freedreno/a6xx: Write multiple regs for SP_VS_OUT_REG and SP_VS_VPC_DST_REG

Compute the number of writes up front.

Signed-off-by: Kristian H. Kristensen <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>

1 files changed, 5 insertions, 6 deletions

diff --git a/src/gallium/drivers/freedreno/a6xx/fd6_program.c b/src/gallium/drivers/freedreno/a6xx/fd6_program.c
index 26be370833d..3a1c0a411f9 100644
--- a/src/gallium/drivers/freedreno/a6xx/fd6_program.c
+++ b/src/gallium/drivers/freedreno/a6xx/fd6_program.c
@@ -352,11 +352,11 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
 		setup_stream_out(state, vs, &l);
 	}
 
-	for (i = 0, j = 0; (i < 16) && (j < l.cnt); i++) {
+	debug_assert(l.cnt < 32);
+	OUT_PKT4(ring, REG_A6XX_SP_VS_OUT_REG(0), DIV_ROUND_UP(l.cnt, 2));
+	for (j = 0; j < l.cnt; ) {
 		uint32_t reg = 0;
 
-		OUT_PKT4(ring, REG_A6XX_SP_VS_OUT_REG(i), 1);
-
 		reg |= A6XX_SP_VS_OUT_REG_A_REGID(l.var[j].regid);
 		reg |= A6XX_SP_VS_OUT_REG_A_COMPMASK(l.var[j].compmask);
 		j++;
@@ -368,11 +368,10 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
 		OUT_RING(ring, reg);
 	}
 
-	for (i = 0, j = 0; (i < 8) && (j < l.cnt); i++) {
+	OUT_PKT4(ring, REG_A6XX_SP_VS_VPC_DST_REG(0), DIV_ROUND_UP(l.cnt, 4));
+	for (j = 0; j < l.cnt; ) {
 		uint32_t reg = 0;
 
-		OUT_PKT4(ring, REG_A6XX_SP_VS_VPC_DST_REG(i), 1);
-
 		reg |= A6XX_SP_VS_VPC_DST_REG_OUTLOC0(l.var[j++].loc);
 		reg |= A6XX_SP_VS_VPC_DST_REG_OUTLOC1(l.var[j++].loc);
 		reg |= A6XX_SP_VS_VPC_DST_REG_OUTLOC2(l.var[j++].loc);