Patchwork [V5,12/14] target-ppc: VSX Stage 4: Add Scalar SP Fused Multiply-Adds

login
register
mail settings
Submitter Tom Musta
Date Jan. 3, 2014, 6:22 p.m.
Message ID <1388773331-787-13-git-send-email-tommusta@gmail.com>
Download mbox | patch
Permalink /patch/306661/
State New
Headers show

Comments

Tom Musta - Jan. 3, 2014, 6:22 p.m.
This patch adds the Single Precision VSX Scalar Fused Multiply-Add
instructions: xsmaddasp, xsmaddmsp, xssubasp, xssubmsp, xsnmaddasp,
xsnmaddmsp, xsnmsubasp, xsnmsubmsp.

The existing VSX_MADD() macro is modified to support rounding of the
intermediate double precision result to single precision.

Signed-off-by: Tom Musta <tommusta@gmail.com>
Reviewed-by: Richard Henderson <address@hidden>
---
V2: Re-implemented per feedback from Richard Henderson.  In order to
avoid double rounding and incorrect results, the operands must be
converted to true single precision values and use the single precision
fused multiply/add routine.

V3: Re-implemented per feedback from Richard Henderson (I did not
fully understand his comment when I implemented V2).

V4: Changed to use helper_frsp (inadvertently re-injected when I used
an earlier patch).  Thanks to Richard Henderson for catching this.

 target-ppc/fpu_helper.c |   82 ++++++++++++++++++++++++++++++----------------
 target-ppc/helper.h     |    8 ++++
 target-ppc/translate.c  |   16 +++++++++
 3 files changed, 77 insertions(+), 29 deletions(-)

Patch

diff --git a/target-ppc/fpu_helper.c b/target-ppc/fpu_helper.c
index 33da462..7e5003a 100644
--- a/target-ppc/fpu_helper.c
+++ b/target-ppc/fpu_helper.c
@@ -2192,7 +2192,7 @@  VSX_TSQRT(xvtsqrtsp, 4, float32, f32, -126, 23)
  *   afrm  - A form (1=A, 0=M)
  *   sfprf - set FPRF
  */
-#define VSX_MADD(op, nels, tp, fld, maddflgs, afrm, sfprf)                    \
+#define VSX_MADD(op, nels, tp, fld, maddflgs, afrm, sfprf, r2sp)              \
 void helper_##op(CPUPPCState *env, uint32_t opcode)                           \
 {                                                                             \
     ppc_vsr_t xt_in, xa, xb, xt_out;                                          \
@@ -2218,8 +2218,18 @@  void helper_##op(CPUPPCState *env, uint32_t opcode)                           \
     for (i = 0; i < nels; i++) {                                              \
         float_status tstat = env->fp_status;                                  \
         set_float_exception_flags(0, &tstat);                                 \
-        xt_out.fld[i] = tp##_muladd(xa.fld[i], b->fld[i], c->fld[i],          \
-                                     maddflgs, &tstat);                       \
+        if (r2sp && (tstat.float_rounding_mode == float_round_nearest_even)) {\
+            /* Avoid double rounding errors by rounding the intermediate */   \
+            /* result to odd.                                            */   \
+            set_float_rounding_mode(float_round_to_zero, &tstat);             \
+            xt_out.fld[i] = tp##_muladd(xa.fld[i], b->fld[i], c->fld[i],      \
+                                       maddflgs, &tstat);                     \
+            xt_out.fld[i] |= (get_float_exception_flags(&tstat) &             \
+                              float_flag_inexact) != 0;                       \
+        } else {                                                              \
+            xt_out.fld[i] = tp##_muladd(xa.fld[i], b->fld[i], c->fld[i],      \
+                                        maddflgs, &tstat);                    \
+        }                                                                     \
         env->fp_status.float_exception_flags |= tstat.float_exception_flags;  \
                                                                               \
         if (unlikely(tstat.float_exception_flags & float_flag_invalid)) {     \
@@ -2242,6 +2252,11 @@  void helper_##op(CPUPPCState *env, uint32_t opcode)                           \
                 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf);     \
             }                                                                 \
         }                                                                     \
+                                                                              \
+        if (r2sp) {                                                           \
+            xt_out.fld[i] = helper_frsp(env, xt_out.fld[i]);                  \
+        }                                                                     \
+                                                                              \
         if (sfprf) {                                                          \
             helper_compute_fprf(env, xt_out.fld[i], sfprf);                   \
         }                                                                     \
@@ -2255,32 +2270,41 @@  void helper_##op(CPUPPCState *env, uint32_t opcode)                           \
 #define NMADD_FLGS float_muladd_negate_result
 #define NMSUB_FLGS (float_muladd_negate_c | float_muladd_negate_result)
 
-VSX_MADD(xsmaddadp, 1, float64, f64, MADD_FLGS, 1, 1)
-VSX_MADD(xsmaddmdp, 1, float64, f64, MADD_FLGS, 0, 1)
-VSX_MADD(xsmsubadp, 1, float64, f64, MSUB_FLGS, 1, 1)
-VSX_MADD(xsmsubmdp, 1, float64, f64, MSUB_FLGS, 0, 1)
-VSX_MADD(xsnmaddadp, 1, float64, f64, NMADD_FLGS, 1, 1)
-VSX_MADD(xsnmaddmdp, 1, float64, f64, NMADD_FLGS, 0, 1)
-VSX_MADD(xsnmsubadp, 1, float64, f64, NMSUB_FLGS, 1, 1)
-VSX_MADD(xsnmsubmdp, 1, float64, f64, NMSUB_FLGS, 0, 1)
-
-VSX_MADD(xvmaddadp, 2, float64, f64, MADD_FLGS, 1, 0)
-VSX_MADD(xvmaddmdp, 2, float64, f64, MADD_FLGS, 0, 0)
-VSX_MADD(xvmsubadp, 2, float64, f64, MSUB_FLGS, 1, 0)
-VSX_MADD(xvmsubmdp, 2, float64, f64, MSUB_FLGS, 0, 0)
-VSX_MADD(xvnmaddadp, 2, float64, f64, NMADD_FLGS, 1, 0)
-VSX_MADD(xvnmaddmdp, 2, float64, f64, NMADD_FLGS, 0, 0)
-VSX_MADD(xvnmsubadp, 2, float64, f64, NMSUB_FLGS, 1, 0)
-VSX_MADD(xvnmsubmdp, 2, float64, f64, NMSUB_FLGS, 0, 0)
-
-VSX_MADD(xvmaddasp, 4, float32, f32, MADD_FLGS, 1, 0)
-VSX_MADD(xvmaddmsp, 4, float32, f32, MADD_FLGS, 0, 0)
-VSX_MADD(xvmsubasp, 4, float32, f32, MSUB_FLGS, 1, 0)
-VSX_MADD(xvmsubmsp, 4, float32, f32, MSUB_FLGS, 0, 0)
-VSX_MADD(xvnmaddasp, 4, float32, f32, NMADD_FLGS, 1, 0)
-VSX_MADD(xvnmaddmsp, 4, float32, f32, NMADD_FLGS, 0, 0)
-VSX_MADD(xvnmsubasp, 4, float32, f32, NMSUB_FLGS, 1, 0)
-VSX_MADD(xvnmsubmsp, 4, float32, f32, NMSUB_FLGS, 0, 0)
+VSX_MADD(xsmaddadp, 1, float64, f64, MADD_FLGS, 1, 1, 0)
+VSX_MADD(xsmaddmdp, 1, float64, f64, MADD_FLGS, 0, 1, 0)
+VSX_MADD(xsmsubadp, 1, float64, f64, MSUB_FLGS, 1, 1, 0)
+VSX_MADD(xsmsubmdp, 1, float64, f64, MSUB_FLGS, 0, 1, 0)
+VSX_MADD(xsnmaddadp, 1, float64, f64, NMADD_FLGS, 1, 1, 0)
+VSX_MADD(xsnmaddmdp, 1, float64, f64, NMADD_FLGS, 0, 1, 0)
+VSX_MADD(xsnmsubadp, 1, float64, f64, NMSUB_FLGS, 1, 1, 0)
+VSX_MADD(xsnmsubmdp, 1, float64, f64, NMSUB_FLGS, 0, 1, 0)
+
+VSX_MADD(xsmaddasp, 1, float64, f64, MADD_FLGS, 1, 1, 1)
+VSX_MADD(xsmaddmsp, 1, float64, f64, MADD_FLGS, 0, 1, 1)
+VSX_MADD(xsmsubasp, 1, float64, f64, MSUB_FLGS, 1, 1, 1)
+VSX_MADD(xsmsubmsp, 1, float64, f64, MSUB_FLGS, 0, 1, 1)
+VSX_MADD(xsnmaddasp, 1, float64, f64, NMADD_FLGS, 1, 1, 1)
+VSX_MADD(xsnmaddmsp, 1, float64, f64, NMADD_FLGS, 0, 1, 1)
+VSX_MADD(xsnmsubasp, 1, float64, f64, NMSUB_FLGS, 1, 1, 1)
+VSX_MADD(xsnmsubmsp, 1, float64, f64, NMSUB_FLGS, 0, 1, 1)
+
+VSX_MADD(xvmaddadp, 2, float64, f64, MADD_FLGS, 1, 0, 0)
+VSX_MADD(xvmaddmdp, 2, float64, f64, MADD_FLGS, 0, 0, 0)
+VSX_MADD(xvmsubadp, 2, float64, f64, MSUB_FLGS, 1, 0, 0)
+VSX_MADD(xvmsubmdp, 2, float64, f64, MSUB_FLGS, 0, 0, 0)
+VSX_MADD(xvnmaddadp, 2, float64, f64, NMADD_FLGS, 1, 0, 0)
+VSX_MADD(xvnmaddmdp, 2, float64, f64, NMADD_FLGS, 0, 0, 0)
+VSX_MADD(xvnmsubadp, 2, float64, f64, NMSUB_FLGS, 1, 0, 0)
+VSX_MADD(xvnmsubmdp, 2, float64, f64, NMSUB_FLGS, 0, 0, 0)
+
+VSX_MADD(xvmaddasp, 4, float32, f32, MADD_FLGS, 1, 0, 0)
+VSX_MADD(xvmaddmsp, 4, float32, f32, MADD_FLGS, 0, 0, 0)
+VSX_MADD(xvmsubasp, 4, float32, f32, MSUB_FLGS, 1, 0, 0)
+VSX_MADD(xvmsubmsp, 4, float32, f32, MSUB_FLGS, 0, 0, 0)
+VSX_MADD(xvnmaddasp, 4, float32, f32, NMADD_FLGS, 1, 0, 0)
+VSX_MADD(xvnmaddmsp, 4, float32, f32, NMADD_FLGS, 0, 0, 0)
+VSX_MADD(xvnmsubasp, 4, float32, f32, NMSUB_FLGS, 1, 0, 0)
+VSX_MADD(xvnmsubmsp, 4, float32, f32, NMSUB_FLGS, 0, 0, 0)
 
 #define VSX_SCALAR_CMP(op, ordered)                                      \
 void helper_##op(CPUPPCState *env, uint32_t opcode)                      \
diff --git a/target-ppc/helper.h b/target-ppc/helper.h
index 84c6ee1..655b670 100644
--- a/target-ppc/helper.h
+++ b/target-ppc/helper.h
@@ -293,6 +293,14 @@  DEF_HELPER_2(xsdivsp, void, env, i32)
 DEF_HELPER_2(xsresp, void, env, i32)
 DEF_HELPER_2(xssqrtsp, void, env, i32)
 DEF_HELPER_2(xsrsqrtesp, void, env, i32)
+DEF_HELPER_2(xsmaddasp, void, env, i32)
+DEF_HELPER_2(xsmaddmsp, void, env, i32)
+DEF_HELPER_2(xsmsubasp, void, env, i32)
+DEF_HELPER_2(xsmsubmsp, void, env, i32)
+DEF_HELPER_2(xsnmaddasp, void, env, i32)
+DEF_HELPER_2(xsnmaddmsp, void, env, i32)
+DEF_HELPER_2(xsnmsubasp, void, env, i32)
+DEF_HELPER_2(xsnmsubmsp, void, env, i32)
 
 DEF_HELPER_2(xvadddp, void, env, i32)
 DEF_HELPER_2(xvsubdp, void, env, i32)
diff --git a/target-ppc/translate.c b/target-ppc/translate.c
index 950c02e..5b68c7e 100644
--- a/target-ppc/translate.c
+++ b/target-ppc/translate.c
@@ -7365,6 +7365,14 @@  GEN_VSX_HELPER_2(xsdivsp, 0x00, 0x03, 0, PPC2_VSX207)
 GEN_VSX_HELPER_2(xsresp, 0x14, 0x01, 0, PPC2_VSX207)
 GEN_VSX_HELPER_2(xssqrtsp, 0x16, 0x00, 0, PPC2_VSX207)
 GEN_VSX_HELPER_2(xsrsqrtesp, 0x14, 0x00, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsmaddasp, 0x04, 0x00, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsmaddmsp, 0x04, 0x01, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsmsubasp, 0x04, 0x02, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsmsubmsp, 0x04, 0x03, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsnmaddasp, 0x04, 0x10, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsnmaddmsp, 0x04, 0x11, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsnmsubasp, 0x04, 0x12, 0, PPC2_VSX207)
+GEN_VSX_HELPER_2(xsnmsubmsp, 0x04, 0x13, 0, PPC2_VSX207)
 
 GEN_VSX_HELPER_2(xvadddp, 0x00, 0x0C, 0, PPC2_VSX)
 GEN_VSX_HELPER_2(xvsubdp, 0x00, 0x0D, 0, PPC2_VSX)
@@ -10179,6 +10187,14 @@  GEN_XX3FORM(xsdivsp, 0x00, 0x03, PPC2_VSX207),
 GEN_XX2FORM(xsresp,  0x14, 0x01, PPC2_VSX207),
 GEN_XX2FORM(xssqrtsp,  0x16, 0x00, PPC2_VSX207),
 GEN_XX2FORM(xsrsqrtesp,  0x14, 0x00, PPC2_VSX207),
+GEN_XX3FORM(xsmaddasp, 0x04, 0x00, PPC2_VSX207),
+GEN_XX3FORM(xsmaddmsp, 0x04, 0x01, PPC2_VSX207),
+GEN_XX3FORM(xsmsubasp, 0x04, 0x02, PPC2_VSX207),
+GEN_XX3FORM(xsmsubmsp, 0x04, 0x03, PPC2_VSX207),
+GEN_XX3FORM(xsnmaddasp, 0x04, 0x10, PPC2_VSX207),
+GEN_XX3FORM(xsnmaddmsp, 0x04, 0x11, PPC2_VSX207),
+GEN_XX3FORM(xsnmsubasp, 0x04, 0x12, PPC2_VSX207),
+GEN_XX3FORM(xsnmsubmsp, 0x04, 0x13, PPC2_VSX207),
 
 GEN_XX3FORM(xvadddp, 0x00, 0x0C, PPC2_VSX),
 GEN_XX3FORM(xvsubdp, 0x00, 0x0D, PPC2_VSX),