KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

radeonsi: shorten slot masks to 32 bits 

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This commit is contained in:
Marek Olšák 2016-04-19 02:21:25 +02:00
parent 0954d5e982
commit 3acaefb1bb
4 changed files with 61 additions and 63 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/gallium/drivers/radeonsi/si_blit.c
View File

@ -243,14 +243,14 @@ si_flush_depth_textures(struct si_context *sctx,
struct si_textures_info *textures) struct si_textures_info *textures)
{ {
unsigned i; unsigned i;
uint64_t mask = textures->depth_texture_mask; unsigned mask = textures->depth_texture_mask;
while (mask) { while (mask) {
struct pipe_sampler_view *view; struct pipe_sampler_view *view;
struct si_sampler_view *sview; struct si_sampler_view *sview;
struct r600_texture *tex; struct r600_texture *tex;
i = u_bit_scan64(&mask); i = u_bit_scan(&mask);
view = textures->views.views[i]; view = textures->views.views[i];
assert(view); assert(view);
@ -326,13 +326,13 @@ si_decompress_sampler_color_textures(struct si_context *sctx,
struct si_textures_info *textures) struct si_textures_info *textures)
{ {
unsigned i; unsigned i;
uint64_t mask = textures->compressed_colortex_mask; unsigned mask = textures->compressed_colortex_mask;
while (mask) { while (mask) {
struct pipe_sampler_view *view; struct pipe_sampler_view *view;
struct r600_texture *tex; struct r600_texture *tex;
i = u_bit_scan64(&mask); i = u_bit_scan(&mask);
view = textures->views.views[i]; view = textures->views.views[i];
assert(view); assert(view);
@ -352,13 +352,13 @@ si_decompress_image_color_textures(struct si_context *sctx,
struct si_images_info *images) struct si_images_info *images)
{ {
unsigned i; unsigned i;
uint64_t mask = images->compressed_colortex_mask; unsigned mask = images->compressed_colortex_mask;
while (mask) { while (mask) {
const struct pipe_image_view *view; const struct pipe_image_view *view;
struct r600_texture *tex; struct r600_texture *tex;
i = u_bit_scan64(&mask); i = u_bit_scan(&mask);
view = &images->views[i]; view = &images->views[i];
assert(view->resource->target != PIPE_BUFFER); assert(view->resource->target != PIPE_BUFFER);

104
src/gallium/drivers/radeonsi/si_descriptors.c
View File

@ -110,7 +110,7 @@ static void si_init_descriptors(struct si_descriptors *desc,
desc->list = CALLOC(num_elements, element_dw_size * 4); desc->list = CALLOC(num_elements, element_dw_size * 4);
desc->element_dw_size = element_dw_size; desc->element_dw_size = element_dw_size;
desc->num_elements = num_elements; desc->num_elements = num_elements;
desc->dirty_mask = num_elements == 64 ? ~0llu : (1llu << num_elements) - 1; desc->dirty_mask = num_elements == 32 ? ~0u : (1u << num_elements) - 1;
desc->shader_userdata_offset = shader_userdata_index * 4; desc->shader_userdata_offset = shader_userdata_index * 4;
if (ce_offset) { if (ce_offset) {
@ -210,8 +210,8 @@ static bool si_upload_descriptors(struct si_context *sctx,
while(desc->dirty_mask) { while(desc->dirty_mask) {
int begin, count; int begin, count;
u_bit_scan_consecutive_range64(&desc->dirty_mask, &begin, u_bit_scan_consecutive_range(&desc->dirty_mask, &begin,
&count); &count);
begin *= desc->element_dw_size; begin *= desc->element_dw_size;
count *= desc->element_dw_size; count *= desc->element_dw_size;
@ -276,11 +276,11 @@ static void si_sampler_view_add_buffer(struct si_context *sctx,
static void si_sampler_views_begin_new_cs(struct si_context *sctx, static void si_sampler_views_begin_new_cs(struct si_context *sctx,
struct si_sampler_views *views) struct si_sampler_views *views)
{ {
uint64_t mask = views->desc.enabled_mask; unsigned mask = views->desc.enabled_mask;
/* Add buffers to the CS. */ /* Add buffers to the CS. */
while (mask) { while (mask) {
int i = u_bit_scan64(&mask); int i = u_bit_scan(&mask);
si_sampler_view_add_buffer(sctx, views->views[i]->texture); si_sampler_view_add_buffer(sctx, views->views[i]->texture);
} }
@ -329,16 +329,16 @@ static void si_set_sampler_view(struct si_context *sctx,
views->sampler_states[slot], 4*4); views->sampler_states[slot], 4*4);
} }
views->desc.enabled_mask |= 1llu << slot; views->desc.enabled_mask |= 1u << slot;
} else { } else {
pipe_sampler_view_reference(&views->views[slot], NULL); pipe_sampler_view_reference(&views->views[slot], NULL);
memcpy(views->desc.list + slot*16, null_texture_descriptor, 8*4); memcpy(views->desc.list + slot*16, null_texture_descriptor, 8*4);
/* Only clear the lower dwords of FMASK. */ /* Only clear the lower dwords of FMASK. */
memcpy(views->desc.list + slot*16 + 8, null_texture_descriptor, 4*4); memcpy(views->desc.list + slot*16 + 8, null_texture_descriptor, 4*4);
views->desc.enabled_mask &= ~(1llu << slot); views->desc.enabled_mask &= ~(1u << slot);
} }
views->desc.dirty_mask |= 1llu << slot; views->desc.dirty_mask |= 1u << slot;
} }
static bool is_compressed_colortex(struct r600_texture *rtex) static bool is_compressed_colortex(struct r600_texture *rtex)
@ -363,8 +363,8 @@ static void si_set_sampler_views(struct pipe_context *ctx,
unsigned slot = start + i; unsigned slot = start + i;
if (!views || !views[i]) { if (!views || !views[i]) {
samplers->depth_texture_mask &= ~(1llu << slot); samplers->depth_texture_mask &= ~(1u << slot);
samplers->compressed_colortex_mask &= ~(1llu << slot); samplers->compressed_colortex_mask &= ~(1u << slot);
si_set_sampler_view(sctx, &samplers->views, slot, NULL); si_set_sampler_view(sctx, &samplers->views, slot, NULL);
continue; continue;
} }
@ -376,18 +376,18 @@ static void si_set_sampler_views(struct pipe_context *ctx,
(struct r600_texture*)views[i]->texture; (struct r600_texture*)views[i]->texture;
if (rtex->is_depth && !rtex->is_flushing_texture) { if (rtex->is_depth && !rtex->is_flushing_texture) {
samplers->depth_texture_mask |= 1llu << slot; samplers->depth_texture_mask |= 1u << slot;
} else { } else {
samplers->depth_texture_mask &= ~(1llu << slot); samplers->depth_texture_mask &= ~(1u << slot);
} }
if (is_compressed_colortex(rtex)) { if (is_compressed_colortex(rtex)) {
samplers->compressed_colortex_mask |= 1llu << slot; samplers->compressed_colortex_mask |= 1u << slot;
} else { } else {
samplers->compressed_colortex_mask &= ~(1llu << slot); samplers->compressed_colortex_mask &= ~(1u << slot);
} }
} else { } else {
samplers->depth_texture_mask &= ~(1llu << slot); samplers->depth_texture_mask &= ~(1u << slot);
samplers->compressed_colortex_mask &= ~(1llu << slot); samplers->compressed_colortex_mask &= ~(1u << slot);
} }
} }
} }
@ -395,19 +395,19 @@ static void si_set_sampler_views(struct pipe_context *ctx,
static void static void
si_samplers_update_compressed_colortex_mask(struct si_textures_info *samplers) si_samplers_update_compressed_colortex_mask(struct si_textures_info *samplers)
{ {
uint64_t mask = samplers->views.desc.enabled_mask; unsigned mask = samplers->views.desc.enabled_mask;
while (mask) { while (mask) {
int i = u_bit_scan64(&mask); int i = u_bit_scan(&mask);
struct pipe_resource *res = samplers->views.views[i]->texture; struct pipe_resource *res = samplers->views.views[i]->texture;
if (res && res->target != PIPE_BUFFER) { if (res && res->target != PIPE_BUFFER) {
struct r600_texture *rtex = (struct r600_texture *)res; struct r600_texture *rtex = (struct r600_texture *)res;
if (is_compressed_colortex(rtex)) { if (is_compressed_colortex(rtex)) {
samplers->compressed_colortex_mask |= 1llu << i; samplers->compressed_colortex_mask |= 1u << i;
} else { } else {
samplers->compressed_colortex_mask &= ~(1llu << i); samplers->compressed_colortex_mask &= ~(1u << i);
} }
} }
} }
@ -457,13 +457,13 @@ si_image_views_begin_new_cs(struct si_context *sctx, struct si_images_info *imag
static void static void
si_disable_shader_image(struct si_images_info *images, unsigned slot) si_disable_shader_image(struct si_images_info *images, unsigned slot)
{ {
if (images->desc.enabled_mask & (1llu << slot)) { if (images->desc.enabled_mask & (1u << slot)) {
pipe_resource_reference(&images->views[slot].resource, NULL); pipe_resource_reference(&images->views[slot].resource, NULL);
images->compressed_colortex_mask &= ~(1 << slot); images->compressed_colortex_mask &= ~(1 << slot);
memcpy(images->desc.list + slot*8, null_image_descriptor, 8*4); memcpy(images->desc.list + slot*8, null_image_descriptor, 8*4);
images->desc.enabled_mask &= ~(1llu << slot); images->desc.enabled_mask &= ~(1u << slot);
images->desc.dirty_mask |= 1llu << slot; images->desc.dirty_mask |= 1u << slot;
} }
} }
@ -543,18 +543,18 @@ si_set_shader_images(struct pipe_context *pipe, unsigned shader,
NULL); NULL);
} }
images->desc.enabled_mask |= 1llu << slot; images->desc.enabled_mask |= 1u << slot;
images->desc.dirty_mask |= 1llu << slot; images->desc.dirty_mask |= 1u << slot;
} }
} }
static void static void
si_images_update_compressed_colortex_mask(struct si_images_info *images) si_images_update_compressed_colortex_mask(struct si_images_info *images)
{ {
uint64_t mask = images->desc.enabled_mask; unsigned mask = images->desc.enabled_mask;
while (mask) { while (mask) {
int i = u_bit_scan64(&mask); int i = u_bit_scan(&mask);
struct pipe_resource *res = images->views[i].resource; struct pipe_resource *res = images->views[i].resource;
if (res && res->target != PIPE_BUFFER) { if (res && res->target != PIPE_BUFFER) {
@ -602,7 +602,7 @@ static void si_bind_sampler_states(struct pipe_context *ctx, unsigned shader,
continue; continue;
memcpy(desc->list + slot * 16 + 12, sstates[i]->val, 4*4); memcpy(desc->list + slot * 16 + 12, sstates[i]->val, 4*4);
desc->dirty_mask |= 1llu << slot; desc->dirty_mask |= 1u << slot;
} }
} }
@ -638,11 +638,11 @@ static void si_release_buffer_resources(struct si_buffer_resources *buffers)
static void si_buffer_resources_begin_new_cs(struct si_context *sctx, static void si_buffer_resources_begin_new_cs(struct si_context *sctx,
struct si_buffer_resources *buffers) struct si_buffer_resources *buffers)
{ {
uint64_t mask = buffers->desc.enabled_mask; unsigned mask = buffers->desc.enabled_mask;
/* Add buffers to the CS. */ /* Add buffers to the CS. */
while (mask) { while (mask) {
int i = u_bit_scan64(&mask); int i = u_bit_scan(&mask);
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
(struct r600_resource*)buffers->buffers[i], (struct r600_resource*)buffers->buffers[i],
@ -835,14 +835,14 @@ void si_set_constant_buffer(struct si_context *sctx,
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
(struct r600_resource*)buffer, (struct r600_resource*)buffer,
buffers->shader_usage, buffers->priority); buffers->shader_usage, buffers->priority);
buffers->desc.enabled_mask |= 1llu << slot; buffers->desc.enabled_mask |= 1u << slot;
} else { } else {
/* Clear the descriptor. */ /* Clear the descriptor. */
memset(buffers->desc.list + slot*4, 0, sizeof(uint32_t) * 4); memset(buffers->desc.list + slot*4, 0, sizeof(uint32_t) * 4);
buffers->desc.enabled_mask &= ~(1llu << slot); buffers->desc.enabled_mask &= ~(1u << slot);
} }
buffers->desc.dirty_mask |= 1llu << slot; buffers->desc.dirty_mask |= 1u << slot;
} }
static void si_pipe_set_constant_buffer(struct pipe_context *ctx, static void si_pipe_set_constant_buffer(struct pipe_context *ctx,
@ -879,7 +879,7 @@ static void si_set_shader_buffers(struct pipe_context *ctx, unsigned shader,
if (!sbuffer || !sbuffer->buffer) { if (!sbuffer || !sbuffer->buffer) {
pipe_resource_reference(&buffers->buffers[slot], NULL); pipe_resource_reference(&buffers->buffers[slot], NULL);
memset(desc, 0, sizeof(uint32_t) * 4); memset(desc, 0, sizeof(uint32_t) * 4);
buffers->desc.enabled_mask &= ~(1llu << slot); buffers->desc.enabled_mask &= ~(1u << slot);
continue; continue;
} }
@ -900,10 +900,9 @@ static void si_set_shader_buffers(struct pipe_context *ctx, unsigned shader,
pipe_resource_reference(&buffers->buffers[slot], &buf->b.b); pipe_resource_reference(&buffers->buffers[slot], &buf->b.b);
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, buf, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, buf,
buffers->shader_usage, buffers->priority); buffers->shader_usage, buffers->priority);
buffers->desc.enabled_mask |= 1llu << slot; buffers->desc.enabled_mask |= 1u << slot;
buffers->desc.dirty_mask |= 1llu << slot; buffers->desc.dirty_mask |= 1u << slot;
} }
} }
/* RING BUFFERS */ /* RING BUFFERS */
@ -991,14 +990,14 @@ void si_set_ring_buffer(struct pipe_context *ctx, uint shader, uint slot,
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
(struct r600_resource*)buffer, (struct r600_resource*)buffer,
buffers->shader_usage, buffers->priority); buffers->shader_usage, buffers->priority);
buffers->desc.enabled_mask |= 1llu << slot; buffers->desc.enabled_mask |= 1u << slot;
} else { } else {
/* Clear the descriptor. */ /* Clear the descriptor. */
memset(buffers->desc.list + slot*4, 0, sizeof(uint32_t) * 4); memset(buffers->desc.list + slot*4, 0, sizeof(uint32_t) * 4);
buffers->desc.enabled_mask &= ~(1llu << slot); buffers->desc.enabled_mask &= ~(1u << slot);
} }
buffers->desc.dirty_mask |= 1llu << slot; buffers->desc.dirty_mask |= 1u << slot;
} }
/* STREAMOUT BUFFERS */ /* STREAMOUT BUFFERS */
@ -1088,26 +1087,25 @@ static void si_set_streamout_targets(struct pipe_context *ctx,
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
(struct r600_resource*)buffer, (struct r600_resource*)buffer,
buffers->shader_usage, buffers->priority); buffers->shader_usage, buffers->priority);
buffers->desc.enabled_mask |= 1llu << bufidx; buffers->desc.enabled_mask |= 1u << bufidx;
} else { } else {
/* Clear the descriptor and unset the resource. */ /* Clear the descriptor and unset the resource. */
memset(buffers->desc.list + bufidx*4, 0, memset(buffers->desc.list + bufidx*4, 0,
sizeof(uint32_t) * 4); sizeof(uint32_t) * 4);
pipe_resource_reference(&buffers->buffers[bufidx], pipe_resource_reference(&buffers->buffers[bufidx],
NULL); NULL);
buffers->desc.enabled_mask &= ~(1llu << bufidx); buffers->desc.enabled_mask &= ~(1u << bufidx);
} }
buffers->desc.dirty_mask |= 1llu << bufidx; buffers->desc.dirty_mask |= 1u << bufidx;
} }
for (; i < old_num_targets; i++) { for (; i < old_num_targets; i++) {
bufidx = SI_VS_STREAMOUT_BUF0 + i; bufidx = SI_VS_STREAMOUT_BUF0 + i;
/* Clear the descriptor and unset the resource. */ /* Clear the descriptor and unset the resource. */
memset(buffers->desc.list + bufidx*4, 0, sizeof(uint32_t) * 4); memset(buffers->desc.list + bufidx*4, 0, sizeof(uint32_t) * 4);
pipe_resource_reference(&buffers->buffers[bufidx], NULL); pipe_resource_reference(&buffers->buffers[bufidx], NULL);
buffers->desc.enabled_mask &= ~(1llu << bufidx); buffers->desc.enabled_mask &= ~(1u << bufidx);
buffers->desc.dirty_mask |= 1llu << bufidx; buffers->desc.dirty_mask |= 1u << bufidx;
} }
} }
static void si_desc_reset_buffer_offset(struct pipe_context *ctx, static void si_desc_reset_buffer_offset(struct pipe_context *ctx,
@ -1171,15 +1169,15 @@ static void si_reset_buffer_resources(struct si_context *sctx,
struct pipe_resource *buf, struct pipe_resource *buf,
uint64_t old_va) uint64_t old_va)
{ {
uint64_t mask = buffers->desc.enabled_mask; unsigned mask = buffers->desc.enabled_mask;
while (mask) { while (mask) {
unsigned i = u_bit_scan64(&mask); unsigned i = u_bit_scan(&mask);
if (buffers->buffers[i] == buf) { if (buffers->buffers[i] == buf) {
si_desc_reset_buffer_offset(&sctx->b.b, si_desc_reset_buffer_offset(&sctx->b.b,
buffers->desc.list + i*4, buffers->desc.list + i*4,
old_va, buf); old_va, buf);
buffers->desc.dirty_mask |= 1llu << i; buffers->desc.dirty_mask |= 1u << i;
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
(struct r600_resource *)buf, (struct r600_resource *)buf,
@ -1271,16 +1269,16 @@ static void si_invalidate_buffer(struct pipe_context *ctx, struct pipe_resource
/* Texture buffers - update bindings. */ /* Texture buffers - update bindings. */
for (shader = 0; shader < SI_NUM_SHADERS; shader++) { for (shader = 0; shader < SI_NUM_SHADERS; shader++) {
struct si_sampler_views *views = &sctx->samplers[shader].views; struct si_sampler_views *views = &sctx->samplers[shader].views;
uint64_t mask = views->desc.enabled_mask; unsigned mask = views->desc.enabled_mask;
while (mask) { while (mask) {
unsigned i = u_bit_scan64(&mask); unsigned i = u_bit_scan(&mask);
if (views->views[i]->texture == buf) { if (views->views[i]->texture == buf) {
si_desc_reset_buffer_offset(ctx, si_desc_reset_buffer_offset(ctx,
views->desc.list + views->desc.list +
i * 16 + 4, i * 16 + 4,
old_va, buf); old_va, buf);
views->desc.dirty_mask |= 1llu << i; views->desc.dirty_mask |= 1u << i;
radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx,
rbuffer, RADEON_USAGE_READ, rbuffer, RADEON_USAGE_READ,
@ -1301,7 +1299,7 @@ static void si_invalidate_buffer(struct pipe_context *ctx, struct pipe_resource
si_desc_reset_buffer_offset( si_desc_reset_buffer_offset(
ctx, images->desc.list + i * 8 + 4, ctx, images->desc.list + i * 8 + 4,
old_va, buf); old_va, buf);
images->desc.dirty_mask |= 1llu << i; images->desc.dirty_mask |= 1u << i;
radeon_add_to_buffer_list( radeon_add_to_buffer_list(
&sctx->b, &sctx->b.gfx, rbuffer, &sctx->b, &sctx->b.gfx, rbuffer,

4
src/gallium/drivers/radeonsi/si_pipe.h
View File

@ -137,8 +137,8 @@ struct si_cs_shader_state {
struct si_textures_info { struct si_textures_info {
struct si_sampler_views views; struct si_sampler_views views;
uint64_t depth_texture_mask; /* which textures are depth */ uint32_t depth_texture_mask; /* which textures are depth */
uint64_t compressed_colortex_mask; uint32_t compressed_colortex_mask;
}; };
struct si_images_info { struct si_images_info {

4
src/gallium/drivers/radeonsi/si_state.h
View File

@ -197,10 +197,10 @@ struct si_descriptors {
unsigned ce_offset; unsigned ce_offset;
/* The i-th bit is set if that element is enabled (non-NULL resource). */ /* The i-th bit is set if that element is enabled (non-NULL resource). */
uint64_t enabled_mask; unsigned enabled_mask;
/* elements of the list that are changed and need to be uploaded */ /* elements of the list that are changed and need to be uploaded */
uint64_t dirty_mask; unsigned dirty_mask;
/* Whether the CE ram is dirty and needs to be reinitialized entirely /* Whether the CE ram is dirty and needs to be reinitialized entirely
* before we can do partial updates. */ * before we can do partial updates. */