i965: Fix shared local memory size for Gen9+.

Skylake changes the representation of shared local memory size:

 Size   | 0 kB | 1 kB | 2 kB | 4 kB | 8 kB | 16 kB | 32 kB | 64 kB |
 -------------------------------------------------------------------
 Gen7-8 |    0 | none | none |    1 |    2 |     4 |     8 |    16 |
 -------------------------------------------------------------------
 Gen9+  |    0 |    1 |    2 |    3 |    4 |     5 |     6 |     7 |

The old formula would substantially underallocate the amount of space.
This fixes GPU hangs on Skylake when running with full thread counts.

v2: Fix the Vulkan driver too, use a helper function, and fix the table
    in the comments and commit message.

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

src/intel/vulkan/gen7_cmd_buffer.c

@@ -241,15 +241,7 @@ flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
       }
    }
 
-   assert(prog_data->total_shared <= 64 * 1024);
+   const uint32_t slm_size = encode_slm_size(GEN_GEN, prog_data->total_shared);
-   uint32_t slm_size = 0;
-   if (prog_data->total_shared > 0) {
-      /* slm_size is in 4k increments, but must be a power of 2. */
-      slm_size = 4 * 1024;
-      while (slm_size < prog_data->total_shared)
-         slm_size <<= 1;
-      slm_size /= 4 * 1024;
-   }
 
    struct anv_state state =
       anv_state_pool_emit(&device->dynamic_state_pool,

src/intel/vulkan/gen8_cmd_buffer.c

@@ -326,15 +326,7 @@ flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
       }
    }
 
-   assert(prog_data->total_shared <= 64 * 1024);
+   const uint32_t slm_size = encode_slm_size(GEN_GEN, prog_data->total_shared);
-   uint32_t slm_size = 0;
-   if (prog_data->total_shared > 0) {
-      /* slm_size is in 4k increments, but must be a power of 2. */
-      slm_size = 4 * 1024;
-      while (slm_size < prog_data->total_shared)
-         slm_size <<= 1;
-      slm_size /= 4 * 1024;
-   }
 
    struct anv_state state =
       anv_state_pool_emit(&device->dynamic_state_pool,

src/mesa/drivers/dri/i965/brw_compiler.h

@@ -26,6 +26,7 @@
 #include <stdio.h>
 #include "brw_device_info.h"
 #include "main/mtypes.h"
+#include "main/macros.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -831,6 +832,38 @@ brw_compile_cs(const struct brw_compiler *compiler, void *log_data,
                unsigned *final_assembly_size,
                char **error_str);
 
+static inline uint32_t
+encode_slm_size(const struct brw_device_info *devinfo, uint32_t bytes)
+{
+   uint32_t slm_size = 0;
+
+   /* Shared Local Memory is specified as powers of two, and encoded in
+    * INTERFACE_DESCRIPTOR_DATA with the following representations:
+    *
+    * Size   | 0 kB | 1 kB | 2 kB | 4 kB | 8 kB | 16 kB | 32 kB | 64 kB |
+    * -------------------------------------------------------------------
+    * Gen7-8 |    0 | none | none |    1 |    2 |     4 |     8 |    16 |
+    * -------------------------------------------------------------------
+    * Gen9+  |    0 |    1 |    2 |    3 |    4 |     5 |     6 |     7 |
+    */
+   assert(bytes <= 64 * 1024);
+
+   if (bytes > 0) {
+      /* Shared Local Memory Size is specified as powers of two. */
+      slm_size = util_next_power_of_two(bytes);
+
+      if (devinfo->gen >= 9) {
+         /* Use a minimum of 1kB; turn an exponent of 10 (1024 kB) into 1. */
+         slm_size = ffs(MAX2(slm_size, 1024)) - 10;
+      } else {
+         /* Use a minimum of 4kB; convert to the pre-Gen9 representation. */
+         slm_size = MAX2(slm_size, 4096) / 4096;
+      }
+   }
+
+   return slm_size;
+}
+
 #ifdef __cplusplus
 } /* extern "C" */
 #endif

src/mesa/drivers/dri/i965/gen7_cs_state.c

@@ -45,6 +45,7 @@ brw_upload_cs_state(struct brw_context *brw)
    struct brw_stage_state *stage_state = &brw->cs.base;
    struct brw_cs_prog_data *cs_prog_data = brw->cs.prog_data;
    struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
+   const struct brw_device_info *devinfo = brw->intelScreen->devinfo;
 
    if (INTEL_DEBUG & DEBUG_SHADER_TIME) {
       brw->vtbl.emit_buffer_surface_state(
@@ -147,15 +148,7 @@ brw_upload_cs_state(struct brw_context *brw)
       SET_FIELD(cs_prog_data->threads, MEDIA_GPGPU_THREAD_COUNT);
    assert(cs_prog_data->threads <= brw->max_cs_threads);
 
-   assert(prog_data->total_shared <= 64 * 1024);
+   const uint32_t slm_size = encode_slm_size(devinfo, prog_data->total_shared);
-   uint32_t slm_size = 0;
-   if (prog_data->total_shared > 0) {
-      /* slm_size is in 4k increments, but must be a power of 2. */
-      slm_size = 4 * 1024;
-      while (slm_size < prog_data->total_shared)
-         slm_size <<= 1;
-      slm_size /= 4 * 1024;
-   }
 
    desc[dw++] =
       SET_FIELD(cs_prog_data->uses_barrier, MEDIA_BARRIER_ENABLE) |