8569e..81f3aba02f 100644 --- a/src/compute/skc/platforms/cl_12/runtime_cl_12.c +++ b/src/compute/skc/platforms/cl_12/runtime_cl_12.c @@ -31,7 +31,7 @@ // // -static +static void skc_block_pool_create(struct skc_runtime * const runtime, cl_command_queue cq) { @@ -41,7 +41,7 @@ skc_block_pool_create(struct skc_runtime * const runtime, cl_command_queue cq) // create block extent skc_extent_pdrw_alloc(runtime, &runtime->block_pool.blocks, - runtime->block_pool.size->pool_size * + runtime->block_pool.size->pool_size * runtime->config->block.bytes); // allocate block pool ids @@ -84,7 +84,7 @@ skc_block_pool_create(struct skc_runtime * const runtime, cl_command_queue cq) cl(ReleaseKernel(k1)); } -static +static void skc_block_pool_dispose(struct skc_runtime * const runtime) { @@ -105,7 +105,7 @@ skc_runtime_yield(struct skc_runtime * const runtime) } static -void +void skc_runtime_wait(struct skc_runtime * const runtime) { skc_scheduler_wait(runtime->scheduler); @@ -122,7 +122,7 @@ skc_runtime_cl_12_create(struct skc_context * const context, { // allocate the runtime struct skc_runtime * const runtime = malloc(sizeof(*runtime)); - + // save off CL objects runtime->cl.context = context_cl; runtime->cl.device_id = device_id_cl; @@ -135,7 +135,7 @@ skc_runtime_cl_12_create(struct skc_context * const context, sizeof(align_bits), &align_bits, NULL)); - + runtime->cl.align_bytes = align_bits / 8; // create device @@ -183,7 +183,7 @@ skc_runtime_cl_12_create(struct skc_context * const context, context->yield = skc_runtime_yield; context->wait = skc_runtime_wait; - + context->path_builder = skc_path_builder_cl_12_create; context->path_retain = skc_runtime_path_host_retain; context->path_release = skc_runtime_path_host_release; @@ -196,7 +196,7 @@ skc_runtime_cl_12_create(struct skc_context * const context, context->composition = skc_composition_cl_12_create; context->styling = skc_styling_cl_12_create; - + context->surface = skc_surface_cl_12_create; // block on pool creation @@ -234,7 +234,7 @@ skc_runtime_cl_12_dispose(struct skc_context * const context) skc_block_pool_dispose(context->runtime); // skc_handle_pool_dispose(context->runtime); - + return SKC_ERR_SUCCESS; } @@ -246,7 +246,7 @@ void skc_runtime_cl_12_debug(struct skc_context * const context) { struct skc_runtime * const runtime = context->runtime; - + // acquire out-of-order cq cl_command_queue cq = skc_runtime_acquire_cq_in_order(runtime); diff --git a/src/compute/skc/platforms/cl_12/styling_cl_12.c b/src/compute/skc/platforms/cl_12/styling_cl_12.c index 8d8d90525d..9621a1f4b5 100644 --- a/src/compute/skc/platforms/cl_12/styling_cl_12.c +++ b/src/compute/skc/platforms/cl_12/styling_cl_12.c @@ -32,7 +32,7 @@ void skc_styling_unmap_complete(skc_grid_t const grid) { struct skc_styling_impl * const impl = skc_grid_get_data(grid); - + impl->state = SKC_STYLING_STATE_SEALED; skc_grid_complete(grid); @@ -65,7 +65,7 @@ skc_styling_grid_pfn_execute(skc_grid_t const grid) skc_extent_phwN_pdrN_unmap(&impl->layers,styling->layers.extent,impl->cq,NULL); skc_extent_phwN_pdrN_unmap(&impl->groups,styling->groups.extent,impl->cq,NULL); - skc_extent_phwN_pdrN_unmap(&impl->extras,styling->extras.extent,impl->cq,&complete); + skc_extent_phwN_pdrN_unmap(&impl->extras,styling->extras.extent,impl->cq,&complete); // set the event cl(SetEventCallback(complete,CL_COMPLETE,skc_styling_unmap_cb,grid)); @@ -97,7 +97,7 @@ skc_styling_pfn_seal(struct skc_styling_impl * const impl) { SKC_SCHEDULER_WAIT_WHILE(scheduler,impl->state != SKC_STYLING_STATE_UNSEALED); } - + // // we're unsealed so we need to seal and start the grid // @@ -121,7 +121,7 @@ void skc_styling_unseal_complete(struct skc_styling_impl * const impl) { struct skc_runtime * const runtime = impl->runtime; - + // we're now unsealed impl->state = SKC_STYLING_STATE_UNSEALED; } @@ -131,7 +131,7 @@ void skc_styling_unseal_cb(cl_event event, cl_int status, struct skc_styling_impl * const impl) { SKC_CL_CB(status); - + // as quickly as possible, enqueue next stage in pipeline to context command scheduler SKC_SCHEDULER_SCHEDULE(impl->runtime->scheduler,skc_styling_unseal_complete,impl); } @@ -166,10 +166,10 @@ skc_styling_pfn_unseal(struct skc_styling_impl * const impl, skc_bool const bloc // if (impl->state == SKC_STYLING_STATE_SEALING) { - // wait if sealing + // wait if sealing SKC_SCHEDULER_WAIT_WHILE(scheduler,impl->state != SKC_STYLING_STATE_SEALED); } - + // wait for rendering locks to be released SKC_SCHEDULER_WAIT_WHILE(scheduler,impl->lock_count > 0); @@ -183,7 +183,7 @@ skc_styling_pfn_unseal(struct skc_styling_impl * const impl, skc_bool const bloc cl_event complete; struct skc_styling * const styling = impl->styling; - + styling->layers.extent = skc_extent_phwN_pdrN_map(&impl->layers,impl->cq,NULL); styling->groups.extent = skc_extent_phwN_pdrN_map(&impl->groups,impl->cq,NULL); styling->extras.extent = skc_extent_phwN_pdrN_map(&impl->extras,impl->cq,&complete); @@ -239,7 +239,7 @@ skc_styling_pfn_release(struct skc_styling_impl * const impl) // release the cq skc_runtime_release_cq_in_order(runtime,impl->cq); - + // free extents skc_extent_phwN_pdrN_free(runtime,&impl->layers); skc_extent_phwN_pdrN_free(runtime,&impl->groups); @@ -301,7 +301,7 @@ skc_styling_cl_12_create(struct skc_context * const context, impl->lock_count = 0; impl->cq = skc_runtime_acquire_cq_in_order(runtime); - + // // The styling object is unique in that the API lets the user // specify resource limits diff --git a/src/compute/skc/platforms/cl_12/styling_cl_12.h b/src/compute/skc/platforms/cl_12/styling_cl_12.h index a319568ee5..e50361f555 100644 --- a/src/compute/skc/platforms/cl_12/styling_cl_12.h +++ b/src/compute/skc/platforms/cl_12/styling_cl_12.h @@ -40,7 +40,7 @@ struct skc_styling_impl { struct skc_styling * styling; struct skc_runtime * runtime; - + SKC_ASSERT_STATE_DECLARE(skc_styling_state_e); skc_int lock_count; // # of wip renders diff --git a/src/compute/skc/platforms/cl_12/surface_cl_12.c b/src/compute/skc/platforms/cl_12/surface_cl_12.c index 24a0c45f3c..c4e205a04b 100644 --- a/src/compute/skc/platforms/cl_12/surface_cl_12.c +++ b/src/compute/skc/platforms/cl_12/surface_cl_12.c @@ -80,7 +80,7 @@ static void skc_surface_debug_clear(struct skc_surface_impl * const impl, skc_framebuffer_t fb, - float const rgba[4], + float const rgba[4], uint32_t const rect[4]) { size_t const origin[3] = { rect[0], rect[1], 0 }; @@ -150,11 +150,11 @@ skc_surface_pfn_release(struct skc_surface_impl * const impl) // // otherwise, release all resources // - + // drain the command queue cl(Finish(impl->cq)); - struct skc_runtime * const runtime = impl->runtime; + struct skc_runtime * const runtime = impl->runtime; // release the kernel cl(ReleaseKernel(impl->kernels.render)); @@ -167,7 +167,7 @@ skc_surface_pfn_release(struct skc_surface_impl * const impl) // release fb // skc_extent_phrN_pdwN_free(runtime,&impl->fb); - + // free surface impl skc_runtime_host_perm_free(runtime,impl); } @@ -200,7 +200,7 @@ skc_surface_render_complete(struct skc_surface_render * const render) // grid is now complete skc_grid_complete(render->grid); - struct skc_surface_impl * const impl = render->impl; + struct skc_surface_impl * const impl = render->impl; struct skc_runtime * const runtime = impl->runtime; // release the surface @@ -341,7 +341,7 @@ skc_surface_pfn_render(struct skc_surface_impl * const impl, // them here. // - // retain and lock the styling and composition + // retain and lock the styling and composition skc_styling_retain_and_lock(styling); skc_composition_retain_and_lock(composition); diff --git a/src/compute/skc/raster_builder.c b/src/compute/skc/raster_builder.c index a0f1fcfdb0..b86ff93d08 100644 --- a/src/compute/skc/raster_builder.c +++ b/src/compute/skc/raster_builder.c @@ -155,7 +155,7 @@ skc_raster_builder_transforms_append(struct skc_raster_builder * const raster_bu static skc_bool skc_raster_builder_clips_append(struct skc_raster_builder * const raster_builder, - union skc_cmd_fill * const cmd, + union skc_cmd_fill * const cmd, skc_raster_clip_weakref_t * const raster_clip_weakref, skc_float const * const raster_clip) { @@ -183,7 +183,7 @@ skc_raster_builder_cmds_append(struct skc_raster_builder * const raster_builder, cmd->cohort = skc_extent_ring_wip_count(&raster_builder->raster_ids.ring); skc_uint const base = skc_extent_ring_wip_index_inc(&raster_builder->fill_cmds.ring); - + raster_builder->fill_cmds.extent[base] = *cmd; #if 0 @@ -292,7 +292,7 @@ skc_raster_add_filled(skc_raster_builder_t raster_builder, // mask off the guard bits path = SKC_TYPED_HANDLE_TO_HANDLE(path); - // + // // build the command... // union skc_cmd_fill cmd; diff --git a/src/compute/skc/scheduler.cpp b/src/compute/skc/scheduler.cpp index 1099deb1ca..8171f64002 100644 --- a/src/compute/skc/scheduler.cpp +++ b/src/compute/skc/scheduler.cpp @@ -63,8 +63,8 @@ struct skc_scheduler_command void * data; skc_scheduler_command_pfn pfn; skc_scheduler_command_state state; - char const * name; -}; + char const * name; +}; #if 0 struct skc_scheduler_command diff --git a/src/compute/skc/scheduler.h b/src/compute/skc/scheduler.h index 334cb0f69c..78da50be2a 100644 --- a/src/compute/skc/scheduler.h +++ b/src/compute/skc/scheduler.h @@ -18,12 +18,12 @@ #include "types.h" // -// +// // #define SKC_SCHEDULER_SCHEDULE(s,c,d) skc_scheduler_schedule(s,c,d,#c) -#ifndef NDEBUG +#ifndef NDEBUG #include <stdio.h> @@ -70,7 +70,7 @@ skc_scheduler_create(struct skc_runtime * const runtime, skc_uint const size); void skc_scheduler_dispose(struct skc_runtime * const runtime, struct skc_scheduler * const scheduler); - + // // diff --git a/src/compute/skc/styling.h b/src/compute/skc/styling.h index 7f20c1e489..310a739a07 100644 --- a/src/compute/skc/styling.h +++ b/src/compute/skc/styling.h @@ -27,7 +27,7 @@ struct skc_styling void (* seal )(struct skc_styling_impl * const impl); void (* unseal )(struct skc_styling_impl * const impl, skc_bool const block); void (* release)(struct skc_styling_impl * const impl); - + skc_int ref_count; struct { diff --git a/src/compute/skc/styling_types.h b/src/compute/skc/styling_types.h index 07d9a10ca3..ee0e7aa7e3 100644 --- a/src/compute/skc/styling_types.h +++ b/src/compute/skc/styling_types.h @@ -127,7 +127,7 @@ SKC_STATIC_ASSERT(sizeof(struct skc_group_node) == sizeof(skc_uint2) * 3); // 6 union skc_gradient_slope { skc_float slope; - skc_half color_pair[2]; + skc_half color_pair[2]; }; SKC_STATIC_ASSERT(sizeof(union skc_gradient_slope) == sizeof(skc_float)); @@ -174,7 +174,7 @@ SKC_STATIC_ASSERT(sizeof(union skc_gradient_vector) == sizeof(skc_float4)); // +------------+---------------------------------+--------+-------------------------+ // // COLOR LAYOUT: -// +// // R[0]R[1], R[1]R[2], ... R[N-1]R[N-1] // G[0]G[1], G[1]G[2], ... G[N-1]G[N-1] // B[0]B[1], B[1]B[2], ... B[N-1]B[N-1] @@ -203,7 +203,7 @@ SKC_STATIC_ASSERT(sizeof(union skc_gradient_vector) == sizeof(skc_float4)); // +------------+--------------------------------+--------+ // // COLOR LAYOUT: -// +// // R[0]R[1], R[1]R[2], ... R[N-1]R[N-1] <-------------------------- FIXME -- USE HERB'S SINGLE FMA REPRESENTATION // G[0]G[1], G[1]G[2], ... G[N-1]G[N-1] <-------------------------- FIXME -- USE HERB'S SINGLE FMA REPRESENTATION // B[0]B[1], B[1]B[2], ... B[N-1]B[N-1] <-------------------------- FIXME -- USE HERB'S SINGLE FMA REPRESENTATION @@ -222,7 +222,7 @@ SKC_STATIC_ASSERT(sizeof(union skc_gradient_vector) == sizeof(skc_float4)); // This will require that the colors are laid out in sampler-friendly // order: // -// RGBA[0]RGBA[1], RGBA[1]RGBA[2], ..., RGBA[N-1]RGBA[N-1] +// RGBA[0]RGBA[1], RGBA[1]RGBA[2], ..., RGBA[N-1]RGBA[N-1] // // @@ -247,4 +247,4 @@ SKC_STATIC_ASSERT(sizeof(union skc_gradient_vector) == sizeof(skc_float4)); // // - + diff --git a/src/compute/skc/suballocator.c b/src/compute/skc/suballocator.c index 382e818c27..c6b9ec1681 100644 --- a/src/compute/skc/suballocator.c +++ b/src/compute/skc/suballocator.c @@ -178,7 +178,7 @@ skc_suballocator_subbuf_alloc(struct skc_suballocator * const suballocator, } assert(suballocator->rem.avail > 0); - + // return origin return avail->origin; } @@ -189,7 +189,7 @@ skc_suballocator_subbuf_alloc(struct skc_suballocator * const suballocator, skc_uint spare_idx = suballocator->count - spare_rem; skc_subbuf_id_t const spare_id = ids[spare_idx]; struct skc_subbuf * const spare = suballocator->subbufs + spare_id; - + assert(spare->inuse == 0); // simple -- we're popping the top-of-stack of spares @@ -199,7 +199,7 @@ skc_suballocator_subbuf_alloc(struct skc_suballocator * const suballocator, *subbuf_id = spare_id; SKC_SUBALLOCATOR_DEBUG_ALLOC(suballocator,spare_id,size_ru); - + // get prev struct skc_subbuf * const prev = avail->prev; diff --git a/src/compute/skc/suballocator.h b/src/compute/skc/suballocator.h index 7c3a6e863c..df938e06f0 100644 --- a/src/compute/skc/suballocator.h +++ b/src/compute/skc/suballocator.h @@ -39,7 +39,7 @@ typedef skc_ushort skc_subbuf_id_t; typedef skc_uint skc_subbuf_size_t; // <4GB -// typedef size_t skc_subbuf_size_t; // >4GB +// typedef size_t skc_subbuf_size_t; // >4GB // // @@ -49,7 +49,7 @@ struct skc_subbuf { struct skc_subbuf * prev; struct skc_subbuf * next; - + skc_subbuf_size_t size; skc_subbuf_size_t origin; diff --git a/src/compute/skc/tile.h b/src/compute/skc/tile.h index d6d3c7f2f1..d07c4932f3 100644 --- a/src/compute/skc/tile.h +++ b/src/compute/skc/tile.h @@ -414,7 +414,7 @@ #define SKC_TTCK_LO_BITS_PREFIX 1 #define SKC_TTCK_LO_OFFSET_PREFIX SKC_TTCK_LO_BITS_ID #define SKC_TTCK_LO_MASK_PREFIX SKC_BITS_TO_MASK_AT(SKC_TTCK_LO_BITS_PREFIX,SKC_TTCK_LO_OFFSET_PREFIX) - + #define SKC_TTCK_LO_BITS_ID_PREFIX (SKC_TTCK_LO_BITS_ID + SKC_TTCK_LO_BITS_PREFIX) #define SKC_TTCK_LO_MASK_ID_PREFIX SKC_BITS_TO_MASK(SKC_TTCK_LO_BITS_ID_PREFIX) diff --git a/src/compute/skc/util.c b/src/compute/skc/util.c index fe368880ce..bf8a069016 100644 --- a/src/compute/skc/util.c +++ b/src/compute/skc/util.c @@ -12,7 +12,7 @@ // // -#include <intrin.h> +#include <intrin.h> // // diff --git a/src/compute/skc/weakref.c b/src/compute/skc/weakref.c index e8a1d55f08..1867adea84 100644 --- a/src/compute/skc/weakref.c +++ b/src/compute/skc/weakref.c @@ -56,27 +56,27 @@ SKC_STATIC_ASSERT(sizeof(union skc_weakref) == sizeof(skc_ulong)); // // -void +void skc_weakref_epoch_init(skc_epoch_t * const epoch) { *epoch = SKC_WEAKREF_EPOCH_INIT; } -void +void skc_weakref_epoch_inc(skc_epoch_t * const epoch) { *epoch += SKC_WEAKREF_EPOCH_ONE; } -void -skc_weakref_init(skc_weakref_t * const weakref, +void +skc_weakref_init(skc_weakref_t * const weakref, skc_epoch_t * const epoch, skc_uint const index) { *weakref = *epoch | (index & SKC_WEAKREF_INDEX_MASK); } -bool +bool skc_weakref_is_invalid(skc_weakref_t const * const weakref, skc_epoch_t const * const epoch) { diff --git a/src/compute/skc/weakref.h b/src/compute/skc/weakref.h index 9d09f8e5a6..c6ce6490f8 100644 --- a/src/compute/skc/weakref.h +++ b/src/compute/skc/weakref.h @@ -25,20 +25,20 @@ typedef skc_ulong skc_epoch_t; // // -void +void skc_weakref_epoch_init(skc_epoch_t * const epoch); -void +void skc_weakref_epoch_inc(skc_epoch_t * const epoch); -void -skc_weakref_init(skc_weakref_t * const weakref, +void +skc_weakref_init(skc_weakref_t * const weakref, skc_epoch_t * const epoch, skc_uint const index); -bool +bool skc_weakref_is_invalid(skc_weakref_t const * const weakref, - skc_epoch_t const * const epoch); + skc_epoch_t const * const epoch); skc_uint skc_weakref_index(skc_weakref_t const * const weakref); diff --git a/src/compute/ts/transform_stack.c b/src/compute/ts/transform_stack.c index 42a5a1e567..2766490f03 100644 --- a/src/compute/ts/transform_stack.c +++ b/src/compute/ts/transform_stack.c @@ -41,7 +41,7 @@ union skc_transform_stack_3x3_u { skc_transform_float_t a9[9]; - + struct { skc_transform_float_t sx; skc_transform_float_t shx; @@ -62,7 +62,7 @@ struct skc_transform_stack { uint32_t size; uint32_t count; - + skc_transform_weakref_t * weakrefs; union skc_transform_stack_3x3_u * transforms; }; @@ -78,7 +78,7 @@ skc_transform_stack_3x3_u_copy(union skc_transform_stack_3x3_u * const __r { for (int32_t ii=0; ii<9; ii++) dst->a9[ii] = src->a9[ii]; -} +} // // C = A * B @@ -89,7 +89,7 @@ void skc_transform_stack_3x3_u_multiply(union skc_transform_stack_3x3_u * const __restrict C, union skc_transform_stack_3x3_u const * const __restrict A, union skc_transform_stack_3x3_u const * const __restrict B) -{ +{ C->sx = A->sx * B->sx + A->shx * B->shy + A->tx * B->w0; C->shx = A->sx * B->shx + A->shx * B->sy + A->tx * B->w1; C->tx = A->sx * B->tx + A->shx * B->ty + A->tx * B->w2; @@ -167,7 +167,7 @@ skc_transform_stack_save(struct skc_transform_stack * const ts) } void -skc_transform_stack_restore(struct skc_transform_stack * const ts, +skc_transform_stack_restore(struct skc_transform_stack * const ts, uint32_t const restore) { ts->count = restore; @@ -177,7 +177,7 @@ skc_transform_stack_restore(struct skc_transform_stack * const ts, // // -static +static union skc_transform_stack_3x3_u * skc_transform_stack_tos(struct skc_transform_stack * const ts) { @@ -241,7 +241,7 @@ skc_transform_stack_top_weakref(struct skc_transform_stack * const ts) // // -void +void skc_transform_stack_drop(struct skc_transform_stack * const ts) { assert(ts->count >= 1); @@ -249,11 +249,11 @@ skc_transform_stack_drop(struct skc_transform_stack * const ts) ts->count -= 1; } -void +void skc_transform_stack_dup(struct skc_transform_stack * const ts) { skc_transform_stack_ensure(ts); - + union skc_transform_stack_3x3_u * const tos = skc_transform_stack_tos(ts); skc_transform_stack_3x3_u_copy(tos+1,tos); @@ -269,18 +269,18 @@ skc_transform_stack_dup(struct skc_transform_stack * const ts) void skc_transform_stack_push_matrix(struct skc_transform_stack * const ts, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const shx, skc_transform_float_t const tx, - skc_transform_float_t const shy, - skc_transform_float_t const sy, + skc_transform_float_t const shy, + skc_transform_float_t const sy, skc_transform_float_t const ty, - skc_transform_float_t const w0, - skc_transform_float_t const w1, + skc_transform_float_t const w0, + skc_transform_float_t const w1, skc_transform_float_t const w2) { skc_transform_stack_ensure(ts); - + union skc_transform_stack_3x3_u * t = ts->transforms + ts->count; t->sx = sx; @@ -310,10 +310,10 @@ skc_transform_stack_push_identity(struct skc_transform_stack * const ts) void skc_transform_stack_push_affine(struct skc_transform_stack * const ts, skc_transform_float_t const sx, - skc_transform_float_t const shx, + skc_transform_float_t const shx, skc_transform_float_t const tx, - skc_transform_float_t const shy, - skc_transform_float_t const sy, + skc_transform_float_t const shy, + skc_transform_float_t const sy, skc_transform_float_t const ty) { skc_transform_stack_push_matrix(ts, @@ -330,23 +330,23 @@ skc_transform_stack_push_translate(struct skc_transform_stack * const ts, skc_transform_stack_push_matrix(ts, 1.0, 0.0, tx, 0.0, 1.0, ty, - 0.0, 0.0, 1.0); + 0.0, 0.0, 1.0); } void skc_transform_stack_push_scale(struct skc_transform_stack * const ts, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const sy) { skc_transform_stack_push_matrix(ts, sx, 0.0, 0.0, 0.0, sy, 0.0, - 0.0, 0.0, 1.0); + 0.0, 0.0, 1.0); } void skc_transform_stack_push_shear(struct skc_transform_stack * const ts, - skc_transform_float_t const shx, + skc_transform_float_t const shx, skc_transform_float_t const shy) { skc_transform_stack_push_matrix(ts, @@ -412,7 +412,7 @@ skc_transform_stack_push_rotate_xy2(struct skc_transform_stack * const ts, void skc_transform_stack_push_rotate_xy(struct skc_transform_stack * const ts, skc_transform_float_t const theta, - skc_transform_float_t const cx, + skc_transform_float_t const cx, skc_transform_float_t const cy) { skc_transform_stack_push_rotate_xy2(ts,theta,cx,cy,cx,cy); @@ -421,7 +421,7 @@ skc_transform_stack_push_rotate_xy(struct skc_transform_stack * const ts, void skc_transform_stack_push_rotate_scale_xy(struct skc_transform_stack * const ts, skc_transform_float_t const theta, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const sy, skc_transform_float_t const cx, skc_transform_float_t const cy) @@ -465,7 +465,7 @@ skc_transform_stack_push_quad_to_unit(struct skc_transform_stack * const ts, // // dividing by the determinant (just w2?) // - skc_transform_float_t const det = + skc_transform_float_t const det = DIAG2(T,0,4) + DIAG3(T,3,7,2) + DIAG3(T,6,1,5) - @@ -478,7 +478,7 @@ skc_transform_stack_push_quad_to_unit(struct skc_transform_stack * const ts, skc_transform_float_t const rcp = 1.0 / det; - const union skc_transform_stack_3x3_u A = + const union skc_transform_stack_3x3_u A = { +DET3(T,4,5,7) * rcp, -DET3(T,1,2,7) * rcp, @@ -495,7 +495,7 @@ skc_transform_stack_push_quad_to_unit(struct skc_transform_stack * const ts, #else // just the adjoint can result in large values - const union skc_transform_stack_3x3_u A = + const union skc_transform_stack_3x3_u A = { +DET3(T,4,5,7), -DET3(T,1,2,7), @@ -513,7 +513,7 @@ skc_transform_stack_push_quad_to_unit(struct skc_transform_stack * const ts, #endif skc_transform_stack_3x3_u_copy(T,&A); - + return true; } @@ -536,7 +536,7 @@ skc_transform_stack_push_unit_to_quad(struct skc_transform_stack * const ts, skc_transform_float_t const x3 = (skc_transform_float_t)X(quad_dst,3); skc_transform_float_t const y3 = (skc_transform_float_t)Y(quad_dst,3); - + skc_transform_float_t sx = x1 - x0; skc_transform_float_t shy = y1 - y0; @@ -547,7 +547,7 @@ skc_transform_stack_push_unit_to_quad(struct skc_transform_stack * const ts, skc_transform_float_t const dy3 = -shy - dy2; // if both zero then quad_dst is a parallelogram and affine - if ((dx3 == 0.0) && (dy3 == 0.0)) + if ((dx3 == 0.0) && (dy3 == 0.0)) { skc_transform_float_t const shx = x2 - x1; skc_transform_float_t const sy = y2 - y1; @@ -556,7 +556,7 @@ skc_transform_stack_push_unit_to_quad(struct skc_transform_stack * const ts, sx, shx, x0, shy, sy, y0, 0.0, 0.0, 1.0); - } + } else { skc_transform_float_t const dx1 = x1 - x2; @@ -600,7 +600,7 @@ skc_transform_stack_push_quad_to_quad(struct skc_transform_stack * const ts, if (skc_transform_stack_push_quad_to_unit(ts,quad_src) == false) return false; - if (skc_transform_stack_push_unit_to_quad(ts,quad_dst) == false) + if (skc_transform_stack_push_unit_to_quad(ts,quad_dst) == false) return false; union skc_transform_stack_3x3_u * const U2Q = skc_transform_stack_tos(ts); @@ -631,21 +631,21 @@ skc_transform_stack_push_rect_to_quad(struct skc_transform_stack * const ts, const float y1, const float quad_dst[8]) { - if (skc_transform_stack_push_unit_to_quad(ts,quad_dst) == false) + if (skc_transform_stack_push_unit_to_quad(ts,quad_dst) == false) return false; - const union skc_transform_stack_3x3_u R2U = + const union skc_transform_stack_3x3_u R2U = { - SKC_TRANSFORM_RCP((skc_transform_float_t)(x1-x0)), - 0.0f, + SKC_TRANSFORM_RCP((skc_transform_float_t)(x1-x0)), + 0.0f, (skc_transform_float_t)-x0, - 0.0f, + 0.0f, SKC_TRANSFORM_RCP((skc_transform_float_t)(y1-y0)), (skc_transform_float_t)-y0, - 0.0f, - 0.0f, + 0.0f, + 0.0f, 1.0, }; @@ -682,7 +682,7 @@ skc_transform_stack_concat(struct skc_transform_stack * const ts) if (ts->count <= 1) return; - // get A and B + // get A and B union skc_transform_stack_3x3_u * const B = skc_transform_stack_tos(ts); const union skc_transform_stack_3x3_u * const A = B - 1; @@ -702,7 +702,7 @@ skc_transform_stack_concat(struct skc_transform_stack * const ts) // // -void +void skc_transform_stack_transform_affine(struct skc_transform_stack * const ts, skc_transform_float_t const x_pre, skc_transform_float_t const y_pre, diff --git a/src/compute/ts/transform_stack.h b/src/compute/ts/transform_stack.h index 275f529db5..95589d8889 100644 --- a/src/compute/ts/transform_stack.h +++ b/src/compute/ts/transform_stack.h @@ -70,21 +70,21 @@ skc_transform_stack_top_weakref(struct skc_transform_stack * const ts); // // -void +void skc_transform_stack_drop(struct skc_transform_stack * const ts); -void +void skc_transform_stack_dup(struct skc_transform_stack * const ts); // // // -void +void skc_transform_stack_transform_affine(struct skc_transform_stack * const ts, - skc_transform_float_t const x_pre, + skc_transform_float_t const x_pre, skc_transform_float_t const y_pre, - skc_transform_float_t * const x_post, + skc_transform_float_t * const x_post, skc_transform_float_t * const y_post); // @@ -93,14 +93,14 @@ skc_transform_stack_transform_affine(struct skc_transform_stack * const ts, void skc_transform_stack_push_matrix(struct skc_transform_stack * const ts, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const shx, skc_transform_float_t const tx, skc_transform_float_t const shy, - skc_transform_float_t const sy, + skc_transform_float_t const sy, skc_transform_float_t const ty, - skc_transform_float_t const w0, - skc_transform_float_t const w1, + skc_transform_float_t const w0, + skc_transform_float_t const w1, skc_transform_float_t const w2); void @@ -108,26 +108,26 @@ skc_transform_stack_push_identity(struct skc_transform_stack * const ts); void skc_transform_stack_push_affine(struct skc_transform_stack * const ts, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const shx, skc_transform_float_t const tx, - skc_transform_float_t const shy, + skc_transform_float_t const shy, skc_transform_float_t const sy, skc_transform_float_t const ty); void skc_transform_stack_push_translate(struct skc_transform_stack * const ts, - skc_transform_float_t const tx, + skc_transform_float_t const tx, skc_transform_float_t const ty); void skc_transform_stack_push_scale(struct skc_transform_stack * const ts, - skc_transform_float_t const sx, + skc_transform_float_t const sx, skc_transform_float_t const sy); void skc_transform_stack_push_shear(struct skc_transform_stack * const ts, - skc_transform_float_t const shx, + skc_transform_float_t const shx, skc_transform_float_t const shy); @@ -148,7 +148,7 @@ skc_transform_stack_push_rotate_xy2(struct skc_transform_stack * const ts, skc_transform_float_t const theta, skc_transform_float_t const cx, skc_transform_float_t const cy, - skc_transform_float_t const tx, + skc_transform_float_t const tx, skc_transform_float_t const ty); void |