freedreno/a3xx: vtx formats

Add support for more vertex buffer formats.

Signed-off-by: Rob Clark <[email protected]>

2 files changed, 79 insertions, 63 deletions

diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
index 4c6b5c15aaa..878d6ff6325 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
@@ -327,9 +327,12 @@ fd3_emit_vertex_bufs(struct fd_ringbuffer *ring,
 		if (vp->inputs[i].compmask) {
 			struct pipe_resource *prsc = vbufs[i].prsc;
 			struct fd_resource *rsc = fd_resource(prsc);
-			enum a3xx_vtx_fmt fmt = fd3_pipe2vtx(vbufs[i].format);
+			enum pipe_format pfmt = vbufs[i].format;
+			enum a3xx_vtx_fmt fmt = fd3_pipe2vtx(pfmt);
 			bool switchnext = (i != last);
-			uint32_t fs = util_format_get_blocksize(vbufs[i].format);
+			uint32_t fs = util_format_get_blocksize(pfmt);
+
+			debug_assert(fmt != ~0);
 
 			OUT_PKT0(ring, REG_A3XX_VFD_FETCH(j), 2);
 			OUT_RING(ring, A3XX_VFD_FETCH_INSTR_0_FETCHSIZE(fs - 1) |
@@ -343,6 +346,7 @@ fd3_emit_vertex_bufs(struct fd_ringbuffer *ring,
 			OUT_RING(ring, A3XX_VFD_DECODE_INSTR_CONSTFILL |
 					A3XX_VFD_DECODE_INSTR_WRITEMASK(vp->inputs[i].compmask) |
 					A3XX_VFD_DECODE_INSTR_FORMAT(fmt) |
+					A3XX_VFD_DECODE_INSTR_SWAP(fd3_pipe2swap(pfmt)) |
 					A3XX_VFD_DECODE_INSTR_REGID(vp->inputs[i].regid) |
 					A3XX_VFD_DECODE_INSTR_SHIFTCNT(fs) |
 					A3XX_VFD_DECODE_INSTR_LASTCOMPVALID |
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_util.c b/src/gallium/drivers/freedreno/a3xx/fd3_util.c
index 682b47d0836..43f30ba526d 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_util.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_util.c
@@ -37,70 +37,44 @@ fd3_pipe2vtx(enum pipe_format format)
 {
 	switch (format) {
 	/* 8-bit buffers. */
-	case PIPE_FORMAT_A8_UNORM:
-	case PIPE_FORMAT_I8_UNORM:
-	case PIPE_FORMAT_L8_UNORM:
 	case PIPE_FORMAT_R8_UNORM:
-	case PIPE_FORMAT_L8_SRGB:
 		return VFMT_NORM_UBYTE_8;
 
-	case PIPE_FORMAT_A8_SNORM:
-	case PIPE_FORMAT_I8_SNORM:
-	case PIPE_FORMAT_L8_SNORM:
 	case PIPE_FORMAT_R8_SNORM:
 		return VFMT_NORM_BYTE_8;
 
-	case PIPE_FORMAT_A8_UINT:
-	case PIPE_FORMAT_I8_UINT:
-	case PIPE_FORMAT_L8_UINT:
 	case PIPE_FORMAT_R8_UINT:
 		return VFMT_UBYTE_8;
 
-	case PIPE_FORMAT_A8_SINT:
-	case PIPE_FORMAT_I8_SINT:
-	case PIPE_FORMAT_L8_SINT:
 	case PIPE_FORMAT_R8_SINT:
 		return VFMT_BYTE_8;
 
 	/* 16-bit buffers. */
 	case PIPE_FORMAT_R16_UNORM:
-	case PIPE_FORMAT_A16_UNORM:
-	case PIPE_FORMAT_L16_UNORM:
-	case PIPE_FORMAT_I16_UNORM:
 	case PIPE_FORMAT_Z16_UNORM:
 		return VFMT_NORM_USHORT_16;
 
 	case PIPE_FORMAT_R16_SNORM:
-	case PIPE_FORMAT_A16_SNORM:
-	case PIPE_FORMAT_L16_SNORM:
-	case PIPE_FORMAT_I16_SNORM:
 		return VFMT_NORM_SHORT_16;
 
 	case PIPE_FORMAT_R16_UINT:
-	case PIPE_FORMAT_A16_UINT:
-	case PIPE_FORMAT_L16_UINT:
-	case PIPE_FORMAT_I16_UINT:
 		return VFMT_USHORT_16;
 
 	case PIPE_FORMAT_R16_SINT:
-	case PIPE_FORMAT_A16_SINT:
-	case PIPE_FORMAT_L16_SINT:
-	case PIPE_FORMAT_I16_SINT:
 		return VFMT_SHORT_16;
 
-	case PIPE_FORMAT_L8A8_UNORM:
+	case PIPE_FORMAT_R16_FLOAT:
+		return VFMT_FLOAT_16;
+
 	case PIPE_FORMAT_R8G8_UNORM:
 		return VFMT_NORM_UBYTE_8_8;
 
-	case PIPE_FORMAT_L8A8_SNORM:
 	case PIPE_FORMAT_R8G8_SNORM:
 		return VFMT_NORM_BYTE_8_8;
 
-	case PIPE_FORMAT_L8A8_UINT:
 	case PIPE_FORMAT_R8G8_UINT:
 		return VFMT_UBYTE_8_8;
 
-	case PIPE_FORMAT_L8A8_SINT:
 	case PIPE_FORMAT_R8G8_SINT:
 		return VFMT_BYTE_8_8;
 
@@ -121,42 +95,62 @@ fd3_pipe2vtx(enum pipe_format format)
 	case PIPE_FORMAT_A8B8G8R8_UNORM:
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
 	case PIPE_FORMAT_B8G8R8A8_UNORM:
-	case PIPE_FORMAT_B8G8R8X8_UNORM:
 	case PIPE_FORMAT_R8G8B8A8_UNORM:
-	case PIPE_FORMAT_R8G8B8X8_UNORM:
-	case PIPE_FORMAT_X8B8G8R8_UNORM:
-	case PIPE_FORMAT_X8R8G8B8_UNORM:
-	case PIPE_FORMAT_A8B8G8R8_SRGB:
-	case PIPE_FORMAT_B8G8R8A8_SRGB:
 		return VFMT_NORM_UBYTE_8_8_8_8;
 
 	case PIPE_FORMAT_R8G8B8A8_SNORM:
-	case PIPE_FORMAT_R8G8B8X8_SNORM:
 		return VFMT_NORM_BYTE_8_8_8_8;
 
 	case PIPE_FORMAT_R8G8B8A8_UINT:
-	case PIPE_FORMAT_R8G8B8X8_UINT:
 		return VFMT_UBYTE_8_8_8_8;
 
 	case PIPE_FORMAT_R8G8B8A8_SINT:
-	case PIPE_FORMAT_R8G8B8X8_SINT:
 		return VFMT_BYTE_8_8_8_8;
 
-/* TODO probably need gles3 blob drivers to find the 32bit int formats:
-	case PIPE_FORMAT_R32_UINT:
-	case PIPE_FORMAT_R32_SINT:
-	case PIPE_FORMAT_A32_UINT:
-	case PIPE_FORMAT_A32_SINT:
-	case PIPE_FORMAT_L32_UINT:
-	case PIPE_FORMAT_L32_SINT:
-	case PIPE_FORMAT_I32_UINT:
-	case PIPE_FORMAT_I32_SINT:
-*/
+	case PIPE_FORMAT_R16G16_SSCALED:
+		return VFMT_SHORT_16_16;
+
+	case PIPE_FORMAT_R16G16_FLOAT:
+		return VFMT_FLOAT_16_16;
+
+	case PIPE_FORMAT_R16G16_UINT:
+		return VFMT_USHORT_16_16;
+
+	case PIPE_FORMAT_R16G16_UNORM:
+		return VFMT_NORM_USHORT_16_16;
+
+	case PIPE_FORMAT_R16G16_SNORM:
+		return VFMT_NORM_SHORT_16_16;
+
+	case PIPE_FORMAT_R10G10B10A2_UNORM:
+		return VFMT_NORM_UINT_10_10_10_2;
+
+	case PIPE_FORMAT_R10G10B10A2_SNORM:
+		return VFMT_NORM_INT_10_10_10_2;
+
+	case PIPE_FORMAT_R10G10B10A2_USCALED:
+		return VFMT_UINT_10_10_10_2;
+
+	case PIPE_FORMAT_R10G10B10A2_SSCALED:
+		return VFMT_INT_10_10_10_2;
+
+	/* 48-bit buffers. */
+	case PIPE_FORMAT_R16G16B16_FLOAT:
+		return VFMT_FLOAT_16_16_16;
+
+	case PIPE_FORMAT_R16G16B16_SSCALED:
+		return VFMT_SHORT_16_16_16;
+
+	case PIPE_FORMAT_R16G16B16_UINT:
+		return VFMT_USHORT_16_16_16;
+
+	case PIPE_FORMAT_R16G16B16_SNORM:
+		return VFMT_NORM_SHORT_16_16_16;
+
+	case PIPE_FORMAT_R16G16B16_UNORM:
+		return VFMT_NORM_USHORT_16_16_16;
 
 	case PIPE_FORMAT_R32_FLOAT:
-	case PIPE_FORMAT_A32_FLOAT:
-	case PIPE_FORMAT_L32_FLOAT:
-	case PIPE_FORMAT_I32_FLOAT:
 	case PIPE_FORMAT_Z32_FLOAT:
 		return VFMT_FLOAT_32;
 
@@ -177,23 +171,14 @@ fd3_pipe2vtx(enum pipe_format format)
 		return VFMT_SHORT_16_16_16_16;
 
 	case PIPE_FORMAT_R32G32_FLOAT:
-	case PIPE_FORMAT_L32A32_FLOAT:
 		return VFMT_FLOAT_32_32;
 
 	case PIPE_FORMAT_R32G32_FIXED:
 		return VFMT_FIXED_32_32;
 
 	case PIPE_FORMAT_R16G16B16A16_FLOAT:
-	case PIPE_FORMAT_R16G16B16X16_FLOAT:
 		return VFMT_FLOAT_16_16_16_16;
 
-/* TODO probably need gles3 blob drivers to find the 32bit int formats:
-	case PIPE_FORMAT_R32G32_SINT:
-	case PIPE_FORMAT_R32G32_UINT:
-	case PIPE_FORMAT_L32A32_UINT:
-	case PIPE_FORMAT_L32A32_SINT:
-*/
-
 	/* 96-bit buffers. */
 	case PIPE_FORMAT_R32G32B32_FLOAT:
 		return VFMT_FLOAT_32_32_32;
@@ -203,7 +188,6 @@ fd3_pipe2vtx(enum pipe_format format)
 
 	/* 128-bit buffers. */
 	case PIPE_FORMAT_R32G32B32A32_FLOAT:
-	case PIPE_FORMAT_R32G32B32X32_FLOAT:
 		return VFMT_FLOAT_32_32_32_32;
 
 	case PIPE_FORMAT_R32G32B32A32_FIXED:
@@ -214,6 +198,20 @@ fd3_pipe2vtx(enum pipe_format format)
 	case PIPE_FORMAT_R32G32B32A32_UNORM:
 	case PIPE_FORMAT_R32G32B32A32_SINT:
 	case PIPE_FORMAT_R32G32B32A32_UINT:
+
+	case PIPE_FORMAT_R32_UINT:
+	case PIPE_FORMAT_R32_SINT:
+	case PIPE_FORMAT_A32_UINT:
+	case PIPE_FORMAT_A32_SINT:
+	case PIPE_FORMAT_L32_UINT:
+	case PIPE_FORMAT_L32_SINT:
+	case PIPE_FORMAT_I32_UINT:
+	case PIPE_FORMAT_I32_SINT:
+
+	case PIPE_FORMAT_R32G32_SINT:
+	case PIPE_FORMAT_R32G32_UINT:
+	case PIPE_FORMAT_L32A32_UINT:
+	case PIPE_FORMAT_L32A32_SINT:
 */
 
 	default:
@@ -358,8 +356,22 @@ fd3_pipe2swap(enum pipe_format format)
 	switch (format) {
 	case PIPE_FORMAT_B8G8R8A8_UNORM:
 	case PIPE_FORMAT_B8G8R8X8_UNORM:
+	case PIPE_FORMAT_B8G8R8A8_SRGB:
+	case PIPE_FORMAT_B8G8R8X8_SRGB:
 		return WXYZ;
 
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+	case PIPE_FORMAT_A8R8G8B8_SRGB:
+	case PIPE_FORMAT_X8R8G8B8_SRGB:
+		return ZYXW;
+
+	case PIPE_FORMAT_A8B8G8R8_UNORM:
+	case PIPE_FORMAT_X8B8G8R8_UNORM:
+	case PIPE_FORMAT_A8B8G8R8_SRGB:
+	case PIPE_FORMAT_X8B8G8R8_SRGB:
+		return XYZW;
+
 	case PIPE_FORMAT_R8G8B8A8_UNORM:
 	case PIPE_FORMAT_R8G8B8X8_UNORM:
 	case PIPE_FORMAT_Z24X8_UNORM: