qf/quakeforge
0
0
Fork 0
mirror of https://git.code.sf.net/p/quake/quakeforge synced 2024-11-11 07:42:18 +00:00
Code Issues Projects Releases Packages Wiki Activity
quakeforge/libs/gamecode/test/test-unsigned.c

573 lines
20 KiB
C
Raw Normal View History

[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
#include "head.c"
#include "QF/mathlib.h"
static pr_uivec4_t uint_cmpop_init[] = {
{ 5, -5, 5, -5},
{ 5, 5, -5, -5},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
};
static pr_uivec4_t uint_cmpop_expect[] = {
{ 5, -5, 5, -5},
{ 5, 5, -5, -5},
{ 0, 0, 0, 0}, // no unsigned EQ (redundant)
{ 0, 0, -1, 0},
{ 0, -1, 0, 0},
{ 0, 0, 0, 0}, // no unsigned NE (redundant)
{ -1, -1, 0, -1},
{ -1, 0, -1, -1},
};
static dstatement_t uint_cmpop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -1, 32 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_u_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_u_1), 0, 4, 16 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_u_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_u_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -8, 0, 0 },
};
static dstatement_t uint_cmpop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 32 }, // index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 32, -2, 32 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 32 },
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 32, 1 },
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_u_2), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_u_2), 0, 4, 16 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_u_2), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_u_2), 0, 4, 28 },
{ OP(1, 1, 1, OP_JUMP_A), -8, 0, 0 },
};
static dstatement_t uint_cmpop_3a_statements[] = {
// no unsigned EQ (redundant)
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_u_3), 0, 4, 12 },
{ OP(1, 1, 1, OP_LT_u_1), 3, 7, 15 },
{ OP(1, 1, 1, OP_GT_u_3), 0, 4, 16 },
{ OP(1, 1, 1, OP_GT_u_1), 3, 7, 19 },
// no unsigned NE (redundant)
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_u_3), 0, 4, 24 },
{ OP(1, 1, 1, OP_GE_u_1), 3, 7, 27 },
{ OP(1, 1, 1, OP_LE_u_3), 0, 4, 28 },
{ OP(1, 1, 1, OP_LE_u_1), 3, 7, 31 },
};
static dstatement_t uint_cmpop_3b_statements[] = {
// no unsigned EQ (redundant)
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_u_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_LT_u_3), 1, 5, 13 },
{ OP(1, 1, 1, OP_GT_u_1), 0, 4, 16 },
{ OP(1, 1, 1, OP_GT_u_3), 1, 5, 17 },
// no unsigned NE (redundant)
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_u_1), 0, 4, 24 },
{ OP(1, 1, 1, OP_GE_u_3), 1, 5, 25 },
{ OP(1, 1, 1, OP_LE_u_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_LE_u_3), 1, 5, 29 },
};
static dstatement_t uint_cmpop_4_statements[] = {
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_u_4), 0, 4, 12 },
{ OP(1, 1, 1, OP_GT_u_4), 0, 4, 16 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_u_4), 0, 4, 24 },
{ OP(1, 1, 1, OP_LE_u_4), 0, 4, 28 },
};
static pr_ulvec4_t ulong_cmpop_init[] = {
{ 5, -5, 5, -5},
{ 5, 5, -5, -5},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
{ 0, 0, 0, 0},
};
static pr_ulvec4_t ulong_cmpop_expect[] = {
{ 5, -5, 5, -5},
{ 5, 5, -5, -5},
{ 0, 0, 0, 0}, // no unsigned EQ (redundant)
{ 0, 0, -1, 0},
{ 0, -1, 0, 0},
{ 0, 0, 0, 0}, // no unsigned NE (redundant)
{ -1, -1, 0, -1},
{ -1, 0, -1, -1},
};
static dstatement_t ulong_cmpop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 8, 0, 64 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 64, -2, 64 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 64 },
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 64, 1 },
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_U_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_GT_U_1), 0, 8, 32 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_U_1), 0, 8, 48 },
{ OP(1, 1, 1, OP_LE_U_1), 0, 8, 56 },
{ OP(1, 1, 1, OP_JUMP_A), -8, 0, 0 },
};
static dstatement_t ulong_cmpop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 8, 0, 64 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 64, -4, 64 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 64 },
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 64, 1 },
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_U_2), 0, 8, 24 },
{ OP(1, 1, 1, OP_GT_U_2), 0, 8, 32 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_U_2), 0, 8, 48 },
{ OP(1, 1, 1, OP_LE_U_2), 0, 8, 56 },
{ OP(1, 1, 1, OP_JUMP_A), -8, 0, 0 },
};
static dstatement_t ulong_cmpop_3a_statements[] = {
// no unsigned EQ (redundant)
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_U_3), 0, 8, 24 },
{ OP(1, 1, 1, OP_LT_U_1), 6, 14, 30 },
{ OP(1, 1, 1, OP_GT_U_3), 0, 8, 32 },
{ OP(1, 1, 1, OP_GT_U_1), 6, 14, 38 },
// no unsigned NE (redundant)
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_U_3), 0, 8, 48 },
{ OP(1, 1, 1, OP_GE_U_1), 6, 14, 54 },
{ OP(1, 1, 1, OP_LE_U_3), 0, 8, 56 },
{ OP(1, 1, 1, OP_LE_U_1), 6, 14, 62 },
};
static dstatement_t ulong_cmpop_3b_statements[] = {
// no unsigned EQ (redundant)
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_U_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_LT_U_3), 2, 10, 26 },
{ OP(1, 1, 1, OP_GT_U_1), 0, 8, 32 },
{ OP(1, 1, 1, OP_GT_U_3), 2, 10, 34 },
// no unsigned NE (redundant)
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_U_1), 0, 8, 48 },
{ OP(1, 1, 1, OP_GE_U_3), 2, 10, 50 },
{ OP(1, 1, 1, OP_LE_U_1), 0, 8, 56 },
{ OP(1, 1, 1, OP_LE_U_3), 2, 10, 58 },
};
static dstatement_t ulong_cmpop_4_statements[] = {
// no unsigned EQ (redundant)
{ OP(1, 1, 1, OP_LT_U_4), 0, 8, 24 },
{ OP(1, 1, 1, OP_GT_U_4), 0, 8, 32 },
// no unsigned NE (redundant)
{ OP(1, 1, 1, OP_GE_U_4), 0, 8, 48 },
{ OP(1, 1, 1, OP_LE_U_4), 0, 8, 56 },
};
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
static pr_uivec4_t uint_shiftop_init[] = {
{ 0x12345678, 0x9abcdef0, 0x80000001, 0xaaaa5555 },
{ 12, 16, 9, 1 },
{ 20, 16, 23, 31 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
};
static pr_uivec4_t uint_shiftop_expect[] = {
{ 0x12345678, 0x9abcdef0, 0x80000001, 0xaaaa5555 },
{ 12, 16, 9, 1 },//a
{ 20, 16, 23, 31 },//b
{ 0x45678000, 0xdef00000, 0x00000200, 0x5554aaaa },//shl a
{ 0x67800000, 0xdef00000, 0x00800000, 0x80000000 },//shl b
{ 0x00012345, 0x00009abc, 0x00400000, 0x55552aaa },//shr a
{ 0x00000123, 0x00009abc, 0x00000100, 0x00000001 },//shr b
{ 0x00012345, 0xffff9abc, 0xffc00000, 0xd5552aaa },//asr a
{ 0x00000123, 0xffff9abc, 0xffffff00, 0xffffffff },//asr b
};
static dstatement_t uint_shiftop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 36 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 36, -1, 36 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 36 },
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 36, 1 },
{ OP(1, 1, 1, OP_SHL_I_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_SHL_I_1), 0, 8, 16 },
{ OP(1, 1, 1, OP_SHR_u_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_SHR_u_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_ASR_I_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_ASR_I_1), 0, 8, 32 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t uint_shiftop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 4, 0, 36 }, // index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 36, -2, 36 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 36 },
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 36, 1 },
{ OP(1, 1, 1, OP_SHL_I_2), 0, 4, 12 },
{ OP(1, 1, 1, OP_SHL_I_2), 0, 8, 16 },
{ OP(1, 1, 1, OP_SHR_u_2), 0, 4, 20 },
{ OP(1, 1, 1, OP_SHR_u_2), 0, 8, 24 },
{ OP(1, 1, 1, OP_ASR_I_2), 0, 4, 28 },
{ OP(1, 1, 1, OP_ASR_I_2), 0, 8, 32 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t uint_shiftop_3a_statements[] = {
{ OP(1, 1, 1, OP_SHL_I_3), 0, 4, 12 },
{ OP(1, 1, 1, OP_SHL_I_1), 3, 7, 15 },
{ OP(1, 1, 1, OP_SHL_I_3), 0, 8, 16 },
{ OP(1, 1, 1, OP_SHL_I_1), 3, 11, 19 },
{ OP(1, 1, 1, OP_SHR_u_3), 0, 4, 20 },
{ OP(1, 1, 1, OP_SHR_u_1), 3, 7, 23 },
{ OP(1, 1, 1, OP_SHR_u_3), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHR_u_1), 3, 11, 27 },
{ OP(1, 1, 1, OP_ASR_I_3), 0, 4, 28 },
{ OP(1, 1, 1, OP_ASR_I_1), 3, 7, 31 },
{ OP(1, 1, 1, OP_ASR_I_3), 0, 8, 32 },
{ OP(1, 1, 1, OP_ASR_I_1), 3, 11, 35 },
};
static dstatement_t uint_shiftop_3b_statements[] = {
{ OP(1, 1, 1, OP_SHL_I_1), 0, 4, 12 },
{ OP(1, 1, 1, OP_SHL_I_3), 1, 5, 13 },
{ OP(1, 1, 1, OP_SHL_I_1), 0, 8, 16 },
{ OP(1, 1, 1, OP_SHL_I_3), 1, 9, 17 },
{ OP(1, 1, 1, OP_SHR_u_1), 0, 4, 20 },
{ OP(1, 1, 1, OP_SHR_u_3), 1, 5, 21 },
{ OP(1, 1, 1, OP_SHR_u_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHR_u_3), 1, 9, 25 },
{ OP(1, 1, 1, OP_ASR_I_1), 0, 4, 28 },
{ OP(1, 1, 1, OP_ASR_I_3), 1, 5, 29 },
{ OP(1, 1, 1, OP_ASR_I_1), 0, 8, 32 },
{ OP(1, 1, 1, OP_ASR_I_3), 1, 9, 33 },
};
static dstatement_t uint_shiftop_4_statements[] = {
{ OP(1, 1, 1, OP_SHL_I_4), 0, 4, 12 },
{ OP(1, 1, 1, OP_SHL_I_4), 0, 8, 16 },
{ OP(1, 1, 1, OP_SHR_u_4), 0, 4, 20 },
{ OP(1, 1, 1, OP_SHR_u_4), 0, 8, 24 },
{ OP(1, 1, 1, OP_ASR_I_4), 0, 4, 28 },
{ OP(1, 1, 1, OP_ASR_I_4), 0, 8, 32 },
};
static pr_ulvec4_t ulong_shiftop_init[] = {
{ UINT64_C(0x123456789abcdef0), UINT64_C(0x9abcdef012345678),
UINT64_C(0x8000000180000001), UINT64_C(0xaaaa5555aaaa5555) },
{ 12, 16, 9, 1 },
{ 52, 48, 55, 63 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
};
static pr_ulvec4_t ulong_shiftop_expect[] = {
{ UINT64_C(0x123456789abcdef0), UINT64_C(0x9abcdef012345678),
UINT64_C(0x8000000180000001), UINT64_C(0xaaaa5555aaaa5555) },
{ 12, 16, 9, 1 },//a
{ 52, 48, 55, 63 },//b
{ UINT64_C(0x456789abcdef0000), UINT64_C(0xdef0123456780000),
UINT64_C(0x0000030000000200), UINT64_C(0x5554aaab5554aaaa) },//shl a
{ UINT64_C(0xef00000000000000), UINT64_C(0x5678000000000000),
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
UINT64_C(0x0080000000000000), UINT64_C(0x8000000000000000) },//shl b
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{ UINT64_C(0x000123456789abcd), UINT64_C(0x00009abcdef01234),
UINT64_C(0x0040000000c00000), UINT64_C(0x55552aaad5552aaa) },//shr a
{ UINT64_C(0x0000000000000123), UINT64_C(0x0000000000009abc),
UINT64_C(0x0000000000000100), UINT64_C(0x0000000000000001) },//shr b
{ UINT64_C(0x000123456789abcd), UINT64_C(0xffff9abcdef01234),
UINT64_C(0xffc0000000c00000), UINT64_C(0xd5552aaad5552aaa) },//asr a
{ UINT64_C(0x0000000000000123), UINT64_C(0xffffffffffff9abc),
UINT64_C(0xffffffffffffff00), UINT64_C(0xffffffffffffffff) },//asr b
};
static dstatement_t ulong_shiftop_1_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 8, 0, 72 }, // init index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 72, -2, 72 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 72 },
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 72, 1 },
{ OP(1, 1, 1, OP_SHL_L_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHL_L_1), 0, 16, 32 },
{ OP(1, 1, 1, OP_SHR_U_1), 0, 8, 40 },
{ OP(1, 1, 1, OP_SHR_U_1), 0, 16, 48 },
{ OP(1, 1, 1, OP_ASR_L_1), 0, 8, 56 },
{ OP(1, 1, 1, OP_ASR_L_1), 0, 16, 64 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t ulong_shiftop_2_statements[] = {
{ OP(0, 0, 0, OP_LEA_A), 8, 0, 72 }, // index
//loop:
{ OP(0, 0, 0, OP_LEA_C), 72, -4, 72 }, // dec index
[gamecode] Rearrange several instructions ANY/ALL/NONE have been temporarily removed until I implement the HOPS (horizontal operations) sub-instructions, which will all both 32-bit and 64-bit operands and several other operations (eg, horizontal add). All the fancy addressing modes for the conditional branch instructions have been permanently removed: I decided the gain was too little for the cost (24 instructions vs 6). JUMP and CALL retain their addressing modes, though. Other instructions have been shuffled around a little to fill most of the holes in the upper block of 256 instructions: just a single small 7-instruction hole. Rearrangements in the actual engine are mostly just to keep the code organized. The only real changes were the various IF statements and dealing with the resulting changes in their addressing.
2022-01-16 05:22:04 +00:00
{ OP(0, 0, 0, OP_IFAE), 2, 0, 72 },
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{ OP(0, 0, 0, OP_BREAK), 0, 0, 0 },
{ OP(0, 0, 0, OP_WITH), 4, 72, 1 },
{ OP(1, 1, 1, OP_SHL_L_2), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHL_L_2), 0, 16, 32 },
{ OP(1, 1, 1, OP_SHR_U_2), 0, 8, 40 },
{ OP(1, 1, 1, OP_SHR_U_2), 0, 16, 48 },
{ OP(1, 1, 1, OP_ASR_L_2), 0, 8, 56 },
{ OP(1, 1, 1, OP_ASR_L_2), 0, 16, 64 },
{ OP(1, 1, 1, OP_JUMP_A), -10, 0, 0 },
};
static dstatement_t ulong_shiftop_3a_statements[] = {
{ OP(1, 1, 1, OP_SHL_L_3), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHL_L_1), 6, 14, 30 },
{ OP(1, 1, 1, OP_SHL_L_3), 0, 16, 32 },
{ OP(1, 1, 1, OP_SHL_L_1), 6, 22, 38 },
{ OP(1, 1, 1, OP_SHR_U_3), 0, 8, 40 },
{ OP(1, 1, 1, OP_SHR_U_1), 6, 14, 46 },
{ OP(1, 1, 1, OP_SHR_U_3), 0, 16, 48 },
{ OP(1, 1, 1, OP_SHR_U_1), 6, 22, 54 },
{ OP(1, 1, 1, OP_ASR_L_3), 0, 8, 56 },
{ OP(1, 1, 1, OP_ASR_L_1), 6, 14, 62 },
{ OP(1, 1, 1, OP_ASR_L_3), 0, 16, 64 },
{ OP(1, 1, 1, OP_ASR_L_1), 6, 22, 70 },
};
static dstatement_t ulong_shiftop_3b_statements[] = {
{ OP(1, 1, 1, OP_SHL_L_1), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHL_L_3), 2, 10, 26 },
{ OP(1, 1, 1, OP_SHL_L_1), 0, 16, 32 },
{ OP(1, 1, 1, OP_SHL_L_3), 2, 18, 34 },
{ OP(1, 1, 1, OP_SHR_U_1), 0, 8, 40 },
{ OP(1, 1, 1, OP_SHR_U_3), 2, 10, 42 },
{ OP(1, 1, 1, OP_SHR_U_1), 0, 16, 48 },
{ OP(1, 1, 1, OP_SHR_U_3), 2, 18, 50 },
{ OP(1, 1, 1, OP_ASR_L_1), 0, 8, 56 },
{ OP(1, 1, 1, OP_ASR_L_3), 2, 10, 58 },
{ OP(1, 1, 1, OP_ASR_L_1), 0, 16, 64 },
{ OP(1, 1, 1, OP_ASR_L_3), 2, 18, 66 },
};
static dstatement_t ulong_shiftop_4_statements[] = {
{ OP(1, 1, 1, OP_SHL_L_4), 0, 8, 24 },
{ OP(1, 1, 1, OP_SHL_L_4), 0, 16, 32 },
{ OP(1, 1, 1, OP_SHR_U_4), 0, 8, 40 },
{ OP(1, 1, 1, OP_SHR_U_4), 0, 16, 48 },
{ OP(1, 1, 1, OP_ASR_L_4), 0, 8, 56 },
{ OP(1, 1, 1, OP_ASR_L_4), 0, 16, 64 },
};
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
test_t tests[] = {
{
.desc = "uint cmpop 1",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_cmpop_init,uint_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (uint_cmpop_1_statements),
.statements = uint_cmpop_1_statements,
.init_globals = (pr_int_t *) uint_cmpop_init,
.expect_globals = (pr_int_t *) uint_cmpop_expect,
},
{
.desc = "uint cmpop 2",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_cmpop_init,uint_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (uint_cmpop_2_statements),
.statements = uint_cmpop_2_statements,
.init_globals = (pr_int_t *) uint_cmpop_init,
.expect_globals = (pr_int_t *) uint_cmpop_expect,
},
{
.desc = "uint cmpop 3a",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_cmpop_init,uint_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (uint_cmpop_3a_statements),
.statements = uint_cmpop_3a_statements,
.init_globals = (pr_int_t *) uint_cmpop_init,
.expect_globals = (pr_int_t *) uint_cmpop_expect,
},
{
.desc = "uint cmpop 3b",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_cmpop_init,uint_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (uint_cmpop_3b_statements),
.statements = uint_cmpop_3b_statements,
.init_globals = (pr_int_t *) uint_cmpop_init,
.expect_globals = (pr_int_t *) uint_cmpop_expect,
},
{
.desc = "uint cmpop 4",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_cmpop_init,uint_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (uint_cmpop_4_statements),
.statements = uint_cmpop_4_statements,
.init_globals = (pr_int_t *) uint_cmpop_init,
.expect_globals = (pr_int_t *) uint_cmpop_expect,
},
{
.desc = "ulong cmpop 1",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_cmpop_init,ulong_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (ulong_cmpop_1_statements),
.statements = ulong_cmpop_1_statements,
.init_globals = (pr_int_t *) ulong_cmpop_init,
.expect_globals = (pr_int_t *) ulong_cmpop_expect,
},
{
.desc = "ulong cmpop 2",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_cmpop_init,ulong_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (ulong_cmpop_2_statements),
.statements = ulong_cmpop_2_statements,
.init_globals = (pr_int_t *) ulong_cmpop_init,
.expect_globals = (pr_int_t *) ulong_cmpop_expect,
},
{
.desc = "ulong cmpop 3a",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_cmpop_init,ulong_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (ulong_cmpop_3a_statements),
.statements = ulong_cmpop_3a_statements,
.init_globals = (pr_int_t *) ulong_cmpop_init,
.expect_globals = (pr_int_t *) ulong_cmpop_expect,
},
{
.desc = "ulong cmpop 3b",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_cmpop_init,ulong_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (ulong_cmpop_3b_statements),
.statements = ulong_cmpop_3b_statements,
.init_globals = (pr_int_t *) ulong_cmpop_init,
.expect_globals = (pr_int_t *) ulong_cmpop_expect,
},
{
.desc = "ulong cmpop 4",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_cmpop_init,ulong_cmpop_expect),
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
.num_statements = num_statements (ulong_cmpop_4_statements),
.statements = ulong_cmpop_4_statements,
.init_globals = (pr_int_t *) ulong_cmpop_init,
.expect_globals = (pr_int_t *) ulong_cmpop_expect,
},
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
{
.desc = "uint shiftop 1",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_shiftop_init,uint_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (uint_shiftop_1_statements),
.statements = uint_shiftop_1_statements,
.init_globals = (pr_int_t *) uint_shiftop_init,
.expect_globals = (pr_int_t *) uint_shiftop_expect,
},
{
.desc = "uint shiftop 2",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_shiftop_init,uint_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (uint_shiftop_2_statements),
.statements = uint_shiftop_2_statements,
.init_globals = (pr_int_t *) uint_shiftop_init,
.expect_globals = (pr_int_t *) uint_shiftop_expect,
},
{
.desc = "uint shiftop 3a",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_shiftop_init,uint_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (uint_shiftop_3a_statements),
.statements = uint_shiftop_3a_statements,
.init_globals = (pr_int_t *) uint_shiftop_init,
.expect_globals = (pr_int_t *) uint_shiftop_expect,
},
{
.desc = "uint shiftop 3b",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_shiftop_init,uint_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (uint_shiftop_3b_statements),
.statements = uint_shiftop_3b_statements,
.init_globals = (pr_int_t *) uint_shiftop_init,
.expect_globals = (pr_int_t *) uint_shiftop_expect,
},
{
.desc = "uint shiftop 4",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(uint_shiftop_init,uint_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (uint_shiftop_4_statements),
.statements = uint_shiftop_4_statements,
.init_globals = (pr_int_t *) uint_shiftop_init,
.expect_globals = (pr_int_t *) uint_shiftop_expect,
},
{
.desc = "ulong shiftop 1",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_shiftop_init,ulong_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (ulong_shiftop_1_statements),
.statements = ulong_shiftop_1_statements,
.init_globals = (pr_int_t *) ulong_shiftop_init,
.expect_globals = (pr_int_t *) ulong_shiftop_expect,
},
{
.desc = "ulong shiftop 2",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_shiftop_init,ulong_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (ulong_shiftop_2_statements),
.statements = ulong_shiftop_2_statements,
.init_globals = (pr_int_t *) ulong_shiftop_init,
.expect_globals = (pr_int_t *) ulong_shiftop_expect,
},
{
.desc = "ulong shiftop 3a",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_shiftop_init,ulong_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (ulong_shiftop_3a_statements),
.statements = ulong_shiftop_3a_statements,
.init_globals = (pr_int_t *) ulong_shiftop_init,
.expect_globals = (pr_int_t *) ulong_shiftop_expect,
},
{
.desc = "ulong shiftop 3b",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_shiftop_init,ulong_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (ulong_shiftop_3b_statements),
.statements = ulong_shiftop_3b_statements,
.init_globals = (pr_int_t *) ulong_shiftop_init,
.expect_globals = (pr_int_t *) ulong_shiftop_expect,
},
{
.desc = "ulong shiftop 4",
.extra_globals = 4 * 1,
[gamecode] Make num_globals() calculate actual globals It calculating only the size of the array (which was often 4 or 8 globals per element) proved to be a pain when I forgot to alter the size for the new scale tests. Fixing the size calculation even found a bug in the shiftop tests.
2022-01-14 10:43:34 +00:00
.num_globals = num_globals(ulong_shiftop_init,ulong_shiftop_expect),
[gamecode] Add tests for shiftops They're in test-unsigned because 2/3 of them are unsigned.
2022-01-12 01:24:59 +00:00
.num_statements = num_statements (ulong_shiftop_4_statements),
.statements = ulong_shiftop_4_statements,
.init_globals = (pr_int_t *) ulong_shiftop_init,
.expect_globals = (pr_int_t *) ulong_shiftop_expect,
},
[gamecode] Add tests for unsigned comparisons And fix the implementation: I had used the wrong macro.
2022-01-11 04:00:54 +00:00
};
#include "main.c"
Reference in a new issue Copy permalink