v3dv: add support for TFU jobs in v71

This includes update the simulator.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25450>

include/drm-uapi/v3d_drm.h

@@ -319,6 +319,11 @@ struct drm_v3d_submit_tfu {
 
 	/* Pointer to an array of ioctl extensions*/
 	__u64 extensions;
+
+	struct {
+		__u32 ioc;
+		__u32 pad;
+	} v71;
 };
 
 /* Submits a compute shader for dispatch.  This job will block on any

src/broadcom/common/v3d_tfu.h

@@ -48,4 +48,27 @@
 #define V3D33_TFU_ICFG_FORMAT_UIF_NO_XOR 14
 #define V3D33_TFU_ICFG_FORMAT_UIF_XOR 15
 
+/* Disable level 0 write, just write following mipmaps */
+#define V3D71_TFU_IOC_DIMTW (1 << 0)
+#define V3D71_TFU_IOC_FORMAT_SHIFT              12
+#define V3D71_TFU_IOC_FORMAT_LINEARTILE          3
+#define V3D71_TFU_IOA_FORMAT_UBLINEAR_1_COLUMN   4
+#define V3D71_TFU_IOA_FORMAT_UBLINEAR_2_COLUMN   5
+#define V3D71_TFU_IOA_FORMAT_UIF_NO_XOR          6
+#define V3D71_TFU_IOA_FORMAT_UIF_XOR             7
+
+#define V3D71_TFU_IOC_STRIDE_SHIFT              16
+#define V3D71_TFU_IOC_NUMMM_SHIFT                4
+
+#define V3D71_TFU_ICFG_OTYPE_SHIFT              16
+#define V3D71_TFU_ICFG_IFORMAT_SHIFT            23
+#define V3D71_TFU_ICFG_FORMAT_RASTER             0
+#define V3D71_TFU_ICFG_FORMAT_SAND_128           1
+#define V3D71_TFU_ICFG_FORMAT_SAND_256           2
+#define V3D71_TFU_ICFG_FORMAT_LINEARTILE        11
+#define V3D71_TFU_ICFG_FORMAT_UBLINEAR_1_COLUMN 12
+#define V3D71_TFU_ICFG_FORMAT_UBLINEAR_2_COLUMN 13
+#define V3D71_TFU_ICFG_FORMAT_UIF_NO_XOR        14
+#define V3D71_TFU_ICFG_FORMAT_UIF_XOR           15
+
 #endif

src/broadcom/simulator/v3dx_simulator.c

@@ -182,26 +182,36 @@ v3d_flush_caches(struct v3d_hw *v3d)
         v3d_flush_l2t(v3d);
 }
 
+#if V3D_VERSION < 71
+#define TFU_REG(NAME) V3D_TFU_ ## NAME
+#else
+#define TFU_REG(NAME) V3D_IFC_ ## NAME
+#endif
+
+
 int
 v3dX(simulator_submit_tfu_ioctl)(struct v3d_hw *v3d,
                                  struct drm_v3d_submit_tfu *args)
 {
-        int last_vtct = V3D_READ(V3D_TFU_CS) & V3D_TFU_CS_CVTCT_SET;
+        int last_vtct = V3D_READ(TFU_REG(CS)) & V3D_TFU_CS_CVTCT_SET;
 
-        V3D_WRITE(V3D_TFU_IIA, args->iia);
+        V3D_WRITE(TFU_REG(IIA), args->iia);
-        V3D_WRITE(V3D_TFU_IIS, args->iis);
+        V3D_WRITE(TFU_REG(IIS), args->iis);
-        V3D_WRITE(V3D_TFU_ICA, args->ica);
+        V3D_WRITE(TFU_REG(ICA), args->ica);
-        V3D_WRITE(V3D_TFU_IUA, args->iua);
+        V3D_WRITE(TFU_REG(IUA), args->iua);
-        V3D_WRITE(V3D_TFU_IOA, args->ioa);
+        V3D_WRITE(TFU_REG(IOA), args->ioa);
-        V3D_WRITE(V3D_TFU_IOS, args->ios);
+#if V3D_VERSION >= 71
-        V3D_WRITE(V3D_TFU_COEF0, args->coef[0]);
+        V3D_WRITE(TFU_REG(IOC), args->v71.ioc);
-        V3D_WRITE(V3D_TFU_COEF1, args->coef[1]);
+#endif
-        V3D_WRITE(V3D_TFU_COEF2, args->coef[2]);
+        V3D_WRITE(TFU_REG(IOS), args->ios);
-        V3D_WRITE(V3D_TFU_COEF3, args->coef[3]);
+        V3D_WRITE(TFU_REG(COEF0), args->coef[0]);
+        V3D_WRITE(TFU_REG(COEF1), args->coef[1]);
+        V3D_WRITE(TFU_REG(COEF2), args->coef[2]);
+        V3D_WRITE(TFU_REG(COEF3), args->coef[3]);
 
-        V3D_WRITE(V3D_TFU_ICFG, args->icfg);
+        V3D_WRITE(TFU_REG(ICFG), args->icfg);
 
-        while ((V3D_READ(V3D_TFU_CS) & V3D_TFU_CS_CVTCT_SET) == last_vtct) {
+        while ((V3D_READ(TFU_REG(CS)) & V3D_TFU_CS_CVTCT_SET) == last_vtct) {
                 v3d_hw_tick(v3d);
         }
 

src/broadcom/vulkan/v3dvx_meta_common.c

@@ -950,6 +950,7 @@ v3dX(meta_emit_tfu_job)(struct v3dv_cmd_buffer *cmd_buffer,
 
    tfu.iia |= src_offset;
 
+#if V3D_VERSION <= 42
    if (src_tiling == V3D_TILING_RASTER) {
       tfu.icfg = V3D33_TFU_ICFG_FORMAT_RASTER << V3D33_TFU_ICFG_FORMAT_SHIFT;
    } else {
@@ -958,12 +959,46 @@ v3dX(meta_emit_tfu_job)(struct v3dv_cmd_buffer *cmd_buffer,
                    V3D33_TFU_ICFG_FORMAT_SHIFT;
    }
    tfu.icfg |= format_plane->tex_type << V3D33_TFU_ICFG_TTYPE_SHIFT;
+#endif
+#if V3D_VERSION >= 71
+   if (src_tiling == V3D_TILING_RASTER) {
+      tfu.icfg = V3D71_TFU_ICFG_FORMAT_RASTER << V3D71_TFU_ICFG_IFORMAT_SHIFT;
+   } else {
+      tfu.icfg = (V3D71_TFU_ICFG_FORMAT_LINEARTILE +
+                  (src_tiling - V3D_TILING_LINEARTILE)) <<
+                   V3D71_TFU_ICFG_IFORMAT_SHIFT;
+   }
+   tfu.icfg |= format_plane->tex_type << V3D71_TFU_ICFG_OTYPE_SHIFT;
+#endif
 
    tfu.ioa = dst_offset;
 
+#if V3D_VERSION <= 42
    tfu.ioa |= (V3D33_TFU_IOA_FORMAT_LINEARTILE +
                (dst_tiling - V3D_TILING_LINEARTILE)) <<
                 V3D33_TFU_IOA_FORMAT_SHIFT;
+#endif
+
+#if V3D_VERSION >= 71
+   tfu.v71.ioc = (V3D71_TFU_IOC_FORMAT_LINEARTILE +
+                  (dst_tiling - V3D_TILING_LINEARTILE)) <<
+                   V3D71_TFU_IOC_FORMAT_SHIFT;
+
+   switch (dst_tiling) {
+   case V3D_TILING_UIF_NO_XOR:
+   case V3D_TILING_UIF_XOR:
+      tfu.v71.ioc |=
+         (dst_padded_height_or_stride / (2 * v3d_utile_height(dst_cpp))) <<
+         V3D71_TFU_IOC_STRIDE_SHIFT;
+      break;
+   case V3D_TILING_RASTER:
+      tfu.v71.ioc |= (dst_padded_height_or_stride / dst_cpp) <<
+                      V3D71_TFU_IOC_STRIDE_SHIFT;
+      break;
+   default:
+      break;
+   }
+#endif
 
    switch (src_tiling) {
    case V3D_TILING_UIF_NO_XOR:
@@ -980,6 +1015,7 @@ v3dX(meta_emit_tfu_job)(struct v3dv_cmd_buffer *cmd_buffer,
    /* The TFU can handle raster sources but always produces UIF results */
    assert(dst_tiling != V3D_TILING_RASTER);
 
+#if V3D_VERSION <= 42
    /* If we're writing level 0 (!IOA_DIMTW), then we need to supply the
     * OPAD field for the destination (how many extra UIF blocks beyond
     * those necessary to cover the height).
@@ -991,6 +1027,7 @@ v3dX(meta_emit_tfu_job)(struct v3dv_cmd_buffer *cmd_buffer,
                       uif_block_h;
       tfu.icfg |= icfg << V3D33_TFU_ICFG_OPAD_SHIFT;
    }
+#endif
 
    v3dv_cmd_buffer_add_tfu_job(cmd_buffer, &tfu);
 }