freedreno/a6xx: Re-write the function-of-doom

Re-write the program state building to align more closely with turnip,
for easier comparision.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24999>

src/freedreno/common/fd6_pack.h

@@ -180,4 +180,12 @@ __reg_iova(const struct fd_reg_pair *reg)
       ring->cur = __p;                                                         \
    } while (0)
 
+#define OUT_BUF(ring, dwords, sizedwords)                                      \
+   do {                                                                        \
+      uint32_t *__p = ring->cur;                                               \
+      memcpy(__p, dwords, 4 * sizedwords);                                     \
+      __p += sizedwords;                                                       \
+      ring->cur = __p;                                                         \
+   } while (0)
+
 #endif /* FD6_PACK_H */

src/gallium/drivers/freedreno/a6xx/fd6_emit.cc

@@ -806,9 +806,11 @@ fd6_emit_restore(struct fd_batch *batch, struct fd_ringbuffer *ring)
    WRITE(REG_A6XX_UCHE_UNKNOWN_0E12, screen->info->a6xx.magic.UCHE_UNKNOWN_0E12);
    WRITE(REG_A6XX_UCHE_CLIENT_PF, screen->info->a6xx.magic.UCHE_CLIENT_PF);
    WRITE(REG_A6XX_RB_UNKNOWN_8E01, screen->info->a6xx.magic.RB_UNKNOWN_8E01);
+   WRITE(REG_A6XX_SP_UNKNOWN_A9A8, 0);
    WRITE(REG_A6XX_SP_MODE_CONTROL,
          A6XX_SP_MODE_CONTROL_CONSTANT_DEMOTION_ENABLE | 4);
    WRITE(REG_A6XX_VFD_ADD_OFFSET, A6XX_VFD_ADD_OFFSET_VERTEX);
+   WRITE(REG_A6XX_VPC_UNKNOWN_9107, 0);
    WRITE(REG_A6XX_RB_UNKNOWN_8811, 0x00000010);
    WRITE(REG_A6XX_PC_MODE_CNTL, screen->info->a6xx.magic.PC_MODE_CNTL);