r600: fix textureSize queries with tbos

piglit doesn't care, but I'm quite confident that the size actually bound
as range should be reported and not the base size of the resource (and
some quick piglit test hacking confirms this).
Also, the array in the constant buffer looks overallocated by a factor of 4.
For eg, also decrease the size by another factor of 2 by using the same
constant slot for both buffer size (required for txq for TBOs) and the number
of layers for cube arrays, as these are mutually exclusive. Could of course use
some more logic and only actually do this for the samplers/images/buffers where
it's required rather than for all, but ah well...

Reviewed-by: Dave Airlie <airlied@redhat.com>

diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index c9247e1c28618cca04c6c2bc71daa7bd0cab5039..e28882b2e5d7fee8ae4891d2134ce3deaf6d7b58 100644 (file)
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -6955,9 +6955,9 @@ static int r600_do_buffer_txq(struct r600_shader_ctx *ctx, int reg_idx, int offs
        alu.op = ALU_OP1_MOV;
        alu.src[0].sel = R600_SHADER_BUFFER_INFO_SEL;
        if (ctx->bc->chip_class >= EVERGREEN) {
-               /* channel 0 or 2 of each word */
-               alu.src[0].sel += (id / 2);
-               alu.src[0].chan = (id % 2) * 2;
+               /* with eg each dword is either buf size or number of cubes */
+               alu.src[0].sel += id / 4;
+               alu.src[0].chan = id % 4;
        } else {
                /* r600 we have them at channel 2 of the second dword */
                alu.src[0].sel += (id * 2) + 1;
@@ -7615,9 +7615,9 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
 
                alu.src[0].sel = R600_SHADER_BUFFER_INFO_SEL;
                if (ctx->bc->chip_class >= EVERGREEN) {
-                       /* channel 1 or 3 of each word */
-                       alu.src[0].sel += (id / 2);
-                       alu.src[0].chan = ((id % 2) * 2) + 1;
+                       /* with eg each dword is either buf size or number of cubes */
+                       alu.src[0].sel += id / 4;
+                       alu.src[0].chan = id % 4;
                } else {
                        /* r600 we have them at channel 2 of the second dword */
                        alu.src[0].sel += (id * 2) + 1;
@@ -8782,9 +8782,9 @@ static int tgsi_resq(struct r600_shader_ctx *ctx)
                alu.op = ALU_OP1_MOV;
 
                alu.src[0].sel = R600_SHADER_BUFFER_INFO_SEL;
-               /* channel 1 or 3 of each word */
-               alu.src[0].sel += (id / 2);
-               alu.src[0].chan = ((id % 2) * 2) + 1;
+               /* with eg each dword is either buf size or number of cubes */
+               alu.src[0].sel += id / 4;
+               alu.src[0].chan = id % 4;
                alu.src[0].kc_bank = R600_BUFFER_INFO_CONST_BUFFER;
                tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst);
                alu.last = 1;

diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index e5a5a333679e1a61e403c1aa29ced4eab689c17b..e7fa1bbf5761c511fb468eccb32074ce75da3d7c 100644 (file)
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -902,7 +902,6 @@ struct r600_pipe_shader_selector *r600_create_shader_state_tokens(struct pipe_co
                                                                  unsigned pipe_shader_type)
 {
        struct r600_pipe_shader_selector *sel = CALLOC_STRUCT(r600_pipe_shader_selector);
-       int i;
 
        sel->type = pipe_shader_type;
        sel->tokens = tgsi_dup_tokens(tokens);
@@ -1326,7 +1325,7 @@ static void r600_setup_buffer_constants(struct r600_context *rctx, int shader_ty
        samplers->views.dirty_buffer_constants = FALSE;
 
        bits = util_last_bit(samplers->views.enabled_mask);
-       array_size = bits * 8 * sizeof(uint32_t) * 4;
+       array_size = bits * 8 * sizeof(uint32_t);
 
        constants = r600_alloc_buf_consts(rctx, shader_type, array_size, &base_offset);
 
@@ -1349,15 +1348,16 @@ static void r600_setup_buffer_constants(struct r600_context *rctx, int shader_ty
                        } else
                                constants[offset + 4] = 0;
 
-                       constants[offset + 5] = samplers->views.views[i]->base.texture->width0 / util_format_get_blocksize(samplers->views.views[i]->base.format);
+                       constants[offset + 5] = samplers->views.views[i]->base.u.buf.size /
+                                           util_format_get_blocksize(samplers->views.views[i]->base.format);
                        constants[offset + 6] = samplers->views.views[i]->base.texture->array_size / 6;
                }
        }
 
 }
 
-/* On evergreen we store two values
- * 1. buffer size for TXQ
+/* On evergreen we store one value
+ * 1. buffer size for TXQ or
  * 2. number of cube layers in a cube map array.
  */
 void eg_setup_buffer_constants(struct r600_context *rctx, int shader_type)
@@ -1396,25 +1396,33 @@ void eg_setup_buffer_constants(struct r600_context *rctx, int shader_type)
        img_bits = bits;
        if (buffers)
                bits += util_last_bit(buffers->enabled_mask);
-       array_size = bits * 2 * sizeof(uint32_t) * 4;
+       array_size = bits * sizeof(uint32_t);
 
        constants = r600_alloc_buf_consts(rctx, shader_type, array_size,
                                          &base_offset);
 
        for (i = 0; i < sview_bits; i++) {
                if (samplers->views.enabled_mask & (1 << i)) {
-                       uint32_t offset = (base_offset / 4) + i * 2;
-                       constants[offset] = samplers->views.views[i]->base.texture->width0 / util_format_get_blocksize(samplers->views.views[i]->base.format);
-                       constants[offset + 1] = samplers->views.views[i]->base.texture->array_size / 6;
+                       uint32_t offset = (base_offset / 4) + i;
+                       if (samplers->views.views[i]->base.target == PIPE_BUFFER) {
+                               constants[offset] = samplers->views.views[i]->base.u.buf.size /
+                                               util_format_get_blocksize(samplers->views.views[i]->base.format);
+                       } else {
+                               constants[offset] = samplers->views.views[i]->base.texture->array_size / 6;
+                       }
                }
        }
        if (images) {
                for (i = sview_bits; i < img_bits; i++) {
                        int idx = i - sview_bits;
                        if (images->enabled_mask & (1 << idx)) {
-                               uint32_t offset = (base_offset / 4) + i * 2;
-                               constants[offset] = images->views[i].base.resource->width0 / util_format_get_blocksize(images->views[i].base.format);
-                               constants[offset + 1] = images->views[i].base.resource->array_size / 6;
+                               uint32_t offset = (base_offset / 4) + i;
+                               if (images->views[i].base.resource->target == PIPE_BUFFER) {
+                                       constants[offset] = images->views[i].base.u.buf.size /
+                                                       util_format_get_blocksize(images->views[i].base.format);
+                               } else {
+                                       constants[offset] = images->views[i].base.resource->array_size / 6;
+                               }
                        }
                }
        }
@@ -1422,9 +1430,10 @@ void eg_setup_buffer_constants(struct r600_context *rctx, int shader_type)
                for (i = img_bits; i < bits; i++) {
                        int idx = i - img_bits;
                        if (buffers->enabled_mask & (1 << idx)) {
-                               uint32_t offset = (base_offset / 4) + i * 2;
-                               constants[offset] = buffers->views[i].base.resource->width0 / util_format_get_blocksize(buffers->views[i].base.format);
-                               constants[offset + 1] = 0;
+                               uint32_t offset = (base_offset / 4) + i;
+                               assert(buffers->views[i].base.resource->target == PIPE_BUFFER);
+                               constants[offset] = buffers->views[i].base.u.buf.size /
+                                               util_format_get_blocksize(buffers->views[i].base.format);
                        }
                }
        }