[Mesa-dev] [PATCH 2/2] WIP tgsi: add PK2H/UP2H support

[Ilia Mirkin]

It seems like there's something horribly wrong with the
util_float_to_half function. In a standalone compilation it works fine
for -6.10203e-05, generating 0x8400, but inside mesa it ends up with
0x8000. The result of the magic.f multiplication is 0. No idea why.

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
---

See above comments for why I didn't include this at all. Should you figure out
what was going wrong, note that this can only be enabled in softpipe if the
whole pipeline is using tgsi_exec (or if the gallivm patch is fixed/upstreamed).

Feel free to take this over and/or modify as necessary.

 src/gallium/auxiliary/tgsi/tgsi_exec.c   | 44 ++++++++++++++++++++++++++++++--
 src/gallium/drivers/softpipe/sp_screen.c |  2 +-
 2 files changed, 43 insertions(+), 3 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index f67c162..12a477b 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -58,6 +58,7 @@
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 #include "tgsi_exec.h"
+#include "util/u_half.h"
 #include "util/u_memory.h"
 #include "util/u_math.h"
 
@@ -3058,6 +3059,45 @@ exec_dp2(struct tgsi_exec_machine *mach,
 }
 
 static void
+exec_pk2h(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[2], dst;
+
+   fetch_source(mach, &arg[0], &inst->Src[0], TGSI_CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[0], TGSI_CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   for (chan = 0; chan < TGSI_QUAD_SIZE; chan++) {
+      dst.u[chan] = util_float_to_half(arg[0].f[chan]) |
+         (util_float_to_half(arg[1].f[chan]) << 16);
+   }
+   for (chan = 0; chan < TGSI_NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst, &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_UINT);
+      }
+   }
+}
+
+static void
+exec_up2h(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg, dst[2];
+
+   fetch_source(mach, &arg, &inst->Src[0], TGSI_CHAN_X, TGSI_EXEC_DATA_UINT);
+   for (chan = 0; chan < 4; chan++) {
+      dst[0].f[chan] = util_half_to_float(arg.u[chan] & 0xffff);
+      dst[1].f[chan] = util_half_to_float(arg.u[chan] >> 16);
+   }
+   for (chan = 0; chan < TGSI_NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst[chan & 1], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
 exec_scs(struct tgsi_exec_machine *mach,
          const struct tgsi_full_instruction *inst)
 {
@@ -4339,7 +4379,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_PK2H:
-      assert (0);
+      exec_pk2h(mach, inst);
       break;
 
    case TGSI_OPCODE_PK2US:
@@ -4425,7 +4465,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_UP2H:
-      assert (0);
+      exec_up2h(mach, inst);
       break;
 
    case TGSI_OPCODE_UP2US:
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index e74044b..d4526ef 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -236,6 +236,7 @@ softpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_CLIP_HALFZ:
    case PIPE_CAP_TEXTURE_FLOAT_LINEAR:
    case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
+   case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
       return 1;
    case PIPE_CAP_VERTEXID_NOBASE:
       return 0;
@@ -252,7 +253,6 @@ softpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
    case PIPE_CAP_CLEAR_TEXTURE:
    case PIPE_CAP_DRAW_PARAMETERS:
-   case PIPE_CAP_TGSI_PACK_HALF_FLOAT:
    case PIPE_CAP_MULTI_DRAW_INDIRECT:
    case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
       return 0;
-- 
2.4.10