drm/radeon/kms: add drm blit support for evergreen
authorAlex Deucher <alexdeucher@gmail.com>
Thu, 9 Sep 2010 15:33:36 +0000 (11:33 -0400)
committerDave Airlie <airlied@redhat.com>
Wed, 6 Oct 2010 01:46:30 +0000 (11:46 +1000)
This patch implements blit support for bo moves using
the 3D engine.  It uses the same method as r6xx/r7xx:
- store the base state in an IB
- emit variable state and vertex buffers to do the blit

This allows the hw to move bos using the 3D engine and allows
full use of vram beyond the pci aperture size.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drivers/gpu/drm/radeon/Makefile
drivers/gpu/drm/radeon/evergreen.c
drivers/gpu/drm/radeon/evergreen_blit_kms.c [new file with mode: 0644]
drivers/gpu/drm/radeon/evergreen_blit_shaders.c [new file with mode: 0644]
drivers/gpu/drm/radeon/evergreen_blit_shaders.h [new file with mode: 0644]
drivers/gpu/drm/radeon/evergreend.h
drivers/gpu/drm/radeon/r600_blit_kms.c
drivers/gpu/drm/radeon/radeon.h
drivers/gpu/drm/radeon/radeon_asic.c
drivers/gpu/drm/radeon/radeon_asic.h

index aebe00875041448351bb129b7a131baab66f4e9e..6cae4f2028d27544467f2c2c3d605788b9f60df8 100644 (file)
@@ -65,7 +65,7 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \
        rs400.o rs600.o rs690.o rv515.o r520.o r600.o rv770.o radeon_test.o \
        r200.o radeon_legacy_tv.o r600_cs.o r600_blit.o r600_blit_shaders.o \
        r600_blit_kms.o radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \
-       evergreen.o evergreen_cs.o
+       evergreen.o evergreen_cs.o evergreen_blit_shaders.o evergreen_blit_kms.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
index 4c82cc8302714df9df381849df8a94ad76aa2c1b..aee61ae2440227eb51a01e6ff0302adddd2074ea 100644 (file)
@@ -2062,26 +2062,13 @@ static int evergreen_startup(struct radeon_device *rdev)
                        return r;
        }
        evergreen_gpu_init(rdev);
-#if 0
-       if (!rdev->r600_blit.shader_obj) {
-               r = r600_blit_init(rdev);
-               if (r) {
-                       DRM_ERROR("radeon: failed blitter (%d).\n", r);
-                       return r;
-               }
-       }
 
-       r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
-       if (unlikely(r != 0))
-               return r;
-       r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
-                       &rdev->r600_blit.shader_gpu_addr);
-       radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+       r = evergreen_blit_init(rdev);
        if (r) {
-               DRM_ERROR("failed to pin blit object %d\n", r);
-               return r;
+               evergreen_blit_fini(rdev);
+               rdev->asic->copy = NULL;
+               dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
        }
-#endif
 
        /* allocate wb buffer */
        r = radeon_wb_init(rdev);
@@ -2139,23 +2126,43 @@ int evergreen_resume(struct radeon_device *rdev)
 
 int evergreen_suspend(struct radeon_device *rdev)
 {
-#if 0
        int r;
-#endif
+
        /* FIXME: we should wait for ring to be empty */
        r700_cp_stop(rdev);
        rdev->cp.ready = false;
        evergreen_irq_suspend(rdev);
        radeon_wb_disable(rdev);
        evergreen_pcie_gart_disable(rdev);
-#if 0
+
        /* unpin shaders bo */
        r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
        if (likely(r == 0)) {
                radeon_bo_unpin(rdev->r600_blit.shader_obj);
                radeon_bo_unreserve(rdev->r600_blit.shader_obj);
        }
-#endif
+
+       return 0;
+}
+
+int evergreen_copy_blit(struct radeon_device *rdev,
+                       uint64_t src_offset, uint64_t dst_offset,
+                       unsigned num_pages, struct radeon_fence *fence)
+{
+       int r;
+
+       mutex_lock(&rdev->r600_blit.mutex);
+       rdev->r600_blit.vb_ib = NULL;
+       r = evergreen_blit_prepare_copy(rdev, num_pages * RADEON_GPU_PAGE_SIZE);
+       if (r) {
+               if (rdev->r600_blit.vb_ib)
+                       radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
+               mutex_unlock(&rdev->r600_blit.mutex);
+               return r;
+       }
+       evergreen_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * RADEON_GPU_PAGE_SIZE);
+       evergreen_blit_done_copy(rdev, fence);
+       mutex_unlock(&rdev->r600_blit.mutex);
        return 0;
 }
 
@@ -2286,7 +2293,7 @@ int evergreen_init(struct radeon_device *rdev)
 
 void evergreen_fini(struct radeon_device *rdev)
 {
-       /*r600_blit_fini(rdev);*/
+       evergreen_blit_fini(rdev);
        r700_cp_fini(rdev);
        r600_irq_fini(rdev);
        radeon_wb_fini(rdev);
diff --git a/drivers/gpu/drm/radeon/evergreen_blit_kms.c b/drivers/gpu/drm/radeon/evergreen_blit_kms.c
new file mode 100644 (file)
index 0000000..ce1ae4a
--- /dev/null
@@ -0,0 +1,776 @@
+/*
+ * Copyright 2010 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *     Alex Deucher <alexander.deucher@amd.com>
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "radeon_drm.h"
+#include "radeon.h"
+
+#include "evergreend.h"
+#include "evergreen_blit_shaders.h"
+
+#define DI_PT_RECTLIST        0x11
+#define DI_INDEX_SIZE_16_BIT  0x0
+#define DI_SRC_SEL_AUTO_INDEX 0x2
+
+#define FMT_8                 0x1
+#define FMT_5_6_5             0x8
+#define FMT_8_8_8_8           0x1a
+#define COLOR_8               0x1
+#define COLOR_5_6_5           0x8
+#define COLOR_8_8_8_8         0x1a
+
+/* emits 17 */
+static void
+set_render_target(struct radeon_device *rdev, int format,
+                 int w, int h, u64 gpu_addr)
+{
+       u32 cb_color_info;
+       int pitch, slice;
+
+       h = ALIGN(h, 8);
+       if (h < 8)
+               h = 8;
+
+       cb_color_info = ((format << 2) | (1 << 24));
+       pitch = (w / 8) - 1;
+       slice = ((w * h) / 64) - 1;
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 15));
+       radeon_ring_write(rdev, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, gpu_addr >> 8);
+       radeon_ring_write(rdev, pitch);
+       radeon_ring_write(rdev, slice);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, cb_color_info);
+       radeon_ring_write(rdev, (1 << 4));
+       radeon_ring_write(rdev, (w - 1) | ((h - 1) << 16));
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+}
+
+/* emits 5dw */
+static void
+cp_set_surface_sync(struct radeon_device *rdev,
+                   u32 sync_type, u32 size,
+                   u64 mc_addr)
+{
+       u32 cp_coher_size;
+
+       if (size == 0xffffffff)
+               cp_coher_size = 0xffffffff;
+       else
+               cp_coher_size = ((size + 255) >> 8);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SURFACE_SYNC, 3));
+       radeon_ring_write(rdev, sync_type);
+       radeon_ring_write(rdev, cp_coher_size);
+       radeon_ring_write(rdev, mc_addr >> 8);
+       radeon_ring_write(rdev, 10); /* poll interval */
+}
+
+/* emits 11dw + 1 surface sync = 16dw */
+static void
+set_shaders(struct radeon_device *rdev)
+{
+       u64 gpu_addr;
+
+       /* VS */
+       gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 3));
+       radeon_ring_write(rdev, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, gpu_addr >> 8);
+       radeon_ring_write(rdev, 2);
+       radeon_ring_write(rdev, 0);
+
+       /* PS */
+       gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 4));
+       radeon_ring_write(rdev, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, gpu_addr >> 8);
+       radeon_ring_write(rdev, 1);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 2);
+
+       gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+       cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
+}
+
+/* emits 10 + 1 sync (5) = 15 */
+static void
+set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
+{
+       u32 sq_vtx_constant_word2, sq_vtx_constant_word3;
+
+       /* high addr, stride */
+       sq_vtx_constant_word2 = ((upper_32_bits(gpu_addr) & 0xff) | (16 << 8));
+       /* xyzw swizzles */
+       sq_vtx_constant_word3 = (0 << 3) | (1 << 6) | (2 << 9) | (3 << 12);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 8));
+       radeon_ring_write(rdev, 0x580);
+       radeon_ring_write(rdev, gpu_addr & 0xffffffff);
+       radeon_ring_write(rdev, 48 - 1); /* size */
+       radeon_ring_write(rdev, sq_vtx_constant_word2);
+       radeon_ring_write(rdev, sq_vtx_constant_word3);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, SQ_TEX_VTX_VALID_BUFFER << 30);
+
+       if (rdev->family == CHIP_CEDAR)
+               cp_set_surface_sync(rdev,
+                                   PACKET3_TC_ACTION_ENA, 48, gpu_addr);
+       else
+               cp_set_surface_sync(rdev,
+                                   PACKET3_VC_ACTION_ENA, 48, gpu_addr);
+
+}
+
+/* emits 10 */
+static void
+set_tex_resource(struct radeon_device *rdev,
+                int format, int w, int h, int pitch,
+                u64 gpu_addr)
+{
+       u32 sq_tex_resource_word0, sq_tex_resource_word1;
+       u32 sq_tex_resource_word4, sq_tex_resource_word7;
+
+       if (h < 1)
+               h = 1;
+
+       sq_tex_resource_word0 = (1 << 0); /* 2D */
+       sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 6) |
+                                 ((w - 1) << 18));
+       sq_tex_resource_word1 = ((h - 1) << 0);
+       /* xyzw swizzles */
+       sq_tex_resource_word4 = (0 << 16) | (1 << 19) | (2 << 22) | (3 << 25);
+
+       sq_tex_resource_word7 = format | (SQ_TEX_VTX_VALID_TEXTURE << 30);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_RESOURCE, 8));
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, sq_tex_resource_word0);
+       radeon_ring_write(rdev, sq_tex_resource_word1);
+       radeon_ring_write(rdev, gpu_addr >> 8);
+       radeon_ring_write(rdev, gpu_addr >> 8);
+       radeon_ring_write(rdev, sq_tex_resource_word4);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, sq_tex_resource_word7);
+}
+
+/* emits 12 */
+static void
+set_scissors(struct radeon_device *rdev, int x1, int y1,
+            int x2, int y2)
+{
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+       radeon_ring_write(rdev, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, (x1 << 0) | (y1 << 16));
+       radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+       radeon_ring_write(rdev, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
+       radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+       radeon_ring_write(rdev, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+       radeon_ring_write(rdev, (x1 << 0) | (y1 << 16) | (1 << 31));
+       radeon_ring_write(rdev, (x2 << 0) | (y2 << 16));
+}
+
+/* emits 10 */
+static void
+draw_auto(struct radeon_device *rdev)
+{
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+       radeon_ring_write(rdev, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_START) >> 2);
+       radeon_ring_write(rdev, DI_PT_RECTLIST);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_INDEX_TYPE, 0));
+       radeon_ring_write(rdev, DI_INDEX_SIZE_16_BIT);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_NUM_INSTANCES, 0));
+       radeon_ring_write(rdev, 1);
+
+       radeon_ring_write(rdev, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
+       radeon_ring_write(rdev, 3);
+       radeon_ring_write(rdev, DI_SRC_SEL_AUTO_INDEX);
+
+}
+
+/* emits 20 */
+static void
+set_default_state(struct radeon_device *rdev)
+{
+       u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2, sq_gpr_resource_mgmt_3;
+       u32 sq_thread_resource_mgmt, sq_thread_resource_mgmt_2;
+       u32 sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2, sq_stack_resource_mgmt_3;
+       int num_ps_gprs, num_vs_gprs, num_temp_gprs;
+       int num_gs_gprs, num_es_gprs, num_hs_gprs, num_ls_gprs;
+       int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads;
+       int num_hs_threads, num_ls_threads;
+       int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries;
+       int num_hs_stack_entries, num_ls_stack_entries;
+       u64 gpu_addr;
+       int dwords;
+
+       switch (rdev->family) {
+       case CHIP_CEDAR:
+       default:
+               num_ps_gprs = 93;
+               num_vs_gprs = 46;
+               num_temp_gprs = 4;
+               num_gs_gprs = 31;
+               num_es_gprs = 31;
+               num_hs_gprs = 23;
+               num_ls_gprs = 23;
+               num_ps_threads = 96;
+               num_vs_threads = 16;
+               num_gs_threads = 16;
+               num_es_threads = 16;
+               num_hs_threads = 16;
+               num_ls_threads = 16;
+               num_ps_stack_entries = 42;
+               num_vs_stack_entries = 42;
+               num_gs_stack_entries = 42;
+               num_es_stack_entries = 42;
+               num_hs_stack_entries = 42;
+               num_ls_stack_entries = 42;
+               break;
+       case CHIP_REDWOOD:
+               num_ps_gprs = 93;
+               num_vs_gprs = 46;
+               num_temp_gprs = 4;
+               num_gs_gprs = 31;
+               num_es_gprs = 31;
+               num_hs_gprs = 23;
+               num_ls_gprs = 23;
+               num_ps_threads = 128;
+               num_vs_threads = 20;
+               num_gs_threads = 20;
+               num_es_threads = 20;
+               num_hs_threads = 20;
+               num_ls_threads = 20;
+               num_ps_stack_entries = 42;
+               num_vs_stack_entries = 42;
+               num_gs_stack_entries = 42;
+               num_es_stack_entries = 42;
+               num_hs_stack_entries = 42;
+               num_ls_stack_entries = 42;
+               break;
+       case CHIP_JUNIPER:
+               num_ps_gprs = 93;
+               num_vs_gprs = 46;
+               num_temp_gprs = 4;
+               num_gs_gprs = 31;
+               num_es_gprs = 31;
+               num_hs_gprs = 23;
+               num_ls_gprs = 23;
+               num_ps_threads = 128;
+               num_vs_threads = 20;
+               num_gs_threads = 20;
+               num_es_threads = 20;
+               num_hs_threads = 20;
+               num_ls_threads = 20;
+               num_ps_stack_entries = 85;
+               num_vs_stack_entries = 85;
+               num_gs_stack_entries = 85;
+               num_es_stack_entries = 85;
+               num_hs_stack_entries = 85;
+               num_ls_stack_entries = 85;
+               break;
+       case CHIP_CYPRESS:
+       case CHIP_HEMLOCK:
+               num_ps_gprs = 93;
+               num_vs_gprs = 46;
+               num_temp_gprs = 4;
+               num_gs_gprs = 31;
+               num_es_gprs = 31;
+               num_hs_gprs = 23;
+               num_ls_gprs = 23;
+               num_ps_threads = 128;
+               num_vs_threads = 20;
+               num_gs_threads = 20;
+               num_es_threads = 20;
+               num_hs_threads = 20;
+               num_ls_threads = 20;
+               num_ps_stack_entries = 85;
+               num_vs_stack_entries = 85;
+               num_gs_stack_entries = 85;
+               num_es_stack_entries = 85;
+               num_hs_stack_entries = 85;
+               num_ls_stack_entries = 85;
+               break;
+       }
+
+       if (rdev->family == CHIP_CEDAR)
+               sq_config = 0;
+       else
+               sq_config = VC_ENABLE;
+
+       sq_config |= (EXPORT_SRC_C |
+                     CS_PRIO(0) |
+                     LS_PRIO(0) |
+                     HS_PRIO(0) |
+                     PS_PRIO(0) |
+                     VS_PRIO(1) |
+                     GS_PRIO(2) |
+                     ES_PRIO(3));
+
+       sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) |
+                                 NUM_VS_GPRS(num_vs_gprs) |
+                                 NUM_CLAUSE_TEMP_GPRS(num_temp_gprs));
+       sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) |
+                                 NUM_ES_GPRS(num_es_gprs));
+       sq_gpr_resource_mgmt_3 = (NUM_HS_GPRS(num_hs_gprs) |
+                                 NUM_LS_GPRS(num_ls_gprs));
+       sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) |
+                                  NUM_VS_THREADS(num_vs_threads) |
+                                  NUM_GS_THREADS(num_gs_threads) |
+                                  NUM_ES_THREADS(num_es_threads));
+       sq_thread_resource_mgmt_2 = (NUM_HS_THREADS(num_hs_threads) |
+                                    NUM_LS_THREADS(num_ls_threads));
+       sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) |
+                                   NUM_VS_STACK_ENTRIES(num_vs_stack_entries));
+       sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) |
+                                   NUM_ES_STACK_ENTRIES(num_es_stack_entries));
+       sq_stack_resource_mgmt_3 = (NUM_HS_STACK_ENTRIES(num_hs_stack_entries) |
+                                   NUM_LS_STACK_ENTRIES(num_ls_stack_entries));
+
+       /* emit an IB pointing at default state */
+       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);
+
+       /* disable dyn gprs */
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+       radeon_ring_write(rdev, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ - PACKET3_SET_CONFIG_REG_START) >> 2);
+       radeon_ring_write(rdev, 0);
+
+       /* SQ config */
+       radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 11));
+       radeon_ring_write(rdev, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_START) >> 2);
+       radeon_ring_write(rdev, sq_config);
+       radeon_ring_write(rdev, sq_gpr_resource_mgmt_1);
+       radeon_ring_write(rdev, sq_gpr_resource_mgmt_2);
+       radeon_ring_write(rdev, sq_gpr_resource_mgmt_3);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, 0);
+       radeon_ring_write(rdev, sq_thread_resource_mgmt);
+       radeon_ring_write(rdev, sq_thread_resource_mgmt_2);
+       radeon_ring_write(rdev, sq_stack_resource_mgmt_1);
+       radeon_ring_write(rdev, sq_stack_resource_mgmt_2);
+       radeon_ring_write(rdev, sq_stack_resource_mgmt_3);
+}
+
+static inline uint32_t i2f(uint32_t input)
+{
+       u32 result, i, exponent, fraction;
+
+       if ((input & 0x3fff) == 0)
+               result = 0; /* 0 is a special case */
+       else {
+               exponent = 140; /* exponent biased by 127; */
+               fraction = (input & 0x3fff) << 10; /* cheat and only
+                                                     handle numbers below 2^^15 */
+               for (i = 0; i < 14; i++) {
+                       if (fraction & 0x800000)
+                               break;
+                       else {
+                               fraction = fraction << 1; /* keep
+                                                            shifting left until top bit = 1 */
+                               exponent = exponent - 1;
+                       }
+               }
+               result = exponent << 23 | (fraction & 0x7fffff); /* mask
+                                                                   off top bit; assumed 1 */
+       }
+       return result;
+}
+
+int evergreen_blit_init(struct radeon_device *rdev)
+{
+       u32 obj_size;
+       int r, dwords;
+       void *ptr;
+       u32 packet2s[16];
+       int num_packet2s = 0;
+
+       /* pin copy shader into vram if already initialized */
+       if (rdev->r600_blit.shader_obj)
+               goto done;
+
+       mutex_init(&rdev->r600_blit.mutex);
+       rdev->r600_blit.state_offset = 0;
+
+       rdev->r600_blit.state_len = evergreen_default_size;
+
+       dwords = rdev->r600_blit.state_len;
+       while (dwords & 0xf) {
+               packet2s[num_packet2s++] = PACKET2(0);
+               dwords++;
+       }
+
+       obj_size = dwords * 4;
+       obj_size = ALIGN(obj_size, 256);
+
+       rdev->r600_blit.vs_offset = obj_size;
+       obj_size += evergreen_vs_size * 4;
+       obj_size = ALIGN(obj_size, 256);
+
+       rdev->r600_blit.ps_offset = obj_size;
+       obj_size += evergreen_ps_size * 4;
+       obj_size = ALIGN(obj_size, 256);
+
+       r = radeon_bo_create(rdev, NULL, obj_size, true, RADEON_GEM_DOMAIN_VRAM,
+                               &rdev->r600_blit.shader_obj);
+       if (r) {
+               DRM_ERROR("evergreen failed to allocate shader\n");
+               return r;
+       }
+
+       DRM_DEBUG("evergreen blit allocated bo %08x vs %08x ps %08x\n",
+                 obj_size,
+                 rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset);
+
+       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;
+       }
+
+       memcpy_toio(ptr + rdev->r600_blit.state_offset,
+                   evergreen_default_state, rdev->r600_blit.state_len * 4);
+
+       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, evergreen_vs, evergreen_vs_size * 4);
+       memcpy(ptr + rdev->r600_blit.ps_offset, evergreen_ps, evergreen_ps_size * 4);
+       radeon_bo_kunmap(rdev->r600_blit.shader_obj);
+       radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+
+done:
+       r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+       if (unlikely(r != 0))
+               return r;
+       r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+                         &rdev->r600_blit.shader_gpu_addr);
+       radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+       if (r) {
+               dev_err(rdev->dev, "(%d) pin blit object failed\n", r);
+               return r;
+       }
+       return 0;
+}
+
+void evergreen_blit_fini(struct radeon_device *rdev)
+{
+       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);
+}
+
+static int evergreen_vb_ib_get(struct radeon_device *rdev)
+{
+       int r;
+       r = radeon_ib_get(rdev, &rdev->r600_blit.vb_ib);
+       if (r) {
+               DRM_ERROR("failed to get IB for vertex buffer\n");
+               return r;
+       }
+
+       rdev->r600_blit.vb_total = 64*1024;
+       rdev->r600_blit.vb_used = 0;
+       return 0;
+}
+
+static void evergreen_vb_ib_put(struct radeon_device *rdev)
+{
+       radeon_fence_emit(rdev, rdev->r600_blit.vb_ib->fence);
+       radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
+}
+
+int evergreen_blit_prepare_copy(struct radeon_device *rdev, int size_bytes)
+{
+       int r;
+       int ring_size, line_size;
+       int max_size;
+       /* loops of emits + fence emit possible */
+       int dwords_per_loop = 74, num_loops;
+
+       r = evergreen_vb_ib_get(rdev);
+       if (r)
+               return r;
+
+       /* 8 bpp vs 32 bpp for xfer unit */
+       if (size_bytes & 3)
+               line_size = 8192;
+       else
+               line_size = 8192 * 4;
+
+       max_size = 8192 * line_size;
+
+       /* 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 += 36; /* shaders + def state */
+       ring_size += 10; /* fence emit for VB IB */
+       ring_size += 5; /* done copy */
+       ring_size += 10; /* fence emit for done copy */
+       r = radeon_ring_lock(rdev, ring_size);
+       if (r)
+               return r;
+
+       set_default_state(rdev); /* 20 */
+       set_shaders(rdev); /* 16 */
+       return 0;
+}
+
+void evergreen_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence)
+{
+       int r;
+
+       if (rdev->r600_blit.vb_ib)
+               evergreen_vb_ib_put(rdev);
+
+       if (fence)
+               r = radeon_fence_emit(rdev, fence);
+
+       radeon_ring_unlock_commit(rdev);
+}
+
+void evergreen_kms_blit_copy(struct radeon_device *rdev,
+                            u64 src_gpu_addr, u64 dst_gpu_addr,
+                            int size_bytes)
+{
+       int max_bytes;
+       u64 vb_gpu_addr;
+       u32 *vb;
+
+       DRM_DEBUG("emitting copy %16llx %16llx %d %d\n", src_gpu_addr, dst_gpu_addr,
+                 size_bytes, rdev->r600_blit.vb_used);
+       vb = (u32 *)(rdev->r600_blit.vb_ib->ptr + rdev->r600_blit.vb_used);
+       if ((size_bytes & 3) || (src_gpu_addr & 3) || (dst_gpu_addr & 3)) {
+               max_bytes = 8192;
+
+               while (size_bytes) {
+                       int cur_size = size_bytes;
+                       int src_x = src_gpu_addr & 255;
+                       int dst_x = dst_gpu_addr & 255;
+                       int h = 1;
+                       src_gpu_addr = src_gpu_addr & ~255;
+                       dst_gpu_addr = dst_gpu_addr & ~255;
+
+                       if (!src_x && !dst_x) {
+                               h = (cur_size / max_bytes);
+                               if (h > 8192)
+                                       h = 8192;
+                               if (h == 0)
+                                       h = 1;
+                               else
+                                       cur_size = max_bytes;
+                       } else {
+                               if (cur_size > max_bytes)
+                                       cur_size = max_bytes;
+                               if (cur_size > (max_bytes - dst_x))
+                                       cur_size = (max_bytes - dst_x);
+                               if (cur_size > (max_bytes - src_x))
+                                       cur_size = (max_bytes - src_x);
+                       }
+
+                       if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
+                               WARN_ON(1);
+                       }
+
+                       vb[0] = i2f(dst_x);
+                       vb[1] = 0;
+                       vb[2] = i2f(src_x);
+                       vb[3] = 0;
+
+                       vb[4] = i2f(dst_x);
+                       vb[5] = i2f(h);
+                       vb[6] = i2f(src_x);
+                       vb[7] = i2f(h);
+
+                       vb[8] = i2f(dst_x + cur_size);
+                       vb[9] = i2f(h);
+                       vb[10] = i2f(src_x + cur_size);
+                       vb[11] = i2f(h);
+
+                       /* src 10 */
+                       set_tex_resource(rdev, FMT_8,
+                                        src_x + cur_size, h, src_x + cur_size,
+                                        src_gpu_addr);
+
+                       /* 5 */
+                       cp_set_surface_sync(rdev,
+                                           PACKET3_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
+
+
+                       /* dst 17 */
+                       set_render_target(rdev, COLOR_8,
+                                         dst_x + cur_size, h,
+                                         dst_gpu_addr);
+
+                       /* scissors 12 */
+                       set_scissors(rdev, dst_x, 0, dst_x + cur_size, h);
+
+                       /* 15 */
+                       vb_gpu_addr = rdev->r600_blit.vb_ib->gpu_addr + rdev->r600_blit.vb_used;
+                       set_vtx_resource(rdev, vb_gpu_addr);
+
+                       /* draw 10 */
+                       draw_auto(rdev);
+
+                       /* 5 */
+                       cp_set_surface_sync(rdev,
+                                           PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA,
+                                           cur_size * h, dst_gpu_addr);
+
+                       vb += 12;
+                       rdev->r600_blit.vb_used += 12 * 4;
+
+                       src_gpu_addr += cur_size * h;
+                       dst_gpu_addr += cur_size * h;
+                       size_bytes -= cur_size * h;
+               }
+       } else {
+               max_bytes = 8192 * 4;
+
+               while (size_bytes) {
+                       int cur_size = size_bytes;
+                       int src_x = (src_gpu_addr & 255);
+                       int dst_x = (dst_gpu_addr & 255);
+                       int h = 1;
+                       src_gpu_addr = src_gpu_addr & ~255;
+                       dst_gpu_addr = dst_gpu_addr & ~255;
+
+                       if (!src_x && !dst_x) {
+                               h = (cur_size / max_bytes);
+                               if (h > 8192)
+                                       h = 8192;
+                               if (h == 0)
+                                       h = 1;
+                               else
+                                       cur_size = max_bytes;
+                       } else {
+                               if (cur_size > max_bytes)
+                                       cur_size = max_bytes;
+                               if (cur_size > (max_bytes - dst_x))
+                                       cur_size = (max_bytes - dst_x);
+                               if (cur_size > (max_bytes - src_x))
+                                       cur_size = (max_bytes - src_x);
+                       }
+
+                       if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
+                               WARN_ON(1);
+                       }
+
+                       vb[0] = i2f(dst_x / 4);
+                       vb[1] = 0;
+                       vb[2] = i2f(src_x / 4);
+                       vb[3] = 0;
+
+                       vb[4] = i2f(dst_x / 4);
+                       vb[5] = i2f(h);
+                       vb[6] = i2f(src_x / 4);
+                       vb[7] = i2f(h);
+
+                       vb[8] = i2f((dst_x + cur_size) / 4);
+                       vb[9] = i2f(h);
+                       vb[10] = i2f((src_x + cur_size) / 4);
+                       vb[11] = i2f(h);
+
+                       /* src 10 */
+                       set_tex_resource(rdev, FMT_8_8_8_8,
+                                        (src_x + cur_size) / 4,
+                                        h, (src_x + cur_size) / 4,
+                                        src_gpu_addr);
+                       /* 5 */
+                       cp_set_surface_sync(rdev,
+                                           PACKET3_TC_ACTION_ENA, (src_x + cur_size * h), src_gpu_addr);
+
+                       /* dst 17 */
+                       set_render_target(rdev, COLOR_8_8_8_8,
+                                         (dst_x + cur_size) / 4, h,
+                                         dst_gpu_addr);
+
+                       /* scissors 12  */
+                       set_scissors(rdev, (dst_x / 4), 0, (dst_x + cur_size / 4), h);
+
+                       /* Vertex buffer setup 15 */
+                       vb_gpu_addr = rdev->r600_blit.vb_ib->gpu_addr + rdev->r600_blit.vb_used;
+                       set_vtx_resource(rdev, vb_gpu_addr);
+
+                       /* draw 10 */
+                       draw_auto(rdev);
+
+                       /* 5 */
+                       cp_set_surface_sync(rdev,
+                                           PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA,
+                                           cur_size * h, dst_gpu_addr);
+
+                       /* 74 ring dwords per loop */
+                       vb += 12;
+                       rdev->r600_blit.vb_used += 12 * 4;
+
+                       src_gpu_addr += cur_size * h;
+                       dst_gpu_addr += cur_size * h;
+                       size_bytes -= cur_size * h;
+               }
+       }
+}
+
diff --git a/drivers/gpu/drm/radeon/evergreen_blit_shaders.c b/drivers/gpu/drm/radeon/evergreen_blit_shaders.c
new file mode 100644 (file)
index 0000000..5d50450
--- /dev/null
@@ -0,0 +1,359 @@
+/*
+ * Copyright 2010 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *     Alex Deucher <alexander.deucher@amd.com>
+ */
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+
+/*
+ * evergreen cards need to use the 3D engine to blit data which requires
+ * quite a bit of hw state setup.  Rather than pull the whole 3D driver
+ * (which normally generates the 3D state) into the DRM, we opt to use
+ * statically generated state tables.  The regsiter state and shaders
+ * were hand generated to support blitting functionality.  See the 3D
+ * driver or documentation for descriptions of the registers and
+ * shader instructions.
+ */
+
+const u32 evergreen_default_state[] =
+{
+       0xc0012800, /* CONTEXT_CONTROL */
+       0x80000000,
+       0x80000000,
+
+       0xc0016900,
+       0x0000023b,
+       0x00000000, /* SQ_LDS_ALLOC_PS */
+
+       0xc0066900,
+       0x00000240,
+       0x00000000, /* SQ_ESGS_RING_ITEMSIZE */
+       0x00000000,
+       0x00000000,
+       0x00000000,
+       0x00000000,
+       0x00000000,
+
+       0xc0046900,
+       0x00000247,
+       0x00000000, /* SQ_GS_VERT_ITEMSIZE */
+       0x00000000,
+       0x00000000,
+       0x00000000,
+
+       0xc0026f00,
+       0x00000000,
+       0x00000000, /* SQ_VTX_BASE_VTX_LOC */
+       0x00000000,
+
+       0xc0026900,
+       0x00000010,
+       0x00000000, /* DB_Z_INFO */
+       0x00000000, /* DB_STENCIL_INFO */
+
+
+       0xc0016900,
+       0x00000200,
+       0x00000000, /* DB_DEPTH_CONTROL */
+
+       0xc0066900,
+       0x00000000,
+       0x00000060, /* DB_RENDER_CONTROL */
+       0x00000000, /* DB_COUNT_CONTROL */
+       0x00000000, /* DB_DEPTH_VIEW */
+       0x0000002a, /* DB_RENDER_OVERRIDE */
+       0x00000000, /* DB_RENDER_OVERRIDE2 */
+       0x00000000, /* DB_HTILE_DATA_BASE */
+
+       0xc0026900,
+       0x0000000a,
+       0x00000000, /* DB_STENCIL_CLEAR */
+       0x00000000, /* DB_DEPTH_CLEAR */
+
+       0xc0016900,
+       0x000002dc,
+       0x0000aa00, /* DB_ALPHA_TO_MASK */
+
+       0xc0016900,
+       0x00000080,
+       0x00000000, /* PA_SC_WINDOW_OFFSET */
+
+       0xc00d6900,
+       0x00000083,
+       0x0000ffff, /* PA_SC_CLIPRECT_RULE */
+       0x00000000, /* PA_SC_CLIPRECT_0_TL */
+       0x20002000, /* PA_SC_CLIPRECT_0_BR */
+       0x00000000,
+       0x20002000,
+       0x00000000,
+       0x20002000,
+       0x00000000,
+       0x20002000,
+       0xaaaaaaaa, /* PA_SC_EDGERULE */
+       0x00000000, /* PA_SU_HARDWARE_SCREEN_OFFSET */
+       0x0000000f, /* CB_TARGET_MASK */
+       0x0000000f, /* CB_SHADER_MASK */
+
+       0xc0226900,
+       0x00000094,
+       0x80000000, /* PA_SC_VPORT_SCISSOR_0_TL */
+       0x20002000, /* PA_SC_VPORT_SCISSOR_0_BR */
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x80000000,
+       0x20002000,
+       0x00000000, /* PA_SC_VPORT_ZMIN_0 */
+       0x3f800000, /* PA_SC_VPORT_ZMAX_0 */
+
+       0xc0016900,
+       0x000000d4,
+       0x00000000, /* SX_MISC */
+
+       0xc0026900,
+       0x00000292,
+       0x00000000, /* PA_SC_MODE_CNTL_0 */
+       0x00000000, /* PA_SC_MODE_CNTL_1 */
+
+       0xc0106900,
+       0x00000300,
+       0x00000000, /* PA_SC_LINE_CNTL */
+       0x00000000, /* PA_SC_AA_CONFIG */
+       0x00000005, /* PA_SU_VTX_CNTL */
+       0x3f800000, /* PA_CL_GB_VERT_CLIP_ADJ */
+       0x3f800000, /* PA_CL_GB_VERT_DISC_ADJ */
+       0x3f800000, /* PA_CL_GB_HORZ_CLIP_ADJ */
+       0x3f800000, /* PA_CL_GB_HORZ_DISC_ADJ */
+       0x00000000, /* PA_SC_AA_SAMPLE_LOCS_0 */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /* PA_SC_AA_SAMPLE_LOCS_7 */
+       0xffffffff, /* PA_SC_AA_MASK */
+
+       0xc00d6900,
+       0x00000202,
+       0x00cc0010, /* CB_COLOR_CONTROL */
+       0x00000210, /* DB_SHADER_CONTROL */
+       0x00010000, /* PA_CL_CLIP_CNTL */
+       0x00000004, /* PA_SU_SC_MODE_CNTL */
+       0x00000100, /* PA_CL_VTE_CNTL */
+       0x00000000, /* PA_CL_VS_OUT_CNTL */
+       0x00000000, /* PA_CL_NANINF_CNTL */
+       0x00000000, /* PA_SU_LINE_STIPPLE_CNTL */
+       0x00000000, /* PA_SU_LINE_STIPPLE_SCALE */
+       0x00000000, /* PA_SU_PRIM_FILTER_CNTL */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /* SQ_DYN_GPR_RESOURCE_LIMIT_1 */
+
+       0xc0066900,
+       0x000002de,
+       0x00000000, /* PA_SU_POLY_OFFSET_DB_FMT_CNTL */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+
+       0xc0016900,
+       0x00000229,
+       0x00000000, /* SQ_PGM_START_FS */
+
+       0xc0016900,
+       0x0000022a,
+       0x00000000, /* SQ_PGM_RESOURCES_FS */
+
+       0xc0096900,
+       0x00000100,
+       0x00ffffff, /* VGT_MAX_VTX_INDX */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /* SX_ALPHA_TEST_CONTROL */
+       0x00000000, /* CB_BLEND_RED */
+       0x00000000, /* CB_BLEND_GREEN */
+       0x00000000, /* CB_BLEND_BLUE */
+       0x00000000, /* CB_BLEND_ALPHA */
+
+       0xc0026900,
+       0x000002a8,
+       0x00000000, /* VGT_INSTANCE_STEP_RATE_0 */
+       0x00000000, /*  */
+
+       0xc0026900,
+       0x000002ad,
+       0x00000000, /* VGT_REUSE_OFF */
+       0x00000000, /*  */
+
+       0xc0116900,
+       0x00000280,
+       0x00000000, /* PA_SU_POINT_SIZE */
+       0x00000000, /* PA_SU_POINT_MINMAX */
+       0x00000008, /* PA_SU_LINE_CNTL */
+       0x00000000, /* PA_SC_LINE_STIPPLE */
+       0x00000000, /* VGT_OUTPUT_PATH_CNTL */
+       0x00000000, /* VGT_HOS_CNTL */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /* VGT_GS_MODE */
+
+       0xc0016900,
+       0x000002a1,
+       0x00000000, /* VGT_PRIMITIVEID_EN */
+
+       0xc0016900,
+       0x000002a5,
+       0x00000000, /* VGT_MULTI_PRIM_IB_RESET_EN */
+
+       0xc0016900,
+       0x000002d5,
+       0x00000000, /* VGT_SHADER_STAGES_EN */
+
+       0xc0026900,
+       0x000002e5,
+       0x00000000, /* VGT_STRMOUT_CONFIG */
+       0x00000000, /*  */
+
+       0xc0016900,
+       0x000001e0,
+       0x00000000, /* CB_BLEND0_CONTROL */
+
+       0xc0016900,
+       0x000001b1,
+       0x00000000, /* SPI_VS_OUT_CONFIG */
+
+       0xc0016900,
+       0x00000187,
+       0x00000000, /* SPI_VS_OUT_ID_0 */
+
+       0xc0016900,
+       0x00000191,
+       0x00000100, /* SPI_PS_INPUT_CNTL_0 */
+
+       0xc00b6900,
+       0x000001b3,
+       0x20000001, /* SPI_PS_IN_CONTROL_0 */
+       0x00000000, /* SPI_PS_IN_CONTROL_1 */
+       0x00000000, /* SPI_INTERP_CONTROL_0 */
+       0x00000000, /* SPI_INPUT_Z */
+       0x00000000, /* SPI_FOG_CNTL */
+       0x00100000, /* SPI_BARYC_CNTL */
+       0x00000000, /* SPI_PS_IN_CONTROL_2 */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+       0x00000000, /*  */
+
+       0xc0036e00, /* SET_SAMPLER */
+       0x00000000,
+       0x00000012,
+       0x00000000,
+       0x00000000,
+};
+
+const u32 evergreen_vs[] =
+{
+       0x00000004,
+       0x80800400,
+       0x0000a03c,
+       0x95000688,
+       0x00004000,
+       0x15200688,
+       0x00000000,
+       0x00000000,
+       0x3c000000,
+       0x67961001,
+       0x00080000,
+       0x00000000,
+       0x1c000000,
+       0x67961000,
+       0x00000008,
+       0x00000000,
+};
+
+const u32 evergreen_ps[] =
+{
+       0x00000003,
+       0xa00c0000,
+       0x00000008,
+       0x80400000,
+       0x00000000,
+       0x95200688,
+       0x00380400,
+       0x00146b10,
+       0x00380000,
+       0x20146b10,
+       0x00380400,
+       0x40146b00,
+       0x80380000,
+       0x60146b00,
+       0x00000000,
+       0x00000000,
+       0x00000010,
+       0x000d1000,
+       0xb0800000,
+       0x00000000,
+};
+
+const u32 evergreen_ps_size = ARRAY_SIZE(evergreen_ps);
+const u32 evergreen_vs_size = ARRAY_SIZE(evergreen_vs);
+const u32 evergreen_default_size = ARRAY_SIZE(evergreen_default_state);
diff --git a/drivers/gpu/drm/radeon/evergreen_blit_shaders.h b/drivers/gpu/drm/radeon/evergreen_blit_shaders.h
new file mode 100644 (file)
index 0000000..bb8d6c7
--- /dev/null
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2009 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef EVERGREEN_BLIT_SHADERS_H
+#define EVERGREEN_BLIT_SHADERS_H
+
+extern const u32 evergreen_ps[];
+extern const u32 evergreen_vs[];
+extern const u32 evergreen_default_state[];
+
+extern const u32 evergreen_ps_size, evergreen_vs_size;
+extern const u32 evergreen_default_size;
+
+#endif
index 9b7532dd30f754d4ad2e3e02915006ae823538f2..319aa9752d404e88d17a88b5d98a0c3a317e4ea4 100644 (file)
 #define SQ_ALU_CONST_CACHE_LS_14                       0x28f78
 #define SQ_ALU_CONST_CACHE_LS_15                       0x28f7c
 
+#define PA_SC_SCREEN_SCISSOR_TL                         0x28030
+#define PA_SC_GENERIC_SCISSOR_TL                        0x28240
+#define PA_SC_WINDOW_SCISSOR_TL                         0x28204
+#define VGT_PRIMITIVE_TYPE                              0x8958
+
 #define DB_DEPTH_CONTROL                               0x28800
 #define DB_DEPTH_VIEW                                  0x28008
 #define DB_HTILE_DATA_BASE                             0x28014
index 90394df63009a0a05ca9c46a85536a66a1e6170c..2a4747d9747c2134e67fec892363a0e40ae27ee4 100644 (file)
@@ -565,7 +565,7 @@ void r600_blit_fini(struct radeon_device *rdev)
        radeon_bo_unref(&rdev->r600_blit.shader_obj);
 }
 
-int r600_vb_ib_get(struct radeon_device *rdev)
+static int r600_vb_ib_get(struct radeon_device *rdev)
 {
        int r;
        r = radeon_ib_get(rdev, &rdev->r600_blit.vb_ib);
@@ -579,7 +579,7 @@ int r600_vb_ib_get(struct radeon_device *rdev)
        return 0;
 }
 
-void r600_vb_ib_put(struct radeon_device *rdev)
+static void r600_vb_ib_put(struct radeon_device *rdev)
 {
        radeon_fence_emit(rdev, rdev->r600_blit.vb_ib->fence);
        radeon_ib_free(rdev, &rdev->r600_blit.vb_ib);
@@ -683,17 +683,6 @@ void r600_kms_blit_copy(struct radeon_device *rdev,
 
                        if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
                                WARN_ON(1);
-
-#if 0
-                               r600_vb_ib_put(rdev);
-
-                               r600_nomm_put_vb(dev);
-                               r600_nomm_get_vb(dev);
-                               if (!dev_priv->blit_vb)
-                                       return;
-                               set_shaders(dev);
-                               vb = r600_nomm_get_vb_ptr(dev);
-#endif
                        }
 
                        vb[0] = i2f(dst_x);
@@ -778,17 +767,6 @@ void r600_kms_blit_copy(struct radeon_device *rdev,
                        if ((rdev->r600_blit.vb_used + 48) > rdev->r600_blit.vb_total) {
                                WARN_ON(1);
                        }
-#if 0
-                       if ((rdev->blit_vb->used + 48) > rdev->blit_vb->total) {
-                               r600_nomm_put_vb(dev);
-                               r600_nomm_get_vb(dev);
-                               if (!rdev->blit_vb)
-                                       return;
-
-                               set_shaders(dev);
-                               vb = r600_nomm_get_vb_ptr(dev);
-                       }
-#endif
 
                        vb[0] = i2f(dst_x / 4);
                        vb[1] = 0;
index 8e5cb2c4fa7efad623cb9c90e948794bd527ecfa..2edd52ece226bf38f363b9432ef2e88aa3d5c5d2 100644 (file)
@@ -1130,6 +1130,12 @@ void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence)
 void r600_kms_blit_copy(struct radeon_device *rdev,
                        u64 src_gpu_addr, u64 dst_gpu_addr,
                        int size_bytes);
+/* evergreen blit */
+int evergreen_blit_prepare_copy(struct radeon_device *rdev, int size_bytes);
+void evergreen_blit_done_copy(struct radeon_device *rdev, struct radeon_fence *fence);
+void evergreen_kms_blit_copy(struct radeon_device *rdev,
+                            u64 src_gpu_addr, u64 dst_gpu_addr,
+                            int size_bytes);
 
 static inline uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg)
 {
@@ -1471,6 +1477,8 @@ extern void r700_cp_stop(struct radeon_device *rdev);
 extern void r700_cp_fini(struct radeon_device *rdev);
 extern void evergreen_disable_interrupt_state(struct radeon_device *rdev);
 extern int evergreen_irq_set(struct radeon_device *rdev);
+extern int evergreen_blit_init(struct radeon_device *rdev);
+extern void evergreen_blit_fini(struct radeon_device *rdev);
 
 /* radeon_acpi.c */ 
 #if defined(CONFIG_ACPI) 
index 25e1dd1977917ad4151cdaf599add440ef14d625..64fb89ecbf74728893bccb0dfcf442688c904688 100644 (file)
@@ -726,9 +726,9 @@ static struct radeon_asic evergreen_asic = {
        .get_vblank_counter = &evergreen_get_vblank_counter,
        .fence_ring_emit = &r600_fence_ring_emit,
        .cs_parse = &evergreen_cs_parse,
-       .copy_blit = NULL,
-       .copy_dma = NULL,
-       .copy = NULL,
+       .copy_blit = &evergreen_copy_blit,
+       .copy_dma = &evergreen_copy_blit,
+       .copy = &evergreen_copy_blit,
        .get_engine_clock = &radeon_atom_get_engine_clock,
        .set_engine_clock = &radeon_atom_set_engine_clock,
        .get_memory_clock = &radeon_atom_get_memory_clock,
index 6d3b055c02fd41cea2b89c58fdaaafe008b8cb34..740988244143ce2105ba6fb04964a87c9eff644e 100644 (file)
@@ -254,11 +254,6 @@ void r600_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 int r600_cs_parse(struct radeon_cs_parser *p);
 void r600_fence_ring_emit(struct radeon_device *rdev,
                          struct radeon_fence *fence);
-int r600_copy_dma(struct radeon_device *rdev,
-                 uint64_t src_offset,
-                 uint64_t dst_offset,
-                 unsigned num_pages,
-                 struct radeon_fence *fence);
 int r600_irq_process(struct radeon_device *rdev);
 int r600_irq_set(struct radeon_device *rdev);
 bool r600_gpu_is_lockup(struct radeon_device *rdev);
@@ -304,6 +299,9 @@ int evergreen_resume(struct radeon_device *rdev);
 bool evergreen_gpu_is_lockup(struct radeon_device *rdev);
 int evergreen_asic_reset(struct radeon_device *rdev);
 void evergreen_bandwidth_update(struct radeon_device *rdev);
+int evergreen_copy_blit(struct radeon_device *rdev,
+                       uint64_t src_offset, uint64_t dst_offset,
+                       unsigned num_pages, struct radeon_fence *fence);
 void evergreen_hpd_init(struct radeon_device *rdev);
 void evergreen_hpd_fini(struct radeon_device *rdev);
 bool evergreen_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);