qf/quakeforge
0
0
Fork 0
mirror of https://git.code.sf.net/p/quake/quakeforge synced 2024-11-30 08:00:51 +00:00
Code Issues Projects Releases Packages Wiki Activity

[gamecode] Clean up the conversion instructions

Browse source 
Float bit-ops as well.

Also, add q*v4 and v4*q instructions. There are currently 48 free
opcodes, and I might remove the scale instructions, but they could be
useful as expanding a single float to a vector would take 3 instructions
(copy to temp, swizzle-expand temp, multiply, vs just scale).
This commit is contained in:
Bill Currie 2022-01-03 22:54:34 +09:00
parent b6f9b68434
commit 6de1ba6901
2 changed files with 53 additions and 42 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

60
include/QF/pr_comp.h
View file

@ -484,11 +484,11 @@ typedef enum {
OP_GT_F_1, OP_GT_F_2, OP_GT_F_3, OP_GT_F_4, OP_GT_F_1, OP_GT_F_2, OP_GT_F_3, OP_GT_F_4,
OP_GT_L_1, OP_GT_L_2, OP_GT_L_3, OP_GT_L_4, OP_GT_L_1, OP_GT_L_2, OP_GT_L_3, OP_GT_L_4,
OP_GT_D_1, OP_GT_D_2, OP_GT_D_3, OP_GT_D_4, OP_GT_D_1, OP_GT_D_2, OP_GT_D_3, OP_GT_D_4,
// 0 1011 convert between signed integral and double(XXX how useful as vec?) // 0 1011
OP_CONV_ID_1, OP_CONV_ID_2, OP_CONV_ID_3, OP_CONV_ID_4, OP_spare_1, OP_spare_2, OP_spare_3, OP_spare_4,
OP_CONV_DI_1, OP_CONV_DI_2, OP_CONV_DI_3, OP_CONV_DI_4, OP_spare_5, OP_spare_6, OP_spare_7, OP_spare_8,
OP_CONV_LD_1, OP_CONV_LD_2, OP_CONV_LD_3, OP_CONV_LD_4, OP_spare_9, OP_spare_10, OP_spare_11, OP_spare_12,
OP_CONV_DL_1, OP_CONV_DL_2, OP_CONV_DL_3, OP_CONV_DL_4, OP_spare_13, OP_spare_14, OP_spare_15, OP_spare_16,
// comparison // comparison
// 0 1100 != // 0 1100 !=
OP_NE_I_1, OP_NE_I_2, OP_NE_I_3, OP_NE_I_4, OP_NE_I_1, OP_NE_I_2, OP_NE_I_3, OP_NE_I_4,
@ -505,11 +505,11 @@ typedef enum {
OP_LE_F_1, OP_LE_F_2, OP_LE_F_3, OP_LE_F_4, OP_LE_F_1, OP_LE_F_2, OP_LE_F_3, OP_LE_F_4,
OP_LE_L_1, OP_LE_L_2, OP_LE_L_3, OP_LE_L_4, OP_LE_L_1, OP_LE_L_2, OP_LE_L_3, OP_LE_L_4,
OP_LE_D_1, OP_LE_D_2, OP_LE_D_3, OP_LE_D_4, OP_LE_D_1, OP_LE_D_2, OP_LE_D_3, OP_LE_D_4,
// 0 1111 convert between signed integral sizes (XXX how useful as vec?) // 0 1111
OP_CONV_IL_1, OP_CONV_IL_2, OP_CONV_IL_3, OP_CONV_IL_4, OP_spare_17, OP_spare_18, OP_spare_19, OP_spare_20,
OP_CONV_LI_1, OP_CONV_LI_2, OP_CONV_LI_3, OP_CONV_LI_4, OP_spare_21, OP_spare_22, OP_spare_23, OP_spare_24,
OP_CONV_uU_1, OP_CONV_uU_2, OP_CONV_uU_3, OP_CONV_uU_4, OP_spare_25, OP_spare_26, OP_spare_27, OP_spare_28,
OP_CONV_Uu_1, OP_CONV_Uu_2, OP_CONV_Uu_3, OP_CONV_Uu_4, OP_spare_29, OP_spare_30, OP_spare_31, OP_spare_32,
// 1 0000 c = a * b // 1 0000 c = a * b
OP_MUL_I_1, OP_MUL_I_2, OP_MUL_I_3, OP_MUL_I_4, OP_MUL_I_1, OP_MUL_I_2, OP_MUL_I_3, OP_MUL_I_4,
@ -556,16 +556,16 @@ typedef enum {
OP_BITOR_I_1, OP_BITOR_I_2, OP_BITOR_I_3, OP_BITOR_I_4, OP_BITOR_I_1, OP_BITOR_I_2, OP_BITOR_I_3, OP_BITOR_I_4,
OP_BITXOR_I_1, OP_BITXOR_I_2, OP_BITXOR_I_3, OP_BITXOR_I_4, OP_BITXOR_I_1, OP_BITXOR_I_2, OP_BITXOR_I_3, OP_BITXOR_I_4,
OP_BITNOT_I_1, OP_BITNOT_I_2, OP_BITNOT_I_3, OP_BITNOT_I_4, OP_BITNOT_I_1, OP_BITNOT_I_2, OP_BITNOT_I_3, OP_BITNOT_I_4,
// 1 1001 < unsigned (float logic and bit ops mixed in) // 1 1001 < unsigned with swizzle and scale mixed in
OP_LT_u_1, OP_LT_u_2, OP_LT_u_3, OP_LT_u_4, OP_LT_u_1, OP_LT_u_2, OP_LT_u_3, OP_LT_u_4,
OP_BITAND_F, OP_BITOR_F, OP_BITXOR_F, OP_BITNOT_F, OP_SWIZZLE_F, OP_SCALE_F_2, OP_SCALE_F_3, OP_SCALE_F_4,
OP_LT_U_1, OP_LT_U_2, OP_LT_U_3, OP_LT_U_4, OP_LT_U_1, OP_LT_U_2, OP_LT_U_3, OP_LT_U_4,
OP_AND_F, OP_OR_F, OP_XOR_F, OP_NOT_F, OP_SWIZZLE_D, OP_SCALE_D_2, OP_SCALE_D_3, OP_SCALE_D_4,
// 1 1010 > unsigned // 1 1010 > unsigned and conversions
OP_GT_u_1, OP_GT_u_2, OP_GT_u_3, OP_GT_u_4, OP_GT_u_1, OP_GT_u_2, OP_GT_u_3, OP_GT_u_4,
OP_spare, OP_NOT_D, OP_NOT_V, OP_NOT_Q, OP_CONV_IF_1, OP_CONV_LD_1, OP_CONV_uF_1, OP_CONV_UD_1,
OP_GT_U_1, OP_GT_U_2, OP_GT_U_3, OP_GT_U_4, OP_GT_U_1, OP_GT_U_2, OP_GT_U_3, OP_GT_U_4,
OP_EQ_V, OP_EQ_Q, OP_NE_V, OP_NE_Q, OP_CONV_FI_1, OP_CONV_DL_1, OP_CONV_Fu_1, OP_CONV_DU_1,
// 1 1011 lea, with, etc // 1 1011 lea, with, etc
OP_LEA_A, OP_LEA_B, OP_LEA_C, OP_LEA_D, OP_LEA_A, OP_LEA_B, OP_LEA_C, OP_LEA_D,
OP_LEA_E, OP_ANY_2, OP_ANY_3, OP_ANY_4, OP_LEA_E, OP_ANY_2, OP_ANY_3, OP_ANY_4,
@ -576,21 +576,23 @@ typedef enum {
OP_OR_I_1, OP_OR_I_2, OP_OR_I_3, OP_OR_I_4, OP_OR_I_1, OP_OR_I_2, OP_OR_I_3, OP_OR_I_4,
OP_XOR_I_1, OP_XOR_I_2, OP_XOR_I_3, OP_XOR_I_4, OP_XOR_I_1, OP_XOR_I_2, OP_XOR_I_3, OP_XOR_I_4,
OP_NOT_I_1, OP_NOT_I_2, OP_NOT_I_3, OP_NOT_I_4, OP_NOT_I_1, OP_NOT_I_2, OP_NOT_I_3, OP_NOT_I_4,
// 1 1101 >= unsigned with float shifts and moves mixed in // 1 1101 >= unsigned with q v4 mul and moves mixed in
OP_GE_u_1, OP_GE_u_2, OP_GE_u_3, OP_GE_u_4, OP_GE_u_1, OP_GE_u_2, OP_GE_u_3, OP_GE_u_4,
OP_SHL_F, OP_MOVE_I, OP_MOVE_P, OP_MOVE_PI, OP_MUL_QV4_F, OP_MOVE_I, OP_MOVE_P, OP_MOVE_PI,
OP_GE_U_1, OP_GE_U_2, OP_GE_U_3, OP_GE_U_4, OP_GE_U_1, OP_GE_U_2, OP_GE_U_3, OP_GE_U_4,
OP_SHR_F, OP_MEMSET_I, OP_MEMSET_P, OP_MEMSET_PI, OP_MUL_QV4_D, OP_MEMSET_I, OP_MEMSET_P, OP_MEMSET_PI,
// 1 1110 <= unsigned with scale and swizzle mixed in // 1 1110 <= unsigned with v4 q mul and conversion mixed in
OP_LE_u_1, OP_LE_u_2, OP_LE_u_3, OP_LE_u_4, OP_LE_u_1, OP_LE_u_2, OP_LE_u_3, OP_LE_u_4,
OP_SWIZZLE_F, OP_SCALE_F_2, OP_SCALE_F_3, OP_SCALE_F_4, OP_MUL_V4Q_F, OP_CONV_IL_1, OP_CONV_uU_1, OP_CONV_FD_1,
OP_LE_U_1, OP_LE_U_2, OP_LE_U_3, OP_LE_U_4, OP_LE_U_1, OP_LE_U_2, OP_LE_U_3, OP_LE_U_4,
OP_SWIZZLE_D, OP_SCALE_D_2, OP_SCALE_D_3, OP_SCALE_D_4, OP_MUL_V4Q_D, OP_CONV_LI_1, OP_CONV_Uu_1, OP_CONV_DF_1,
// 1 1111 convert between integral and float (XXX how useful as vec?) // 1 1111
OP_CONV_IF_1, OP_CONV_IF_2, OP_CONV_IF_3, OP_CONV_IF_4, OP_spare_33, OP_spare_34, OP_spare_35, OP_spare_36,
OP_CONV_FI_1, OP_CONV_FI_2, OP_CONV_FI_3, OP_CONV_FI_4, OP_spare_37, OP_spare_38, OP_spare_39, OP_spare_40,
OP_CONV_FD_1, OP_CONV_FD_2, OP_CONV_FD_3, OP_CONV_FD_4, OP_spare_41, OP_spare_42, OP_spare_43, OP_spare_44,
OP_CONV_DF_1, OP_CONV_DF_2, OP_CONV_DF_3, OP_CONV_DF_4, OP_spare_45, OP_spare_46, OP_spare_47, OP_spare_48,
} pr_opcode_e; } pr_opcode_e;
#define OP_A_SHIFT (9) #define OP_A_SHIFT (9)
#define OP_B_SHIFT (11) #define OP_B_SHIFT (11)

35
libs/gamecode/pr_exec.c
View file

@ -3228,14 +3228,20 @@ pr_exec_ruamoko (progs_t *pr, int exitdepth)
OP_uop_T (BITNOT, I, int, ivec2, ivec4, ~); OP_uop_T (BITNOT, I, int, ivec2, ivec4, ~);
// 1 1001 // 1 1001
OP_op_T (LT, u, uint, uivec2, uivec4, <); OP_op_T (LT, u, uint, uivec2, uivec4, <);
//FIXME float ops case OP_SWIZZLE_F:
OPC(ivec4) = pr_swizzle_f (OPA(ivec4), st->b);
break;
//FIXME scale ops
OP_op_T (LT, U, ulong, ulvec2, ulvec4, <); OP_op_T (LT, U, ulong, ulvec2, ulvec4, <);
//FIXME float ops case OP_SWIZZLE_D:
OPC(lvec4) = pr_swizzle_d (OPA(lvec4), st->b);
break;
//FIXME scale ops
// 1 1010 // 1 1010
OP_op_T (GT, u, uint, uivec2, uivec4, >); OP_op_T (GT, u, uint, uivec2, uivec4, >);
//FIXME misc ops //FIXME conversion ops
OP_op_T (GT, U, ulong, ulvec2, ulvec4, >); OP_op_T (GT, U, ulong, ulvec2, ulvec4, >);
//FIXME misc ops //FIXME conversion ops
// 1 1011 // 1 1011
case OP_LEA_A: case OP_LEA_A:
case OP_LEA_B: case OP_LEA_B:
@ -3337,7 +3343,9 @@ pr_exec_ruamoko (progs_t *pr, int exitdepth)
OP_not_n (NOT, ivec4, 4, +); OP_not_n (NOT, ivec4, 4, +);
// 1 1101 // 1 1101
OP_op_T (GE, u, uint, uivec2, uivec4, >=); OP_op_T (GE, u, uint, uivec2, uivec4, >=);
//FIXME float shift case OP_MUL_QV4_F:
OPC(vec4) = qvmulf (OPA(vec4), OPB(vec4));
break;
case OP_MOVE_I: case OP_MOVE_I:
memmove (op_c, op_a, st->b * sizeof (pr_type_t)); memmove (op_c, op_a, st->b * sizeof (pr_type_t));
break; break;
@ -3350,7 +3358,9 @@ pr_exec_ruamoko (progs_t *pr, int exitdepth)
st->b * sizeof (pr_type_t)); st->b * sizeof (pr_type_t));
break; break;
OP_op_T (GE, U, ulong, ulvec2, ulvec4, >=); OP_op_T (GE, U, ulong, ulvec2, ulvec4, >=);
//FIXME float shift case OP_MUL_QV4_D:
OPC(dvec4) = qvmuld (OPA(dvec4), OPB(dvec4));
break;
case OP_MEMSET_I: case OP_MEMSET_I:
pr_memset (op_c, OPA(int), st->b); pr_memset (op_c, OPA(int), st->b);
break; break;
@ -3362,17 +3372,16 @@ pr_exec_ruamoko (progs_t *pr, int exitdepth)
break; break;
// 1 1110 // 1 1110
OP_op_T (LE, u, uint, uivec2, uivec4, <=); OP_op_T (LE, u, uint, uivec2, uivec4, <=);
case OP_SWIZZLE_F: case OP_MUL_V4Q_F:
OPC(ivec4) = pr_swizzle_f (OPA(ivec4), st->b); OPC(vec4) = vqmulf (OPA(vec4), OPB(vec4));
break; break;
//FIXME misc ops
OP_op_T (LE, U, ulong, ulvec2, ulvec4, <=); OP_op_T (LE, U, ulong, ulvec2, ulvec4, <=);
case OP_SWIZZLE_D: case OP_MUL_V4Q_D:
OPC(lvec4) = pr_swizzle_d (OPA(lvec4), st->b); OPC(dvec4) = vqmuld (OPA(dvec4), OPB(dvec4));
break; break;
//FIXME misc ops
// 1 1111 // 1 1111
//FIXME conversion 3
default: default:
PR_RunError (pr, "Bad opcode o%03o", st->op & OP_MASK); PR_RunError (pr, "Bad opcode o%03o", st->op & OP_MASK);