X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fgpu%2Fdrm%2Fradeon%2Fr600_blit_kms.c;h=d13622ae74e9799d75045cd282d2797071a23d21;hb=0fcbe9473ac9c53463a61c9c83db8293bee15d12;hp=bbb0d615ac1c273e989c42880e7d86247ba7d33b;hpb=119e20dc149581db3064661b2e659f308f97b663;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/radeon/r600_blit_kms.c index bbb0d61..d13622a 100644 --- a/drivers/gpu/drm/radeon/r600_blit_kms.c +++ b/drivers/gpu/drm/radeon/r600_blit_kms.c @@ -25,7 +25,7 @@ set_render_target(struct radeon_device *rdev, int format, u32 cb_color_info; int pitch, slice; - h = (h + 7) & ~7; + h = ALIGN(h, 8); if (h < 8) h = 8; @@ -368,7 +368,7 @@ set_default_state(struct radeon_device *rdev) if ((rdev->family == CHIP_RV610) || (rdev->family == CHIP_RV620) || (rdev->family == CHIP_RS780) || - (rdev->family == CHIP_RS780) || + (rdev->family == CHIP_RS880) || (rdev->family == CHIP_RV710)) sq_config = 0; else @@ -396,15 +396,13 @@ set_default_state(struct radeon_device *rdev) NUM_ES_STACK_ENTRIES(num_es_stack_entries)); /* emit an IB pointing at default state */ - dwords = (rdev->r600_blit.state_len + 0xf) & ~0xf; + dwords = ALIGN(rdev->r600_blit.state_len, 0x10); gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset; radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); radeon_ring_write(rdev, gpu_addr & 0xFFFFFFFC); radeon_ring_write(rdev, upper_32_bits(gpu_addr) & 0xFF); radeon_ring_write(rdev, dwords); - radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0)); - radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT); /* SQ config */ radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 6)); radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); @@ -449,6 +447,10 @@ int r600_blit_init(struct radeon_device *rdev) u32 packet2s[16]; int num_packet2s = 0; + /* don't reinitialize blit */ + if (rdev->r600_blit.shader_obj) + return 0; + mutex_init(&rdev->r600_blit.mutex); rdev->r600_blit.state_offset = 0; if (rdev->family >= CHIP_RV770) @@ -473,31 +475,25 @@ int r600_blit_init(struct radeon_device *rdev) obj_size += r6xx_ps_size * 4; obj_size = ALIGN(obj_size, 256); - r = radeon_object_create(rdev, NULL, obj_size, - true, RADEON_GEM_DOMAIN_VRAM, - false, &rdev->r600_blit.shader_obj); + r = radeon_bo_create(rdev, NULL, obj_size, true, RADEON_GEM_DOMAIN_VRAM, + &rdev->r600_blit.shader_obj); if (r) { DRM_ERROR("r600 failed to allocate shader\n"); return r; } - r = radeon_object_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->r600_blit.shader_gpu_addr); - if (r) { - DRM_ERROR("failed to pin blit object %d\n", r); - return r; - } - - DRM_DEBUG("r6xx blit allocated bo @ 0x%16llx %08x vs %08x ps %08x\n", - rdev->r600_blit.shader_gpu_addr, obj_size, + DRM_DEBUG("r6xx blit allocated bo %08x vs %08x ps %08x\n", + obj_size, rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset); - r = radeon_object_kmap(rdev->r600_blit.shader_obj, &ptr); + r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); + if (unlikely(r != 0)) + return r; + r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr); if (r) { DRM_ERROR("failed to map blit object %d\n", r); return r; } - if (rdev->family >= CHIP_RV770) memcpy_toio(ptr + rdev->r600_blit.state_offset, r7xx_default_state, rdev->r600_blit.state_len * 4); @@ -507,19 +503,28 @@ int r600_blit_init(struct radeon_device *rdev) if (num_packet2s) memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4), packet2s, num_packet2s * 4); - - memcpy(ptr + rdev->r600_blit.vs_offset, r6xx_vs, r6xx_vs_size * 4); memcpy(ptr + rdev->r600_blit.ps_offset, r6xx_ps, r6xx_ps_size * 4); - - radeon_object_kunmap(rdev->r600_blit.shader_obj); + radeon_bo_kunmap(rdev->r600_blit.shader_obj); + radeon_bo_unreserve(rdev->r600_blit.shader_obj); return 0; } void r600_blit_fini(struct radeon_device *rdev) { - radeon_object_unpin(rdev->r600_blit.shader_obj); - radeon_object_unref(&rdev->r600_blit.shader_obj); + int r; + + if (rdev->r600_blit.shader_obj == NULL) + return; + /* If we can't reserve the bo, unref should be enough to destroy + * it when it becomes idle. + */ + r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); + if (!r) { + radeon_bo_unpin(rdev->r600_blit.shader_obj); + radeon_bo_unreserve(rdev->r600_blit.shader_obj); + } + radeon_bo_unref(&rdev->r600_blit.shader_obj); } int r600_vb_ib_get(struct radeon_device *rdev) @@ -538,37 +543,48 @@ int r600_vb_ib_get(struct radeon_device *rdev) void r600_vb_ib_put(struct radeon_device *rdev) { - mutex_lock(&rdev->ib_pool.mutex); radeon_fence_emit(rdev, rdev->r600_blit.vb_ib->fence); - list_add_tail(&rdev->r600_blit.vb_ib->list, &rdev->ib_pool.scheduled_ibs); - mutex_unlock(&rdev->ib_pool.mutex); radeon_ib_free(rdev, &rdev->r600_blit.vb_ib); } int r600_blit_prepare_copy(struct radeon_device *rdev, int size_bytes) { int r; - int ring_size; + int ring_size, line_size; int max_size; + /* loops of emits 64 + fence emit possible */ + int dwords_per_loop = 76, num_loops; + + r = r600_vb_ib_get(rdev); + if (r) + return r; + + /* set_render_target emits 2 extra dwords on rv6xx */ + if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) + dwords_per_loop += 2; /* 8 bpp vs 32 bpp for xfer unit */ if (size_bytes & 3) - max_size = 8192*8192; + line_size = 8192; else - max_size = 8192*8192*4; + line_size = 8192*4; - r = r600_vb_ib_get(rdev); - WARN_ON(r); + max_size = 8192 * line_size; - /* loops of emits 64 + fence emit possible */ - ring_size = ((size_bytes + max_size) / max_size) * 78; + /* major loops cover the max size transfer */ + num_loops = ((size_bytes + max_size) / max_size); + /* minor loops cover the extra non aligned bits */ + num_loops += ((size_bytes % line_size) ? 1 : 0); + /* calculate number of loops correctly */ + ring_size = num_loops * dwords_per_loop; /* set default + shaders */ ring_size += 40; /* shaders + def state */ - ring_size += 3; /* fence emit for VB IB */ + ring_size += 10; /* fence emit for VB IB */ ring_size += 5; /* done copy */ - ring_size += 3; /* fence emit for done copy */ + ring_size += 10; /* fence emit for done copy */ r = radeon_ring_lock(rdev, ring_size); - WARN_ON(r); + if (r) + return r; set_default_state(rdev); /* 14 */ set_shaders(rdev); /* 26 */ @@ -579,13 +595,6 @@ void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence) { int r; - radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0)); - radeon_ring_write(rdev, CACHE_FLUSH_AND_INV_EVENT); - /* wait for 3D idle clean */ - radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); - radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit); - if (rdev->r600_blit.vb_ib) r600_vb_ib_put(rdev); @@ -769,7 +778,7 @@ void r600_kms_blit_copy(struct radeon_device *rdev, /* dst 23 */ set_render_target(rdev, COLOR_8_8_8_8, - dst_x + cur_size, h, + (dst_x + cur_size) / 4, h, dst_gpu_addr); /* scissors 12 */