July 2023 - Wine-commits - list.winehq.org

Nikolay Sivov : vkd3d-shader/hlsl: Support evaluated expressions for sample count in multisampled textures declarations.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: vkd3d Branch: master Commit: a65c0b0e22195cfe4bdeb760e74bc6a86a46e8b1 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/a65c0b0e22195cfe4bdeb760e74bc… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Wed Jun 28 11:28:49 2023 +0200 vkd3d-shader/hlsl: Support evaluated expressions for sample count in multisampled textures declarations. Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- libs/vkd3d-shader/hlsl.y | 20 +++++++++++++++----- tests/hlsl/texture-load.shader_test | 3 ++- 2 files changed, 17 insertions(+), 6 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index 80ba8e11..3dd8dd3a 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -1139,20 +1139,28 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str { struct hlsl_ir_constant *constant; struct hlsl_ir_node *node; + struct hlsl_block expr; unsigned int ret = 0; bool progress; - if (!add_implicit_conversion(ctx, &block->instrs, node_from_list(&block->instrs), + if (!hlsl_clone_block(ctx, &expr, &ctx->static_initializers)) + return 0; + hlsl_block_add_block(&expr, block); + + if (!add_implicit_conversion(ctx, &expr.instrs, node_from_list(&expr.instrs), hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), loc)) + { + hlsl_block_cleanup(&expr); return 0; + } do { - progress = hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, block, NULL); - progress |= hlsl_copy_propagation_execute(ctx, block); + progress = hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, &expr, NULL); + progress |= hlsl_copy_propagation_execute(ctx, &expr); } while (progress); - node = node_from_list(&block->instrs); + node = node_from_list(&expr.instrs); if (node->type == HLSL_IR_CONSTANT) { constant = hlsl_ir_constant(node); @@ -1164,6 +1172,8 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str "Failed to evaluate constant expression %d.", node->type); } + hlsl_block_cleanup(&expr); + return ret; } @@ -5675,7 +5685,7 @@ arrays: uint32_t *new_array; unsigned int size; - hlsl_clone_block(ctx, &block, &ctx->static_initializers); + hlsl_block_init(&block); list_move_tail(&block.instrs, $2); size = evaluate_static_expression_as_uint(ctx, &block, &@2); diff --git a/tests/hlsl/texture-load.shader_test b/tests/hlsl/texture-load.shader_test index 4893e4d8..362e1d2e 100644 --- a/tests/hlsl/texture-load.shader_test +++ b/tests/hlsl/texture-load.shader_test @@ -37,7 +37,8 @@ probe (0, 1) rgba (0.5, 0.7, 0.6, 0.8) probe (1, 1) rgba (0.8, 0.0, 0.7, 1.0) [pixel shader] -Texture2DMS<float4, 1> t; +static const int size = 2; +Texture2DMS<float4, size - 1> t; float4 main(float4 pos : sv_position) : sv_target {

1 0

Nikolay Sivov : vkd3d-shader/hlsl: Disallow certain instruction types from constant expressions.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: vkd3d Branch: master Commit: 06040d2a3034eafbe839951f77b6a7abda32f7ce URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/06040d2a3034eafbe839951f77b6a… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Wed Jun 28 11:57:54 2023 +0200 vkd3d-shader/hlsl: Disallow certain instruction types from constant expressions. Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- libs/vkd3d-shader/hlsl.y | 24 +++++++++++++++++++++++- tests/hlsl/array-size-expr.shader_test | 15 +++++++++++++++ tests/hlsl/invalid.shader_test | 2 +- 3 files changed, 39 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index 3dd8dd3a..82dd19eb 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -1143,6 +1143,28 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str unsigned int ret = 0; bool progress; + LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry) + { + switch (node->type) + { + case HLSL_IR_CONSTANT: + case HLSL_IR_EXPR: + case HLSL_IR_SWIZZLE: + case HLSL_IR_LOAD: + case HLSL_IR_INDEX: + continue; + case HLSL_IR_CALL: + case HLSL_IR_IF: + case HLSL_IR_LOOP: + case HLSL_IR_JUMP: + case HLSL_IR_RESOURCE_LOAD: + case HLSL_IR_RESOURCE_STORE: + case HLSL_IR_STORE: + hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, + "Expected literal expression."); + } + } + if (!hlsl_clone_block(ctx, &expr, &ctx->static_initializers)) return 0; hlsl_block_add_block(&expr, block); @@ -1169,7 +1191,7 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str else { hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, - "Failed to evaluate constant expression %d.", node->type); + "Failed to evaluate constant expression."); } hlsl_block_cleanup(&expr); diff --git a/tests/hlsl/array-size-expr.shader_test b/tests/hlsl/array-size-expr.shader_test index ac774be1..1fd4e262 100644 --- a/tests/hlsl/array-size-expr.shader_test +++ b/tests/hlsl/array-size-expr.shader_test @@ -51,3 +51,18 @@ float4 main() : sv_target [test] draw quad probe all rgba (2, 3, 6, 1) + +% Additional level of indirection +[pixel shader todo] +static const float array[8] = {1, 2, 3, 4, 5, 6, 7, 8}; +static const int idx = 2; +static const float array2[array[idx]] = {1, 2, 3}; + +float4 main() : sv_target +{ + return float4(array[1], array2[2], array[5], array[0]); +} + +[test] +todo draw quad +probe all rgba (2, 3, 6, 1) diff --git a/tests/hlsl/invalid.shader_test b/tests/hlsl/invalid.shader_test index 61033811..ad062652 100644 --- a/tests/hlsl/invalid.shader_test +++ b/tests/hlsl/invalid.shader_test @@ -72,7 +72,7 @@ float4 main() : sv_target return 0; } -[pixel shader fail todo] +[pixel shader fail] float4 main() : sv_target { int x;

1 0

Zebediah Figura : vkd3d-shader/d3dbc: Set the source count for DEF instructions to 1 (Valgrind).
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: vkd3d Branch: master Commit: e390bc35e2c9b0a2110370f916033eea2366317e URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/e390bc35e2c9b0a2110370f916033… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Wed Jun 28 16:14:27 2023 -0500 vkd3d-shader/d3dbc: Set the source count for DEF instructions to 1 (Valgrind). We emit one source with multiple components. --- libs/vkd3d-shader/d3dbc.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d-shader/d3dbc.c index e8a6236b..92c78422 100644 --- a/libs/vkd3d-shader/d3dbc.c +++ b/libs/vkd3d-shader/d3dbc.c @@ -260,9 +260,9 @@ static const struct vkd3d_sm1_opcode_info vs_opcode_table[] = /* Declarations */ {VKD3D_SM1_OP_DCL, 0, 2, VKD3DSIH_DCL}, /* Constant definitions */ - {VKD3D_SM1_OP_DEF, 1, 4, VKD3DSIH_DEF}, + {VKD3D_SM1_OP_DEF, 1, 1, VKD3DSIH_DEF}, {VKD3D_SM1_OP_DEFB, 1, 1, VKD3DSIH_DEFB}, - {VKD3D_SM1_OP_DEFI, 1, 4, VKD3DSIH_DEFI}, + {VKD3D_SM1_OP_DEFI, 1, 1, VKD3DSIH_DEFI}, /* Control flow */ {VKD3D_SM1_OP_REP, 0, 1, VKD3DSIH_REP, {2, 0}, {~0u, ~0u}}, {VKD3D_SM1_OP_ENDREP, 0, 0, VKD3DSIH_ENDREP, {2, 0}, {~0u, ~0u}}, @@ -327,9 +327,9 @@ static const struct vkd3d_sm1_opcode_info ps_opcode_table[] = /* Declarations */ {VKD3D_SM1_OP_DCL, 0, 2, VKD3DSIH_DCL}, /* Constant definitions */ - {VKD3D_SM1_OP_DEF, 1, 4, VKD3DSIH_DEF}, + {VKD3D_SM1_OP_DEF, 1, 1, VKD3DSIH_DEF}, {VKD3D_SM1_OP_DEFB, 1, 1, VKD3DSIH_DEFB}, - {VKD3D_SM1_OP_DEFI, 1, 4, VKD3DSIH_DEFI}, + {VKD3D_SM1_OP_DEFI, 1, 1, VKD3DSIH_DEFI}, /* Control flow */ {VKD3D_SM1_OP_REP, 0, 1, VKD3DSIH_REP, {2, 1}, {~0u, ~0u}}, {VKD3D_SM1_OP_ENDREP, 0, 0, VKD3DSIH_ENDREP, {2, 1}, {~0u, ~0u}},

1 0

Zebediah Figura : vkd3d-shader/hlsl: Consistently use the destination width in constant folding instructions (Valgrind).
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: vkd3d Branch: master Commit: 717cd3f5bbbfdad8e3b3c03485625ac1a95ccf20 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/717cd3f5bbbfdad8e3b3c03485625… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Wed Jun 28 16:14:55 2023 -0500 vkd3d-shader/hlsl: Consistently use the destination width in constant folding instructions (Valgrind). --- libs/vkd3d-shader/hlsl_constant_ops.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_constant_ops.c b/libs/vkd3d-shader/hlsl_constant_ops.c index e8330319..01c438ae 100644 --- a/libs/vkd3d-shader/hlsl_constant_ops.c +++ b/libs/vkd3d-shader/hlsl_constant_ops.c @@ -80,7 +80,7 @@ static bool fold_cast(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, return false; } - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (src->node.data_type->base_type) { @@ -160,7 +160,7 @@ static bool fold_neg(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, assert(type == src->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (type) { @@ -240,7 +240,7 @@ static bool fold_add(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons assert(type == src1->node.data_type->base_type); assert(type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (type) { @@ -425,7 +425,7 @@ static bool fold_equal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, co assert(dst_type->base_type == HLSL_TYPE_BOOL); assert(src1->node.data_type->base_type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (src1->node.data_type->base_type) { @@ -461,7 +461,7 @@ static bool fold_gequal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, c assert(dst_type->base_type == HLSL_TYPE_BOOL); assert(src1->node.data_type->base_type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (src1->node.data_type->base_type) { @@ -500,7 +500,7 @@ static bool fold_less(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, con assert(dst_type->base_type == HLSL_TYPE_BOOL); assert(src1->node.data_type->base_type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (src1->node.data_type->base_type) { @@ -659,7 +659,7 @@ static bool fold_mul(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons assert(type == src1->node.data_type->base_type); assert(type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (type) { @@ -693,7 +693,7 @@ static bool fold_nequal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, c assert(dst_type->base_type == HLSL_TYPE_BOOL); assert(src1->node.data_type->base_type == src2->node.data_type->base_type); - for (k = 0; k < 4; ++k) + for (k = 0; k < dst_type->dimx; ++k) { switch (src1->node.data_type->base_type) {

1 0

Zebediah Figura : vkd3d-shader/hlsl: Allocate anonymous constants after uniforms.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: vkd3d Branch: master Commit: 8a2b357855a87f220bd6dd702d09b8b20aa3102f URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/8a2b357855a87f220bd6dd702d09b… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Mon Nov 7 17:45:35 2022 -0600 vkd3d-shader/hlsl: Allocate anonymous constants after uniforms. --- libs/vkd3d-shader/hlsl_codegen.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index b980ed56..e265a36e 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -3359,8 +3359,6 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_ir_functi struct register_allocator allocator = {0}; struct hlsl_ir_var *var; - allocate_const_registers_recurse(ctx, &entry_func->body, &allocator); - LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) { if (var->is_uniform && var->last_read) @@ -3377,6 +3375,8 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_ir_functi } } + allocate_const_registers_recurse(ctx, &entry_func->body, &allocator); + vkd3d_free(allocator.allocations); }

1 0

Matteo Bruni : wined3d: Increase the CS queue size to 16 MiB on 64-bit architectures.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: wine Branch: master Commit: 203cbab6444922079e5e39d0bb68f6562b1ba19d URL: https://gitlab.winehq.org/wine/wine/-/commit/203cbab6444922079e5e39d0bb68f6… Author: Matteo Bruni <mbruni(a)codeweavers.com> Date: Wed Mar 8 18:40:56 2017 +0100 wined3d: Increase the CS queue size to 16 MiB on 64-bit architectures. With the buffer / texture mapping acceleration bits working properly, we can get many outstanding commands and, to avoid blocking the client thread while waiting for free space, we need a larger queue. 16 MiB was always enough in my testing (8 MiB wasn't in a few cases). Keep a lower limit on 32-bit since we're often times address space-starved there. Tweaked by Zebediah Figura. --- dlls/wined3d/wined3d_private.h | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h index 001534e8c38..be576fcfbde 100644 --- a/dlls/wined3d/wined3d_private.h +++ b/dlls/wined3d/wined3d_private.h @@ -5074,7 +5074,11 @@ enum wined3d_push_constants }; #define WINED3D_CS_QUERY_POLL_INTERVAL 10u +#if defined(_WIN64) +#define WINED3D_CS_QUEUE_SIZE 0x1000000u +#else #define WINED3D_CS_QUEUE_SIZE 0x400000u +#endif #define WINED3D_CS_SPIN_COUNT 10000000u #define WINED3D_CS_QUEUE_MASK (WINED3D_CS_QUEUE_SIZE - 1)

1 0

Matteo Bruni : wined3d: Block in wined3d_cs_emit_present() when we reach the latency limit.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: wine Branch: master Commit: 30c3697017220cc22ccdda7a3d86e7f4956551b3 URL: https://gitlab.winehq.org/wine/wine/-/commit/30c3697017220cc22ccdda7a3d86e7… Author: Matteo Bruni <mbruni(a)codeweavers.com> Date: Fri Apr 20 03:37:13 2018 +0200 wined3d: Block in wined3d_cs_emit_present() when we reach the latency limit. Avoid busy spinning for a potentially long time. --- dlls/wined3d/cs.c | 25 ++++++++++++++++++++++++- dlls/wined3d/wined3d_private.h | 3 ++- 2 files changed, 26 insertions(+), 2 deletions(-) diff --git a/dlls/wined3d/cs.c b/dlls/wined3d/cs.c index 6a09ec2a35a..14f2ca9bb4a 100644 --- a/dlls/wined3d/cs.c +++ b/dlls/wined3d/cs.c @@ -724,6 +724,8 @@ static void wined3d_cs_exec_present(struct wined3d_cs *cs, const void *data) } InterlockedDecrement(&cs->pending_presents); + if (InterlockedCompareExchange(&cs->waiting_for_present, FALSE, TRUE)) + SetEvent(cs->present_event); } void wined3d_cs_emit_present(struct wined3d_cs *cs, struct wined3d_swapchain *swapchain, @@ -734,6 +736,8 @@ void wined3d_cs_emit_present(struct wined3d_cs *cs, struct wined3d_swapchain *sw unsigned int i; LONG pending; + wined3d_not_from_cs(cs); + op = wined3d_device_context_require_space(&cs->c, sizeof(*op), WINED3D_CS_QUEUE_DEFAULT); op->opcode = WINED3D_CS_OP_PRESENT; op->dst_window_override = dst_window_override; @@ -757,8 +761,17 @@ void wined3d_cs_emit_present(struct wined3d_cs *cs, struct wined3d_swapchain *sw * ahead of the worker thread. */ while (pending >= swapchain->max_frame_latency) { - YieldProcessor(); + InterlockedExchange(&cs->waiting_for_present, TRUE); + pending = InterlockedCompareExchange(&cs->pending_presents, 0, 0); + if (pending >= swapchain->max_frame_latency || !InterlockedCompareExchange(&cs->waiting_for_present, FALSE, TRUE)) + { + TRACE_(d3d_perf)("Reached latency limit (%u frames), blocking to wait.\n", swapchain->max_frame_latency); + wined3d_mutex_unlock(); + WaitForSingleObject(cs->present_event, INFINITE); + wined3d_mutex_lock(); + TRACE_(d3d_perf)("Woken up from the wait.\n"); + } } } @@ -3472,11 +3485,18 @@ struct wined3d_cs *wined3d_cs_create(struct wined3d_device *device, heap_free(cs->data); goto fail; } + if (!(cs->present_event = CreateEventW(NULL, FALSE, FALSE, NULL))) + { + ERR("Failed to create command stream present event.\n"); + heap_free(cs->data); + goto fail; + } if (!(GetModuleHandleExW(GET_MODULE_HANDLE_EX_FLAG_FROM_ADDRESS, (const WCHAR *)wined3d_cs_run, &cs->wined3d_module))) { ERR("Failed to get wined3d module handle.\n"); + CloseHandle(cs->present_event); CloseHandle(cs->event); heap_free(cs->data); goto fail; @@ -3486,6 +3506,7 @@ struct wined3d_cs *wined3d_cs_create(struct wined3d_device *device, { ERR("Failed to create wined3d command stream thread.\n"); FreeLibrary(cs->wined3d_module); + CloseHandle(cs->present_event); CloseHandle(cs->event); heap_free(cs->data); goto fail; @@ -3508,6 +3529,8 @@ void wined3d_cs_destroy(struct wined3d_cs *cs) { wined3d_cs_emit_stop(cs); CloseHandle(cs->thread); + if (!CloseHandle(cs->present_event)) + ERR("Closing present event failed.\n"); if (!CloseHandle(cs->event)) ERR("Closing event failed.\n"); } diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h index be576fcfbde..c71ac59270b 100644 --- a/dlls/wined3d/wined3d_private.h +++ b/dlls/wined3d/wined3d_private.h @@ -5131,8 +5131,9 @@ struct wined3d_cs struct list query_poll_list; BOOL queries_flushed; - HANDLE event; + HANDLE event, present_event; LONG waiting_for_event; + LONG waiting_for_present; LONG pending_presents; };

1 0

Matteo Bruni : wined3d: Call YieldProcessor() when spinning waiting for CS commands.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: wine Branch: master Commit: ecdeab593bdb42c6c8854cdef9c59daba9ba6ce1 URL: https://gitlab.winehq.org/wine/wine/-/commit/ecdeab593bdb42c6c8854cdef9c59d… Author: Matteo Bruni <mbruni(a)codeweavers.com> Date: Sun May 6 03:14:21 2018 +0200 wined3d: Call YieldProcessor() when spinning waiting for CS commands. Yielding to other threads might help to get new packets and (very slightly) reduce contention on the queue head and tail pointers. --- dlls/wined3d/cs.c | 1 + 1 file changed, 1 insertion(+) diff --git a/dlls/wined3d/cs.c b/dlls/wined3d/cs.c index 14f2ca9bb4a..2cd53cd7a5b 100644 --- a/dlls/wined3d/cs.c +++ b/dlls/wined3d/cs.c @@ -3421,6 +3421,7 @@ static DWORD WINAPI wined3d_cs_run(void *ctx) queue = &cs->queue[WINED3D_CS_QUEUE_DEFAULT]; if (wined3d_cs_queue_is_empty(cs, queue)) { + YieldProcessor(); if (++spin_count >= WINED3D_CS_SPIN_COUNT && list_empty(&cs->query_poll_list)) wined3d_cs_wait_event(cs); continue;

1 0

Matteo Bruni : wined3d: Make the CS thread yield CPU when it can't block waiting.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: wine Branch: master Commit: fcccca153a1db3d1151c62521b0670ecb4194619 URL: https://gitlab.winehq.org/wine/wine/-/commit/fcccca153a1db3d1151c62521b0670… Author: Matteo Bruni <mbruni(a)codeweavers.com> Date: Thu Nov 15 22:39:45 2018 +0100 wined3d: Make the CS thread yield CPU when it can't block waiting. It can't block because there are pending queries. --- dlls/wined3d/cs.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/dlls/wined3d/cs.c b/dlls/wined3d/cs.c index 2cd53cd7a5b..04e37f5e712 100644 --- a/dlls/wined3d/cs.c +++ b/dlls/wined3d/cs.c @@ -3422,8 +3422,13 @@ static DWORD WINAPI wined3d_cs_run(void *ctx) if (wined3d_cs_queue_is_empty(cs, queue)) { YieldProcessor(); - if (++spin_count >= WINED3D_CS_SPIN_COUNT && list_empty(&cs->query_poll_list)) - wined3d_cs_wait_event(cs); + if (++spin_count >= WINED3D_CS_SPIN_COUNT) + { + if (list_empty(&cs->query_poll_list)) + wined3d_cs_wait_event(cs); + else + Sleep(0); + } continue; } }

1 0

Matteo Bruni : wined3d: Reduce CS spin count to 2000.
by Alexandre Julliard 04 Jul '23

04 Jul '23

Module: wine Branch: master Commit: f032ccc2715a5525601f741337a7bc7df6873c9d URL: https://gitlab.winehq.org/wine/wine/-/commit/f032ccc2715a5525601f741337a7bc… Author: Matteo Bruni <mbruni(a)codeweavers.com> Date: Sun May 6 03:09:44 2018 +0200 wined3d: Reduce CS spin count to 2000. Over the last few Wine releases we greatly reduced the need for the application thread to wait for replies from the CS thread. Compared to the time when the command stream was initially introduced, it's now quite likely that, when the command queues become empty, they are going to stay like that for a while (e.g. the game is throttling the framerate or is busy doing some CPU work on its part before generating more commands). As a first step, reduce the spin count to reduce the busy waiting in wined3d_cs_run(). In my tests this significantly reduces CPU usage, with minimally decreased performance in a few rare cases but also significantly improved performance in many cases (notably with integrated GPUs where freeing the CPU directly allows more power to be allocated to the GPU). --- dlls/wined3d/wined3d_private.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h index c71ac59270b..efdcff41e58 100644 --- a/dlls/wined3d/wined3d_private.h +++ b/dlls/wined3d/wined3d_private.h @@ -5079,7 +5079,7 @@ enum wined3d_push_constants #else #define WINED3D_CS_QUEUE_SIZE 0x400000u #endif -#define WINED3D_CS_SPIN_COUNT 10000000u +#define WINED3D_CS_SPIN_COUNT 2000u #define WINED3D_CS_QUEUE_MASK (WINED3D_CS_QUEUE_SIZE - 1) C_ASSERT(!(WINED3D_CS_QUEUE_SIZE & (WINED3D_CS_QUEUE_SIZE - 1)));

1 0