i965: Let the caller of brw_set_dp_write/read_message control the target cache.

brw_set_dp_read_message already had a target_cache argument, but its
interpretation was rather convoluted (on Gen6 the render cache was
used if the caller asked for it, otherwise it was ignored using the
sampler cache instead), and the constant cache wasn't representable at
all.  brw_set_dp_write_message used the data cache on Gen7+ except for
RENDER_TARGET_WRITE messages, in which case it would use the render
cache.  On Gen6 the render cache was always used.

Instead of the above, provide the shared unit SFID that the caller
expects will be used.  Makes no functional changes.

v3: Non-trivial rebase.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 737a335ab55465d0e66c1210d5a8aefa0e70549b..c44896bc0d5f4ab27790a97315845b9a7dc74823 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -233,6 +233,7 @@ void brw_set_dp_write_message(struct brw_codegen *p,
                              unsigned binding_table_index,
                              unsigned msg_control,
                              unsigned msg_type,
+                              unsigned target_cache,
                              unsigned msg_length,
                              bool header_present,
                              unsigned last_render_target,

diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index ca04221686b3e346dda5984a68a0295c69c8ae49..72b6df6555e4a1fd42ab7a53b663ed40c7d6bfb7 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -706,6 +706,7 @@ brw_set_dp_write_message(struct brw_codegen *p,
                         unsigned binding_table_index,
                         unsigned msg_control,
                         unsigned msg_type,
+                         unsigned target_cache,
                         unsigned msg_length,
                         bool header_present,
                         unsigned last_render_target,
@@ -714,20 +715,8 @@ brw_set_dp_write_message(struct brw_codegen *p,
                         unsigned send_commit_msg)
 {
    const struct gen_device_info *devinfo = p->devinfo;
-   unsigned sfid;
-
-   if (devinfo->gen >= 7) {
-      /* Use the Render Cache for RT writes; otherwise use the Data Cache */
-      if (msg_type == GEN6_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE)
-        sfid = GEN6_SFID_DATAPORT_RENDER_CACHE;
-      else
-        sfid = GEN7_SFID_DATAPORT_DATA_CACHE;
-   } else if (devinfo->gen == 6) {
-      /* Use the render cache for all write messages. */
-      sfid = GEN6_SFID_DATAPORT_RENDER_CACHE;
-   } else {
-      sfid = BRW_SFID_DATAPORT_WRITE;
-   }
+   const unsigned sfid = (devinfo->gen >= 6 ? target_cache :
+                          BRW_SFID_DATAPORT_WRITE);
 
    brw_set_message_descriptor(p, insn, sfid, msg_length, response_length,
                              header_present, end_of_thread);
@@ -753,26 +742,8 @@ brw_set_dp_read_message(struct brw_codegen *p,
                        unsigned response_length)
 {
    const struct gen_device_info *devinfo = p->devinfo;
-   unsigned sfid;
-
-   if (devinfo->gen >= 7) {
-      if (target_cache == BRW_DATAPORT_READ_TARGET_RENDER_CACHE)
-         sfid = GEN6_SFID_DATAPORT_RENDER_CACHE;
-      else if (target_cache == BRW_DATAPORT_READ_TARGET_DATA_CACHE)
-         sfid = GEN7_SFID_DATAPORT_DATA_CACHE;
-      else if (target_cache == BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE)
-         sfid = GEN6_SFID_DATAPORT_SAMPLER_CACHE;
-      else
-         unreachable("Invalid target cache");
-
-   } else if (devinfo->gen == 6) {
-      if (target_cache == BRW_DATAPORT_READ_TARGET_RENDER_CACHE)
-        sfid = GEN6_SFID_DATAPORT_RENDER_CACHE;
-      else
-        sfid = GEN6_SFID_DATAPORT_SAMPLER_CACHE;
-   } else {
-      sfid = BRW_SFID_DATAPORT_READ;
-   }
+   const unsigned sfid = (devinfo->gen >= 6 ? target_cache :
+                          BRW_SFID_DATAPORT_READ);
 
    brw_set_message_descriptor(p, insn, sfid, msg_length, response_length,
                              header_present, false);
@@ -2073,6 +2044,10 @@ void brw_oword_block_write_scratch(struct brw_codegen *p,
                                   unsigned offset)
 {
    const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+       devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
+       BRW_DATAPORT_READ_TARGET_RENDER_CACHE);
    uint32_t msg_type;
 
    if (devinfo->gen >= 6)
@@ -2161,6 +2136,7 @@ void brw_oword_block_write_scratch(struct brw_codegen *p,
                                brw_scratch_surface_idx(p),
                               msg_control,
                               msg_type,
+                               target_cache,
                               mlen,
                               true, /* header_present */
                               0, /* not a render target */
@@ -2210,9 +2186,10 @@ brw_oword_block_read_scratch(struct brw_codegen *p,
        num_regs == 2 ? BRW_DATAPORT_OWORD_BLOCK_4_OWORDS :
        num_regs == 4 ? BRW_DATAPORT_OWORD_BLOCK_8_OWORDS : 0);
    assert(msg_control);
-   const unsigned target_cache = devinfo->gen >= 7 ?
-      BRW_DATAPORT_READ_TARGET_DATA_CACHE :
-      BRW_DATAPORT_READ_TARGET_RENDER_CACHE;
+   const unsigned target_cache =
+      (devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+       devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
+       BRW_DATAPORT_READ_TARGET_RENDER_CACHE);
 
    {
       brw_push_insn_state(p);
@@ -2300,6 +2277,10 @@ void brw_oword_block_read(struct brw_codegen *p,
                          uint32_t bind_table_index)
 {
    const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+       devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_SAMPLER_CACHE :
+       BRW_DATAPORT_READ_TARGET_DATA_CACHE);
 
    /* On newer hardware, offset is in units of owords. */
    if (devinfo->gen >= 6)
@@ -2340,7 +2321,7 @@ void brw_oword_block_read(struct brw_codegen *p,
                           bind_table_index,
                           BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW,
                           BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ,
-                          BRW_DATAPORT_READ_TARGET_DATA_CACHE,
+                          target_cache,
                           1, /* msg_length */
                            true, /* header_present */
                           1); /* response_length (1 reg, 2 owords!) */
@@ -2361,6 +2342,9 @@ void brw_fb_WRITE(struct brw_codegen *p,
                   bool header_present)
 {
    const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
+       BRW_DATAPORT_READ_TARGET_RENDER_CACHE);
    brw_inst *insn;
    unsigned msg_type;
    struct brw_reg dest, src0;
@@ -2397,6 +2381,7 @@ void brw_fb_WRITE(struct brw_codegen *p,
                            binding_table_index,
                            msg_control,
                            msg_type,
+                            target_cache,
                            msg_length,
                            header_present,
                            last_render_target,
@@ -2425,7 +2410,7 @@ gen9_fb_READ(struct brw_codegen *p,
    brw_set_dp_read_message(p, insn, binding_table_index,
                            per_sample << 5 | msg_subtype,
                            GEN9_DATAPORT_RC_RENDER_TARGET_READ,
-                           BRW_DATAPORT_READ_TARGET_RENDER_CACHE,
+                           GEN6_SFID_DATAPORT_RENDER_CACHE,
                            msg_length, true /* header_present */,
                            response_length);
    brw_inst_set_rt_slot_group(devinfo, insn,
@@ -2888,6 +2873,11 @@ brw_svb_write(struct brw_codegen *p,
               unsigned binding_table_index,
               bool   send_commit_msg)
 {
+   const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+       devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
+       BRW_DATAPORT_READ_TARGET_RENDER_CACHE);
    brw_inst *insn;
 
    gen6_resolve_implied_move(p, &src0, msg_reg_nr);
@@ -2900,6 +2890,7 @@ brw_svb_write(struct brw_codegen *p,
                             binding_table_index,
                             0, /* msg_control: ignored */
                             GEN6_DATAPORT_WRITE_MESSAGE_STREAMED_VB_WRITE,
+                            target_cache,
                             1, /* msg_length */
                             true, /* header_present */
                             0, /* last_render_target: ignored */

diff --git a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
index bb184792cf607b8998841654c9f5098aa9c3e032..ef7ce78b8030a4cf9716593024170dcb7d8674cb 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_generator.cpp
@@ -1140,8 +1140,9 @@ generate_scratch_read(struct brw_codegen *p,
    else
       msg_type = BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ;
 
-   const unsigned target_cache = devinfo->gen >= 7 ?
-      BRW_DATAPORT_READ_TARGET_DATA_CACHE :
+   const unsigned target_cache =
+      devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+      devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
       BRW_DATAPORT_READ_TARGET_RENDER_CACHE;
 
    /* Each of the 8 channel enables is considered for whether each
@@ -1169,6 +1170,10 @@ generate_scratch_write(struct brw_codegen *p,
                        struct brw_reg index)
 {
    const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 7 ? GEN7_SFID_DATAPORT_DATA_CACHE :
+       devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_RENDER_CACHE :
+       BRW_DATAPORT_READ_TARGET_RENDER_CACHE);
    struct brw_reg header = brw_vec8_grf(0, 0);
    bool write_commit;
 
@@ -1228,6 +1233,7 @@ generate_scratch_write(struct brw_codegen *p,
                             brw_scratch_surface_idx(p),
                            BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD,
                            msg_type,
+                            target_cache,
                            3, /* mlen */
                            true, /* header present */
                            false, /* not a render target write */
@@ -1245,6 +1251,9 @@ generate_pull_constant_load(struct brw_codegen *p,
                             struct brw_reg offset)
 {
    const struct gen_device_info *devinfo = p->devinfo;
+   const unsigned target_cache =
+      (devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_SAMPLER_CACHE :
+       BRW_DATAPORT_READ_TARGET_DATA_CACHE);
    assert(index.file == BRW_IMMEDIATE_VALUE &&
          index.type == BRW_REGISTER_TYPE_UD);
    uint32_t surf_index = index.ud;
@@ -1290,7 +1299,7 @@ generate_pull_constant_load(struct brw_codegen *p,
                           surf_index,
                           BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD,
                           msg_type,
-                          BRW_DATAPORT_READ_TARGET_DATA_CACHE,
+                           target_cache,
                           2, /* mlen */
                            true, /* header_present */
                           1 /* rlen */);