X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=isa%2Fmacros%2Fscalar%2Ftest_macros.h;h=58e389fe36386a4f4d52ad595668768565e285ce;hb=49f478416cec9f73801fdc5dedf9955494be8c66;hp=11645fee5b283b969efb0a5fbb1ebd71259ba426;hpb=acc35f32c788fdcbb854f8fb6d74026050ced273;p=riscv-tests.git diff --git a/isa/macros/scalar/test_macros.h b/isa/macros/scalar/test_macros.h index 11645fe..58e389f 100644 --- a/isa/macros/scalar/test_macros.h +++ b/isa/macros/scalar/test_macros.h @@ -44,9 +44,9 @@ test_ ## testnum: \ #define SEXT_IMM(x) ((x) | (-(((x) >> 11) & 1) << 11)) #define TEST_IMM_OP( testnum, inst, result, val1, imm ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x1, MASK_XLEN(val1); \ - inst x3, x1, SEXT_IMM(imm); \ + inst x30, x1, SEXT_IMM(imm); \ ) #define TEST_IMM_SRC1_EQ_DEST( testnum, inst, result, val1, imm ) \ @@ -59,20 +59,20 @@ test_ ## testnum: \ TEST_CASE( testnum, x6, result, \ li x4, 0; \ 1: li x1, MASK_XLEN(val1); \ - inst x3, x1, SEXT_IMM(imm); \ + inst x30, x1, SEXT_IMM(imm); \ TEST_INSERT_NOPS_ ## nop_cycles \ - addi x6, x3, 0; \ + addi x6, x30, 0; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ ) #define TEST_IMM_SRC1_BYPASS( testnum, nop_cycles, inst, result, val1, imm ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x4, 0; \ 1: li x1, MASK_XLEN(val1); \ TEST_INSERT_NOPS_ ## nop_cycles \ - inst x3, x1, SEXT_IMM(imm); \ + inst x30, x1, SEXT_IMM(imm); \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -94,9 +94,9 @@ test_ ## testnum: \ #----------------------------------------------------------------------- #define TEST_R_OP( testnum, inst, result, val1 ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x1, val1; \ - inst x3, x1; \ + inst x30, x1; \ ) #define TEST_R_SRC1_EQ_DEST( testnum, inst, result, val1 ) \ @@ -109,9 +109,9 @@ test_ ## testnum: \ TEST_CASE( testnum, x6, result, \ li x4, 0; \ 1: li x1, val1; \ - inst x3, x1; \ + inst x30, x1; \ TEST_INSERT_NOPS_ ## nop_cycles \ - addi x6, x3, 0; \ + addi x6, x30, 0; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -122,10 +122,10 @@ test_ ## testnum: \ #----------------------------------------------------------------------- #define TEST_RR_OP( testnum, inst, result, val1, val2 ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x1, MASK_XLEN(val1); \ li x2, MASK_XLEN(val2); \ - inst x3, x1, x2; \ + inst x30, x1, x2; \ ) #define TEST_RR_SRC1_EQ_DEST( testnum, inst, result, val1, val2 ) \ @@ -153,35 +153,35 @@ test_ ## testnum: \ li x4, 0; \ 1: li x1, MASK_XLEN(val1); \ li x2, MASK_XLEN(val2); \ - inst x3, x1, x2; \ + inst x30, x1, x2; \ TEST_INSERT_NOPS_ ## nop_cycles \ - addi x6, x3, 0; \ + addi x6, x30, 0; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ ) #define TEST_RR_SRC12_BYPASS( testnum, src1_nops, src2_nops, inst, result, val1, val2 ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x4, 0; \ 1: li x1, MASK_XLEN(val1); \ TEST_INSERT_NOPS_ ## src1_nops \ li x2, MASK_XLEN(val2); \ TEST_INSERT_NOPS_ ## src2_nops \ - inst x3, x1, x2; \ + inst x30, x1, x2; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ ) #define TEST_RR_SRC21_BYPASS( testnum, src1_nops, src2_nops, inst, result, val1, val2 ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ li x4, 0; \ 1: li x2, MASK_XLEN(val2); \ TEST_INSERT_NOPS_ ## src1_nops \ li x1, MASK_XLEN(val1); \ TEST_INSERT_NOPS_ ## src2_nops \ - inst x3, x1, x2; \ + inst x30, x1, x2; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -216,17 +216,17 @@ test_ ## testnum: \ #----------------------------------------------------------------------- #define TEST_LD_OP( testnum, inst, result, offset, base ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ la x1, base; \ - inst x3, offset(x1); \ + inst x30, offset(x1); \ ) #define TEST_ST_OP( testnum, load_inst, store_inst, result, offset, base ) \ - TEST_CASE( testnum, x3, result, \ + TEST_CASE( testnum, x30, result, \ la x1, base; \ li x2, result; \ store_inst x2, offset(x1); \ - load_inst x3, offset(x1); \ + load_inst x30, offset(x1); \ ) #define TEST_LD_DEST_BYPASS( testnum, nop_cycles, inst, result, offset, base ) \ @@ -234,9 +234,9 @@ test_ ## testnum: \ li TESTNUM, testnum; \ li x4, 0; \ 1: la x1, base; \ - inst x3, offset(x1); \ + inst x30, offset(x1); \ TEST_INSERT_NOPS_ ## nop_cycles \ - addi x6, x3, 0; \ + addi x6, x30, 0; \ li x29, result; \ bne x6, x29, fail; \ addi x4, x4, 1; \ @@ -249,9 +249,9 @@ test_ ## testnum: \ li x4, 0; \ 1: la x1, base; \ TEST_INSERT_NOPS_ ## nop_cycles \ - inst x3, offset(x1); \ + inst x30, offset(x1); \ li x29, result; \ - bne x3, x29, fail; \ + bne x30, x29, fail; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -265,9 +265,9 @@ test_ ## testnum: \ la x2, base; \ TEST_INSERT_NOPS_ ## src2_nops \ store_inst x1, offset(x2); \ - load_inst x3, offset(x2); \ + load_inst x30, offset(x2); \ li x29, result; \ - bne x3, x29, fail; \ + bne x30, x29, fail; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -281,9 +281,9 @@ test_ ## testnum: \ li x1, result; \ TEST_INSERT_NOPS_ ## src2_nops \ store_inst x1, offset(x2); \ - load_inst x3, offset(x2); \ + load_inst x30, offset(x2); \ li x29, result; \ - bne x3, x29, fail; \ + bne x30, x29, fail; \ addi x4, x4, 1; \ li x5, 2; \ bne x4, x5, 1b \ @@ -374,6 +374,11 @@ test_ ## testnum: \ # Tests floating-point instructions #----------------------------------------------------------------------- +#define qNaNf 0f:7fc00000 +#define sNaNf 0f:7f800001 +#define qNaN 0d:7ff8000000000000 +#define sNaN 0d:7ff0000000000001 + #define TEST_FP_OP_S_INTERNAL( testnum, flags, result, val1, val2, val3, code... ) \ test_ ## testnum: \ li TESTNUM, testnum; \ @@ -418,6 +423,35 @@ test_ ## testnum: \ .result; \ .popsection +// TODO: assign a separate mem location for the comparison address? +#define TEST_FP_OP_D32_INTERNAL( testnum, flags, result, val1, val2, val3, code... ) \ +test_ ## testnum: \ + li TESTNUM, testnum; \ + la a0, test_ ## testnum ## _data ;\ + fld f0, 0(a0); \ + fld f1, 8(a0); \ + fld f2, 16(a0); \ + lw a3, 24(a0); \ + lw t1, 28(a0); \ + code; \ + fsflags a1, x0; \ + li a2, flags; \ + bne a0, a3, fail; \ + bne t1, t2, fail; \ + bne a1, a2, fail; \ + .pushsection .data; \ + .align 3; \ + test_ ## testnum ## _data: \ + .double val1; \ + .double val2; \ + .double val3; \ + .result; \ + .popsection + +#define TEST_FCVT_S_D32( testnum, result, val1 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, 0, double result, val1, 0.0, 0.0, \ + fcvt.s.d f3, f0; fcvt.d.s f3, f3; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0)) + #define TEST_FCVT_S_D( testnum, result, val1 ) \ TEST_FP_OP_D_INTERNAL( testnum, 0, double result, val1, 0.0, 0.0, \ fcvt.s.d f3, f0; fcvt.d.s f3, f3; fmv.x.d a0, f3) @@ -430,6 +464,11 @@ test_ ## testnum: \ TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, 0.0, 0.0, \ inst f3, f0; fmv.x.s a0, f3) +#define TEST_FP_OP1_D32( testnum, inst, flags, result, val1 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, 0.0, 0.0, \ + inst f3, f0; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0)) +// ^: store computation result in address from a0, load high-word into t2 + #define TEST_FP_OP1_D( testnum, inst, flags, result, val1 ) \ TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, 0.0, 0.0, \ inst f3, f0; fmv.x.d a0, f3) @@ -438,6 +477,11 @@ test_ ## testnum: \ TEST_FP_OP_S_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \ inst f3, f0; fmv.x.s a0, f3) +#define TEST_FP_OP1_D32_DWORD_RESULT( testnum, inst, flags, result, val1 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \ + inst f3, f0; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0)) +// ^: store computation result in address from a0, load high-word into t2 + #define TEST_FP_OP1_D_DWORD_RESULT( testnum, inst, flags, result, val1 ) \ TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \ inst f3, f0; fmv.x.d a0, f3) @@ -446,6 +490,11 @@ test_ ## testnum: \ TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, val2, 0.0, \ inst f3, f0, f1; fmv.x.s a0, f3) +#define TEST_FP_OP2_D32( testnum, inst, flags, result, val1, val2 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, val2, 0.0, \ + inst f3, f0, f1; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0)) +// ^: store computation result in address from a0, load high-word into t2 + #define TEST_FP_OP2_D( testnum, inst, flags, result, val1, val2 ) \ TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, val2, 0.0, \ inst f3, f0, f1; fmv.x.d a0, f3) @@ -454,6 +503,11 @@ test_ ## testnum: \ TEST_FP_OP_S_INTERNAL( testnum, flags, float result, val1, val2, val3, \ inst f3, f0, f1, f2; fmv.x.s a0, f3) +#define TEST_FP_OP3_D32( testnum, inst, flags, result, val1, val2, val3 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, double result, val1, val2, val3, \ + inst f3, f0, f1, f2; fsd f3, 0(a0); lw t2, 4(a0); lw a0, 0(a0)) +// ^: store computation result in address from a0, load high-word into t2 + #define TEST_FP_OP3_D( testnum, inst, flags, result, val1, val2, val3 ) \ TEST_FP_OP_D_INTERNAL( testnum, flags, double result, val1, val2, val3, \ inst f3, f0, f1, f2; fmv.x.d a0, f3) @@ -462,22 +516,41 @@ test_ ## testnum: \ TEST_FP_OP_S_INTERNAL( testnum, flags, word result, val1, 0.0, 0.0, \ inst a0, f0, rm) +#define TEST_FP_INT_OP_D32( testnum, inst, flags, result, val1, rm ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \ + inst a0, f0, f1; li t2, 0) + #define TEST_FP_INT_OP_D( testnum, inst, flags, result, val1, rm ) \ TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, 0.0, 0.0, \ inst a0, f0, rm) -#define TEST_FP_CMP_OP_S( testnum, inst, result, val1, val2 ) \ - TEST_FP_OP_S_INTERNAL( testnum, 0, word result, val1, val2, 0.0, \ +#define TEST_FP_CMP_OP_S( testnum, inst, flags, result, val1, val2 ) \ + TEST_FP_OP_S_INTERNAL( testnum, flags, word result, val1, val2, 0.0, \ inst a0, f0, f1) -#define TEST_FP_CMP_OP_D( testnum, inst, result, val1, val2 ) \ - TEST_FP_OP_D_INTERNAL( testnum, 0, dword result, val1, val2, 0.0, \ +#define TEST_FP_CMP_OP_D32( testnum, inst, flags, result, val1, val2 ) \ + TEST_FP_OP_D32_INTERNAL( testnum, flags, dword result, val1, val2, 0.0, \ + inst a0, f0, f1; li t2, 0) + +#define TEST_FP_CMP_OP_D( testnum, inst, flags, result, val1, val2 ) \ + TEST_FP_OP_D_INTERNAL( testnum, flags, dword result, val1, val2, 0.0, \ inst a0, f0, f1) #define TEST_FCLASS_S(testnum, correct, input) \ TEST_CASE(testnum, a0, correct, li a0, input; fmv.s.x fa0, a0; \ fclass.s a0, fa0) +#define TEST_FCLASS_D32(testnum, correct, input) \ + TEST_CASE(testnum, a0, correct, \ + la a0, test_ ## testnum ## _data ;\ + fld fa0, 0(a0); \ + fclass.d a0, fa0) \ + .pushsection .data; \ + .align 3; \ + test_ ## testnum ## _data: \ + .dword input; \ + .popsection + #define TEST_FCLASS_D(testnum, correct, input) \ TEST_CASE(testnum, a0, correct, li a0, input; fmv.d.x fa0, a0; \ fclass.d a0, fa0) @@ -498,6 +571,28 @@ test_ ## testnum: \ .float result; \ .popsection +#define TEST_INT_FP_OP_D32( testnum, inst, result, val1 ) \ +test_ ## testnum: \ + li TESTNUM, testnum; \ + la a0, test_ ## testnum ## _data ;\ + lw a3, 0(a0); \ + lw a4, 4(a0); \ + li a1, val1; \ + inst f0, a1; \ + \ + fsd f0, 0(a0); \ + lw a1, 4(a0); \ + lw a0, 0(a0); \ + \ + fsflags x0; \ + bne a0, a3, fail; \ + bne a1, a4, fail; \ + .pushsection .data; \ + .align 3; \ + test_ ## testnum ## _data: \ + .double result; \ + .popsection + #define TEST_INT_FP_OP_D( testnum, inst, result, val1 ) \ test_ ## testnum: \ li TESTNUM, testnum; \ @@ -514,6 +609,25 @@ test_ ## testnum: \ .double result; \ .popsection +// We need some special handling here to allow 64-bit comparison in 32-bit arch +// TODO: find a better name and clean up when intended for general usage? +#define TEST_CASE_D32( testnum, testreg1, testreg2, correctval, code... ) \ +test_ ## testnum: \ + code; \ + la x31, test_ ## testnum ## _data ; \ + lw x29, 0(x31); \ + lw x31, 4(x31); \ + li TESTNUM, testnum; \ + bne testreg1, x29, fail;\ + bne testreg2, x31, fail;\ + .pushsection .data; \ + .align 3; \ + test_ ## testnum ## _data: \ + .dword correctval; \ + .popsection + +// ^ x30 is used in some other macros, to avoid issues we use x31 for upper word + #----------------------------------------------------------------------- # Pass and fail code (assumes test num is in TESTNUM) #-----------------------------------------------------------------------