drm/amdgpu/atom: add support for new div32 opcodes (v3)

Better precision than the regular div opcode.

v2: drop 64 bit divide
v3: fix op handling.  This actually is a 64 bit divide.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/atom.c

@@ -685,6 +685,27 @@ static void atom_op_div(atom_exec_context *ctx, int *ptr, int arg)
 	}
 }
 
+static void atom_op_div32(atom_exec_context *ctx, int *ptr, int arg)
+{
+	uint64_t val64;
+	uint8_t attr = U8((*ptr)++);
+	uint32_t dst, src;
+	SDEBUG("   src1: ");
+	dst = atom_get_dst(ctx, arg, attr, ptr, NULL, 1);
+	SDEBUG("   src2: ");
+	src = atom_get_src(ctx, attr, ptr);
+	if (src != 0) {
+		val64 = dst;
+		val64 |= ((uint64_t)ctx->ctx->divmul[1]) << 32;
+		do_div(val64, src);
+		ctx->ctx->divmul[0] = lower_32_bits(val64);
+		ctx->ctx->divmul[1] = upper_32_bits(val64);
+	} else {
+		ctx->ctx->divmul[0] = 0;
+		ctx->ctx->divmul[1] = 0;
+	}
+}
+
 static void atom_op_eot(atom_exec_context *ctx, int *ptr, int arg)
 {
 	/* functionally, a nop */
@@ -1176,7 +1197,9 @@ static struct {
 	atom_op_debug, 0}, {
 	atom_op_processds, 0}, {
 	atom_op_mul32, ATOM_ARG_PS}, {
-	atom_op_mul32, ATOM_ARG_WS},
+	atom_op_mul32, ATOM_ARG_WS}, {
+	atom_op_div32, ATOM_ARG_PS}, {
+	atom_op_div32, ATOM_ARG_WS},
 };
 
 static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t * params)

drivers/gpu/drm/amd/amdgpu/atom.h

@@ -60,7 +60,7 @@
 #define ATOM_CT_PS_MASK		0x7F
 #define ATOM_CT_CODE_PTR	6
 
-#define ATOM_OP_CNT		125
+#define ATOM_OP_CNT		127
 #define ATOM_OP_EOT		91
 
 #define ATOM_CASE_MAGIC		0x63