[Mesa-dev] [PATCH 01/11] i965: Switch to absolute addressing for constant buffer 0.

[Kenneth Graunke]

By default, 3DSTATE_CONSTANT_* Constant Buffer 0 is relative to dynamic
state base address.  This makes it unusable for pushing UBOs.  I'd like
to be able to use all four push buffers.

There is a bit in the INSTPM register (or CS_DEBUG_MODE2 on Skylake)
which controls whether buffer 0 is relative to dynamic state base
address, or simply a normal pointer.  Setting that gives us full
flexibility.

We can't currently write this on Haswell and earlier, and will need
to update the kernel command parser, and then do the whole version
checking song and dance.
---
 src/intel/compiler/brw_compiler.h            |  6 ++++++
 src/mesa/drivers/dri/i965/brw_defines.h      |  6 ++++++
 src/mesa/drivers/dri/i965/brw_state_upload.c | 24 ++++++++++++++++++++++++
 src/mesa/drivers/dri/i965/intel_screen.c     |  1 +
 4 files changed, 37 insertions(+)

diff --git a/src/intel/compiler/brw_compiler.h b/src/intel/compiler/brw_compiler.h
index 3f383403883..e4c22e31177 100644
--- a/src/intel/compiler/brw_compiler.h
+++ b/src/intel/compiler/brw_compiler.h
@@ -100,6 +100,12 @@ struct brw_compiler {
     * This can negatively impact performance.
     */
    bool precise_trig;
+
+   /**
+    * Is 3DSTATE_CONSTANT_*'s Constant Buffer 0 relative to Dynamic State
+    * Base Address?  (If not, it's a normal GPU address.)
+    */
+   bool constant_buffer_0_is_relative;
 };
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index a4794c6a1d2..2a8dbf8cb9a 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -1680,4 +1680,10 @@ enum brw_pixel_shader_coverage_mask_mode {
 # define GEN8_L3CNTLREG_ALL_ALLOC_SHIFT    25
 # define GEN8_L3CNTLREG_ALL_ALLOC_MASK     INTEL_MASK(31, 25)
 
+#define INSTPM                             0x20c0
+# define INSTPM_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE (1 << 6)
+
+#define CS_DEBUG_MODE2                     0x20d8 /* Gen9+ */
+# define CSDBG2_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE (1 << 4)
+
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 5e82c1b4ce8..e14ed02e2e1 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -90,6 +90,30 @@ brw_upload_initial_gpu_state(struct brw_context *brw)
       OUT_BATCH(0);
       ADVANCE_BATCH();
    }
+
+   /* Set the "CONSTANT_BUFFER Address Offset Disable" bit, so
+    * 3DSTATE_CONSTANT_XS buffer 0 is an absolute address.
+    *
+    * On Gen6-7.5, we use an execbuf parameter to do this for us.
+    * However, the kernel ignores that when execlists are in use.
+    * Fortunately, we can just write the registers from userspace
+    * on Gen8+, and they're context saved/restored.
+    */
+   if (brw->gen >= 9) {
+      BEGIN_BATCH(3);
+      OUT_BATCH(MI_LOAD_REGISTER_IMM | (3 - 2));
+      OUT_BATCH(CS_DEBUG_MODE2);
+      OUT_BATCH(REG_MASK(CSDBG2_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE) |
+                CSDBG2_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE);
+      ADVANCE_BATCH();
+   } else if (brw->gen == 8) {
+      BEGIN_BATCH(3);
+      OUT_BATCH(MI_LOAD_REGISTER_IMM | (3 - 2));
+      OUT_BATCH(INSTPM);
+      OUT_BATCH(REG_MASK(INSTPM_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE) |
+                INSTPM_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE);
+      ADVANCE_BATCH();
+   }
 }
 
 static inline const struct brw_tracked_state *
diff --git a/src/mesa/drivers/dri/i965/intel_screen.c b/src/mesa/drivers/dri/i965/intel_screen.c
index c75f2125d4b..5f2c512fc51 100644
--- a/src/mesa/drivers/dri/i965/intel_screen.c
+++ b/src/mesa/drivers/dri/i965/intel_screen.c
@@ -2307,6 +2307,7 @@ __DRIconfig **intelInitScreen2(__DRIscreen *dri_screen)
    screen->compiler = brw_compiler_create(screen, devinfo);
    screen->compiler->shader_debug_log = shader_debug_log_mesa;
    screen->compiler->shader_perf_log = shader_perf_log_mesa;
+   screen->compiler->constant_buffer_0_is_relative = devinfo->gen < 8;
    screen->program_id = 1;
 
    screen->has_exec_fence =
-- 
2.13.2