drm/amdgpu/atom: add support for new mul32 opcodes (v2)

Better precision than the regular mul opcode.

v2: handle big endian properly.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c
index 885d3d3b52138bc961d46df9e44653d02000a02f..474357d019b219f90010200a615a196f17b39655 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/atom.c
+++ b/drivers/gpu/drm/amd/amdgpu/atom.c
@@ -788,6 +788,20 @@ static void atom_op_mul(atom_exec_context *ctx, int *ptr, int arg)
        ctx->ctx->divmul[0] = dst * src;
 }
 
+static void atom_op_mul32(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);
+       val64 = (uint64_t)dst * (uint64_t)src;
+       ctx->ctx->divmul[0] = lower_32_bits(val64);
+       ctx->ctx->divmul[1] = upper_32_bits(val64);
+}
+
 static void atom_op_nop(atom_exec_context *ctx, int *ptr, int arg)
 {
        /* nothing */
@@ -1160,7 +1174,9 @@ static struct {
        atom_op_shr, ATOM_ARG_PLL}, {
        atom_op_shr, ATOM_ARG_MC}, {
        atom_op_debug, 0}, {
-       atom_op_processds, 0},
+       atom_op_processds, 0}, {
+       atom_op_mul32, ATOM_ARG_PS}, {
+       atom_op_mul32, ATOM_ARG_WS},
 };
 
 static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t * params)

diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h
index 09d0f8230708a4ff08947600d6913f9a5f9c5b0d..a940eb3bb512281a549363f84a2114116ff3f416 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/atom.h
+++ b/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            123
+#define ATOM_OP_CNT            125
 #define ATOM_OP_EOT            91
 
 #define ATOM_CASE_MAGIC                0x63